Fundamentals of .Sfrill
vt,

A. T. WELFORD
M.4., Sc.D.
Professor of Psychology in the University
of Adelaide
Formerly Fellow of St lohn's College, Cambridge
and University Lecturer in Experimental Psychology

METHUEN & CO LTD
r r New Fetter Lane London EC+

First published 1968 by Methuen and Co Ltd

e

1968 A. T, Welfurd
Printed in Ctreat Britain by

Butler & Tanner Ltd
Frome and London

SBN 416 o3ooo 9

Distributed in the USA
by Barutes €* Noble fnc.

Contents

I

Preface

9

Introdustion

II

Sensory-motor shill and tlu cyberrutic apfroach
Signal dctection, vigilartrn afid dccision
Mental skill and conceptuctl frameworks
The present research position

II Simple Decisions

27

lYebef s Law, threslwlds and, disdmirrution

Tlw dccision-tlrcory approach
The tirne required for discrimination

III Identification and Choice

60

Hick' s inforuation-theory law
Conceptual modcls

Factors affecting thc extent to sthich reaction-time rises
ssith number of alternatioes
Separating idmtifuation from choice

Fwtlw studies of prceptual selcction

Mingle-Channel Operation
The effect of a signal during the reaction-timc to a praious

signal

Tfu effect of a signal dur@ tlrc movement made in response
to a preuious signal
Sorne arnmalous resuhs
Altematizte tlwories
The speed of continuous prformance

Meast ing'mefltal load'
,l

5

ro5

Cmtutts

6

V Movement

r37

Smsory control
Time required for feedback to be effectioe
Relationships between speed and accurecy

VI Economy of Decision

r6r

Economy in prception
Recoding and translation

Higher units of perfotmance

VII

Short-Term Retention

r97

Basic facts of immediate memory
Sources of lhnitation
Measuring thc capacity of thc store
Locating the store
Short-tqm retmtion as a factor in complex performance

VIII Effeas of Loading

240^

N eur ornrs cular f atigue

Mental fatigtrc
Srress,

ffirt and arousal

Vigilance

'Fatigue' at work

IX Acquisition of Sk'rll

286

Sources of limitation in learning and recall
Kruwledge of thc results of acdon

Airns and, incentioes
Clwges of prformance during long practice

X Individuals and Social Groups
Personality

Motives
Social relationships and behaoiow

3r7

Contmts

7

References

337

Appendix

3e5

Tables:

r. Crossmant's Confwiott Ftmction
2. Normal daliates and udinaus to calanlate d' aild B
3. Logrfu wlnle rutmbers up to r@
,)
4. p tos,i *u ? tosg(i

*

5. Calaiation of single-chmnel fficts in paced tasks

Index of names

4o,9

Index of subiects

42r

Acknowledgements

The author would also like to thank all authors and publishers who have
given permission to quote copyright material from iournals and books:
Amqican Joumal of Prychology, British Joumal of Psychology, Butterworttr & Co. Ltd., Ergonomics, Ergonomics Research Soci.ty, H.M.S.O.,

Jourual of Experiruntal Psychology, Mind, Nature, Phil'ips Technical
Raian, Psyclwhsy Reuieut, Quarterly Journal of Experimental Psychology, Soc. Scl. Fennica Commentationes Humonorum Littqarum,
Iohn Wiley and Sons, fnc.

Preface

Studies of hunan performance made during the past twenty-five years
have profoundly affected thought and practice in experimental psycholory both in the laboratory and in many fields of application. The
present volume attempts a broad survey and appraisal of the main
ideas which have emerged from this work. In writing it, three points
especially have been borne in mind: firstlp although the research
surveyed did not aim at establishing a'grand theory' of human behaviour, certain principles seem to recur in several contexts so that
different facets of performance can be linked together more closely than
they have been hitherto. Secondly, much of the work has involved
mathematical treatments which have been clifficult to master for those
whose mathematical education, like the author's, stopped at the age of
flflteen or sixteen. It was felt to be essential to discuss some of these
treatments, but an effort has been made to do so in a way that requires
no mathematical knowledge beyond O-level in the British General
Certificate of Education together with an elementary course on statistics
as usually taken by psychology srudents. Thirdlp many of the sflrdies
surveyed have been of interest not only to psychologists but also to
those engaged in various branches of operational research and industrial
work sttrdy, for whom psychological terms are often a barrier to

understanding. Technical terminology has therefore been kept, as
far as possible, within the bounds of that which is courmon to any
broad scientifis tlaining.
The amount of work published in the areas with which this book is
concerned has increased very greatly during the last decade, snd Se
task of surveyrng it has proved a formidable orre. I am well aware that
much has been omitted and that some treatments will be regarded as
controversial. f hope that anyone whose work has been inadequately
represented will accept my apologies, and will get in touch with me to

point out my shortcomings.
This book owes a great deal to discussion with colleagues too
numerous to mention. Especial thanks are, however, due to C. G.
Cameron, E. R. F. W'. Crossman, B. Fowler, D. A. Kassum, A. A.
Knight, D. McNicol, f. Ryan and D. Vickers for permission to cite
9

ro

Preface

unpublished data for the first time, and to W'. E. Hick for help in
checking the derivation of Table A5. Thanks are also due to JVIrs A. K.
Copeman for her unusually competent translation of manuscript into
typescript. Drafting was begun in 1964 while I was.holdi"g a Commonwealth Visiting Professorship at the University of Adelaide, and the
book was finished shortly before returning there from Cambridge in
a more permanent capacity.

A. T. WELFORD
St ]ohn's College,
Cambridge
z4 November ry62

I
Introduction

The scientific study of skill may perhaps be taken as dating from r8zo
when the asuonomer Bessel began to examine the differences between
his colleagues and himself in the recordirg of star-uansit times. The
observer had to note the time on a clock accurately to a second and
then to count seconds by the ticks of the pendulum while he watched
the star cross the graticule line of a telescope, estimating the time of
crossing to the nearest tenth of a second. It was a complex task and it
is not surprising that errors were made or that some observers were more

skilful than others in avoiding them, so that substantial differences
appeared when one was compared \rith another. Important pioneering

studies of various aspects of skilled performance continued through the

ninglgqrth century and in the early years of the present one. Every
sntdent of psychology has heard of the psychologrct Bryan and his
engineer collaborator Harter whose sttrdies (rBgT, lSgg) of learning
Morse code still make interesting readi.g. The researches on movement
by Woodworth (t8gg), the mathematical teacher who turned psychologist, are still ftrndamental. The intensive work by Book (lgo8) on

tlpewriting although less known is nevertheless classical.
During the r92os and r93os this type of snrdy was swamped by other
lines of interest. The work on skill that continued was usudly done
outside psychological laboratories as a part of industrial 'work srudy'.
Its separation from the main stream of psychology was unfornrnate,
leaving work-snrdy without theoretical foundations and depriving experimental psychology of some stimulating lines of research.
The Second World Var, however, brought a resurgence of interest.
just as in Bessel's era perfection of the chronoscope and the improvement it brought in the measruement of time foctrsed attention on htrman
errors of observation, so in the early r94os technical developments cf
weapons, high-speed aircraft, radar and other devices reached a point
at which the limitations of man and machine working together were no
longer mainly in the machine but in the hrunan operator. The full

II

rz

Fundamentals of Skill

potentialities of the machine could sometimes be realised if the operator
was rigorously selected and highly uained, but often it was clear that
no amount of selection or training could ensure adequate performance.
The need was therefore imperative for an understanding of the fastors
making for ease or difficulty in the operation of elaborate equipment,
and this in its turn called for knowledge of many facets of human
capacity and performance. It was ttrus that experimental psychologists
came, together with physiologists and anatomists, to take up work
alongside the engineers responsible for developing equipment.
The partnership was destined to be of far-reaching importance to
both sides. The engineers came to recognise the necessity of considering
the potentialities and limitations of the human operator if their machines
were to be used to best advantsge, and their psychological colleagues
set new criteria of adequacy in design which, although not always convenient for the engineer, were clearly important. The lessons did not
end with the war but have continued for service equipment since, and
have gradually been gaining currency in industry both in the desrgn
of work, machinery and machine tools and also in the design of products such as cars and household equipment.
On the psychological side it was significant that much of the initial
research was done by people who came from their trniversity laboratories
to play their part in the war effort. These men brought to their studies
an interest in fundamentals, and recognised at once that a great deal
of ad lnc rcsearch on particular pieces of equipment could be by-passed
if the essential 'key' features of human skill shown by expert operators
could be understood. Previous psychological theory turned out on the
whole to be of little help, but the application of certain mathematical
and engineering concepts proved remarkably fruitful, and has wrought
something of a revolution in thought about many aspects of human
performance. There trer indeed, few important developments of theory
in post-war experimental psychology which do not owe something to
this type of approach.
The researches made have come to be called studies of skill, but the
word is not used in quite the same sense as it is in indusuy. An industrial
worker is said to be skilled when he is qualified to carry out uade or
craft work involving knowledge, iudgment and manual deftness, usually
acquired as a result of long training, whereas an unskilled man is not
expected to do anything that cannot be learnt in a relatively short time.
The industrial definition of skill is thus a formal one in terms of uainitg.
The psychological use of the term is wider, concerned with all the
factors which go to make up a competent, expert, rapid and accurate

Introduction

13

performance. Skill in this sense thus attaches, to a greater or lesser
extent, to any performance and is not limited to manual operations but
covers a wide range of mental activities as well.
An early and infuential srgu of the developing thought of the war
years was the insistence by Craik (tg+f) that the brain should not be
conceived either as a vast telephone exchange of reflex arcs or as a
vaguely defined field of interaaing forces. Rather, he urged, it must
be thought of as a computer receiving inputs from uumy sources and
gsrnfining them to produce an output which is unique to each particular
occasion although nevertheless laurftrl. The effects of this thinking might
have been far less, however, had not three books appeared at the end
of the forties. The first was Wiener's Cybernetics (lg+8) which set out
ideas, clothed with appropriate mathematics, for considering man as a
self-regulating mechaniSmr and which outlined the basic concepts of
'information theory'. The second was Shannon and Weaver's Tlu
Mathematical Tluory of Commun'ication (tg+g) which gave an e4panded
treatment of information theory includi.g a set of theorems. The third
was Wald's Statistical Decision Functions (tgSo) which paved the way

for the application in recent years of statistical decision theory to
human perforrrance. These ways of thinking have not been without
their critics, but it seems fair to claim that, whether or not they have
provided significant answers to psychological questions, they have
given a vigorous impenrs to the treatment of psychological processes in
precise, quantitative terms, ond have brought together many areas of
study previously thought to be trnrelated.
The aim of this book is to sketch some of the main lines of this
thinking. We shall in the rest of the present chapter consider, as
examples, three areas of research each of which illusuates a fundamental

principle of the general approach and emphasizes certain features of
capacity which have come in for intensive snrdy. In subsequent chapters
we shall take up these several feanrres one by one in more detail.
SENSORY.MOTOR SKILL AND THE CYBERNETIC APPROACH

One of the lines of work developed in the forties was the uacking of
moving targets - a problem derived from grrnlaying. Tracking in many
ways epitomises sensory-motor performance and is similar to various
everyday tasks such as steering a vehicle. Precise snrdy is, however,
difficult in real-tfe situations so that the task was incorporated into a
number of laboratory experiments aimed at the study of its essential
features and designed to ascertain some of the important characteristics

14

Fundamentals of Skill

of the human link in systems where man and machine interact. Let us
look at one of these experiments by way of example.
A track, drawn on a strip of paper, passes vertically downwards past
a window as shown in Fig. r.r. The uack moves irregularly from side
to side of the paper and the subiect attempts to follow it by moving a
pen from side to side by means of a steering wheel. He observes any
discrepancy between the positions of uack and pen and takes action to
bring the two into alignment. Any remaining discrepancy due to the
correction not being adequate or to subsequent movement of the track
leads to further action, and so on. Subject and machine together thus
form a closed-loop, error-actuated servo system in which misalignments
COVER WHICH CAN BE
RAISED OR LOWERED TO

+

- VARY THE AMOUNT OF

TRACK SEEN AHEAD OF

THE PEN

tsss

TRACK DRAWN ON PAPER

- PULLED DOWNWARDS
PAST WINDOW

BALL-POINT PEN SET IN
SSSTPERSPEX
SHEET WHTCH
CAN BE MOVED FROM
SIDE TO SIDE WITH THE
STEERING WHEEL

-----

STEERING WHEEL

Figrrre r.r. Tracking apparatus designed by A. E. Earle and used in
experiments by $TeHord (r95r, 1958), Griew (r958a, r95g) and Crossman
(r96oa).
lead to corrections and are in nrn modified by them. Here is at once a
contrast to much classical experimental psychology: performance is not
being studied as discrete, isolated responses to particular perceived signals,
but as an activity serial intime and involving a constant, interplay between
signal and response. Although it may often be necessary, in order to
simplify problemsr to snrdy discrete reactions, these are absuactions
which leave out of account many significant features of the continuous
performances that are normal in most real-life situations.

A number of interesting affempts were made to apply to tracking
tasks the mathematical techniques developed for non-human seryo
mechanisms; for example to calculate 'transfer functions' to describe
the human link benueen signals displayed and control actions (e.9.
Tustin, 1947, Henderson, t959, &lcRuer and Krendel, 1959, see also

Introduction

15

Poulton, 1966). These eqtrations, although essentially empirical, ernphasised certain key feattrres of htrman performance such as latency,
stability and anticipation which were taken up in further experimenrs
to yield indications of importance not only for uacking but for sensorimotor performance in general.
Reaction time and intermittency

If the uack in Fig. r.r is hidden from view until it reaches the peD,
the subiect almost inevitably tracks a [ttle late due to a reastion time
berween a stimulus entering the eye and the beginning of the responding action, which represents the time taken by various sensory, central
and motor mechanisms to act. Craik (tg41r 1948) raised the question of
whether in a continuous task such as uacking the lag could be explained
on the basis of nerve impulses having to traverse a long chain of synapses
in the brain, but found evidence that this view was too simple. Close

scrutiny of the trace made by the subiect's pen reveals a nnmber of
minor wanderings first to one side of the track and then to the other.
If reaaion time were due simply to the time required for signals to
pass through the various sensory, central and motor mechanisms, these
irregularities should not occru: signals could be received and action
initiated continuouslp and the subiect would quickly attain a smooth
reproducion of the track. Instead he seems to initiate corrections only
at intervals of about '5 sec: in other words the servo acts discontinuously.

Broadly speaking the subject behaves as if somewhere in the brain
he has a computer which periodically samples the incoming dau and
cdculates 'orders' to the motor mechanism. This then runs off the
response 'ballistically' while the computer is taking in and processing
the next sample leading to the next response. The response so triggered
may be complex, embracing several detailed muscular astions in a
phased and co-ordinated sequence. They are to some extent monitored
by feedback from muscles, ioints and tendons - in other words the
motor mechanism itself is a servo-mechanism - but overall monitoring
seems to involve the visual mechanism and 'computer' and to take time
in these in the same way as do siguals which initiate new action. The
'5 sec intervals between successive corrections appear to represent the
sum of a reaction time of about .3 sec and a monitoring time of about
'2 sec.
If this is so, the times taken to process data and to monitor action set
an upper limit to the amount of data that can be handled in a grven time.
Further evidence from this view comes from the effects of speeding up

16

Fundam,entals of Skill

the track. At low speeds the total movement of the subiea's pen is,
owing to the minor irregularities already mentioned, a little gteater
than the minimum required to follow the excrusions of the uack precisely. As speed rises above a critical level, however, the subiea swings
shorter and shorter: so that although the correct amount of movement
rises, the amount actually made remains practically constant as shown
in Fig. r.z, The fact that movement falls short of what is required cannot
tr,

s00 5

200
lrJ

=
=
u.

=

o-

=

400 s

t-l'l

G,

G,

u,
(L

F
trl

(,

=
tr

ctrJ

-e---f,or-X

UJ

=

a\'
/,

zuJ

(J r00

=

G,

t.l',

a

o
=
(L

2 ?

z

o
z.
J.-

2oo -a
t-

a

(J
r!

Q SUBJECTS OVER 30

IL

X SUBJECTS UNDER 30

o 50
F
z

O TARGEI - POINTER

u,J

l=
r,

-|
ID

f
a
lr

100 o
F

ztrl
t=
r,

o
=

z

3oo U

o
0

0.4
0.L7
0.6
0.9
SPEED IN MEAN SECONDS PER SWING OF T HE TARGET - POINTER

1.6

=

Figrrre r.z. Average amounts of movement made when uacking at different speeds. (From \Melford, r95rr p. 76, 1958, P. 88.) The apparatus was
somewhat simpler than that shown in Fig. r.r and the task consisted of
keeping a pointer moved by a lever in line with a target-pointer which
swung irregularly from side to side. Each point is the mean for z5 subiects
tracking for at least r min. Note that the older subiects make less movement: this was not accompanied by any consistent change of accuracy and
thus implies a lowering of capacity with age.

be due to sheer inability to make the necessary movements as such,
since the subiea can move his wheel to-and-fro without regard for the
track, very much faster than he does. It seems clearly to be due to the
high speeds not allowing enough time for the necessary control to be
qrercised. We are thus led to view the human mechanisms mediating
between sensory input and motor ouqput as a comrnunication clmnnel of
timited capacity and reaction time as a potentially valuable measure of
this capacity.

Introduction

t7

Other sources of limitation
Performance carr, of course, be limited by the motor mechanisms if the
actual response is made sufficiently laborious. For example, the steering
wheel shown in Fig. r.r took about half a trun to bring the pen from
one extreme position to the other. If it had required, say, ro times this
amount of movement, sheer motor factors would have been a serious
limitation at high speeds of the track. It is clearly idle to argue that one
or other factor always litnits performance: rather the stimulating possi-

bility is raised of manipulating experimental conditions in order to
determine hout nwmy mechanisms, and of what kind, are involved in the
chain from sense organ to muscles.
Following this line we can say at once that sensory limitations have not
normally entered into tracking experiments, although they could do so
if the excursions of the target and of the subjea's pen were sufficiently
reduced in size or if the illumination was very low. The importance of
several prceptual fa*ors has, howeverrbeen demonstrated. For example,

if the track can be seen for a short distance before it reaches the pen that is if the cover shown in Fig. r . r is raised - the time lag due to reaction time is eliminated. So much is, perhaps, obvious: the subiect
looks ahead and responds to the track before it reaches the pen although
how he manages thus to adapt the 'strategy' of his performance is not
at present easy to say. Less obviously when the track can be seen ahead,

the amount of movement by the subject's pen no longer falls short of
that required as speed increases, or does so only at very much higher
speeds than it did with the cover down (Welford, 1958, Crossman,
196o). The maximum effect of raising the cover is not obtained trntil the
whole of one 'swing' of the track is revealed (Poulton, $64), which
suggests that the advantage of being able to see ahead lies in being
able to observe the track in larger 'units': instead of having to observe
each acceleration and deceleration separately, both can be seen together
and form the basis of a single co-ordinated movement. In consequence
fewer 'messages' or 'decisions' have to be passed through the cenual
mechanisms in a gtven time. Those that are passed ffioy, because they
are more complex, take longer, but there is nevertheless a net saving so

that the speed at which the 'load' exceeds the available capacity is
raised.

Similar results are obtained with the subiea unable to seek the uack
ahead if it follows a simple, regular pattern instead of swingrng irregu-

larly. The subjea can, on the basis of what he sees of the uack, learn its
regularity and thus predict its course (Poulton, ry66). He is able, in

r8

Fundamentals of Skill

other words, to respond in larger units because his central computer
can combine present data with constants extracted during previous
experience. The process is by no means always deliberate or conscious,
but the use of regularities observed in sequences to predict events and
deal with larger units of data and action is one of the most ftrndamental
features of skill. The ability to predict and anticipate makes for 'smoothness' and co-ordination of action, and the time saved by dealing in
larger units can often lead to the timing of aaions being more flexible
and less hurried (Bartlefrt 1947). The size and nature of the units that
can be handled are at present little known and pose challenging questions

for future research.
The fornration of such larger units is perhaps only one aspect of a
rnuch more pervasive process. One of the most marked characteristics
of many highly skilled performances is that details of the task are seen
not only as present and irnmediate problems, but are placed in a wider
setting, xnd actions are not designed as individual units but as parts of
an extended activity demanded by the task as a whole.
The chain of mechanisms

During the fonies it was commonly believed that the central conuol
mechanistns could be adequately described in terms of tvvo divisions -

perceptual and motor. It gradually became clear, however, that performance depended not only upon these but also upon the relationships
between them. The time required to respond to a signal depends very
greatly upon the directness and the familiarity of the relationship between signal and response. For example, it seems 'nafirral' that turning
the wheel of Fig. r.r clockwise should move the pen to the dght, and
this arrangement yields more accurate tracking than steering which
works the other way or when the wheel is replaced by, say, a handle
moving up and down. It seems, in short, that the time the central
computer requires to operate depends on the amount of 'work' needed
to relate the data from the signal to the responding astion. The translation from signal to action appears to be an important stage which incorporates the uaditionally recognised choice of ruponse, but also emphasises the complex mediating processes that may be involved.
The disadvantages of unusual relationships can be largely overcome
with practice, and this fact argues that very well-learnt relationships
are somehow 'built into' the brain and enable the computer to be bypassed for routine actions. The establishment of such connections can
be regarded as another mark of skill akin to the extraction of constants,

Introduction

19

the constancies being between display and conuol rather than in
changes of the display done.

\[e may represent the mechanism of sensory-motor performance
EITERNAL OBJECT

s
E

s
E

o
R

G

E

TRANSLATION

N

SHORT

TERM
STORE

F

F-ROM

PERCEPTION
TO ACTION:

cHorcE oF

F
CONTROL

E

OF

c

RESPONSE

A

T

0
R

s

N

s
LONG TERM STORE

Figure r.3. Hypothetical block d;agram of the human sensory-motor
system. Only a few of the many feedback-loops which exist ars shown.
(From !flelford, 1965, p. 6.)
sketched so far in the block diagram of Fig. r.3. Most of the researches
on skill we shall be surveying in later chapters can be regarded as

attempts to refine and extend this diagramr to understand the working
of its various parts and to determine their capacities.

SIGNAL DETECTION, VIGILANCE AND DECTSION

A different area of research was opened up in the forties by problems of
watchkeeping, especially with operators of Asdic and Radar equipment
Both required the deteaion of faint, infrequent signals, and it was fotrnd
that many were missed even by the best watctrkeepers. The attempt to
find out why led to two lines of work, both with implications far beyond
the situations that gave rise to them.
The first was concerned with the fact that the proportion of signals
detected usually fell from the beginning to the later stages of a watch
(Mackworth, r95o). The fall seemed often to be associated with drowsiness and lack of interest, and questions were thus raised about the
conditions under which attention could be maintained and its relation
to motivation, fatigue, boredomr monotony and state of arousal. Many
laboratory studies made dtrring the last 20 years have built up a complex, but reasonably coherent, picnre of the association between, on
the one hand, sensitivity and responsiveness as measured by behaviou,

zo

Fundarnentals of Skill

and on the other, physiological variables such as level of autonomic
activity, and of the relationships of both to various environmental conditions. Ideas in this field have gtown in parallel with those in some
other areas of psychological enquiry, particularly emotion, motivation
and personahty, and it seems clear that the concepts which have been
developed apply over a wide field. At the same time they are leading to
a radical re-appraisal of concepts of fatiBUe, boredom and environmental stress, and have emphasised the importance of snrdying contirued performance as a function of time.
The second tine of work that arose from the problem of signal detection was a new approach to sensory thresholds. It had been known since
the beginnings of classical psychophysics in the 'tineteenth century that
the minimum strength of signal which can be detected varies in an
apparently random manner from moment to moment, but the reasons
why it does so had not been urderstood. A breakthrough was achieved,
however, in the application by Tanner and his associates (Tanner and
Swers, rg14rSwets et al.rr96r) of statistical decision theory. They argued
thatsignals have to be detected against a certain amount of background
noise: Asdic signals have to be distinguished from various unwanted
sounds; the Radar watchkeeper has to observe flashes on his tube against
other irrelevant flashes, or as they came to be termed 'visual noise'.
Besides these external soruces of noise, internal 'neural noise' arises
from random firing in the sensory pathways and brain. The noise level,
whether external or internal, varies from moment to moment, so that
if the level of the sigual greatly exceeds that of the noise the nnro will be
clearly distinguishable, but if the signal level is relatively low, this will
no longer be so and errors may occur. The signal level required to
secure any gtven degree of accuracy will increase with the noise level
so that the discriminability of a signal can be specified in terms of
signal- t o-nois e r atio .

This approach has provided a powerful concepnral tool and metric
for quantitative sturdies not only of sensory thresholds but of a wide
variety of decision-making tasks including discrimination, guessing,
betting risk-taking and, more recently, recovery of data from memory.
The concept of netual noise has been a stimulus to the formulation of
physiological models and has emphasised the possibility of a much closer

tie-up than hitherto betrreen the two disciplines in the snrdy of many
sensory and behavioural phenomena.
The idea that performance is a function of signal-to-noise ratio in the
brain leads to a concept of capacity parallel to, but different from, that
derived from the snrdy of reaction times. It is exemplified in the findings

Intro&rction

2r

summarised by Miller (1956) that if a subiea is given a series of stimuli

to classfi, the number of different classes he can distinguish accurately
is limited and errors are rnade if this number is exceeded. Such a finding

is understandable if the classffication involves the identification of different signal levels and there is a maximum level that the system can
handle: the presence of noise means that a certein separation must be
preserved betrreen classes if identification is to be accurate, and only a
limited number of such separations can be fitted into the available
range. Capacity in this case does not depend on the speed with which
signals can follow each other along a singls shannel, but on the number

of distinct states the brain mecha'Iism concerned can assume at any
one instant. The importance of this tJpe of capacity as setting limits
to skilled performance has not been at all thoroughly explored, but it
seems likely to enter not only into discrimination and classification, but
into the storage of data in memory, especially short-tenn memory.
More generallS the concept of neural noise blurring or distorting
signals provides a plausible way of accounting for certain tJpes of
variability in performance. If signals in the brain are in the form of
trains of nerve impulses - as they obviously are - and messages are
conveyed by their patterning, the occrurence of random impulses could
well transform one message into another upon occasion, and thus lead
to errors. After effects of previous activity in the brain or the neural
effects of insistent thoughts and interests can also, perhaps, act as noise
in this wsy, leading to slips of the tongue and other minor confusions
of everyday life.

MENTAL SKILL AND CONCEPTUAL FRAMEWORKS
Although a distinaion is commonly drawn berween sensory-motor and
mental skills, it is very difrcult to maintain completely. All skilled
performance is mental in the sense that perception, decision, knowledge
and iudgment are required. At the same time all skills involve some
kind of co-ordinated, overt aaivity by hands, organs of speech or other
effectors. In sensory-motor skills the overt actions clearly form an essential part of the performance, and without them the purpose of the
aaivity as a whole would disappear. In mental skills overt actions play
a more incidental prrt, serving rather to gwe expression to the skill
than forming an essential part of it. There are thus many feattrres
coulmon to both sensory-motor and mental skills, while each seryes to
emphasize and illustrate some features more than others.
Let us consider, as an example of a mental skill, performance at a

22

Fundamentals of Skill

problem-solving task used in experiments by Bernardelli (see Welford,
1958, pp. zoz-zo4). The apparatus consisted of a number of small
boxes each with a row of six electrical terminals on top connected rurderneath by resistances. Each subiect was given a box, together with a
resistance meter and a circuit diagram which showed the connections
between the terminals on the box, but did not indicate which terminals
on the diagram corresponded to which on the box. The subjea's task
was to deduce which terminals corresponded to which, by means of
readings taken on the meter. A modified version of the task, designed
to avoid the need for subjects to understand electrical circuit diagrams,
consisted of boxes with a row of six buttons on top and a diagram of
the tJpe shown in Fig. r.4. !(Ihen any two buttons were pressed the

I
Figure r.4. Diagram given to subiects in a problem-solving task (a), and
its electrical equivalent (b). (From lUelford, 1958, p. 2o5.)

The experiment was designed by Miss ff. M. Clay and was based on a
type of electrical-circuit problem originated by Carpenter (1946).

connections and distance between them on the diagram were indicated

on a meter. The subiect's task was to deduce which buuons corresponded to which circles on the diagram.

The perfornrances of different subiects varied widely. Some took a
large number of readi.gs, many of them several times and, if they
arrived at a solution at all, did so laboriously and haltingly. Others
proceeded in a much more puposeful and effective manner: each
reading appeared to be taken with some clear oh, either of obtaining
systematic information or of testing some hypothesis. We can see here
the operation of a 'strategy' of perforrnance such as was implied by
looking ahead in the tracking task, but in a very much more elaborate
form. Such suategies develop rapidly with practice, becoming much
clearer cut and more systematic in the later members of a series of
problems. The formulation of an hypothesis Bsy, perhaps, be thought
of as the counterpart in a problem-solving task of the ballistic move-

Introductim

23

ments made in sensory-motor performance: in both the subiea makes
a move ahead of his data, and in both skill can be said to be shown in
the extent to which these moves are rnade without the need subsequently
to correct for errors.
Suategies varied from one individual to another. Some systematically
took all possible readings, wrote them down and then uied to think out
a solution. Others took readings until they had identified one terminal,
then went on to another trntil all had been found. There were furttrer
more subtle variations benreen the performances of the same individual
on different boxes. All these variations, benreen and within individual
performances, emphasise that the same task may be done in many
different ways and that rf, therefore, we measure only overall achievement, such as success or failure in arriving at a solution or time taken
or number of errors made, we miss some of the most important data
available in the performance. For example, the strategies adopted by
people of different sBsr or under different conditions of fatigue, differ
far more than do their overall achievements. There seem often to be
compensatory shifts tn the method and Eranner of performance under
adverse conditions or in states of impairment, which serve to opt'mise
the performance as a whole and to make the best use of capacities and
conditions available at the time (Welford, 1958). Iq therefore, performances of these kinds are to be snrdied adequatelp it is essential to
look not only at aatrall achianemenl but also at the ma?rner in sthich it
wcts attained.

Viewed from a different snglg, this principle is a re-statement of the
point made in the context of uacking tasks, that in skilled performance
particular signals and actions are placed in a wider sening determined
by the task as a whole, and larger trnits of perforurance exert an organis-

ing control over the smaller details of perception and response. The
attainment of this wider co-ordination appears to depend on two
capacities. FirstlS pieces of data arriving at different times may require
to be combined, or a series of actions may have to be taken over a
period of time. In these cases there is a need fot short-term retention
to hold some pieces of data while others are gathered and to kep a
'tally' upon action so that the subiea knows what has been done and
what remains to be done. Intensive experimental snrdies of short-term
retention during recent years have indicated that more than one storage
system is involved. Their identification and the measurement of their
capacities are challenging questions for the futtue.
As anyone faced with a problem-solving task knows to his sorrow,
however, possession of all the relevant facts does not enslue insight.

24

Fundammtals of Skill

Short-term retention may often be a necessary element in problemsolving, but something more is also required. Just what this is seems
to have eluded precise definition, but an important lead was given by
Craik when he suggested that 'the fundamental feattrre of neural
machinery' is 'its power to parallel or model external events' (tg43,
p. 52). He had in mind that the brain's computer is able to extrapolate
from present data and thus test the consequences of various lines of
action or sequences of events, without their having to be acted out. In
one sense this way of looking at the problem merely re-states it in a
dlfferent form. It does, however, imply that thought consists of a
ntrmber of definable computer-operations and that a much clearer picttue of thinking would be obtained if these operations could be identified - in other words if we could spell out the 'computer programme'.
\[e can at present do this to only a very limited extent but we can
recognise a number of categories as important. Arithmetical calculations
and other mathematical and logical procedtres are an obvious case in
point, so are 'sets' and expectancies based on previous events in either
the immediate or the more remote past, temporal and spatial relationships Slving rise to perception of causality (Michotte, 1946, 1963), and
certain qrpes of 'simplicity' or 'economy' in the grouping or 'coding'
of events which enable the maximum data to be accounted for in the
minimtrm tems. Perhaps the most promising lead into the problem
from an experimentd point of view is to regard thinking as an elaborate
version of the translation process from perception to action mentioned
in relation to tracking. Some progress, which we shall review in
Chaptet 6, has been made in analysing 'indirectness', strd thus dim,culty, of relation benreen display and control in terms of spatial rota-

tions, distortions and translations into different symbols. It seems
plausible to regard these as simple prototypes of the more elaborate
and recondite operations of thinking.

THE PRESENT RESEARCH POSITION

The snrdy of skill in its various forms has not led to the formation of
any all-embracing 'grand theory' with a label such as affaches to Gestal-

tism or to the various schools of Learning Theory. These broad
generalisations are really a sign of immaturity: they do less than iustice
to the complexity of the organism and of its functiodtg, so that they
are inevitably either incomplete or so broad as to be of trivial explanatory value. Instead there has been an attempt to build smaller-scale
theories which, as far as possible, are scrutinised to see that they are

Intro&rction

25

not obviously inconsistent, but are left for the futtre to link together
into a more coherent strucnrre. Present achievements can, perhaps, be
broadly summarised trnder four headings:

r. It has been clearly recognised that sophisticated self-regulating
machines can carry out many operations closely analogous to those
performed by hurnan beings and that, in turn, it is an important aid
to clear thinking about htrman behaviour to consider the qpe of machine
that would have the same characteristics. It is a not unfair test of a
psychological theory to challenge its author to desigp a machine that
would behave in the same way as his theory would predict for a human
bei.g.

2. Substantial progress has been made towards a psychology which
is truly quantitative, and thus towards the removal of what had previously been a serious deficiency. A physiologist can speak of human
performance in terms of body-temperatures and pulse-rates and relate
these to foot-potrnds, calories and heat-loads. Hitherto a psychologist
has been virftally unable to measure performance in ways which transcend the particular cases concerned. His ability to do so now is still
fragmentary, but treatments of reaction-times and the detectability of
signals using probability as a metric, avoid much of the criticism often
levelled at psychometric procedures in dre past - that they 'meastrred
the unknown in units of uncertain maguinrde'.
3. This quantitative emphasis has grven fresh impenrs to the detailed
analysis of the various mechanisms which go to make up the sensorymotor chain. Two kinds of approach have been tried at this level,

both potentially very important. The first, which we may term a
'microbehavioural' approach, is to consider extremely detailed behaviour

by regarding, for example, the choice of response in a choice-reaction
task as containing a series of sub-choices carried out according to a
particular strategy. These details are seldom observable directly but
can often be inferred by fining mathematical models or by comparing
perforrnances at subtly different versions of a task. The second approach
considers capacity neurologically in terms of the funaioning of nerve

cells either singly or, more important, in large masses to which statistical concepts can be applied. This has, perhaps, been particularly
prominent in Britain where there has been a long and honoured tradition of close contact benneen experimental psychology and physiology.
In the present state of knowledge both these approaches have their part
to plap and each can be a valuable supplement to the other.
4. Finally there has been a much fuller recognition than ever before
that performance cannot be adequately snrdied in terms of discrete,

26

Fmdarnentals of Skill

isolated reactions. Behaviour tlpically involves a stream of signals and
responses each in part dependent on those that have gone before and
influencing those that come after. Even the simplest activity involves a
complex, phased and co-ordinated set of detailed actions. Serious
attempts have been made, and many more are needed, to grapple with
the complexity that this implies, to snrdy the basic components of per-

formance without removing them from the wider context of the
performance as a whole, to identi$ the main funaional divisions of
centrd processes, to show how each contributes to performance in any
given circtunstances, and to trnderstand how the various contributions
change with time.

II
Simple Decisions

It is uaditional to begrn the snrdy of experimental psycholory with a
treament of prychophysics - the relating of sensory experience ro

physical measures. The tradition is sound because, although the pursuit
of psychophysics for its own sake is indeed an arid academic exercise,
the iudgments involved in psychophysical experiments represent a form
of simple decision-naking which is amenable to relatively precise
measurement and mathematical formulation. Analogies of these simple
decisions are fotrnd in other areas of psychological function and in the
more complex decisions of everyday life, so that the theories developed
in the psychophysical field provide a framework for thought which has
an application far beyond the e4periments on which they are based.

WEBER's LAw, THREsHoLDs AND DISCRIMINATIoN

It was recognised in the nineteenth century that arnoturts of sotrnd or of
light that were physically measurable might nevertheless fail to be
heard or seen even though the subiect was fully alerted to detect them.

To accotrnt for this, there grew up the concept of a 'threshold' or
minimum quantity above which a stimulus had to rise in order to enter
the subiect's percepftal mechanism. It was also early recognised that
the stimulus level required to pass this threshold was not completely
fixed, but varied from instant to instant in an apparently random
manner so that several measurements were required to establish its mean
value. Just why this was so was not clear although it was reasonably

argued that the central effect of a glven physical stimulus may vary
from one moment to another (Cattell, 1893, Solomons, r9oo, Oldfield,
rg55).

It is corlrmonly assumed, following Thurstone (t9z7a, b) that the
central effect of a stimulus, such as the frequency of nerve impulses
generated, is proportional to the logarithm of the physical intensiry,
although the site and mechanism of the transformation are not generally
27

28

Fundamentals of Skill

clear in physiological terms. Such a logarithmic uansformation provides what is probably the simplest and most elegant explanation for the
constancy of the Weber Fraction

ds

t:aconstant

(z.r)

where dS is the least noticeable increase of stimulus intensity from a
given level S measured in physical units. This uaditional way of
stating that the least noticeable increase is a constant ratio, may also
be written
Sc

H:aconstant

(z'z)

where Sc and Sr are the greater and lesser respectively of two
quantities compared, or alternatively

log Sc - log Sz : a constant

(zA)

The implication of this last equation, taken with Thurstone's assumption is, of course, that the discriminability of two stimuli depends upon
a constant difference berween their central effects, whatever their
absolute magninldes.

It can be argued that this holds even though the \U7eber Fraction

rises substantially at low values of S. It has been noted since the tir e

of Fechner and Helmholtz that the rise can be abolished if a small
constant (r) is added to S so that in place of Eq. 2.r we write
ds

ip-aconstant

(z'4)

and in place of Eq. 2.3,

log (Sc + r) - log (Sr * r) - a constant

(2.5)

The quantity r is small so that it affeas the \il7eber Fraction substantially only when S is small and can normally be neglected when S is
large. It is commonly assumed to represent spontaneous activity in the
sense organ so that it does in a very real sense add to S. Its relation to
S and dS and a method of measuring it graphically in the same nnits as
S is shown in Fig. z.r.
The attempt to tie discrimination down into such central effects
as frequencies of nerve impulses is obviously plausible for simple sensory ma$urudes, but is perhaps more questionable when made in
relation to complex perceptual quantities such as lengths of line or
numbers of objeas. Such quantities must, however, somehow be
represented centrally, so that although the precise mode of representa-

Siopb Decisions

29

tion is not clear, treating perceptual magRrtudes in this way is not farfetched. There is, however, need for caution in speciffing S: for instance Ross and Gregory Gg6+) showed that the least noticeable dif:
ference between weights was greater for a set of small werghts than for
a larger set which, although they weighed the same, appeared lighter in
accordance with the size-weight illusion. The subieas in this case

ds

.6-- r-- ->o

s

Figrrre 2.r. Plotting dS against S to measure the small random component

r. Based on a diagram by Gregory (1956).

seemed not to have been iudgrng in terms of weight alone, but of some
integration of weight and size such as density.
There is an obvious extension of Eqs. zq anld 2.5 to above-threshold
discrimination: if we can for practical purposes neglest r in these cases,
we should be able to write, following Eq. 2.3,

Discriminability

- (log Sc - log .Sz) x a constant (2.6)

This point was recognised by Fechner in the mid-nineteenth century
but its main use and testing have come into prominence only recently
with the work of Crossman (tgSS) and of Mtrdock (t96oa). Crossman
proposed as an index of confirsability (C) benreen two qtrantities
C - a constant /(log, Sc - log, Sz)

(z.T)

The use of logarithms to base z makes the unit of confi.rsability the
measure of confusion betrreen two signals, one half the magninrde of
the other. A table of C is gtven in the Appendix. Crossmau tested his
formula with studies of times taken to discriminate and found that rne
was linearly related to C. I[e shall consider these results in more detail
later in this chapter. He extended his treament to cases in which

30

Fundammtals of Skill

there are more than two quantities by suggesting that the average con-

fusability of any one quantity with the others can be measured by
taking the mean C for that quantity with each of the others.
Murdock adopted essentially the same type of approach. He suggested

that the distinctiveness of a quantity from several others may be
measured by taking the total of the log differences between it and each
of the others and expressing this sum as a percentage of the total sum

for all the quantities. This measure he termed Dl, i.e.

Do/, :

) Cros,si - log sr)

(2.8)

22(loe s, - log si)

i:L j:L

He tested his formula against accuracy in categorising sound intensities
and applied it to previous results by Eriksen and Hake (tgSil on
categorising areas of squares. Reasonable matches berween theory and
observation were obtained, especially when it is recognised that other
factors such as biases against using extreme categories may also be
operating. A method of correcting for such biases is suggested by

Doherty (1966) who found that, after correction, Murdock's D scale
provided a good fit for data on categorising lines of different lengths.
Both Crossman and Murdock also proposed that the well-known tendency for recall to be more accurate at the ends of a list of items learnt
by rote can be explained by the greater discriminability of serial positions at the ends of the list than in the middle. Table 2.r sets out a
sample of Murdock's results obtained by working out D% for each
serial position on the assumption that distinctiveness can be measured
in terms of log ordinal position. Very fair fits were obtained by this
method with two samples of rote-learning data.
TABLE 2.r Percentage distinctiamess (D%) for the sezteral positions in a

series of eight items, measured in terms of log ordinal position. Frorn Murdock

(r96oa).

Serialposition r
z
3 4
22.9 r4.o ro.4 9.2

D%

567
9.2 rO.O rr'4

8

12'9

THE DECISION-THEORY APPROACH

It was commonly assumed until recent years in psychophysical sttrdies
that if the threshold was not passed but the subiect had nevertheless

Si*pk Deci-rians

3r
to respond, he guessed at random, and methods of correcting data for
such chance guessing were advocated. There were, however, two facts
which called into question the idea and the procedure based on it.
The first was that the measured threshold depended very much upon
the degree of confidence required of the subject: his threshold was
higher if he had to report a signal only when sure, than if he could do
so even when somewhat doubtftrl. The second fact was that if he was
not sure but was nevertheless forced to guess whether a signal had been
grven or not, his guesses over a number of trials were substantially
better than chance, implying that the signal had to some exrent been
perceived even though the subiea had no confidence that it had
occured.

-

ltl

OFF

CUT

6

I

I

=

qr(
Ou>o
F q.l
J=

NOISE

I

SlGt'lAL
+ NOISE

qi

o
o
G,
o.

rc

Figure z.z. The basic signal-detection model. Based on a diagram by
Tanner and Swets (1954).
Several groups seem to have arrived at similar methods of accotrnting

for these facts round about the same time (e.g. Smith and Wilson,
r953r Munson and IGrlin, 1954), but the approach most thoroughly
worked out has been that of Tatrtrer, Swets and their associates working

on the detection of faint sotrnds against background noise or faint
visual signals on an illtrminated backgrotrnd constituting 'visual noise'
(Tanner and Swets, r954rSwers, 1959, rgfi4rswetsrTanner and Birdsall,
tg6r, Green and Swets, t966).
They consider as a basic situation one in which the subiea is given a
series of trials in each of which there is a brief presentation of either the
backgrotrnd noise plus a signalr or the background noise alone. The
subiect's task is to decide whether or not a signal was present. Alternatively a series of two to eight presentations is made and the subiect has
to say in which one a signal occrured. The authors argue that at each
presentation the subject observes a quantity r which, because of the
noise, is liable to vary randomly in magnitude from trial to trial. We

32

Fundamentals of Skill

can represent the magninrde of r observed over a series of trials in
which no srgnal was present by a disuibution such as that shown for
'Noise Alone' in Fig. 2.2. The noise may be either extemal due to
variability of the stimulation presented to the subject or intemal due to
randomness in the activity of the sense organ, neural pathways and
brain. Such internal noise could arise from several sources such as
spontaneous random firing in any of the sensory or central mechanisms
concerned, from tonic neural activity in the brain (Pinneo, 1966) or
from the after-effects of previous stimuli (Welford, 1965). Its presence
means that signals can be conceived as having to be distinguished from
noise even when external sources of noise are excluded. Any random
variability in the sensitivity of any of the mechanisms involved could
also be, for many purposes at least, taken as part of the internal noise
since the essential result of both variations of sensitivity and of noise is
to add a random component to the cenual effect of a sigual (Atkinson,
1963). The shape of the noise distribution wifl depend on the distributions of the various components making up the noise and on any transformations of physical quantities that take place in the sense-organs or
brain, but for purposes of our present discussion they can be tentatively regarded as notmal.
The quantity r for presentations in which a signal is present is taken
as having a similar distribution to the noise, but with each observation
incteased by the amount of the sigual, as shown in the 'Sigual-plusNoise' curve of Fig. 2.2. The subject is assumed to establish a sttt-off
point r, and to treat any level of r above this point as 'Signal' and any
point below it as noise alone or rather as 'No Signal'. If the signals are
strong enough for the disuibutions to be well separated the discrimination of 'signal' from 'No Signal' can be virtually complete. If, however,
the sigual suength is weaker so that there is overlap betrreen the disuibutions, dissimination cannot always be accurate: part of one or
other distribution or of both will inevitably be on the wrong side of r,
so that errors will be made.
The model represented in Fig. 2.2 treats discrimination in terms of
two parameters d' and p. The former is the distance between corresponding points - say the means - of the two curves measured in
standard deviation units. We can thus write
d,,

:fsar - far

(z.g)

The second parameter 0 is the Un U'nood. ratiothat a central effect of
the magmude represented by ,, is due to signal-plus-noise as opposed

SilnPk

Decisions

33

to noise alone. In other words it is the ratio of the frequencies (f) - that
is the heights of the ordinates - at ffc so that we write

I -& at xc
nf*

(z.ro)

If the distributions of Signal-plus-Noise and Noise Alone are known
or can be assume d, d' and B can be calculated from the proportions of
the two possible classes of conrect response thtrs:
(") YES when a signal is present - represented by the area of the
Sigual-plus-Noise distribution to the right of x, in Fig. 2.2. Ve may
refer to the proportion of responses in the disuibution falling into this
category as pYESslr.
(D) NO when a signd is not present - represented by the area of the
Noise Alone distribution to the left of lcr. \[/e may rder to this proportion of the Noise Alone distribution as pNOry.
If, for example the distributions are normd, tre distance from r, to
the mean of &e Signal-plus-Noise distribution measured in standard
deviation units can be found from a table of the normal probability
integral, such as Fisher and Yatm' (rgf8) Table I or I& by noting the
deviation required to produce pf,Ssr. $imilarly the distance from r,
to the mean of the Noise Alone disuibution can be found by noting the
derriation required to produce pNOr. Assunring the two distributions
are of equal variance, the value of d' will be the sum of these two
deviations. To take a practical example, suppose 2YESsar : '9o and
pNOar : -95t the two deviations would be r'28 and r'64 rCIpeaively
and d' would be 2'92.
The same result could have been obtained using the proportions of
the trro possible classes of error:
(") 'Misses' - that is responses of NO when a signal is in faa present,
represented by the area of the Signal-plus-Noise distribution to the left
of x, (?NOsar).
(D) 'False Positives' - that is replies of YES when no signal is present,
represented by the area of the Noise Alone distribution to the right of x,

@rSr).

The value of P can be calculated, when the disuibutions are normal,

from a table of the ordinates of a normal disuibution such as Fisher
and Yates' Table II. For the example iust quoted these are '176 and
.ro3 respectively so that P
r'Tr. The value of p diminishes as r, is

-

moved to the left.
A table to find d' and P for different values ofPNOsr and2YESar
*rO*n in the Appendix.

34

Fundamentals of Shill

Although we have dealt so far with the absolute detection of siguds,
the model is obviously applicable to situations in which a signal is an
increase or a decrease from a given reference value: the distributions of
Sigual-plus-Noise and of Noise Alone become respeaively Sc (or Sz)plus-noise and S6-plus-noise with Sc (or Sz) and Sn the changed and
the original signal values respectively.
The value of d' is the measure of true discriminability in terrrs of the
ratio betrreen, in the absolute case signal strength or in the comparative case difference between sigual strengths, and the aariability of the
noise level. On the other hand f: as a measure of the cutoff point r,
can vary independently of d' . It. can be thought of as a measure of the
caution exercised by the subiect or of the confidence with which his
iudgments are made. For example a large value of fl, implying a high
cutoff well to the right in Fig. 2.2 means that he is being cautious
about recognising signals and demanding a high degree of confidence
before saying YES. The result will, of course, be that the proportion
of occasions on which a YES response is given (pYESs.n plus pyESar)
will be relatively low and the proportion of NO responses (pNOa,,
plus 2NOsar) will be corresponclingly high. There witl thus be few
false positives and a relatively large number of misses.
The subiea is able to vary his cutoffto some extent at will, and indeed
to make iudgments in terms ofmore than one cutoffby defining different
levels of confidence (e.g. Pollack and Decker, 1958, Swets, 1959, Swets,
et al., t96t, Broadbent and Gregory, t963a} For example, Broadbent
and Gregory presented their subiects with short bursts of noise during

which there might or might not occur a pure tone of lrooo cycles.
Subiects had to rate their confidence as to whether or not a tone was
present on a five-point scale:'Sure; Not quite sure; Uncertain; Not
quite sure not; Strre not'. These can be regarded as a series of five cutoff
points of progressively decreasing severity, and there was in fact a

clear trend for the proportion of misses to decrease and of false positives
to increase from 'Sure' through the successive categories to 'Sure 1ot'.
If pYESsrv is plotted against pYESa,, the points obtained for a range of
criteria lie on a curve of the form shown in Fig. 2.3d, known as a
receioer-operating-charactristic or ROC curye. !(Ihen d,' : O the ROC
curve is a straight diagonal from the bottom Ieft to the top right corner.
As d' incteases, the curve is bowed more and more inio tie top left
corner. When the distributions are normal and of equal varianci, the
ROC curve plotted on double-probability paper becomes a straight line
as shown in Fig. 2.3b.
It should be recognised that the 'noise' need not be noise in the

SWII Decisims

3j

literal sense but any random disturbance which may affea the signal.
For example Winnick et al. (1967) have shown that the detectability of
words shown briefly with a background of random letters in a tachistoscope can be accounted for in signal-detection terms if the random
letters are regarded as 'noise' and the word as signal.
There is now an impressive array of experiments on sigpal detection
in which this model and its assumptions seem to be valid although there
are also cases (e.9. Hohle, 1965) in which they do not, indicating that

irrl l.o
a
t!
G,

fr o't

0.9

2

z.

0.7

3 o.s
zr!

-

o.r
.=
(n

0.5
d'= o
0.3

UJ

iu-

o 0.2

0.t

F

A

=
E

So 0
o
E
(L

.02

0,2 0.4 0.6 .YES,
0.9 1.0 . .02 o.l
PRoBABILITY oF

0.3 0.5 0.7

WHEN No SIGNAL IS PRESENT

0.9. 0-98

Figrrre 2.3. Receiver-Operating-Characteristic (ROC) curves.
(A) with linear axes, (B) with normal probability a:(es.
there is further work to be done to determine its range of application.
What has been achieved so far, however, makes it clear that the approach
is a powerful one with, as we shall see in later chapters, widespread
implications.

Placing the cutoff point

Although analysis of discrimination in terms of d' and B enables the
effects of caution to be distinguished from those of true disctiminability,
it does not really solve the problem of thresholds but merely shifts it to
the question of how the cutoff is fixed. Clearly the cutoff point can be
affected by instructions to the subject to be cautious or otherwise, in
other words by the relative importance of the two tlpes of correct
response and the two tlpes of error. The obieaively optimum cutoff
point c:rn be calculated if the rewards attached to each of the correct

36

Fundamentals of Skill

responses and the costs of the two types of error are known and are
measurable in the same units:

Optimu m

B:

U

Value of correct NO * Cost of incorrect YES /r r r\
,.
/\
\-"'
Value
correct

of

YES + Cost of incorrect NO

L

'

In order to fix his cutoff optimally, however, the subject would have to
know or to be able to assess all these quantities, and also to relate them
together. The evidence from a number of experiments on discrimination (e.9. Ulehla, ry66) and on guessing and betting (e.g. Edwards,
196l) makes it clear that the optimum is seldom if ever achieved. At
least part of the reason probably lies in the fact that the limited span
of short-term retention makes it impossible to assess accurately the way
in which sequences of signals have been constructed: the subjea tends
to give undue weight at each point in the series to the three or four
items presented just before. Part of the reason also may lie in a well
marked tendency to overestimate the probability of rare events and to
underestimate the probability of frequent ones (Howard, 1963). Part
again may lie in subjects' inability to adjust their strategies to take frrll
account of values and costs: for example Pitz and Downing (1967) in a
guessing task found that performance was near optimum when the
values of different guesses were equal, but departed markedly from
optimum when they were not.
Knowledge and experience of the signal sequences, rewards and
punishments can, however, affect the setting of cutoff points to some
extent (e.9. Taub and Myers r tg6t, Katz, 1964). A good example of the
effect of knowledge is mentioned by Laming (r g6z) who required his
subjects to sort packs of cards into two piles according to whether they
bore a longer or shorter line on the face. The numbers of longer and
shorter were varied in different packs from z4 of each to 30 shorter
with 18 longer or 36 shorter with rz longer. The proportions of errors
in which longer was mistaken for shorter and oice oersa were about
equal when the pack containing equal frequencies came first, but when
this followed a pack containing a preponderance of shorter lines, the
mistaking of longer for shorter became much commoner than the mistaking of shorter for longer. The same was true in the packs with
larger numbers of short lines and was more marked as the unbalance
incteased. Subjeas appeared to be adjusting their cutoff point in a
direction which reduced the errors made in responding to the more
frequent signal. Swets and Sewall (rg6f) using a task in which subjects

Simple Decisions

37

detected tones in short bnrsts of noise, showed that changes due to
offering monetary rewards for improved performance left d' unchanged,

implyrng that any changes in the proportions of different types of
response were indeed due to a shift of P rather than to any genuine
improvement of detection.
In the absence of any knowledge of the signal sequence or other collsiderations the subiea must base his cutoff on his knowledge of the
noise disuibution alone. In this case he presumably sets it at a level
which will grve a tolerable - usually very low - likelihood of making
false positives.

What is r?

The exponents of signal-detection theory, as this model has come to be
cdled, have been at pains to stress that much of the usefulness of the
model does not depend on being able to speciff the quantity r which
the observer uses as a basis of his decisions. It is clear, however, that r
must in some way increase monotonically with increase of (a) the
suength of the physical sigual or difference betrreen two signal levels,
(b) cenual effects in the brain, such as the frequencies of nerve impulses
generated and G) confidence on the subiect's part that a signal has
arrived or that Sc has occrtrred rather than Sr,. )c cannot, however,
be a linear function of all these. For instance, any logarithmic or
other transformation betrreen physical stimulus and central effect will
preclude linearity with both (a) and (D) simultaneously.
The idea of scaling r in terms of confidence derives from a corsideration of the faa *rat any point on the r axis can be specified in
rerms of a likelihood ratio fsw/fx. We have already noted one such
value, 0, atthe cutoff point r, and that the value of P diminishes as the
cutoff point is moved to the left in Fig. 2.2. If the observer can use
different cutoff points implyrng different degrees of confidence it is
reasonable to suppose that he can somehow scale the whole r axis in
the same wsy, either linearly with the likelihood ratio or linearly with
some function of it such as its logarithm. Such a scale must presumably
be correlated with level of neural activity, but need not be a linear
function of it. For those who object to inuoducing the subiective term
'confidence' into the discrrssion, essentially the same point can be made
by arguing that the likelihood ratio or some uansformation of it can act
as a funaional quantity in the decision process.
It becomes necessary, however, to grve tc a measure, in terms of
physical signal strength or netual effect, if we want to relate d' to

38

Fundarantals of Skill

sigual strength in aoy quantitative way. If Eq. 2.6 holds, the straight-

forward assumption is that
fl,'

- (log Sc - log Sz)/o rog,s

(z.tz)

so that r is a linear function of central activity, but if this is so the dis-

tributions cannot possibly be normal and are probably of trnequal
variance. They cannot, for example, extend from minus infinity to plus
inf-ity as strictly normal distributions would have to do, but from zero
NOISE

ALONE

x
u-

o

CUTOFF
t

I

t-J

=
S ICNAL

z

+ NOISE

tr I

()
z.
lJ-

o
F

6
o

o
d.

o-

x

Figure 2.4. The basic signal-detection model assuming Poisson-tike

distributions.

to some maximum set by the capacity of the brain mechanism concerned. If the noise is conceived as consisting of random neural im,pulses occurring within a brief period of time, the disuibutions would
be approximately of Poisson form with a variance equal to the mean
and would look like those shown in Fig. 2.4.This difficulty is, however,
not always serious. For example, all our discussion so far has been in
terms of the proportions of the four possible responses - YES or NO
either when a signal was or was not present, and these proportions
would remain the same if the measure along the.r axis was uansformed,
for example by taking the logarithm of the number of nerye impulses
concerned. We are thus in the position of an experimenter wishing to
perform a /-test or analysis of variance on data of non-normal distributions and uneqrnl variances, and can followthe procedure reconrmended

Simplc Decisions

39

in that case of uansforming the measure so as to make the distributions
normal and variances equal. Doing so will in no way invalidate the test
or, in oru case, the computation of d', and so long as we work in terms
of our four proportions we do not need to specify the acnral transformation.
The relation of signal-deteaion measures to those used in traditional
psychophysics has been disctrssed by M. Treisman (lg6+) and by
Treisman and Watts (1966). In particular they outline a way of treating
data obtained by the Constant Method in signal-detection terms.

Multiple dis crimination
Several attempts has been made to extend Signal-detection Theory to
cases where more than one signal has to be discriminated. The cases
st-

sc

sR

z
l!O
our

i3

o

fii,
Atll

xa
x

'n

Figure 2.5. E:rtension of the signal-detestion model to cases where a
decision is required as to whether a signal is greater or less than a standard.
The area between ** and rr represents iudgments of 'equal'.

involved may be divided broadly into four gpes which we shall deal
with in turn. The treaunents are often somewhat speculative, but serve
to indicate the potential rang€ of usefulness of the approach.

r. Ttto gnntities presanted irru,ltarcously, or nearly so. The most
straightforward example of this gpe is when a standard reference signal
is presented with a second signal which may be either greater or less.
The correspondi.g signal-detection model is represented in Fig. 2.5 h
which distributions are shown for the reference signal (Sn) and for
greater and lesser signals (Sc and Sr) and there are two cutoff points
(xe and yr,).
It is not so clear what model is appropriate for cases in which, ssy,
two lines are presented simultaneously one on the left and one on the

40

Fundamentals of Skill

right (e.9. Henmon, 19o6, Vickers, r96il and the subiect has to say
which is longer. fn this case Sc and Sz are in effect presented without
SR. ![e may perhaps assume that one of three procedures is followed:

firstly the subiect may take one of the lines as Sn and decide whether
the other is longer or shorter; secondly he may imagrne some intermediate value between Sc and Sr as Sn and iudge both differences from
this simultaneously; thirdly he might simultaneously iudge Sc with .Sr
as a reference and Sz with Sc as a reference. Of these three possibilities
the first appears to be the simplest as it requires only one decision process whereas the second and third involve two.

z. Seoqal oalrcs of a single oariable presmted one at a time. In the discriminations we have discussed so far the qtrantities to be distinguished
have been present either simultaneously or nearly so. It has, therefore,
been possible to make closely comparative iudgments, and under these
conditions very fine differences can be recognised accurately. When,
however, the quantities are presented separately at different times - in
other words when they have to be iudged absolutely - the minimum
differences which can be reliably distinguished are substantially greater.
Lipsitt and Engen (196r), for example, forurd their subjeas were about
equally accurate at iudging which was the longer of trro lines presented
simultaneously or separated by an interval of r sec, but less accurate
when the two lines were presented 5 sec apart. Again, Pollack (1952),

presenting subieas with tones of varying pitches equidistant on a
logarithmic scale ranglng from roo to Srooo c.p.s. and requiring them

to class$ the tones by assigning numbers, found that only about five or

six classes could be reliably distinguished.
As already mentioned in Chapter r (B. zr) Miller (tgS6) has surveyed
data which indicates that comparatively small numbers of distinguishable classes are also found for iudgments of other quantities such as
loudness (about sk), tastes (about four) and points on a line (about
nine) (see also Spitz, tg67). fn short, capacity for absolute iudgment
appears to be severely limited. It is, of course, true that individuals
possessing 'absolute pitch' can recognise accurately a very much larger

ntrmber of tones (Carpenter, r95r) and similarly fine discriminations
can be made by expert indusuial workers when iuclging colours or
other sensory qualities in the course of their work, but how they do this
is at present a mystery.

If we envisage classification on a single dimension as taking place in a
system such as that of Fig. 2.2, we must assume that the quantity to be
classified $/ill produce an amount of activity in the system which

Sinple Deciions

4t

ranges from a level indistinguishable from noise up to some maximum
representing satnration ofthe system. If so, the nunrber of discriminable
classes will be the ntrmber of Signal-plus-Noise distributions that can be
fitted in between minimum and maximtrm with acceptably low overlap,
as shown in Fig. 2.6.
It is at first sight tempting to assume that this discriminable range is
directly related to the sensory mechanism and its cenual proiection, for
example in the case of pitch, to the signalling from different portions
J

bH
>J
--z uJ
-l
d)>

f;(c

o>
E, Z,
o-:<

14INIMUM

LEVEL OF ACTIVITY IN SYSTEM

Figrrre 2.6. Extension of the signal-detection model to cases where absolute iudgments are required of several values on a single 'dimension'.

of the basilar membrane and the representation of pitch in the auditory
proiection area. Adrian Gg4T, P. 50) cites Tunturi (rg4) as having
reported that in the dog equal octaves are represented by approximately
equal inten ds in this area so that discriminability might depend on
netrrological distance. In other cases we have already suggested that
discriminability is likely to depend on frequency of nerve impulss in
such a way that eqtral incteases in frequency represent equal increments
of discriminability. Such views in their crude form, howeverr are llrldepends upon the total
tenable because they imply that dis@
range of stimuli that can be perceived and should be independent of
the range acnrally presented in any grven circunrstances. Pollack (t952,
r953a) has shown that this is clearly not so: the nrrmber of discriminable
pitches remains about the same whether the stimuli are presented over
a wide range of frequencies or concentrated within a narrow one. The
same is tnre whether the frequencies are all towards one or other end
of the range. Similar results have been obtained for different loudness
levels (Hodge and Pollack, 196z): the information transmitted in classifilrng eight loudness levels was essentially the same whether they were
spread over a rsnge of 28, t4, 7 ot 3'5 db.
The implication is that the decision axis is in some way capable of
adiustment in such a way as to make the whole of its capactty available
for the range of stimuli acnrally presented. How this is done is not clear,
but one simple possibility is to assume fualy, as we have already done,

Fundamentals of Skill

42

S\ETEM SATUPAT
ts
F

U
RAl.lGE

J

W]THIN WHICH

d,

IS POSSIBLE

D

ur

DTSCR IM INAT I ON

l&

o
J

IU

u,

J

SIGNAL N OT

DISTINGUISHABLE

FROM NOISE

A

LoGs

Figrrre 2.7. Effeas on the decision uris of attenuating the signal (A)
before and (B) after a logarithmic transformation. In (A) the effect of
attennation is to increase the absolute values of log S over which discrimination can be made while leaving the range of values trnaffested.
In CB) the main effect of attenuation is to increase the range of values of
Iog S over which discrimination can be made.

that somewhere between the physical stimulus and its central effect
there is a logarithmic transformation; and secondly that the signal can
be attenuated either before or after this transformation has taken place.
Attenuation before the transformation would essentialty subtract a constant amount from log S, and different levels of attenuation would produce a series of relationships between log S and its central effect as
showninFig. z.TA.The effect would be to determine the portion of the
total range of stimuli over which the decision mechanism operated,
without affecting its extent - for example it might shift from low pitches
to high. Attenuation after the logarithmic transformation would alter
the slope of the relation between log S and its cenual effect as shown
in Fig. 2.78, and affect the range over which the decision mechanism
operated - for example it would determine whether discriminative
capacity was spread over a wide range of pitches from high ro low or
was confined to one part of the scale. The fine discriminations attained
in comparative iudgrrents might on this view be due to the possi-

bility, when both quantities are present together, of concentrating
the whole discriminative capacity on a very narrow band of stimulus
values.

The mechanism postulated is not without its difficulries. For example,
Pollack (tgSz) noted that the number of classes discriminated was the

St*?lc Decisions

B

43

LOG S

same whether the frequencies were spread at regular logarithmic intervals over a range or whether they were divided into two groups, half
at the high end and half at the low. Such a finding implies some more
complex adjustment of the decision axis; it might, for example, be
accounted for on the assumption that some adiustment of the range
takes place whenever a sigual is presented so that when, ssy, one of the
high group is presented the subiect adiusts his scale towards the high
end and when one of the low group, to the low end. If so, experiments
such as Pollack's could be linked to those in which subiects, when
asked to classiff qnantities, adjust their iudgments to the range of
quantities presented (e.9. Tresselt and Volkman, l94z) and to those of
Helson (e.g. t947, 1960 and others who have shown that a given
quantity tends to be iudged greater when preceding quantities have
been small than when they have been large. The term 'adaptation'
used to describe such adiustments could fairly be applied also to those
we have postulated.
A second diffiorlty lies in understanding what is adiusted. Hodge and
Pollack (t962) have produced evidence that it is not the discriminability
of signals so much as the assigRnrent of responses that is affeaed by a
change in the range of qtrantities presented, and Parducci (1965) has
suggested that adiusment tends to be such as to make the frequrcies
of the various categories of iudgpent equal. We cannot attempt to
meet these difrculties here; it seems fair to remark however, fitstly
that some mechanism of adiustrrent has to be postulated, and secondly
that this might reasonably be taken to operate on the decision axis
mediating betrn een signal and response.

44

Fundamentals of Skill

3. Quantities varying in more than one dimension simultancously. The
number of distinguishable categories is greater if the quantities do nor
lie along a single stimulus 'dimension' but along two or more dimensions simultaneously, such as pitch and loudness or if points have to
be identified in a square instead of along a line (Kleulmer and Frick,
1953, Pollack, r953a, Pollack and Ficks, 1954, see also Miller, 1956).
X2

//

/
//

/,
/
// /

, '1r'

,u r'

/

-t-

./

I

//

/
//

,/

I

r T
I

I

I

I
I

X1

Figrrre 2.8. Extension of the signal-detection model to cases where signals
vary on two independent 'dimensions'. Based on a diagram by Rodwan
and Hake (1964). The separation of the two distributions on the oblique
axis is greater than on either of the others.

The total number of categories is, however, considerably less than the
product of the number for each dimension separately. For example
Pollack, who found that about six pitches and five degrees of loudness could be discriminated, found that only about nine simultaneous
diffferences of both pitch and loudness could be reliably distinguished
instead of the 3o (i.e. 6 x S) which might have been expected.

SWk Decisions

45

Several snrdies have indicated that when subieas make iudgments in
terms of rwo or more characteristics or 'dimensions' they detect differences of each independently and that the total amonnt of differentiation
they achieve results from a combination of these. It is therefore possible
to predia the accuracy of discrimination for two or more dimensions
together from the accuracy attained on each separately. The overall
result will, of course, depend on whether the dimensions are correlated
and on whether accurate response requires correct iudgmeut on all
dimensions simultaneously, or on any one or more aloner or whether
s,

52

HAS ARRIVED

S3

S4

s5

Figure 2.g.Extension of .n. ,;;l-detection model to cases where there
are several possible signals each on an independent 'dimension'.
some more complex synthesis is required such as when iudging personal
characteristics from faces.

Rodwan and Hake (rg6+) have suggested, on the basis of their
experiments, that synthetic two-dimensional discrimination can be
conceived as in Fig. 2.8: they assume that subiects combine the discriminations on the two dimensions with an overall result as indicated
by the circular 'distributions' and by the oblique decision axis on to
which the other two are proiected. The angle of this new axis depends
on the relative weights given to the two dimensions in the ioint decision.
The authors found that different subiects behaved as if they placed the
oblique axis at substantially the same angle, implyrng that they weighted
different dimensions in roughly similar ways. This model can, of course,
be extended to disc,riminations involving three or more dimensions.

4. Sqserat signak each on a different 'dimension'. When one of several
possible signals, each independently liable to be disnubed by noise, is
presented, the sinration can be represented as in Fig. 2.9. Of the total

46

Fundamentals of Skill

number of items one is correct and has a signal strength of d', while
all the others have a signal suength of 'noise alone', i.e. zeto level. For
simplicity all the distributions are assumed to be of equal variance and
normal although in practical cases this would almost certainly not be
exactly true. This kind of model has been suggested by Green and
Birdsall (1964) for words in a vocabulrry, and would seem applicable to
cases where stimuli may arrive over different sensory channels. The
6

/

5

d

-1'-

-a'J

//
/

I
0

-f
Ht

/,

4

0

-J"tJ

I4
0.1

1

0.2

/'
//'
-/

z/
ar-

-zr/

0.3 0.4

-/

0.5 0.6

0.7
PROBABILITY OF CORRECI RESPONSE

1
0.8 0.9

1.0

Figrrre 2.ro. Value of d' required to produce a given proportion of correct responses from the type of case represented in Fig. 2.g. Based on a
table by Elliot (1964).
subiect's task is to decide which of the possible signals has occurred.
At least two methods of doing this seem plausible:
(o) Green and Birdsall suggest that the subiect in effect examines the
strengths of all the alternatives at a particular instant and chooses the
largest. If so, the greater the number of alternatives, the greater the
chance that one of the noise levels will, at the moment concerned,
exceed the signal level of the correct item and lead to an error being
made. If accnracy is to be kept constant, d' must therefore rise with
the nurnber of alternatives. The values of d' for various proportions
of correct response with numbers of alternatives from z to rrooo
have been tabulated by Elliot (rg6+), and are shown graphically in
Fig. 2.ro.
(b) An alternative method would be to ser a criterion x, and having
done so to examine the alternatives until one was found which exceeded
it. With this procedure, the greater the number of alternatives the

Simple Decisions

4T

greater the number that would, on average, have to be examined before
one exceeding the criterion was found, and the greater the chance that
the criterion would be exceeded by a reading from one of the 'noise
alone' distributions so that a false positive would occtu. Taking p as
the probability of a false positive from any one'noise alone' distribution,

Total probability of false positive from ,?, samples

p)*
- r - (t - (2.r3)

where rn is the number of incorrect alternatives sampled. A table of
r - (r - p)* for various values of p and m has been provided by
Wiener (1965).
If the subject is to keep his false positive rate constant, he will have
to raise his criterion level progressively more and more as the
number of alternatives increases. For example, if he wishes to maintain
the rate at ro/orhewould have to set his criterion level at 2.33 standard
deviation units above the mean of the 'noise alone' distribution if he
had to examine only one possible sigRal, at 2.5T if he had to examine
two signals and at 2.8r if he had to examine four. We may note that the
effect of increase in the number of alternatives is greater for relatively
low or 'risky' criteria than for strict or 'cautious' ones. For example if
the false positive rate had been rco/o, the criterion would have had to
be set at r -28 standard deviation units with only one signal, at r'63
trnits with two and at r.94 with four. The difference betrreen two and
four signals is '48 standard deviation units for the rl, criterion and
.66 for the roo/o. The difference might in practice be gteater since the
absolute change in the false positive rate if no change of criterion was
made would be much more noticeable and serious with a low criterion
than with a high one: the changes from one to four signals would be
from to/o to just under 4% and from too/o to iust over 33o/o.
This kind of procedure could explain a fincling by Broadbent and
Gregory (t963b) which is otherwise difficult to understand. Their subieas watched three fluorescent nrbes flashing rhythmically and had to
report when one flashed brighter than usual, grading their responses
into five categories of confidence. The most cautious criterion was
fotrnd to change little as benn een relatively quiet conditions and loud
ambient noise, whereas the most risky criterion rose so that the two
criteria came closer together.
If, as appears to be the case, d' increases with the duration of a signal
(Green et al., 1957, Egan et al., 1959, Swets, 1959) the same procedure
can also explain results obtained by lohn (1964). His subiects had to
respond by pressing a keyto a light which came on at irregular intervals,

48

Fmdamsntals of Skill

while ignoring occasional sounds presented over earphones. The reaction time to the light was longer when the sounds were relatively loud
than when they were softer. We may suppose that subiects raised their
cutoff level to exclude the louder noise and had therefore to attain a
greater d' in order to be srue of responding to the light.
THE TIME REQUIRED FOR DISCRIMINATION

It has been known from the early days of the study of reaction times
that the time taken to discriminate benreen two signals tends to rise
as they come to resemble one another more closely. Thus, for example,
Henmon (19o6) found that the time required to indicate which of two
lines exposed on a screen was the longer, increased over a series of
differences ranging from ro and 13 rnm to ro and ro.j. The same author
fotrnd similar results for pairs of tones of different pitches sounded one

immediately after the other, and for patches of different colours.
Parallel results have been obtained by Slamecka (tg6l) whose subieas
had to decide which of two words was the more similar in meaning to a
third word. Results which may perhaps be regarded as falling within
the same area have been obtained by \trallace (1956) who found that the
acctuacy with which patterns and picnrres of obiects were identified
rose with the time for which they were exposed, and by several authors
(e.9. Barry, 1964, Carterette and Cole, 1963, Haber and Hershenson,
1965) that accuracy of discrimination or identification increases when
the stimuli or messages are presented more than once. \I[e may note in
passing that both Wallace and Haber and Hershenson found that several
brief exposures were less effective than one longer one of the same total
duration. A tie-up with the signal-detection theory approach exists
directly in the findings already mentioned that d' rises with increased
duration or repetition of signals, and by implication in results obtained

by Pierrel and Murray (tg6l) who forurd that the time required to
discriminate weights from a standard increased, and at the same time
both accurary and confidence fell, as the difference from *re standard
became less.

Attempts to work out a quantitative relationship between reaction
time and discriminability seem to date from a pioneering attempt by
Crossman (1955) who conducted a series of experiments in which subiects sorted specially prepared packs of cards according to numbers of
spots on the cards. Each pack contained equal numbers of cards with
each of nvo different numbers of spots, and the time taken to sort packs
was noted and related to the differences between the numbers. Crossman began by confirming a result obtained by Henmon, that equal

SWe Decisions

49

ratios are discriminated in approximately equal time - packs containing

cards with r and 2, 2 and 4, 3 and 6, 4 and 8, 5 and ro spots all took
about the same time to sort. Any formula or law relating reaction time
to fineness of discrimination must therefore provide for time to rise
as the ratio between the quantities concerned becomes smaller but to
be little affected by changes in absolute magRinrde. In other words,
the time taken for discrimination appears to be a funaion of the Weber
Fraction benneen the two quantities. Crossman considered three possible functions and associated models which we shall discuss in turn
using, however, slighdy different terms from his.

r. Statistical sampling
Accordirg to this approach, the brain is regarded as taking a series of
brief samples of the data presented and averaging them. The samples
are conceived as having a variance due to noise which causes the disuibutions for the two quantities to overlap. Accorclirg to well-known
statistical theorerns we should expect the standard deviation of the mean
of a series of samples to narrow at a rate proportional to the square
root of the number of samples thus:

oN: L

\N

(2,t4)

where o1 is the standard deviation for a single observation and or that
of the mean of N observations. If all samples take the same time we
can write for the distribution associated with each of the quantities to
be discriminated:

oP-Go, x aconstant

(2.r5)

when or is the standard deviation of the mean of a sample obained
after tfune T. We may suppose the subiea to go on taking samples until
the disuibutions have reached a sitical separation, ssy when the overlap
is small enough to produce an acceptably low frequency of errors.
This approach has an obvious affinity to sigual-detection theory: we
can write
or

d'r - d', \n x a constant

\/7, -- ftr" a constant

(2,t6)

(z.r1)

where I, is the time required to achieve a critical separation of the distributions and d' , is the value of d' at time Tr. Evidence in support of

50

Fundarnentals of Skill

this formulation comes from results by Gre en, et al. (tgS) and Swets
et al. (195g). The formerfound that d'rose linearly with {T when T
exceeded about roo msec. The latter, using brief tones presented in
external noise and allowing subiects to observe them several times
fotrnd that d' rose linearly with the square root of the number of
observations. Further supporting evidence comes from an analysis by
Taylor et al. (tg6il of data obtained by Schouten and Bekker (tg6l)
from a two-choice reaction task in which the subject had ro decide
which of two lights had appeared. Taylor et al. calculated from the
relationship benneen error frequencies and reaction time that d'2 rose
linearly with T.
If Eq. 2.r2 holds we can write,

\/Tr:, logSc-logS;z\*\
=d''o!^=o xaconstant

(2.r8)

which implies that if error rate is held the same for different degrees of
discrimination,

Meandiscriminationtime:(

xaconstant (2.r9)

The treatment outlined here assumes that both Sc and S.u are presented together, but it can be extended to the case where only one is
present at a time by assuming that the subject carries traces of both
quantities in memory, or a uace of some average of them (cf. Hughes,
\e64).
A somewhat similar result is reached by a different route with a type
of model based on Wald's (tg+il Sequential Probability Ratio Test,
inuoduced by Stone (t96o) and developed by Laming (lg6z) and by
Shallice and Vickers (r 964). The model deals with the case in which
the subject is presented with only one quantity at a time and is required
to state whether it is the larger or smaller of rwo possible alternatives.
He is assumed to take a series of samples of the input data, each taking
an equal time and liable to be added to or subuacted from by random
noise. He makes a running total of the samples, and when this reaches
a pre-assigned value of probability that the samples came from Sc, he
decides for ,Sc. Similarly if the total reaches a pre-assigned probability
that they came from Sz he decides for that. The procedure envisaged is
illustrated in Fig. 2.rr: the running total approaches the criterion by a

so-called'random-walk'.
Laming assumes that the cenual correlate of the signal, on the basis
of which the decision must be made, is normally distributed with constant variance. As Shallice and Vickers point out, if the model is to be

Si*?k Decisims

5r

used to relate reaction time to different degrees of discrimination, it is
necessary to make an assumption about how the cenual processes are
related to the physical input, and they assume in accordance with
Eq. 2.6 and with Crossman's and Henmon's findings, that the mean
of the cenual activity varies as log S. On this basis they produce the

following equation:
Mean discrimination time

tzoz(r - a - D)log

(r

r -b)

(z.zo)

ab

(log Sc
- log Sr)z
where I is the time for any one sample and a and b arc the proportions
of errors on .Sc and Sz. If errors do not vary from one discrimination to

I

OENTI:Y

AS 5r

START

IOENTIFY
AS SG

Figrrre 2.rr. 'Random-walk' model for decision between two alternatives.

another, Eq. 2.zo reduces to Eq. 2.r9, although the constant has, of
course, a different meaning.

The model might perhaps be extended to the case in which two
quantities are presented at a time by assuning that both are sampled
simulaneously.
We have spelt these models out in some detail as they are attractive in
many ways and are enioyrng a considerable vogue. Unfortunately, however, Eqs. 2.rg and 2.zo do not fit the experimental facts. In a few cases
ploning reaction time against them gtves a reasonable fit to the data,
but in these c:tses alternative funaions derived from other models do
equally well. In other cases the alternative functions clearly fit better.
Either the models or some of the assumptions made in applyrng them
must be wrong.

Fundamentals of Skill

52

2. An information-theory model
Crossman considered but rejected a model based on informationmeasurement which proposed that discrimination time was proportional to the logarithm of the reciprocal of the Weber Fraction thus:
Sc
(z.zr)
lt
Mean discrimination time

tgftxaconstant

-

1.4

1.2

d
IJ.J

a

ul

0.8
- -€--o--o---o-

t=
-

z,
o
F

---

C)
UJ

G,

0.4

PER CENT DIFFERENCE

0,2

504030 2015 l0 7

5

1.32

0
LOG

2

(

Figure z.rz. Times taken to decide which is the longer of two lines shown
simultaneously, plotted against Eq. 2.2r. Data obtained by Birren and
Botwinick (1955). Open circles, subjects aged 6r-9r : filled circles subiects
aged tg-36. Each point is the mean of the medians of 43 older or 30
younger subiects. The medians were each based on at least four readings.
The differences benreen the intercept of the two regression lines and
the intercepts of the two dotted lines are approximately the times to be
expected for a two-choice reaction by subiects in the age groups concerned. The results therefore suggest that reaction tirne was the time
taken by discrimination or by choice, whichever was longer. The additional .27 S sec from zero to the intercept of the regression lines was
probably due to a time delay either in the apparatus or in the execudon
of the reponse. (After Welford, r96oc.)

Simple Decisions

53

Eq. 2.2r does in faa provide a strikingly good fit to Henmon's data for
lines and tones, and also to results obtained by Birren and Botwinick
(rgSS) and Bonrinick et al. (t958) who presented pairs of vertical lines
and required subiects to say whether the longer was on the left or on
the right. Birren and Bonrinick's results are shown in Fig. 2,!2: the
linearity is good if it is assumed that there is a lower limit to the reaction
times concerned, determined perhaps by a minimum time required to
choose which of the two responses - 'left' or 'right' - to make.
However, Eq. 2.2r does not fit further data obtained by Crossman or
any of the data on discrimination times obtained by other authors.

3. Crossman's Confusion Funaion
Crossman's third, ffid favoured, suggestion was that the time taken for
discrimination was, as we have already mentioned, linear with his
Confusion Function so that
Mean discrimination time:

#

X a constant (z.zz)

or in other words, discrimination time is inversely proportional to the
Xfeber Fraction betrreen Sc and Sz and d' increases linearly with time.
Eq. 2.22 gave a good fit to two experiments by Crossman himself. In
one, subiects sorted packs containing equal numbers of cards with one

of two numbers of spots: to/t, rc/5, tz/8, r2/9, to/8 and tz/to.
In the other they sorted fi small canisters by weight - eight lighter
and eight heavier affanged in random order - with, in different trials,
the ratios of z/rz, 4/r2, S/tor 6/g and 6/8. The card-sorting results
are shown in Fig. 2.r3.
The good fit of Eq. 2.22 has been confirmed by McCoy (1963) for
diameters of circles and shades of grey as defined by reflestance values,
presented either as a card-sorting task or with each card shown separately and exposed until the subiea responded by pressing one of trro
microswitches. In these experiments, unlike Crossman's, t\tro quantities
were always presented simultaneously. Further confirmatory evidence

was obtained by Shallice and Vickers (lg6+) using card-sorting tasks
with either one or two lengths of line on each card. These experiments
were undertaken in an attempt to reproduce some of the essential
feanres of the experiments by Birren and Botrnrinick and Bomrinick
et al. in a card-soning form in the hope of discovering why the Information Theory model fitted some results and the Conftrsion Fnnction
others. They were followed by a substantial series of experiments by
Vickers (tg6il using pairs of lines proiected on a screen. In all cases

Fmdamentals of Skill

54

o
r050

o
o
e
lJ.l
CL

L

u)

o
zo
()
uJ
a
J

o

850

o

=

750

ll/s t?,la 12/g i0/8 lLln
0

@

t

Figrrre 2.r3. Crossman's (rgSS) data for sorting cards into tlvo categories
plotted in terms of Eq. 2.22. Each point is the mean time per card for
four subiects each sorting 8o-rzo cards.

t.2

-U l.o
lrl

a

trl

tr

z

o

tr
U

lrl
d,

o

to

20

30

LOG2 56 LOG2 Sg
Figure 2.r4. Times taken to decide which is the longer of two lines shown
simultaneously in the presence of visual noise, plotted against Eq. 2.22,
Data obtained by Vickers (r 967). Each point is the mean of 16 readings
obtained from each of rB subjects.

SWt Decisions

55

the Conftrsion Funaion gave a much better fit except with some very
fine discriminations, for which the observed t'nes were much shorter
than predicted by Eq. 2.22. An example of Vickers' results showing this
effect is given in Fig. 2.r4. These fine discriminations were, however,
associated with substantial ntrmbers of errors whereas the frequencies
of errors in cases where Eq. 2.22 fitted well were low.
The simplest way of accounting for these results is to asstrme rhat
data are accrunulated, or some central state builds up, linearly with
(log Sc
- log Sr,) until a critical level is reached. The model envisaged
can be represented as in Fig. 2.Tr except that the progress is in a
straight line towards one criterion or the other. Thus stated, however,
the model takes no account of noise and makes no allowance for errors:
if data from the signal are accumulated over time, ought not the noise
also to be accunulated, and would this not lead to Eq. z.rg or 2.zo
rather than z.zz? Of various possible methods of overcoming this
difficulty we may briefly consider three:
(") In many cases the random variations which constinrte tre noise
might well be the same for both Sc and Sz when they are presented
together, or for both signal and remembered reference standard when
Sc and Sz are presented separately. If so, any noise which accumulated

would do so equally for both, and an accumulated differutce would be
free of noise.
(b) It is usually assumed that the main sources of noise occur prior
to the accumulation of data, but as regards internal noise this is not
necessarily so: it might be that the main source of noise was subsequent
to the store: acctrmulation might take place in the percepnral mechanism

of Fig. r.3 (p. l9) while the main source of internal noise was in rhe
translation mechanism. Both these possibilities may perhaps enable
Crossman's and subsequent results where the noise was mainly internal,
to be reconciled with those of Green et al. (tgSil and of Swets et al.
(tgSg) who fotrnd d' to rise as the square-root of time, since in their
case the noise was mai.ly external.

(r) Before dealing with our third possibility we'need to consider

what sets the limit to discrimination. ff Eq. 2.22 holds without qualification there should be no limit, given sufficient time, to the fineness
with which discrimination can be made. Clearly this is not so, and the
question arises of why not? It cannot be due to a constant sugh as r
in Eqs. 2.4 and 2.5 because although this would slow discriminations
of very small absolute magninrdes (Steirunan, ry4D it would leave
discrimination of larger magninrdes virnrally unaffected. Nor is it likely

to be due to a timelimit beyond which data cannot be effectively

56

Fundamentals of Skill

accumulated because any such limit would have to differ widely from
one experiment to another to account for various results which have
been obtained (Hughes, 1963, Green et al., 1957, Vickers, tg6l).
A more promising line has been suggested by Vickers following

experiments in which he superimposed on his projected lines I randomly varying pattern of spots by means of a cine-film running at

IDENTIFV
AS SL

START

IDENTIFY
AS SG

Figure 2.r5. Modification of the model shown in Fig. 2.t2 proposed by
Vickers (1967).
16 frames per sec. The film has been described by MacKay (tg6j). The
effect was to reduce the black lines on white ground to somewhat illdefined areas of darker spots within a larger area of lighter spots, all on
a white ground. The numbers of spots making up each line varied
from frame to franre producing a distribution for each ling. Vickers
counted the acnral spots in the line areas for samples of frames with each
of several differences between the lines, and showed that Eq. 2.22
broke down at the point where substantial numbers of frames showed
more spots in the shorter line than in the longer. One plausible suggestion is that the subject adds up the positive and negative quantities
taking their sign and magninrde into account, until a critical total in
favour of one decision or the other is reached, but this would imply a

model like that of Fig. 2.rr and that the data should fit Eq. 2.r9,
which they do not. Vickers' alternative suggestion was that the subiect
accumulates positive and negative quantities separately, deciding in
favour of whichever attains a critical total first, as indicated in Fig. 2.!5.

Simplc Decisions

ST

If this is so Eq. 2.22 no longer applies strictly for nno reasons which,
however, partly offset each other. Firstly, the attainment of the criterion
is slowed owing to time lost when data are accumulating in the'wrong'
store. Secondly, the rate of acctrmulation in the 'correst' store is faster
than the mean (log Sc - log Sz) would suggest. The reason is indicated in Fig. 2.16, and is due to the fact that the average rate for

*H

J=
30
o(L
=(n
EH
I u')

o5
ll-

a

r2
z.A

ruE

6s
-1

Ae
lL=
'rJnff"lllEEr-*i
DIFFEREISE BETWEEN OUANTITIES BEING COMPARED

Fignre 2J6. Diagram illustratiqg how, when two quautities to be discriminated are disttrrbed by noise, the effective mean difference benreen
them may be increased acording to the model illustrated in Fig. 2.t5.
The distribution is of the momentary difrerences betrnreen the qtrantities
being compared. Zero difference is indicated by the leftnrost of the three
vertical lines which ctrt the distribution.
moments when the difference between the ntrmbers of spots in the two
lines is 'correct' is based on instances on one side only of the zero

difference point. The net effea of these two factors is to produce a
leveltiqg off of discrimination time when substantial ntrmbers of errors
are made. We may note that Figs. 2.tS and 2.16 imply that even if
Sc ud Sr are exaaly alike, some decision will wentually be reached
since the average rate of accumulation will still be appreciable - it will,
if the distribution in Fig. 2.16 is norrnal, be as if (log Sc - log Sr)
eqnalled the probable error of that distribution, i.e. '674o.
If reaction times level off with very fine discriminations, the Confusion Ftrnction can account quite well for the results of Birren and
Bonrinick (lgSS) and of Bonninick et al. (1958). The data of Fig. 2.rz
replotted against Eq. 2.22 are shown in Fig. z.r7 to fall into a pattern
similar to that of Vickers' results in Fig. 2,r5. If this is accepted, all
the available data are well fitted by the Confirsion Funstion except
Henmon's (19o6) and those of Green et al. (t95il, Swets et al. (lgSg)

58

Fundarnentals of Skill

and Schouten and Bekker (tg61). Crossman (lgSS) claims that Henmon's results are reasonably fitted by Eq. 2.22 and Vickers has pointed
outthat Henmon's lines could hardlyhave been drawn acclratelyenough
to merit precise treatment. The results of Green et al. are probably
involved with threshold effeas so that they are hardly relevant to our
present discussion, and those of Swets et al. are really for a different
1.4

-"""2

o"4--

OLDER

L,
trl

a

ul

=
F

z
o

a-

- - --,--4'
-

--t'a'

YOUNGER

F

o
trl

E, 0.7

PERCENTAGE DIFFERENCE
0
LOG2 SG- LOG2 SL

Figure 2.t7. The data shown in Fig. 2.r3 plotted against Eq. z.zz.

type of task: instead of grving a single exposure until decision is obtained, the subiea is given a series ofexposrues on each of which he may
be presumed to make a tentative iudgment before arriving at a final decision. Schouten and Bekker's task is also of a different tlpe since their

subiects were making choices benn een clear signals rather than fine
discriminations.
The Confusion Function as represented in Eq. 2.22 seems, therefore,
to be able to account in a general way for all the relevanr data available.
It must be emphasised, however, that it can only provide an approximate, overall statement. There are likely to be several comp[cating
factors in particular circumstances, especially at low absolute stimulus
magninrdes. Other complications are implied in results obtained by
Audley and Wallis (lg6+), l7allis and Audley Gg6+), who found that
responses were quicker to ttre brighter of two relatively intense stimuli

Simple Decisions

59

or the higher of two high tones, but slower to the brighter of two relatively weak stimuli or the higher of trro low tones.

A great many snrdies have shown that reaction time shortens as the
intensity or duration of the signal is increased (e.9. Pi6ron, rgzo, 1936,
Chocholle, r94o, 1943, Raab, t962, Kaswan and Young t965a, b) and
it seems reasonable to suppose that this is trnderlain by similxi processes
to those considered in this chapter. The problem of working out a
coherent scheme of funaional relationships berreen stimulus intensity
and reaction t'me is, however, formidable. Some of the present data
are not precisely quantifiable, many results are insufficient for treating
in the present terms and many also seem likely to be complicated by the
factors which operate at low absolute magnitudes. Added to all this,
high intensities of ambient stimulation may have a facilitatory effect on
response so that constant dS/.S may yield shorter reaction times when S
is relatively intense (Raab and Grossber& 1965). Again, short duation
signals sometimes yield relatively short reastion times (e.9. Botrnrinick
et al., 1958), as if the subiect decided that no further data were forthseming and so that he might as well react at once. We shall, therefore,
not attempt to consider this evidence here. It seems fair to suggest,
however, that if what has been said here is tnre, reaction will normally
occur when d' has been built up enough to reach a critical level. The
essential problem is thus to understand the several factors contributing
to d' and the course of their operation over time.

III
Identification and Choice

Why does it take time to react? Vhat happens during a reaction time?
The answers to these questions which have been grven during the last
20 years are fundamental to much of our present understanding of the
factors determinirrg the speed ond, as we shall see, the accuacy of
performance. Broadly speaking, reaction time as usually measured
includes, first, the time taken by the stimulus to activate the sense
organ and for impulses to travel from it to the brain; second, the cenual
processes concerned with identiffing the signal and initiating a response
to it; and third, the time required to energise the muscles and to produce
an overt recorded response. The first and third components are in
most cases relatively short although there are substantial differences of
reaction time to stimuli applied to different sense organs - for exaurple
visual reactions commonly take some 50 msec longer than auditory.
FIow far this is due to the sense organ and how far to the associated
central mechanisms, however, is at present open to question. In any
case, most studies which have attempted to establish laws about reaction
time have been able to proceed as if the whole time is taken up by
cenual processes, without running into anornalies. This is surprising
and merits more thought than it has hitherto received. Are, for example,
the three components uuly successive in time, or do they overlap so
that cenual processes begin before the signal has been fully received
and motor processes are initiated before the prograrnme for their fine
control is complete?
Dtuing the time taken by the central processes the subiect resolves
uncertainty arising from two sources: firstly, he may not know exactly
wlrcn the signal is coming and therefore when to respond; secondlS in
choicereaction tasks where different responses have to be rnade to
each of several possible signals, he may not know which signal is coming
and therefore which response to make. The fact that choice-reactions
take longer than simple has been known since the pioneer experiments
of Donders (1868) who distinguished what he called the a-reactiotrr
60

Identification and

Choice

6r

with only one possible sigual and response, from the b-reaction having
more than one and thus involving identification and choice. It was also
recognised that reaction-time rose progressively with the number of
possible choices, but why and to what extent were not understood.

HIcK's INFoRMATIoN-THEoRY LAw
An important break into this last problem was made by Hick (r95za)
who proposed, on the basis of his own data and also those of Merkel
(t885), that in making choice-reactions the subjea gains 'information',
in the information-theory sense of the term, at a constant rate.
Merkel had presented his subiects with signals rangrng in different
trials, from one to ten alternatives. The signals consisted of the arabic
numerals r-5 and roman numerals I-V, printed round the edge of a
disc. The subiect waited for each signal with his fingers pressed on ro
keys snd, when a number was illuminated, released the corresponding
k.y. The arabic numerals corresponded in order to the fingers of the
right hand, and the roman to the left. When less than ro choices were
required some of the numerals were omitted.
Hick's own experiments used as a display ro pea-lamps arranged in a
'somewhat irregular circle'. The subject reacted by pressing one of
ro morse keys on which his fingers rested. Choices of less than ro were
again obtained by omining some of the lights. The frequencies of the
various signals for any given degree of choice were carefully balanced
and presented in an irregular order so as to ensure as far as possible
that the subiect should not be able to predict what signal was coming
next. Each light appeared S sec after the completion of the previous
response - an interval too long for the subiect to iudge accurately when
the sigual would appear.
Hick fotrnd that if the number of possible signals is taken as n and
reaction time is plotted against log (n + r), the observed reaaion times
for different numbers of signals lie on a straight line which dso passes
through the origin, as shown in Fig. 3.r. \[e can thus write
Mean choice reaction time - K log (z f

l)

G.r)

where K is a constant. If we work in logarithms to the base 2,
log (z f r) : r when n - r, so that K is the simple reaction time - a
convenient result. A table of these logarithms for whole nunrbers up to
roo is given in the Appendix. The log, unit is known as the brr.
The obvious question arises, why (n + r) and not z? Hick pointed
out that if the subject is uncertain when a signal will appear he is faced

6z

Fmdamentals of Skill

with the task, when it does appear, not only of deciding which it is, but
also of deciding that a signal has ocqured at all: failue to do so will
result in his either reacting when there is no signal present or failing
to react when there is one. The additional task of guarding against such
errors can be conceived as adding one to the number of possible states
of affairs that he has to distinguish - instead of states corresponding to
signals rt 21 3, . . . nhe has to deal with states corresponding to o, Tr 22 Jt
0.6

Li
UJ

(/)
tr.l

=
F
z, 0,2

o

F

c)
u,

G 0.1

0

234

n
6

l0

L0G2 (n+ l)

Figure 3.r. Data from a choice-reaction experiment by Hick (r95za)
plotted in terms of Eq. 3.r. The total number of reactions represented
is over 2r4oo recorded after extensive practice.

. . . n. lf the subiect were in no doubt when a signal was comingr 8sr
for example, if he were to determine the point of time himself at which
the signal light came on, the * r in Eq. 3.r would not be required since
there would be no temporal uncertainty to be resolved. We may denote
the sum of the possibilities includi.g 'no signal' as N, defining N as
the equioalmt total number of equally probable alternatives from which
the subject has to choose, and may then rewrite Eq. 3.r as
Mean choice reaction time - K log N

(3.2)

This formulation we may call Hick's Law.It should be trnderstood that

it is an 'ideal' formula and that time lags in the apparatus or in the

making of a response may add a constant to the time: for example
Costa et al. (lg6S) have shown that 4o-To msec may elapse between
the first recordable electrical activity in the muscles and the making of

Idmtification and Choice

63

a mictoswitch contact. Hick took elaborate care to avoid lags due to
apparahrs in his experiments. Since reaction-time can be a remarkably
precise measure, such accuracy is well repaid in clarity of results.
Hick's approach was quickly extended by Crossman (lgSf) to another
task demanding identification and choice, namely the sorting of playing

/
2,0

a
()

,/

//

/

./

//

LU

(n

o
E,

; r.o
lrl

&

:--

Lu

=
tr

./

FACE DOWNWARDS
FACE UPWARDS

/
/
/
0

0

5

Zn
Figrrre 3,2. Data from a card-sorting experiment by Crossman (lgjg)
The points along each line represent, in order from left to right:
Deding cards in pre-arranged order
Soning into Red/Black.
Pictures /Red plain/Black plain.
LOG

Suits.
Red picttrres/Black picrures/Plain in 4 suits.
Suits, dividing 6 and below from 7 nd above.

Numbers.
Numbers, separating red and black.

cards. The subiect held a well-shuffied pack face down in one hand:
with the other he nrrned up the cards one by one and sorted them into
various classes, working as quickly as possible without making errors.
The number of classes was varied in different trials from z (Red/Blac$
uP to z6 (the 13 numb€rsr dividing red from black). Additional trials

64

Fundamentals of Skill

were given in which the cards were in a pre-arranged order such as
alternate red and black so as to provide a measrue of the time taken

to turn and place the cards when no identification or choice was
required - in other words a measure of movement time. The results are
shown by the upper line in Fig. 3.2. Roughly speaking

G.E)
- Movement time + Klogn
where z is the number of piles. The * r of Eq. 3.t is omitted in this

Time per card

case because, as the pack was always available, there was no trncertainty

about when a fresh signal would appear. The point that falls farthest
away from the line is that for the 13 nurrbers: these seem often to be
easier to deal with than less familiar sets of signals. 'W'e shall return to
this point later.
The question is sometimes raised of whether it is sufficient to estimate movement t'me by dealing on to only two piles: should dealing
not be on to as many piles as there are classes to give a different estimate for each ntrmber of classes? An indication that this elaboration is
unnecessary is contained in the results Crossman obtained with the
pack held face up so that the subiect could see each card immediately
he had removed the previous one. The lower line of Fig. 3.2 shows that
the time per card was either the movement time or roughly Klogn
whichever was the longer. It appears that identification and choice can
overlap with movement, so that the two can develop together. Extra

movement time required with large numbers of piles can thus be
absorbed in the extra time needed for identification and choice.
Creneralisation of Hick's Law

Hick's line of approach is supported by three further findings:
(a) The amount of information transmitted in a choice reaction task
is reduced if the signals are not all of equal frequency. The amount of
information due to uncertainty about which signal will occru can be
worked out by summing the amounts of information conveyed by each
signal weighted according to the probability of its occurrence. We can
therefore write in place of log n in Eq. 3.3

io,rori,
and in place of log (z f

G.+)

t) in Eq. 3.t

iu,,"*(i*,)

(r.s)

Idmtifuation and Clwice

65

where pe is the probability of each signal in the set taken in turn. These
expressions reduce to log n and log (z
r) respectively when all the

f

probabilities are equal. Hyman (tgSf) fouud reductions of about the
expected amounts in average choice reaction times when signal frequencies were unequal, and Crossman (tgSf) found approximately the
expected reduaions in the times to sort packs of cards containing
nnequal frequencies of d,ifferent classes. A table of p log,
,AP
+ and of

p losrG . ,) tr given in the Appendix.
(b) Information transmitted is also reduced when signals tend to
follow one another in recogRisable sequences or when any signal is
followed by any other more often than expected by chance, even though
the overall sigual frequencies are equal. This is really an extension of the
foregoing case: the probabilities of different signals are functions of previous signals, and are thus unequal at any given point in the series although the inequalities even out over the series as a whole. Hyman
(lgSl) found the expected shortening of average reaction times in these
cases.

(c) The amorurt of information gained is reduced if the subject
makes errors. A convenient method of cdculating the amount gained
when effors are present is to make a table with, say, a column for each
signal and a row for each response, including o in each case where
appropriate, and to enter the responses made to each signal in the
corresponding cells. We can then write

Information gained

:zIr"bs*+ > pnbs*-)a",roe* 6.6)
where ps is the probability of signals in each column, ?Rthe probability
of responses in each row and Psathe proportion of signal-response pairs
observed in each cell. The summation X is made over each coltrmn,
row or individual cell respectively.* This equation is a flrndamental one

for calculating the information uansmitted from signal to response.
* Hick has given the following formula which provides a convenient practical
method of calculation:
Inforrration gained : log M +

fiz"r(1"" "t,#)

where M is the number of readingsr.fs is the total of the signals in eadr coltmn
taken in turn, .fa the total of the resporures in each row taken in turn and.fs*
the nnmber of readings in each cell taken individually. Hick (r95za) sets out

alexample from his experiments.

66

Fundamentals of Skill

\[hen there are no errors so that each signal always leads to its own
particular response, the equation reduces to Eq. 3.4, or when signal
frequencies are all equal, to Eq. 3.2. It is also a means of providing an
information measure of discrimination (Garner and Hake, r95r) by

using, in the simple case, thefourprobabilities of 'Correct Yes', 'Correct

No', 'Miss' and 'False Positive'mentioned on p. 33, or in the multiple
case discussed on p. 4o, the several categories of signal provided and
response allowed.

Hick found that the shortening of reaction times when substantial
numbers of errors were made was by approximately the amounts expected. By the same tokens we should expect the -F r in Eq.
3.1 to be
reduced if premature responses occurred or responses were omitted.
We have noted in the previous chapter that Schouten and Bekker
Gg6l)
also found a very clear increase of errors as reaction time shortened,
and that Taylor et al. (tg6il analysed this relationship in rerms of a
linear increase of d'2 with time. They linked this with the linear gain
of information with time implied in Hick's tlpe of model through the
fact that both can be treated in terms of signal-to-noise ratio. Evidence
regardiqg the effects of both errors and sequential probabiliry comes
from experiments by Moray and Taylor (rgS8) and Triesman (1965)
who showed that when a subjea had to repeat words played to him by
tape-recorder, the accuracy of performance increased as the words
approximated more and more to connected English, to an extent which
implied a constant rate of information-transfer. Howell and Kreidler
(lg6f) and Fitts (1966), who compared groups of subjecrs performing
choice reaction tasks with instmctions for speed, for acc,uracy or for
both, confirmed that overall rates of information gain were not significantly ffierent for the three gpes of instruction although the
balance berween speed and accuracy was shifted as the instructions
required, although Howell and Kreidler noted that instructions for
speed produced a greater gain in errors than speed. Fitts observed that
subjects who made very large numbers of errors seemed, to gain information at lower rates than the rest. In part this may be due to Fitts
having ignored the * r in Eq. 3.r; in part it may be that Eq.
3.6 tends to
underestimate performarrce. It does so because a subject may well have

gained some information while still making an error, but unless he
makes the same error several times he will get little credit for this

information. Hick in a private communication has pointed out that
Eq. 3.6 assumes that all errors are equally 'bad' and suggests that this
may not in fact be correct: in other words the information gained is
not a complete measure of performance.

Identification and

Choice

67

In so far as the information measure is adequate, it provides a valtrable
means of combining speed and accuracy into a single score, snd emphasises the important fact that times for different tasks are comparable
only if errors are held constant, snd conversely that error rates can be
compared only if times are held constant.

Guarding against false reactions

The * r in Hick's formulation has not always proved easy to understand and an alternative equation proposed by Hyman (rgSf) and also
by Bricker (rgSSa) has often been preferred. They proposed in place
of Eq. 3.r,
Choice reaction time - a + blogn

(l.l)

where a is the simple reaction time and D log n represents the increase
over the simple reaction time due to the need for identification and
choice. A few sets of results such as those of Suci et al. (t96o) are about
equally well fitted by both Hick's and Hynran's formulae, but in most
cases Hick's fits better. This is tnre not only of Merkel's and Hick's

data but also of Griew's (1958b, c), Brown's (196o), Hilgendorf's
(1966) and Hyman's own, some of which are shown in Fig. 3.3.
As can be seen from Fig. 3.3, the reason why Hyman's formula fits
less well than Hick's is that it underestimates simple reaction times: for
degrees of choice above fotr the two formulae fit about equally well.
It is conceivable that Hyman's approach is basically correct but that
there is a minimtrm time required by some stage betrn een reception of
a signal and response to it, so that observed reaction time is either this
minimum or the time required for identification and choice, whichever
is greater.
The main difference benreen Hick's and Hyman's approaches seems
to lie, however, in their treament of the effects of preventing premature
or false responses if there is uncertainty about when signals will appear.
According to Hym"o these effects should be dealt with separately from
those of uncertainty about which signal has arrived. flick's approach
implies that temporal trncertainty affects the probability at any instant
of 'signal' as opposed to 'no sigual' and that the f r is not necessarily
a fixed quantity. We should, therefore, rewrite Eq. 3.1 as

Klog(n * no)
G.8)
where zo is the effect of temporal uncertainty expressed in terms of n.
It might vary from zero if the subject could estimate exactly when the
next signal will appear, to substantially more than * r if he were
Choice reaction time -

68

Fundammtals of Skill

deliberately misled as to the time of its arrival. Values between o and * r

ought to be found when the time of appearance is reasonably, but not
completely, prediaable.
Some evidence that this is so is contained in the results of another
experiment by Crossman (1956). In this the subject sat facing a panel
0.8

/
Oz

o /o

0.6

/
./

(.)
trJ
a

A

s 0.4
tr
z.

o

tr
o

llJ
G,

o-2

./

,/

./
/ ./

./

,/

,/

/o

/o /

6

./

zO

,/o

o

B

n
3

7

0

,-olrn oRLoG2(n+l)

3

Figure 3.3. Data from a choice-reaction experiment by Hyman (rgSl)
plotted against (A) Eq. 3.7, and (B) Eq. 3.r. Thetotalnumberof reactions

represented is about 2orooo: 5rooo from each of four subiects. The signals
were lights, two at each corner of a square, rnd the responses were the
syllables BUN, BOO, BEE . . . BATE spoken by the subiect and corresponding to the signals rt 2t J1 . . . 8.

of signal lights numbered r-8 beneath which was a row of eight pushbuttons. As soon as he raised his hand from a key, one of the lights
came on and he pressed the corresponding button. In different trials
lr 21 4 or all 8 lights and buttons were used. For one condition('s5rmbolic') the lights were in scattered positions so that the subject had to
use the number symbol to translate from signal to response. For the
other ('non-symbolic') the signal lights were directly above the buttons
so that no symbolic translation was required. The recorded time when
only one light was used was really a movement time since it was not

Idcntifuation and Clwice

6g

necessary to observe the display. Crossman had hoped to eliminate temporal uncertainty and the need to guard against false reactions by having
the subiect bring on the signal for himself, but the results suggest

this was not wholly achieved. They are shown in Fig. 3.4 plotted
against log (n + .45). It can be seen that the regression lines for both
conditions converge on the movement-time point: if the results are
plotted against logn, the lines cut the axis at different points above the
movement-time point. A possible reason for the failtrre to eliminate no
is that it is very difficult to make apparatus so reliable that the signal

L)
TU

a

ttl
F
=

z

o

o 0.4
tr,

e

0-2
0

3

LOG2 (n + 'eS)

Figrrre 3.4. Data from an experiment by Crossman (rgS6) comparing
performances with symbolic and non-slmrbolic displays. Each point is
the mean of 16o readings - 40 from each of four subjects. Open circles
: sJrmbolic displays, filled circles : nor-symbolic, X : rDovement time.
always appears as it should: even a very few occasions on which the
subject raised his hand from the key without a light appearing would be
enough to introduce appreciable uncertainty.
Several snrdies of simple reaction-time (e.9. Woodrow, r9r4, Klemmer, 1956, Karlin, 1959, Bonrinick and Brinley, tg6za, b, Aiken and
Lichtenstein, ry6$ have controlled temporal uncertainty by precedirg

each sigual with a warning and varying the interval ('foreperiod')
between them. All these sftdies are agreed that if the foreperiod remains
constant over a series of trids, reaction time lengthens as the foreperiod

is increased from somewhere between .3 and 2 sec upwards. The
increase has been shown to continue for foreperiods up to at least 3zo

70

Fundamentals of Skill

sec (Bevan et al.r 1965). With foreperiods shorter than about .5 sec other

complications enter which we shall consider in the next chapter. If the
foreperiod is randomly varied from trial to trial the reaction times with
all foreperiods tend to approximate to that which would be obtained
with the longest if they were given regularly. Klemmer GgSil has suggested that temporal uncertainty arises not only from variability in the
length of the foreperiod but also from the subjea's inability to estimate
time consistently, and he has shown that these two sources can be
treated together in a single information measlue. This measure is, however, d.ifferently conceived from that of Hick's Law, dealing with
accuracy of time estimation rather than guarding against false reactions,
and it is not appropriate to try to tie them together. A more direct
study of the effects of grrarding against false reactions is that of Drazin
(t961) who not only varied the foreperiod but also the probability with
which a signal acnrally followed. He showed that reaction time increased
as this probability fell.

These findings do not in themselves distinguish benreen Hick's and
Hyman's formtrlae, but they suggest a way of doing so by comparing
the effests of different foreperiods or probabilities of signal for various
degrees of choice. According to Hyman, the only effect of these variables
should be on a tnEq. 3.7 atdthis should be the same for all degrees of
choice. According to Hick the effect should be on z0 in Eq. 3.8 and
should thus diminish as the degree of choice becomes larger. There are
some indications from experiments by Boons and Bertelson (196r),
Fowler (tg6+) and Broadbent and Gregory (rg6S) that the effect does
diminish as the degree of choice increases, although only in Fowler's
case was the result statistically significant and his observations were
only on a single subiect. Further indications in favour of Hick's formula
are, however, forthcoming from an experiment by Bertelson and Barzeele (tg6S) who found that in a nro-choice task with unequal frequencies of the two signals, the effect of lengthening the foreperiod was much
greater upon the more than upon the less frequent responses.

CONCEPTUAL MODELS

Mathematical formulae such as Hick's Law are only a first step towards
an understanding of performance. They provide a sunrmary statement
of a complex process of observation, identification, choice and reaction
which must be analysed in detail if a clear picture is to be obtained.
Several models to account for this process have been proposed, all essentially implying decision processes of the kinds we have considered in the

Idmtifuation, and Choice

Tr

previous chapter. Each has its advantages and difficulties, and it seems
likely that there is no one model which applies in all circumstances.
They fall into three main tlpes which we shall consider in turn.

r. Serial classification models
Perhaps the most readily conceived model is one of those examined
by Hick (t95za). The subiect is thought of as making a series of subdecisions each taking approximately the same time. V/ith the first he
identifies the signal and the response to it as lying within one half of
the totd possibilities; with the second as lying within one half of this
ha$ and so on trntil the specific signal and response have been found.
He is not able to make his divisions into exact halves unless N is an
exact power of zrbut can do so approximately in other cases. fn any case
the model would still glve approximately the correct result so long as
the subiect started by rejecting broad classes of possibility and then
went on to reiect finer classes \Mithin a broad class chosen.
If each sub-decision is conceived as an all-or-none affair this model is,
however, trnsatisfaaory in three ways. Firstly, it must assume that errors
are due to the process of elimination not being carried far enough,
thereby saving one or more sub-decisions and the t'ne these would
take, but risking effor because the final choice is made at random
among the possibilities that remain: iust how the final random selection
is made remains unspecified. Secondy, the model does not adequately
account for reduction of reaction time when frequencies of different
signals and responses are unequal. So long as there are three or more
possibilities, some reduction would result if the subiect eliminated half
the probability rather than half the set of signals in each sub-decision.
For example, if there were five possible signds and one appeared four
times as frequently as each of the others, time would be saved if the
first sub-decision was benreen this one and the rest. Such a procedure
would also account for expectancy effects by assuming that these represented a biasing of the subiective probabilities of the different signals.
When, however, there are only two possibilities, as in some of Crossman's (lgSg) packs of cards, one subdecision should be required in all
cases whether or not sigual frequencies are equal, and thus no reduction of
reaction time would be predicted with unequal frequencies. Thirdly, the
model does not account for the variances of the reaction times observed
by Hick for different degrees of choice: if the total variance depends
in part on the number of sub-decisions involved, it should be relatively
low for degrees of choice which are exact powers of 2 so that all signals

T2

Fundamentals of Skill

require the same number of sub-decisions, and higher for other degrees
of choice where different choices will involve different ntrmbers. Hick
found that variance increased with degree of choice more evenly than
would be expected on this model.
These difficulties are avoided in a modification of the model suggested
by Fitts et al. (1963) and Fitts (1966) and implied also by Taylor et al.
(rg61).They regard the details of each sub-decision as roughly those
of the sequential sampling model outlined in the previous chapter and
illustrated in Fig. 2.r2. If this is so, it is easy to envisage that witr uobalanced frequencies or expectancies of two responses or classes of
response the subject might begin his 'random-walk' nearer to the more
frequent criterion, either by moving the criterion or by biasing the
starting point. Either of these methods would also tend to smooth the
increase of variance that occurs with rise in the number of choices. The
gain of speed at the expense of accuracy noted by Hick and by Fitts
is accotrnted for by moving the criteria closer together. Essentially the
same would also apply if the sub-decisions followed Vickers' modified
model illusuated in Fig. 2.l.6. This is an elegant and flexible approach:
perhaps its main disadvantage is that it is too flextble and can account
for almost any results provided appropriate assumptions are made
about the strifting of criteria relative to the starting point.
A more rigorous treatment of the effects of unequal frequencies and
expectancies is to assume that if dtrring a sub-decision the subject finds
the signal and corresponding response are not inthe halfhe is examining
he checks that they are b, the other half before proceeding. Each dicho-

tomous sub-decision might thus require either one or two 'inspections'
accordi.g to whether or not the first inspection was successful. Every
inspection would represent a decision process of the tlpe considered in
Chapter 2. If we assume that each inspection takes an equal time, the
average time required for a binary decision between equally probable
random alternatives will be made up of So% cases where one inspection-

time was sufficient and So% when two were needed. The mean time
would thus be r.5.
If signal frequencies were not equal, time could be saved even if
there were ody two alternatives, by trying the more likely sigual first
and so increasing the chance that the first inspection would be successful.
If, for example, one alternative occurred three times as frequently as

the other, trying the more frequent first would reduce the average
number of inspections required from r.5o to r.25. The effect of frequency unbalance would be limited by the fact that the ratio of the
times to respond to the more and the less frequent altenratives in a

Idmtifuation attd

Choice

73

binary choice could never exceed r : 2. Two points follow from this,
both of which are in line with Hyman's results. Firstly, the effects of
exrreme frequency unbalance will be less than those predicted on infor-

mation analysis. Hyman fotrnd the reaction times for trn o-choice tasks
\ilith siguals presented in the ratios r : 9 and r: q were almost identical.
Secondly, the times will tend to be longer for more frequent responses
and shorter for less frequent than expression 3'4 or 3'5 would predict.
Hyman specifically notes the occrurence of this in his results.
This model has the surprising implication that binary decisions are
not always the most efficient procedtrre: for choices between two and
eight it is better to make a serial ssamination of each possibility in
turn. This procedure requires one inspection for each possibility examined up to and including the correct one, so that a subiect willr oD
average, arrive at the correct choice in (N + r)/2 inspections. Beyond
eight choices the method of serial exanrination is inferior to dichotomisation. Beyond six choices fiuther economies can be achieved by
dividing the possibilities into threes or fours and exploring these serially
rather than dichotomising. The optimum strategy is thus serial exploration up to six with division into threes and fours for higher degrees of
choice. The various suategies are compared in Table 3.1 and in Fig. 3.5.
The optimum curve in Fig. 3.5 looks very much like an exaggeration of
Hyman's slightly S-shaped curve in Fig. 3.3. The fact that this shape is
not reflected in Merkel's or Hick's data is not a crtrcial argument
TABLE 3. r Average rurmbers of inspections required aith differmt strategies
to decide ctmong n + r (: N) alternatiaes

l'5 log,

log,

(z*t)
: logs N

r
z
3

4
5
6
7
8

:

r'oo
r'58
2'OO

z'32
z'SB
2'8r
3'OO
3'r7

*r)

(n
r.5 logs N

Suict

Running

dichoto- along line
mising i.e. (N *t)/z

'Optimum'
strategy

r'5o

r'50

r.5

r.5

2'38

2'44

2'O

2.O

3'OO

3'OO

2.5

2.5

3'48
3'89
4'22

3'55

3'O

3.O

3'94

3'5

3'5

4'25

4'O

3'7*

4'5O

4'50.

4'5

4'75

4'80

5'O

4'Ot
4'Ot

* Divided into 4 + 3.

t Divided into 4 + 4.
tDividedintog*3+3.

74

Fundammtals of Skill

against the model. The attainment of maximum efficiency is likely to
be rare as it involves the maintenance of a strategy in the face of many

temptations to vary it. It may well take time to develop fully, for
example Fitts et al. (lg6f) showed that when frequencies were unbalanced, the relative shortening of reaction times to the more frequent
signals and relative lengthening of times to the less frequent gradually
increased with practice.
Some kind of serial classification procedrue seems evident in subjects

a

4

z
o

o

(J
lr,
(L
(n

o

oo

o

=
IL

o

o

c,

u,

(D

o

z

=
f
z.

z
ul

=
n

345

7

2

3

I

0
0r

1

LOG2(n+l)

Figure 3.5. Average numbers of inspections required with different

strategies in Table 3.r. The continuous line is r.5 logr(n + r), the open
circles are for the strictest possible dichotomising strategy, and the filled
circles are for the 'optimum' strategy.

sorting cards of different classes on to piles: they move each card to the
appropriate part of the table and then make more detailed searches and
finer movements until the correct pile is located. How far it occurs in
conventional reaction-time tasks is more difficult to say since subdecisions may not show in overt sub-responses. It is perhaps significant,
however, that Leonard (rgS8) was able in a six-choice reaction to give
two indications - in which of two groups of three the signal would
occur, and which of these three was in fact the signal - about 'r sec

apart without lengthening the total reaction time beyond that for a

straightforward six-choice task.

Idmtifrcation and Choice

Ts

The precise strategy adopted is likely in any case to vary with the
way in which signals and responses are arranged. If there xrer say, very
large ntrmbers of lights in a display which are removed at some distance
from their correspondi.g keys and the lights and keys are not split up
into groups, counting from one end may be the only method of ensuring

an accurate response; so that if all signals occur equally often in a
random order, mean reaction time should be linearly related to
(N + t)/z and thus to N. What seems to be an illusuation of this is
given by Morikiyo and Iida (tg6il who compared reaction times for

2, 4 and 8 choices when the siguals were either from lamps in a horizontal row or numbers on a 'Nixie' tube, snd responses were made by
different fingers. Sfith the latter display Eq. 3.T held well (temporal
uncertainty was low), but with the straight-line display reaction time
increased approximately linearly with N. The same was true with the
linear display when responses were all made with one index finger, but
with the 'Nixie' tube display and this method of response the increase
was somewhat less with degree of choice than Eq. 3.7 would predict.
Why this is so is not clear, but Crossman (rgS6) has pointed out, and
Ihight (196) has confirmed, that when only a single finger is used
the differences of reaction time for different positions exceed those for
different degrees of choice. Average times for several positions may thus
depend on precise details of layout and procedure. Knight found that
in a serial task when each response brought on the next qignal, reaction
time was approximately a + b log D where D was the distance from the
previous response k"y to the new one. When signals were presented at
discrete intervals of time instead of seria[y, the same formula held with
D representing the distance from the centre of the row of response keys.
Similar considerations may well apply in card-sorting tasks, so that
sorting times would be affected by the layout of the piles on to which
the different classes are placed. Suppose, for exlmple, that 16 classes
are arranged in a 4 x 4 square: an efficient strategy would be to search
down the rows until the appropriate one was fotrnd and then along that
row until the correct individual pile was reached. Strith such a strategy
the average number of inspeaions would be (+ + r)/z for identifying
the appropriate row and the same again for ident$ing the individual
pile within the row. More generally we could write
Mean choice reaction time :

K(fr + r)

G.g)

This would apply exaaly only when {nwas a whole nrunber, but would
hold approximately when it was not. The author has found Eq. 3.9 to
grve a good fit to some card-sorting results obtained in undergraduate

76

Fundamentals of Skill

practiel classes although the conditions in these were not sufficiently
controlled for the results to carry much weight. Of more interest is the
fact that Eq. 3.9 gles a good fit to Hick's (tg5za) results for two choices
and above and a reasonable fit to data from other authors, although not
quite as good as that provided by Eq. 3.2.
Taking a broad view of the evidence and theoretical considerations,
we may think of Eq. 3.2, and thus of Hick's Law, as an 'ideal' statement
of a serial choice model which is never likely to be attained precisely
in practice, but is approximated in various different ways.

2. Simultaneous scanning models
The models discussed so far have attempted to break down the process

of identification and choice into a series of subsidiary actions. The
models now to be considered posnrlate a single continuous process. The
earliest of them was proposed by Hick (t95zb) who noted that Eq.3.l

could be rewritten

r/(n * r) : e-rtrc

6.ro)

where e is the base of the nattral logarithms and I the reaction-time.
He argued that we can imagine the total range of sensory input as

divided into (n * l) phases and that the left-hand side of Eq. 3.ro
can be regarded as the a priorf probability of any one phase. The righthand side of the equation is the probability of a phase lasting as long as
time T when it has a fixed probability df /K of vanishing at any instant
dT. A decision to release the response is made as soon as the signal
has lasted long enough for the probability of its having continued for
that length of time to be less than the prior probability. Subsequent
work in which several signals have been associated with each response
has indicated, as we shall see later, that choice reaction time depends
more upon the nrunber of possible responses than possible sigRals, but
Hick's model could cover this point by assuming that all signals leading
to a given response are treated as one phase.

A different type of simultaneous scanning model is that of Christie
and Luce (lgS6) discussed by Rapoport (r9Sg) and Laming Q966). It
posnrlates that the incoming signal is compared simultaneously with
all possible identifications, and the decision to respond is taken when
all the comparisons have been completed. Since the time taken by each
comparison is subiect to random variation, the larger the number to be
made the greater the likelihood of one taking a long time and thus
leading to a slow reaction. According to the version of this model set
out by Laming

Identification and
Mean choice reaction time

Choice

: K>
".t"

TT

(c > -r) G.rr)

where R, R, . . . R. are the ordinal ntrmbers of the various responses
and /( and C are constants. Eq. 3.rr can provide a good fit to Hick's
data, with C - r and correspondirg in a sense to no in Eq. 3.8.
A mixed simultaneous and successive model has been proposed by
Stone (196o) who suggested that the subiea might acctrmulate data
about all possible inputs for a given length of time, snd then compare
all possible pairs of inputs, to ascertain which was the largest, and
decide for this. Such a model has clear affinities with the discrimination

model of Fig. 2.9. It provides reasonably suaight-line relationships
berween reaction time and log n, but requires at least one additional
parameter. It must in fairness be said that Stone doubted whether his
model was applicable to cases where the different signals were easily
discriminable from one another, although in this his claim seems to
have been over modest.
It is extremely difficult to draw clear distinctions berneen the various
successive and simultaneous scanning models that have been proposed.
They :ue so flexible and several fit at least portions of the available data
so well that it is impossible to prefer one to another on mathematical
grotrnds alone: distinction needs to be made in terms of their parsimony in accounting for a wide range of experimental results. The
successive models at present have the advantage of being more thoroughly worked out in respect of unbalanced frequency and expectation
effects and of detailed behaviour when carrying out choice tasks, but
this is not to say that corresponding elaborations of the simultan@us
models would not prove successful (see e.g. Laming, t96z). One possibly significant point is that the simultaneous models have been conceived as a means of handling idmtifi,cation of rignal, whereas the
successive models seem more applicable to choice of response: perhaps
this is where their principal respective applications lie.

3. Neurological considerations
The processes postulated by the foregoing models must be underlain by
events in the nervous system, and it seems therefore reasonable to ask

whether any leads could be obtained by considering these events
direaly. The building of detailed neurological models is premafire in
the present state of knowledge, but it is reasonable to sketch one in
broad oudine.

T8

Fundamentals of Skill

Let us consider the making of responses in Hick's (t95za) experiments, in which the subject had from one to ten keys, each under a
different finger. In the simple reaction condition when only one key
was in use, his response could be relatively crude and undifferentiated:
a gross movement of the whole hand, or even of the whole body would
suffice. For two-choice responses with one key operated by each hand,
some differentiation was needed in that one hand had to move while the
other did not. With four possible responses, two by each hand, still
greater specificity was required but the movement could still be a rotation of the hand to put one side down rather than the other. With all
ro choices, there was no alternative to activating one finger to the exclusion of all the others.

=
z
t-

LJ
Ll-

o
J
IJ.J

lJ.l

J

2

ZONES ASSOCIATED WITH DIFFERENT FINGERS

Figure 3.6. Diagrammatic representation of hypothetical build-up of
neutral activity to control the fingers in a choice-reaction experiment
such as Hick's.

It seems obvious that this increasing specificity of response corresponds to a more and more precise localisation of aaivity in the braitr.
Any response must be associated with activity in a restricted set of
nerye cells, and for one response to be differentiated from another, the
activity in these cells must exceed that in cells for other responses by a
critical amount. 'W'e may envisage that a signal for a particular finger
to move will have its maximum effect on the cells associated with that
finger, but that there may be some spread of efflect to surrounding areas
associated with other fingers as suggested in Fig. 3.6. If the aaivity
builds up progressively with time, it might be represented at successive
moments by the series of curves in Fig. 3.6. If so it is clear that any
critical difference between aaivity in one region and in others will be
reached sooner if the specificity of the response is low - for example

Iduttifuation and Choice

Tg

when it can be made with the whole hand - than if it is high as when it
has to be made with one finger and not with adiacent fingers.
It is reasonable to suppose that a response would be triggered not by
relative activity in different areas but by the absolute level in one. If so
there must either, with the general build-up, go a 'backing off'which
keeps activity in unwanted areas below the critical level, or the excitation of any one area must tend to inhibit other areas, somewhat as
has been shown to occru between adiacent points in a sensory area
(Hartline 1938, von Bekesy 1967). The posnrlation of inhibition as well
as exciation is usually a weakness in a theory since, with suitable parameters attached, almost any behaviour cirn be accounted for. In this
case, however, the obiection is not serious since the excitation and
inhibition are presumed always to occur together.
The model proposed here makes the subiect's task in selecting a
response essentially similar to that of discriminating betrreen different
signals as envisaged in Fig. 2.6. Its main limitation is that it is not, at
its present stage of development, quantitative. Also it is perhaps a far
cry from the spatial organisation of the motor cortex to processes of
percepttral identification and the selection of non-nunual responses
such as nonsense syllables or names of letters or digits. The analogy
is not, however, far fetched or unreasonable since perceptual identifications and verbal responses must be based on physical traces which
have some kind of spatial location in the brain. Schouten and Bekker
(rg6il have suggested a somewhat similar approach to percepfiral rasks
in terms of 'percepnral focusing'.
This type of model appears able to account for a wide variety of
evidence. It deals with the effects of unequal frequencies and of expectation by assuming that the subject can activate in advance the area
corresponding to a given response, so that when a signal arrives the
build-up required to produce a decision in favour of this response is
less than that required for others. Evidence in favour of this view is
contained in the results of Davis (r94o) who found that muscle actionpotentials in the responding limb built up during the foreperiod; although since the build-up it assumed to be cenual, such peripheral
effects are not essential to the theory @onrinick and Thompson, ry66).
Ability to hold such a 'preparatory set' over an appreciable period of
time seems to be poor, as evidenced by experiments already cited in
which foreperiod was varied systematically (see also Kartin, 1966). Indeed the rise of reaction-tfune found with long or variable foreperiods
seems inevitably to imply that such preparation cannot be maintained.
We can only speculate upon the reason, but it seems possible that a

8o

Fundamentals of Skill

build-up of activity would tend to spread to adjacent areas if it corrtinued for a long time. The shortening of reaction-time when substantial
numbers of errors are made could be accounted for by a lowering of
critical values leading to a shorter build-up but greater chance of the
wrong response being triggered.
Three more positive lines of evidence in favour of this type of model
serve to distinguish it from the purely mathematical approaches. Firstly,
it accounts well for the otherwise anomalous results obtained by Seibel
(tg6f) who used a task in which all possible patterns of ro signal lights
and ro response keys were used to yield rro23 choices. He compared
this task with one in which five lights and keys were used with one
hand to give 3r choices. It is understandable in terms of the model that
the reaction times in both conditions were very similar - the second
was only about 25 msec faster than the first - since the discriminatory
difficulties involved in selecting patterns of fingers to depress would
differ little once it had become necessary to distinguish each finger
from its immediate neighbolus.
Secondly, it seems reasonable to suppose that any spread of effect

in the motor cortex would affect other areas in proportion to their

distance from the focal pointr so that errors would be related to the
'neural distance' between correct and erroneous responses. There is
some evidence from experiments by Blyth (1963, r96D that this is so.
He found that in a four-choice task in which responses were made by
the trro hands and two feet, the overwhelming majority of errors were
due to the substitution of the wrong limb on the correct side. Occasional
errors were made with the correct limb on the wrong side but never
with the wrong limb on the wrong side. Further experiments by the
same author made it clear that this result did not depend on the layout
of the siguals: it seemed clearly due to responses on the same side being
more readily confused than those on opposite sides.
Thirdly, the build-up of activity we have postulated would probably
take an appreciable time to die away. If sorthe model provides an explanation for the results of experiments (Bertelson, 196r , 1963, Bertelson
and Renkin, 1966, Leonard et al., t966, Hale ry67) which have fotrnd
that, in a rapid series, responses which are the same as those immediately
preceding are made more quickly than those which follow different
responses. Bertelson found that the effect largely disappeared if the
interval benn een the completion of one response and the appearance of
the next signal was increased from .o5 to .5 sec, and Hale fotrnd it much
reduced when the interval was increased from .r to .6 sec, grving time
for any after-effects to die away. Further evidence of a different kind

Idmtification and Choice

8r

but in the same direaion is contained in results obtained by Jeeves
(see !7et[ord, 1958, pp. 9@8, also Jeeves, 196r) who compared delays
of o, 'r and'2 sec between the end of one response and the onset of the
next signal in a serial reaction task, He found that the delays shortened
the reaction times, implying that some after-effea of the previous reaction which disappeared during the delay period could slow down a
subsequent reactiorl - the after-effea of one response seems to have
behaved as noise in relation to the next. Jeeves' results suggest that the
average after-effect lasted about 30 msec for a group of subiects aged
18-33, and about 50 msec for a group aged 58-Tr.
Results which link Bertelson's and Jeeves' findings with Blyth's have
been obtained by Rabbitt (1966b) who compared the times taken in a
serial reaction task somewhat similar to Blyth's. He confirrned that a
reaction was on average fastest when it was identical with the one previous, and slowest when preceded by a reaction with the other limb hand or foot - on the same side: the after-effects which in Blyth's
experiments caused errors, slowed down responses by Rabbitt's subjests.
It seems clear that similarity of response is more important than
similarity of signal: Bertelson (lg6S) who compared sinrations in which
each response was linked with either one or two signds, forurd that
changing from one signal to the other leadirg to the same response had
much less effect than changing from one response to the other. Williams
(1966) in a trno-choice reaction task in which somewhat longer periods
elapsed between each response and the next signal, found that reactions
were acnrally quicker to changed than to repeated signals. However
in one of her experiments in which, instead of each response being
associated with its corresponding signal, one response was made when
a signal was the same as before and the other when it was different, she
found reactions were quicker both with changed signals and also with
repeated resporrses. The implication seems to be that change of signal
and of response may in certain circumstances have opposite effects
which tend in most cases panly to cancel one another. Confiruration
that it is the actual limb moved rather than the control manipulated
that causes the shortening of repeated responses is given by Rabbitt
(tg5S) who used a multi-choice serial reaction task in which each hand
was used for several keys, and these were placed too far apart for manipulation by individual fingers. He showed that responses were made
significantly faster if the preceding response had been rnade by the
same hand than if it had been made by the other. There is, perhaps, I
perceptual equivalent to \Williams' and Rabbitt's findings in the results of
an e{periment by Nickerson (r965a). He displayed pairs of consonants,

8z

Fundamentals of Skill

one after the other, and required subiects to say as quickly as
possible whether the second was the same as or different from the first.
Responses of 'same' were substantially faster than those of 'different'.
Of the various successive, simultaneous and neurological models we
have surveyed, some have, as we have already noted, been clearly aimed
at explaining choice, while others have been conceived essentially in

terms of identification. Their applicability may thus depend on the
extent to which these two processes contribute to choice reaction-time.

This is the question we consider next.
FACTORS AFFECTING THE EXTENT TO WHICH REACTION-TIME
RISES WITH NUMBER OF ALTERNATIVES

The slope in Eq. 3.r for many sttrdies of subiects in their twenties using
conventional light signals and key-pressing or spoken responses, is remarkably constant at about S-T bits per sec (e.g. Merkel, 1885, Hick,
r912a, Hyman, 1953, Brown, 196o). It rises a little with age in some
snrdies (e.9. Griew, r958b, c, 1964, Suci et al., 1960) although not in
all (Crossman and Szafran, 1956, Szafran, 964)., ilrd it is possible that
the incteased slope with age is associated with cardiac deficiency rather

than age as such (Szafran, 1966). Essentially the same rate was found
by Pollack and ]ohnson (t963a) as the maximum for monitoring a series
of binary digits (o or l) briefly flashed one at a time on a screen. However, several sttrdies in which subjects have had to respond verbally
have yielded much higher rates, ranging from 14 to 17 bits per sec for
naming obiects, colours or designs (Morin et al., t965, Oldfield and
Wingfield, 1965, Oldfield, 1966), through about 16-9o bits for repeating digits shown singly (Brainerd et al., 1962, Stone and Callaway,
r96D and about 4o bits for repeating words shown briefly from vocabularies ranging from z to rrooo (Pollack, r963b) up to about roo-16o
bits per sec for repeating letters shown singly (Fitts and Switzer, 1962,

Morin et al., 1965).

It seems that the slope depends to a considerable extent on the relationship between signals and their corresponding responses. The earliest
clear demonsuation of this seems to have been Crossman's (lgS6)
experiment already desc,ribed in which he compared symbolic and nonsymbolic displays in a button-pressing task: the slope for the former
was about five bits per sec and for the latter about 15. Confirmation is
contained in results by Griew (r958b, c, 1964) for a task in which subjeas moved a stylus from a central point to one of zr 4 or 8 'targets'
arranged in a semicircle around it. He compared conditions when each

Identifrcatiut and Choice

83

sigual light was just beyond its corresponding target with those when it
was iust beyond the corresponding target on the opposite side of the
semicircle, and fotrnd the slope in the latter case to be about twice

that in the former. Further striking evidence is provided for a very

different kind of task by Treisman (rg6S) who showed that the accuracy
with which subiects could repeat English or French words fell off much
less as speed of presentation was increased, than it did when they had
to translate from the one language to the other. This was true even for
bilingual subiects.
It has been claimed by Leonard (rgS9) that if the relationship between signal and response is made sufficiently direct - or 'compatible'
as Fitts has termed it - the slope becomes zero. His subieas responded
by pressing the armatures of relays on which their fingers were resting,
and the signals consisted of vibrations in the same annanues. Leonard
found a rise of reaction-time from simple to two-choice conditions, but
none thereafter to four- or eight-choices. Some rise from simple to
rwo-choice conditions is likely in that when only one response is required it can be prepared to an extent that is not possible when any of
two or more may be called for - in the simple case the uanslation
mechanism can be pre-set but in the choice condition it cannot. The
equality of reaction time for choices above two is, however, surprising
and certaiqly calls for an explanation.
It has been further claimed by Mowbray and Rhoades (lgSg) and
Mowbray (196o) that with sufficient praaice or familiarity, reactiontimes for degrees of choice at least up to ro can be brought to the same
level as two-choice, even though the relationships between sigual and
response are not entirely direct. The degree ofindirectness in the former
case was slight - responses were made by keys under the subiects'
frngers and the signal lights were laid out on a panel in the same pattern - and subieas were given extensive practice with two and four
choices. Reaction-times rose appreciably with degree of choice early in
practice but became equal after some 36rooo reactions had been made.
In the latter case a fanriliar relationship was used - the signals were
digits shown singly and the subject responded by speaking the number
concerned. \fith d.ifferent groups, 2, 41 61 8 or all ro digits were used.
Mean reaction times were very similar for all groups.
Unfornrnately a doubt attaches to the interpretation of these results.
In Leonard's and Mowbray and Rhoades' experiments different numbers of fingers were used for different degrees of choice, but one of the
index fingers was common to all degrees and the results for this one
only were analysed. The fingers used with two choices were also used

84

Ftmdammtals of Shill

with four and those used with four were also used with eight. Now if a
subjea followed any kind of serial strategy it seems likely that he
would have started each trial by 'inspecting' one of the two signals and
responses used in the two-choice condition and that he would thus have
achieved very similar times for the corrmon index finger for all degrees

of choice, although if results from all the fingers had been analysed the
times would have been found to rise from two to four choices and
again to eight. Mowbray's (rg60) experiment avoided this criticism by
using a non-manual response and different groups of subieas for
different degrees of choice, but it failed to control the amount of
practice. Each of the signals used in any condition was presented 6o
times to each subiect: so that those who were given two choices had a
total of rzo trials whereas those with ro choices had 600. If there was
any shortening of reaction-time with practice it would become more
marked with higher degrees of choice and exaggerate any tendency
towards equality between different degrees.
A fiuther warning is contained in results obtained by Fitts and
Switzer (tg6z). They found that mean times to repeat digits projected
on a screen rose progressively with number of alternatives when familiar
sub-sets were used such as r, z; r, 2, 3, 4; or r-8, but that results
were less regular with unfamiliar sets such as 2, Ti 41 7i or 4, 51 61 7.
Similar results were obtained when the familiar sub-set A, B, C was
compared with the unfamilisr E, B, P. The latter produced response
times comparable with those when the whole alphabet was used, and
suggests that when a subject has to deal with an unfamiliar sub-set
drawn from a larger familiar set, he cannot rid his mind ofthe unwanted
members of the larger set. In such cases the failure of reaction time to
rise with size of sub-set does not imply a breakdown of Hick's Law in
any fundamental sense - it merely implies that the size of set from the
subiect's point of view is larger than it is from the experimenter's.
In line with the results of these experiments are those obtained for
maximum rates of reading. Pierce and Karlin (tgSil, using words
drawn from the most frequent 5oo in American use found that average
reading time rose from about '26 to'3r sec per word as the vocabulary
rose from 2 to 16 items - an incremental rate of about 6o bits per sec.
There was a further rise to about '4o sec when words were randomly
drawn from a 5rooo-word dictionary but, as the authors point out, these
words were often longer. Average times were, however, closely similar
for vocabularies of 16 to 256 conrmon words and this fact has led the
authors and others since to conclude that maximum reading rate is
independent of the information contained in the items. It seems very

Identifi&tion and Clwice

85

possible, however, that it is difrcult to absuact a restriaed vocabulary

from the totality of common words. Further evidence pointing in the
same direction comes from two experiments by Conrad (t96za). In
the first subjects read, as rapidly as possible, nonsense syllables drawn
randomly from lists of 4, 8, 16 or 32. The times taken increased with
the number of alternatives, but the increase was less for syllables having
a high association-value than for those having a low one. In the second
experiment subieas read lists of nonsense syllables drawn from either 4
or 32 dternatives on three successive days. The times required to read
the lists decreased markedly and the difference of times for the two
lists also decreased although it remained appreciable. By contrast over
the same period, the times taken to read familiar threeletter words
diminished little and *re times for lists drawn from 4 and 3z alternatives were closely similar.
There is an additional difficulty in using maximum reading rates as a
measure of capacity for processing information, in that the rates may
well reach the level at which motor factors and the monitoring of speech
limit performance. Admittedly readirg of random words is, as Pierce
and Karlin showed, slower than repeating a sentence over and over, but
this latter is not an adequate control task since the repeated sentence
\rill almost certainly not be monitored to the same extent as random
words and may well not be so precisely pronounced.
Whatever reservations there may be about the view that compatibility
and familiarity can abolish the rise of reaction time with degree of
choice, there is clear evidence that they can markedly reduce it. Three
sets of experiments may be especially mentioned:
Firstly, Davis et al. (l96la) played nonsense syllables recorded on
tape to their subiects who responded by repeating each syllable as
quickly as possible. The times taken to repeat back syllables drawn at
random from sets of zr4 or 8 rose as predicted by Hick's Law. The slope
flattened rapidly with practice from about 25 to roo bits per sec: the
two-choice time remained about the same while the times for four and
eight choices became markedly shorter. The authors also argued that
for very familiar sets the slope should be low without practice. Comparing digits with letters - a set of ro with one of z6 - they fotrnd a
difference of time which implied an incremental rate of about 70 bits

per sec.

Secondly, Knight (rg6il measured subiects' reaction times for all
degrees of choice from r to 8 on two successive days: the subjects then

practised the 8-choice task for 16 min per day for zo days - five days
per week for fou weeks. After a further test similar to that on the first

86

Fmdamentals of Skill

two days they again practised for a further eight weeks after which a
final test was given. The simple and two-choice reaction times were
little affected by the practice, but those for higher degrees of choice
were all markedly reduced. The average slope from z to 8 choices on
the first day was about 14 bits per sec. After both one month and three
months of practice it had flattened to about 43 bits per sec but still
remained appreciable.

Thirdly, the effects of compatibility on reaction time are shown by
Brainerd et al. (t962) who gave their subieas two-, four- and eightchoice tasks with all possible combinations of nro rypes of signal and
two types of response. The four combinations, snging from least to
most compatible, and the results obtained are shown in Table 3.2. It
TABLE 3.2 Signal-response combinations used and results obtained by
Brainerd et. al. (t962). Figures are in bits per sec
Sigrral

Response

Incremental rates of
information uansmission
Self-paced,

presentation Signals
presented
of signals in
continuous singly
series

Numbers proieaed Pressing keys corresponding to the
on screen
numbers and placed

conveniently under
the subiect's

fingers

5'5

S'6

Keys as above

lights

4'g
9'o

7'4

numbers

go.g

16.4

Lights arranged in Speaking numbers
same pattern as correspond,ing to

keys

Lights as above

Numbers as above Repeating the

5'6

The results are based on a total of z4o subjects - 20 for each of three
degrees of choice in each of the four conditions. Each subiect had a run
of roo seH-paced trials, and 5o trials with sigRals presented singly.

will be seen that the rates vary from values similar to those found by
Hick and Hyman in the less compatible conditions to values similar to
those obtained by Davis et al. with the most compatible.
Taking the evidence as a whole, two conclusions emerge clearly:

Identifuatiut and Choice

87

firstly, the flattening of the slope relating reaction-time ro degree of
choice is associated with increased familiarity or compatibility of the
relation between signal and resPonsa Thus in the experiment by Brainerd

et aI. the highest rate was attained when the familiar set of siguals

(numbers) was paired with its familiar, normal response (speaking). The
next highest was when the unfamiliar set of signals (lights) was paired
with a closely compatible but unfamiliar set of responses (key pressing).
The numbers paired with key pressing or lights with spoken numbers
both yielded relatively low rates. SUre can therefore identi$ the steepness
of the slope as due largely to the involvement of the translation
mechanism (Fig. r.3, p. r9). S7e should perhaps add that the effect
seems to be highly specific to particular conditions: for example Lamb
and Kaufman (lg6S) and Kaufman and Levy (1966) using a highly
compatible arrangement of lights and keys obtained nearly flat slopes
relat'ng times for different degrees of choice when all signals were of
equal frequetrcy, but in two choice tasks with unequally likely alternatives the less frequent response was substantially slower and the more
frequent quicker than would be expected from the equal frequency
results - the opposite of the result obtained by H),orao (lgSf) with his
less compatible task.
Secondly, the flattening of slope represents a true raising of the rate
of information transfer for higher degrees of choice: in other words,
some of the 'work' that would otherwise have to be done by the translation mechanism has been saved. How this is achieved is not known, but
one plausible suggestion is that the connections between various identifications and their corresponding responses become 'built-in' and thus
ready for irnmediate use instead of having to be to some extent worked
out afresh for each uial. Any residual rise of time with increase in the
number of alternative responses might be due to the translation
mechanism still requiring slightly longer time to deal with larger numbers of alternatives. This would follow, for example, if flattening of the
slope implied a change from a selection procedtre involving a series of
sub-decisions to a single decision of the tJpe envisaged in Fig. 2.g
(p. +5). Alternatively it might represent the time taken by the perceptual
mechanism to identi$ the signal. For an answer to this question we
nrn to a further line of evidence.

SEPARATING IDENTIFICATION FROM CHOICE

Perhaps the most direct attempt to separate the times required for
identification and choice is *rat of Hilgendorf (1966) using a task in

88

Fundamerttals of Skill

which the signals were illuminated digits or other symbols shown on a
small screen and the responses were made by pressing qpewriter keys.
Each trial started with the subject pressing on an additional 'home'
key with the palm of the hand. He was instructed that as soon as the
signal appeared he was to identify it and then press the appropriate
key or keys on the typewriter. As soon as he raised his hand from the
'home' key the light illuminating the signal went out so that there was
no chance of continuing to observe the signal once the responding
movement had begrm. Hilgendorf found that the total time to raise the
hand from the 'home' k.y, press the necessary tlpewriter key or keys
and return to the 'home' key was well fined by Eq. 3.r with a slope of
about 5.5 bits per sec - a figure close to Hick's. The time between the
appearance of the signal and raising the hand from the 'home' key was
also well fitted by Eq. 3.r but with a slope of about z7 bits per sec.
These results suggest strongly that identification and choice are
separare processes with the former taking place much more rapidly
than the latter. Further evidence leading to the same conclusion comes
from two ftrrther types of experiment which we shall consider in turn:
Reaction to some only of the signals presented

Donders in his pioneer work (1868) sttrdied not only simple- and
choice- (a- and b-) reactions but a third arrnnBement which he termed
the c-reaction, or as it has come to be called selectioe response. fn this,
two or more signals are presented, one at a time as in a b-reaction task,
but response is made only to one. Donders found c-reaction times were
intermediate betrnreen those for a- and b-reactions and argued that the
difference bet'ween the b- and c-reaction times was the time required
for choice of response, and the difference between the a- and c-reaction
times was the time taken to identify the signal. Subsequent work has
largely vindicated Donders' position, but has also shown that the sinration is more complex than was originally supposed.
We may begrn with an experiment by Broadbent and Gregory $962)
who compared two-choice b- and c-reactions for different levels of
compatibility between signal and response. For the less compatible
conditions c-reactions were quicker than b-reactions, as Donders fotrnd,
but for the highly compatible conditions the times for b- and c-reactions
were about equal - indeed the latter were slightly longer. The obvious
implication is that in the highly compatible conditions identification of
signal took no time at all. This equality cannot, however, be taken at
face value. Forrin and Morin (1966) actually found c-reactions s-ub-

Idffitifuation aild Chuice

8g

stantially lurg* than b-reactions, and suggested that the inhibition of
response to the unwanted items affected the speed of reaction to the
wanted ones. Mowbray (196o) found that c-reactions were quicker than
b-reactions when response had to be made to one of trro possible signals,
but were substantially slouter than b-reactions when response was to one
of eight or ten siguals, snd suggested that this was because the average
interval between responses was longer in the latter case: long intervals,

like long foreperiods in simple reactions, should lead to long reaction
times, and the c-reaction time may therefore be lengthened by temporal
uncertainty effects. In confirmation, he noted (Mowbray 1960 that
c-reactions are affected by the distrihttion of intervals between 'key'
signals to which response has to be made.
Mowbray's suggestions can be tested by manipulating either the rate
at which signals are presented or the relative frequencies of key and
non-key signals so as to keep the mean times between key signals
constant while varying the ntrmber of different non-key signals. Brebner
and Gordon (1962, l960 have shown that when this is done, there is
still a rise of c-reaction time with ntunber of different non-key signals,
indicating that the time required to identi$ a key signal rises with the
number of different non-key siguals. It should be added also that
Nickerson and Feehrer (lg6+) who exposed a series of letters one by one
and required subjects to press a key in response to some but to ignore
others, fotmd that reaction time lengthened ryith increase in the number
of letters to which response had to be made.
Perhaps the clearest confirmation of Donders' results is by Taylor
(t966) who used a two-choice task in which one of trro coloured lights
was presented after a warning slgnal followed by a foreperiod rangng
from'8 to r'j sec. Taylor compared four conditions: (i) b-reactions in
which responses were made to both lights, (ii) c-reactions in which
responses were made to one Ugbt and the other was igRoreq (iii) b-reactions in which only one light was used - one response was given to this
and the other response was given when the warning signal appeared
but no light followed - and (w) c-reactions with response to the one

light and no response to no light. Using only responses to the 'key'
light, the difference between (ii) and (rv) provides a measure of the
need to discriminate benneen signals and the difference betrreen (iii)
and (rv) a measrue of the time taken for choice of response. The sum
of these two differences should equal that benreen (i) and (iv). Taylor
fotrnd that the first two differences were z5 and 22 msec respectively
making a total of 4T as compared with 44 msec observed for the third a very substantial agreement.

Fundamentals of Skill
90
It should be noted that the c-reaction has obvious similarities to
scanning tasks in which subiects search through lists for particular items
- for example a particular letter or word among other letters or words.

The concern of most of the experiments carried out so far is different
from oru present one, but Oostlander and de Swart (t966) have noted
that the time required for searching increases with the size of the set
from which the items have been drawn, in a manner consistent with
Hick's Law. Time has also been shown to rise with the number of
different items being sought in any one scan (Neisser, 1963, Kaplan
and Carvellas, 1965, Kaplan et al., t966, Nickerson, ry66), and when
the definition of a class of words or other signals sought is wide than
when it is naffower (Foster, t962, Neisser and Beller, 1965).
The broad indications from the evidence on c-reactions seem clear
that, even when no choice of response is required, reaction-time rises
with the size of the set from which the signals are drawnr so that it is
difficult to avoid the conclusion that this time is taken in identiSing
the signal. The amount of the increase is usually much less than when
responses have to be made to each signal and one must therefore suppose that percepnral identification proceeds much faster - benreen five
and twenty times as fast - as the selection of response in sinrations
where signals and responses are not higtily compatible or familiar. If
so, it is reasonable to regard the residual slopes forurd in such experiments as those of Davis et al. and Fitts and Switzer where signals and
responses cre highly compatible or familiar, as due to the time taken to
identiS the siguals. It is perhaps fair to argue that there is a biological
advantage in having the percepnral mechanism work substantially
faster than the translation mechanism since it is much more at the
mercy of external events which must be perceived as they occur. To do
this a high peak capacity is needed even if on most occasions it is not
fully used. Action, although it may have to keep approximately in step
with events, usually permits some flexibility of timing and can thus
be adequate with a lower maximum rate of information handling.
Fewer categories of response than of sigual
Several experiments investigating the relative effects on reaction time
of trncertainty regarding signal and response have used tasks in which
more than one response has had to be given to any of nn o or more

siguals. The earliest is that of Crossman (rgSf) who found that the
time taken to sort cards into two piles, one of Red Picnres plus Black
Plain and the other of Black Pictures plus Red Plain, was similar to that

Ident;fication and Choice

9r

for sorting into four suits and substantially greater than for sorting
into two colours. This result was obtained from only one subiect, but
he had been well practised. It seems to favotrr the view that both the
number of signal categories and response categories, or the number of
signal-response contuct'ions, determines sorting time.

On the other hand, Morin et al. (196r), using a conventional choicereaction task obtained indications that only the number of rsponse
categories was important, rt least for well-practised subiects. They
compared the five conditions set out in Fig. 3.7. Of these the first
three are straightforward simple, trro-choice and fotrr-choice arrangements. Condition IV can also be regarded as a suaightforward twochoice benneen circles and squares. The times for thse fou conditions
Signals:

OIOOII

Conditions:

I

II
III

IV

2
2
2

3

I

4

v
2
2
I
Figrrre 3.7. Conditions used in a choice-reaction experiment by Morin
et al. (196r).
The signals were projected from slides onto a small screen. The subiect's responses were made by pressing microswitches on which his
fingers were resting lightly. In each condition ro subieas perfonned
T
blocks of 16 trials on each of z days.
The numbers rt 2t 3 and 4 refer to the responses rnade to the signals
under the various conditions.
rose as exPected, although the increase from simple to two choice was
somewhat larger than usual. Times for both nro-choice conditions II
and IV were closely similar. The crucial question concerned condition
% which was analogous to Crossman's task. The times for this condition were intermediate between those for the two- and fotrr-choice tasks
early in practice, but became virnrally identical to those for nro-choice

after about r8o trials.

In order to pursue the apparent discrepancy of fact between the
results of C,rosslnan and of Morin et dl., Cameron (rg6+) used the
latter's designs in a card-sorting task, reproducing conditions II, III,
IV and V of Fig. 3.7, He found that the times for condition V were a
little lutger than for the four-choice condition III and much longer than
for the two-choice conditions II and fV. Essentially similar results were
rePorted by Fitts and Biederman (rg6S) who repeated conditions II-V
of Fig. 3.T in a task similar to that used by Morin et al. A possible

92

Fundananttab of Skill

explanation of the discrepancy betrnreen the results of the original and
subsequent investigators lies in the fact that the subiects employed by
Morin et al. were strikingly inaccurate, least so in the four-choice
condition and most in condition V. The subieas in the other experimeuts made relatively few effors and their findings seem therefore to
be the more worthy of accePtance.
The results surveyed so far, would be broadly consistent with the
view that choice-reaction time is a function of the number of signalresponse connections, so that in Crossman's task and condition V of
Fig. 3.7 four rather than two connections were involved. To be precise
it should be the probability of such connections rather than their

number which influences reaction time, as is indicated in results

obtained by La Berge and Tweedy (tg64). Their subiects responded
wirh one hand to a green signal and wittr the other to either a red or a
blue. Reaction times to red and blue were influenced by altering their
relative frequencies even though their combined frequencies, and thus
the frequency of the response made to them, remained the same.
Results of experiments by Rabbitt (lgSg) suggest, however, that this
view is too simple. His subjects sorted packs of cards on which letters
or digrts had been stencilled into 2t 4, 6 or 8 piles. With some packs
only one letter or digrt had to be sorted into each pile, with others
21 3r4 or 8. He found that the times taken rose sharply as the number of
symbols per pile increased from one to two, but relatively little thereafter. Rabbitt repeated the experiment using a more conventional reaction-time apparatus \dth which subiects pressed keys in response to
digrts or letters proiected on a screen, and found the same pattern of
results, although only after fairly long practice. Very similar results
have been obtained by Pollack (1963c) who timed subieas classtrying
lists of words into superordinate classes and varying in different lists
the ntrmber of words belonging to each class. The times rose substantially with the number of superordinate classes. fn most cases they also
rose substantially \ilith the nunrber per class from one to two but relatively little thereafter. The same implication is perhaps contained in a
further experiment by Rabbitt (lg6+) who found that when subiects
sorted cards on which there was one relevant symbol and from o to 7
irrelevant, times rose substantially as benreen o and r irrelevant symbols, but little more with larger numbers.
These results suggest a two-stage process consisting of (t) the identification of a letter or digit or word as belonging to a particular group of
symbols which the subiea has been told to place together, and (ii) the
selection of the correct response to it. These two stages seem unlikely

Idmtification and Choice

93

to overlap because trntil the first has been completed the second can
hardly begin. If so we can regard the times for one symbol per pile as a
measure of the speed of (ii) and the differences betrn een the times for
one and two symbols per pile as a measure of the time taken by (i).

The faa that in Rabbitt's (rgSg) experiment times continued to rise a

little when six and eight piles were used with three or four symbols per
pile could be accounted for by assuming that the subiect has some
difficulty in remembering the groups to which the t81z symbols concerned belong. This might imply a limit to the number of items which
carr be 'held ready' in some kind of short-term memory: if so, Rabbitt's
results suggest the number is about 16. Such a two-stage process could
equally apply to Crossman's and the other experiments using condition
V of Fig. 3.7, if account is taken of inacctuacy in the case of Morin
et al. The time taken to identi$ the group to which the signal belongs
has been shown to vary with such conditions as the degree of familiarity
of the grouping (E. Smith 1967), and presrunably also varies according
to the discriminability of the signalsr so that it should not be expected,
except fornritously, to equal the difference betrreen two- and fourchoice reastion times. In an attempt to test the validity of this two-stage
theory by comparing reaction[times when the group to which a signal
belonged was and was not indicated in advance, Forrin and Morin
(tg6il showed that the advance information did in fact save time,
although not to the full extent predicted. If the theory is tenable there
are evidently some complicating factors still to be worked out.
The question has been raised of whether the identification stage
follows a serial classificationl or a simultaneous one orr more specifically, whether particular aspects or 'dimensions' of a sigRal are identified
separately in series or together in parallel. The problem has been
lucidly discussed by Treisman (1966). We may consider it in terms of
an experiment by Nickerson (r967a) who presented signals for brief
intervals and required subjects to press one of two keys to indicate
whether or not the signals satisfied certain criteria. The siguals were of
three shapes: circle, square or triangle; three colours: red, yellow or
blue; and three sizes: large (z in), medium (t in) or small (+ h), in all

possible combinations making z7 in all. In any run, half the signals
satisfied the criterion or criteria and half did not. The criteria used are
set out in Table 3.3 together with the predictions of the two ffierent
theories. Results were on the whole more consistent with series than
$dth parallel analysis although Nickerson noted some anomalies requiring further investigation.
What appears to be clear evidence in favour of serial processing is

94

Furd,aruentals

of Skill

TABLE 3,3 Combinations of criteria used in a classificatory task by Nickerson (r967a), together n)ith predicted results according to ta)o theories of
perceptual analysis
Predicted results if

Predicted results if

Criteria designated

different features are

different features are

as relevant

analysed serially

analysed simultane-

ously
Single

Red I Other

Circular ffeatures
Large J irrelevant

Disjunctiae
Red or circular I
Large or rcd I
Large or circular)

Large or ted or
circular

Quicker than both dis- Slower than disiunctive.
iunctive and conjunc- Quicker than contive.
iunctive according to
Christie and Luce
(tgS6) model.
Slower than single.

Quickerthanbothsingle

Quicker than or equal to and conitrnctive.

coniunctive. Quicker Quicker if both c:riif both ctiteria satis- teria satisfied than if
fied than if only one. only one.
Slower than other dis- Quicker than other dis-

iunctives.

iunctives.

Conjunctiae

Both red and
circular
Both large and red

Slower than single.

Slower than single or
Equal to or slower disiunctive.

than disiunctive.

Both large and
circular
Large and red and
circular

Slower than other conjunctives.

Slower than other cooiunctives.

contained in the results of a series of experiments by Posner and

Mitchell (1967). Subiects were presented with pairs of letters, which
might be either capitals or lower case, &nd had to indicate as rapidly as
possible whether the members of the pair were 'sarle' or 'different'
according, in different runs, to criteria of either physical simil.rity (e.g.
AA) or similarity of name (e.9. Aa) or according to whether both were
vowels or both consonants (e.9. Ae or BC). The reaction times obtained
were such as to imply that the processes involved in these three tlpes of
identification were separate and the times taken were additive.

Identifuation and Choice

95

FURTHER STUDIES OF PERCEPTUAL SELECTION

The principles trnderlying the models we have outlined appear to apply
also to the results of some other experiments on identification which
are not concerned with reaction-times. For example, Fraisse and
Blancheteau Qg6z), whose subiects identified nonsense-syllables drawn
at random from sets known beforehand, found that the minimum time
syllables had to be exposed to secure recognition rose approximately
linearly with the logarithm of the number of syllables in the set from
a mean of tz'9 msec with two alternatives to one of r5.4 msec with
f,flteen - a small but consistent and significant rise. As a further example,
Miller et al. (lgSr) used monosyllables drawn at random from d,ifferent
size sets, again known to the subiects beforehand, and presented them
through earphones in wideband noise which made them difficuk ro
hear. They found that the proportion of syllables heard correaly fell
as the number of syllables in the set incteased from 2 to
4rB, t6r 3zr 256

and 'all possible' syllables. Similar results were obtained by Miller

(rgSil who also showed that the fall became steeper as the signal-tonoise ratio became poorer. This kind of simultaneous relation io both
set size and signal-to-noise ratio fits well with the simultaneous scanning models and with Fitts' approach to serial classification (p.
72)
since in all these cases the subiect's task is conceived as one which will

be affected by both these variables.
Ftrrther evidence in the same direction has been provided by Reid
et al. (t96o) who found that words exposed briefly were more acctgately
perceived the greater the extent to which the category within which they

fell was specified beforehand, and by Binder et al. (1966) who found
that the accuracy with which parts of briefly exposed picnrres of obiects
were recognised increased with the extent to which the picnrres had
been learnt beforehand. Such famitiarity does not seem to affect the
subiea's ability actuatly to see the obiea, since it makes no d.ifference
to his ability to iden@ an object as the 'same' or 'd.ifferent' from one
exposed before: it is rather his ability to idmtify it that is improved

(Robinson et ol., 1964). Long et al. (lg60) have suggested that the
reason for such easier identification is that the smaller the set the more

the subject is able to identi$r and use .key, distinguishing features
between the different items. In other words, he can perceive accgrately

on the basis of a part only of the data. The same argument is used by
Rabbitt (rg6il to explain some of his restilts with choice-reaction tasks.
Evidence in favotu of this view is also contained in the fact that when
irrelevant material remains the same throughout a uial or series of

96

Fundamentals of Skill

trials it has less effect than when it changes (Rabbitt, tg67a, Zeaman
and Denegrc, t967),
Similar indications that the greater the information in the items presented, the stronger the siguals required for adequate identification are
given by experimenters using 'natural' sets of material. Postman and
Bruner Gg+g) found that the exposrue time required to recognise
which of trro words shown denoted a colour, was shorter than that
required to decide which denoted either a colour or a food: the average
times were r9r and zz8 msec respectively. Again Miller et al. (lgSr)
found that words were more accurately recognised when presented in
sentences than alone. This result is to be expected from expression 3.4
since the context of a sentence severely restricts the number of words
likely to occur at any point, so that if the subject can glean anything
of the strucftre and content of the sentence as a whole, he is in effect
recognising each word from a restricted range of possibilities.
Further effects of the sizes of 'natural' sets on identification are
shown in experiments where subjects had to identify material without
any previous indication of what the obiects might be. For example,
several investigators (see Spielberger and Derury, ry63) have shown
that exposure times needed for recognition of common words are
shorter than those needed for words in less frequent use. Turning to
non-verbal material we may note particularly experiments by Verville
and Cameron (1946) and by Wallace (t956). The former used incomplete picttues proieaed on to a screen until identified, or until 9 min
had passed. The latter used geometric designs, silhouette figures and
more complex pictures shown on a band moving behind a slit. The
slit could be varied in width and with it the exposure time. Subiects
were given exposrues until recognition was achieved or until a fixed
number (usually five) had been made. The authors in each case noted

that subieas who tried an identification which they were told was
wrong tended to choose others within the same general class before
moving to other classes: for instance, if they had volunteered the name
of an animal, they would try other animals before moving to, say,
plants or inanimate objects. It looked, in short, as if their categories of
identification were ordered in 'natural' sets of this kind and they tended
to rrur through likely members of one set before moving to other sets.
The relationships observed by Wallace between correct identification
and exposure time are perhaps in line with this view. She forurd that
the proportion of designs identified correctly rose approximately linearly
with the logarithm of the exposure time totalled over all exposures. A

sample of her results is given in Fig. 3.8. The slope varies with the

Idefuation and Clwice

9l

complexity of the material, but seems to be suikingly linear in all cases.
The time relationships in these various experiments are what would
be expected with a serial classification model operating 'in reverse'.
When the total set of possible identifications is known in advance, the
subiect selects first a broad class, then finer sub-classes until one
sufrciently precise is attained. When the total set is not known, he
100

o

75

UJ

tr

E

z

[!

I
J
(J
l-rJ

G,
g.

o
o

(,
o-

O SIMPLE GEOMETRICAL FIGURES

a
o
l!

o

X MORE COMPLEX BLOCK FIGURES
25

+ REPRESENTATIONAL SILHOUETTES

trJ

O PICTORIAL LINE DRAWINGS

(9

F
z.
lr,

()

c.

UJ

(L

0

0'l
cuMULATTvE vrEw,*n ,,,11rN

sEc. ,.ot'ur.or.,

1'0

1'5

Figrrre 3.8. Results from an experiment by l7allace (rgS6) showing the

increase in the percentage of displays identified correctly with increase in

viewing-time.
Each of 16 subiects aged berreen 18 and 3o viewed 4 of each class of
display.

selects a restriaed class and runs through it, then if he is unsuccessful
he tries progressively broader classes until identification is achieved.
The perseverative tendencies shown by Verville and Cameron's and

Wallace's subiects could be plausibly regarded as another manifestation
of the after-effects already discussed in relation to the repetition of
response. It is reasonable to suppose that any spread of effect from one

menrber of a class will tend to affect other members of the class more
than members of other classes: it would thus make identifications
D

98

Fundammtals of Skill

within the original class more readily available than those in other
classes. Perhaps the most extreme form of this process is seen in senile
states. Hurwitz and Alison (rg6S) note that a senile patient asked 'what
is the capital of England' may correctly reply 'London'. ff immediately
after he is asked a different question such as 'what did you have for
breakfast' he is likely again to reply 'London'. Iq however, a few
minutes' gap is left between the questions the answers to the second
will also be correct - the after-effect of the first question and answer
have died down so that the answer 'London' is no longer in the fore-

front.
Selective perception

It has been recognised for a long time that a person reacts to only a
small fraction of the total information that comes in through his sense
organs. There appear to be two ways in which this selection occurs:
firstl5 if we conceive of identification as proceeding by a process of
serial classification, one form of selectivity would be that classification is
carried only so far as is necessary for the task in hand. Thus, for example,
a subiect sorting cards into suits observes the suit-symbols on the cards
and ignores the numbers which on other occasions he might attend to.
Many of the classical experiments on perception, showing that a broad
design is observed while specific details are not, can be explained in

this way.

A second method of selection appears to be that certain classes of
incoming data are 'filtered off' at a relatively early stage of the perceptual process. This conclusion has resulted from a substantial number of
experiments on selective listening. The subjea is required to respond
to one of two or more messages presented simultaneously, and to
iguore the others. The task is in many ways analogous to Donders'
c-reaction with more complex signals and, usually, more complex responses. Early evidence has been surnmarised by Broadbent (1958) to
whom also much of it is due. We shall, therefore, outline only some of
the main indications and more recent work.
When two conversations arrive simultaneously over a loudspeaker,
subiects usually have difficulty in listening to one and ignoring the
other (Broadbent ry5za). The effect has sometimes been thought to
result from masking in the inner ear, but this is not so because on
occasion the two voices can be distinguished reasonably well: the confusion is central. Selection is easier if the wanted and unwanted voices
differ in physical characteristics, as when one is a man's and the other

Iduttifuation and Cltoice

gg

a woman's (Broadbent l95zb)r or when one voice is used for both conversations but one conversation is louder than the other or has the
lower frequencies removed (Egan et al., rg54). Seleaion is also facilitated by spatial separation of the two voices in different loudspeakers,
one on each side of the subject (Poulton, 1953). Similar improvements
are obtained by using stereophonic recordings in which the two voices
appear to come from different positions (Broadbent, ry50 or when

the messages arrive with different relative intensities at the two ears

(Treisman, r964b). These results are in line with those of Hirsch (t95o)
who found that speech was easier to hear in noise if the loudspeakers
conveying the speech and the noise were placed on opposite sides of the
subject or when the speech was to one side and the noise directly in
front. They are also in line with results obtained by Licklider (1948)
who presented speech and noise in earphones and found that hearing
was easier when the speech was in phase and the noise out of phase, or
vice versa, between the two ears.
Taking these results together, they imply that selection can be made
not only on grounds of straightforward physical characteristics such as
pitch and intensity, but on more subtle phase differences bennreen the
two ears, and the various complex physical characteristics which differentiate one person's voice from another.
The clearest separation is obtained, however, if the wanted and
unwanted messages are fed by earphones separately into the two ears,
as in the now classical experiments by Cherry (tg53). In these, subieas repeated ('shadowed') a message played into one ear while ignor-

ing one played into the other. Practically nothing could be reported
from the 'reiected' effr and subiects were unaware of a change of
language or the substitution of a record of speech played baclsilards.
Facts of this kind led Broadbent (rgS8) to posnrlate a 'filter' benreen
the sense-organ and the central mechanisrns responsible for identification, which can block off signals so as to pass only those with certain
physical characteristics or from a particular sense organ. However, it
soon became clear that the mechanism must be somewhat more com-

plex. For example, Treisman (tg6+b) compared a condition similar
to that of Cherry's experiments in which the two messages come one
to each ear, with two other conditions: either one message came to one
ear and the other to bothr or there were three messages, one to each
ear and one to both. She found, as expected, that reieaion of the unwanted message was more difficult in both these alternative conditions
than when one message had exclusive access to each ear. The separation of the messages to different ears did not, however seem to be the

roo

Fundamcntals of Skill

essential difference between the conditions, because in a control experiment in which the unwanted messages were rhythmic, repeated sounds,
the separation of the wanted rnessage from the unwanted was equally
easy in all conditions. Again Moray (lgSg), repeating Cherry's experiments, found that the subject's own name would sometimes 'break
through' to be heard if it was introduced into the unwanted message,
and Peterson and Kroener (rg6+) found that some items from unshadowed messages could be recalled, although recalt of the messages
as wholes was poor.

Broadbent himself suggested (1958, p. 54) that the filter might acr on
classes of words, xod cited evidence by Peters (tg14a., b). Treisman

(t964a) in a series of experiments in which rwo messages by the same
voice were presented each to both ears and subjects repeated back one,
found that more was correctly repeated when the wanted message was
a passage from a novel and the unwanted a statement on biochemistry,
than when both were from the novel. Still more was repeated accurately
when the messages were in different languages or when one was norrsense or speech played in reverse. It is doubtful whether these are truly
separations in terms of 'meaning' since even those subjects fluent in
French did not reject a French translation of the passage from the novel
any less successfully than an irrelevant passage of French. To some
extent it may be in terms of rhythm or phoneric quality and is thus,
perhaps, to be classed with selection based on difflerence of voice Treisman in the same experiment noted that all these aids to rejection
were much less effective than having the wanted message in a woman's
voice and the unwanted in a man's. Whatever the exact basis of separation it is clear, however, that it could not have been by a purely peripheral filter - there must have been some analysis of the data before
the rejection was made.
Treisman (196o) in an experiment in which subiects repeated the
message to one ear and ignored that to the other, found that occa-

sionally words from the unwanted message would break in if they were
especially suitable in the context of the wanted message. The wanted
message was connected narrative and the unwanted a statistical approximation to English. Half-way through a trial the messages were switched
to opposite ears. Subiects were told always to repeat what came into
one ear. At the changeover a few words from the wrong ear tended to
break in without the subfect realising they had done so. One example
quoted by Treisman is given here, the words spoken by the subject are
printed in capitals and the words rejected in lower case. The changeover
point is indicated by the vertical line.

IOI

Idenhrtcafion and Choice

Vanted.'

tfifE GRowL oF THE

Unwanted.' book is she went to

coAt* swim

fast

DT'RING THE

thtrnder TNcREASED sTEADILy and

the

* coAt was a mishearing of GO TO.
Treisman suggested in explanation of these results and of Moray's
finding that a subiect's own name was sometimes heard from the llowanted channel, that the filter does not block unwanted messages corrpletely but merely attenuates them. Several further experimental
results support this view. Lawson (1966) combined the task of shadowing a passage played to one ear while ignoring a passage to the other,
with responding to 'pips' played to either ear. The response consisted
of pressing one of trro keys according to the ear to which the pip had
come. She fotrnd that responses to pips on the unwanted channel tended
to be slower and less accurate than those to the wanted, but substantial
ntrmbers of correct responses were nevertheless made. Treisman and
Geffen (tg6il in a somewhat similar e4periment found that responses
to certain words or classes of word designated beforehand could be
made whether the words occrrred in the wanted or the unwanted message, although very much more frequently with the former (86%)
than with the latter (8%). The difference was greater than in Lawson's
experiment, perhaps because her'pips' were more readily discriminable
than Treisman and Geffen's desigRated words. Looking at their results
in srgnal-detection theory teuns the latter authors found that, on the
basis of correct detections and errors, B remained unchanged as between
the two channels, but d' was substantially greater for the wanted than
for the trmyanted - in other words the criterion was the same for both,
but the effective signal strength of the trnwanted was lower. These
results are similar in their implication to those of Broadbent and
Gregory (r963a) who asked subiects to rate the confidence with which
they judged a tone to be present in a bnrst of noise played into one ear
while a string of digits was played into the other. They found that d'
was substantially higher if the subjea rguored the digits and reported
only the tone, than if he had to report the digts as well. The general
idea that attention to one signal attenuates data from other signals is
also supported by the results of lVebster and Haslerud (tg6+) who
found that attention to either auditory signals or to signals in foveal
vision raised thresholds and slowed reaction times for signals shown in
peripheral vision, and of Yates (t965a) who showed that shadowing a
message played into one ear was less disturbed by white noise in the
other than by another message or by delayed feedback of the same

to2

Fundamentals of Skill

message. It is also supported by recent evidence showing that elearical
responses in the brain (evoked cortical potentials) produced by incom-

ing stimuli are diminished when attention is directed away from them

(Wilkinson 1967).
We may note in passing that Treisman and Geffen's results imply
that the attenuation is in percepnral rather than response processes,
since the response to the designated words was the same - tapping with

a ruler - whichever channel the word was on.
The results we have surveyed seem clearly to imply that it is not

enough to postulate a filter acting only on the input from the various
sense organs and capable of discriminating against simple sensory categories such as pitch or intensity (Deutsch and Deutsch 1963). There
must be some mechanism facilitating or inhibiting categories of identification. 'W'e are faced, therefore, with having to posttrlate at least two
filtering processes or of explaining away the more peripheral in terms
of the more centrd. It is, in general, reasonable to suppose that there
may be two or more stages of filtering, but on the other hand explaining
away the penpheral filter is perhaps not very difficult. Qualities such
as loudness and pitch could be iust as well filtered late as early if it is
assumed that some representation of them remains, as it almost certainly
must, after analysis in the percepnral mechanism. The chief reason for
postulating a peripheral filter is the ease with which data from one ear
can be excluded, but even this would be consistent with filtering after
some analysis if it could be argued that the data enters the percepnral
mechanism 'agged' with the ear from which it has come. Some such
taggtng seerns essential to account for the facts of auditory localisation mere phase or intensity differences between the two ears without tagging
would lead to identification of a signal as 'to the side rather than on the
midline' but would give no indication of whicfr side. If this argument
is accepted, one fi.lter system in the perceptual mechanism would
be enough to cover not only the results of selective listening experiments, but Donders' c-reactions as well. How such a filter works and
how many stages it entails, are matters for future research to decide.
Meanwhile a first hypothesis would be that it operates in the manner
of one of the models we have already discussed in relation to Hick's
Law.
The experiments on selective listening, like those on selective response, can be conceived as having tried to hold the 'setting' of the
filter constant and to study its limitations. They thus represent the
opposite pole from choice-reaction experiments and perceptual identification tasks such as those of Miller et al. (196r) and of Fraisse and

Idmtifuation and Choice

ro3

Blancheteau $962), which can be thought of as sfirdies of the speed at
which the filter can be reset from one item to another.

The mechanism of identification

To explain why it was mainly words which were appropriate in the
context of the wanted message that got through from the unwanted,
Treisman suggested that words are stored in some kind of 'cerebral
dictionary' and that the occurrence of one word brings about some kind
of partial activation of others which would normally tend to follow it.
If one of these words is also partially activated by the attenuated unwanted message, the combined effeas might be sufficient to cause it to
be spoken. The same general princrple seems to be illustrated in an
experiment by Bruce (rgS8) who presented nrelve-word sentences in
noise and prefaced them by a word ptrporting to indicate their topic.
He fonnd that perception was more accurate when the prefatory word
was appropriate than when it was not. Thus the sentence 'f tell you
that our team will win the cup next year' was correctly heard when
prefaced by 'sport', but when prefaced by 'food' became'[ tell you that
our tea will be something to do with beer'. The effect is the verbal
counterpart of the finding by Carmichael et al. (rglz) that perception of
designs seen briefly could be distorted in the direaion of names given
to them beforehand. Ftuther support is given by experiments such as
those of Pollack (rg6f) and Tulving and Gold (rg63). The former
showed that the hearing of words in noise was facilitated by grving one
or more initial letters either before or after presentation. The latter
obtained better recognition of words e4posed briefly if a sentence
leadi.g up to them was shown beforehand.
The princrple that the identification of an item can be influenced by
spread of effect from other preceding, simultaneous or imrnediately
succeeding items and that these effects may surunate from different
sources, is an extremely powerful one with very wide application. For
example it accounts for the frnding by Miller et al. (rgSl) that words
are more readily idenffied in the context of a sentence. It accounts at
the same time for the fact that perception may be easier if the signals
concerned have been received before even though at such a low intensity
that they could not be reported (for a review see Schiff, 196r), thar
signals can be detected at lower levels of intensity following a warning
signal (Howarth and Treisman, 196r) and that familiar signals such as
one's own name are detected at lower intensities than other signals
(Howarth and Ellis, 196r). If similarity implies some kind of neural

ro4

Fund,arnentals of Skill

proximity, the same princrple also provides an obvious framework of
thought in which to account for the fact that errors and intnrsions show
similarities to correct responses, for example the phonetic similarities
between errors and correct responses observed by Bruce (tgS8) for
words in sentences and by Conrad (r964a) for individual letters heard

in noise.
Looking at percepnral identification in broader perspective, we can
think of each obiect seen or event which occrus as pre-activating potential identifications and responses for other likely objects and events

which will, in consequence, when they arrive be reacted to more
quickly than if they had occrured in isolation - in conrmon parlance
the subiect knows what he is looking for and tends to ignore ottrer

things. Perception can thus be conceived as the continual formulation
and checking of a kind of running hlpothesis. In this process, regular
and prediaable chains of events may be dealt with so rapidly and
readily that the subiea is hardly aware of them: less well anticipated
events which involve a revision of the hypothesis will engage his main
attention. The span of such a running hypothesis involves matters of
coding checking and retaining data which we shall consider in more
detail in later chapters.

IV
Single-channel Operation

T7e mentioned in Chapter r that Craik (rg+8) noted how the course
pursued when tracking a moving target did not follow the target motion
smoothlp but showed a series of oscillations, implyrng that correction
of misalignment berreen target and follower was not made continuously
but at discrete intervals of about half a second. In short, the human
operator was perforrring as an intqmittmt-cCImectim seno,Craik pointed

out that this could not be due to the misalignment having to build up to
some sitical value before the subiect could detect it, because the intermittency was not reduced by magniffing the display. Nor could it be
due to any motor limitation, since hand movements of the extent and
nanue required could be made very much more rapldly than two per sec.
The effect, he concluded, must be in the central mechanisms of the
brain. \[e shall in the present chapter look more closely at the evidence
on this point.
Searching for a cause of the intermimency Craik was led to consider
the reasons for the reaction time between the presentation of a signal
and the emergence of a response. He argued:

'We must . . . ask ourselves whether this delay is more likely to
consist of the transmission-time of nerve impulses continuously
uavelling down an imrnensely long chain of nerve-fibres and slmapses
connecting sensory and motor nervesr or of a "condensed" timelag
occurring in one part of the chain. If the first hlpothesis were correct,
there would seem to be no reason why a continuous stream of incoming imFulses should not evoke a continuous stream of motor ones. . . .
If, on the other hand, the time-lag is caused by the building up of
some single "computing" process which then discharges down the
motor nervesr \tre might expect that new sensory imf'ulses entering
the brain while this central computing process was going on would
either disttub it or be hindered from disttubing it by some"switching"
system.

r05

106

Fundamentals of Skill

'These ideas can be tested to some extent by recording the human
response to a series of discrete stimuli presented at various time
intervals, to see whether there is a minimum interval within which
successive stimuli cannot be responded to. Such an experiment is
analogous to physiological investigations of the "refractory phase" of
a nerve or synapse, as pointed out by Telford (r93r). The results of
Telford and of the writer suggest a refractory period of about .5 sec,
such that a stimulus presented within this interval after the precedi.B
one is responded to later, or may be missed' (Craik, 1948, p. r47).

The use of the term 'refractory phase' was unfornrnate because the
analogy is not really at all close. Ap"rt from the gross difference of timescale, the refractory phase of nerves is clearly a recovery phenomenon
whereas the 'psychological refractory period', 8s it has come to be called,
is due to the time occupied by some central process of uanslating from
stimulus to response. However, the idea of testing by experiments using
discrete stimuli has been fruitful indeed, and to these we now tllul.
THE EFFECT OF A SIGNAL DURING THE REACTION-TIME TO A
PREVIOUS SIGNAL

The first experiments desigued to test these ideas used a type of uacking
task. The subiect had to keep a pointer on a line drawn on a paper band
which passed behind a narrow slit. From time to time the line abruptly
changed position and the subiect's reaction time (fR) to begin to follow
it and movement time (TM) to reach the new position could be measured
from the record of the pointer movements (Vince, t948a, r95o). Sflhen
a change of position was well separated in time from the previous
change, TR averaged z5o to 3oo msec. When, however, two changes
(Sr and Sr) occurred close together so that S, came during TR, TR,
was longer than normal. Sfith one class of exceptions which will be
discussed later, the lengthening could be roughly accounted for by
assuming that the cenual mechanisms took the same time to deal with
the data from both & and S, but did not begin to deal with those from
S, until they had finished dealing with those from 51. In other words,
data from S, had to be held in some kind of store until the end of IR,
when the central mechanisms became free. The events envisaged are
shown in Fig. 4.1 and the result can be expressed in the equation

TRz-TRr*TDr-I

(/<IRr)

(4.r)

where / is the interval benreen S, and S, and TD, is the time taken
to 'process' the data from Sr: this was taken in Vince's case to equal fR

Single-channel

Opuation

rcT

to a change of position well clear of others in time (Welford, rg5z).
Eq. 4.r implies that if fRa were plotted against / and IR, had no
variance, the result would be a line sloping at 45 degrees as shown by
the solid line in Fig. 4.2. Since in any practical case IR, always does

s2

sr

t-....-....

TIME

->

Figure 4.r. Lengthening of reaction time to a signal which arrives duing
the reaction-time to a previous signal.
: TD vr.,\.*r : TM . . . . . . : time held in StOfe
I : Signal

-

TR2

N TR I...

I

Figure 4.2. Ideal plots of TR, against f when .S2 comes dtrring TRr.
The solid line shows the results expected if TR, is exactly the same in all
trials: the dotted line shows those expected when IR. varies appreciably
from trial to rial.
have substantial variance, the actual plot expected would be as shown
by the dotted line in Fig. 4.2.
This result has been confirmed many times in subsequent experiments using lights or sounds as siguals and key-pressing responses which
make clearer cut measures of TR possible than Vince was able to attain.

ro8

Fundammtals of Skill

Some of these experiments have followed Vince's in using a continuous
stream of signals (Hick, 1948, Welford, 1959) but mosr have differed
from hers in presenting discrete pairs of signals. The subiect thus knew
that in any one trial he would have only one signal in each of two classes.
This procedure saves time but in some cases makes it difficult to estimate
TDr. If the subject does not know in advance which of the signals will
come first, S, will require a two-choice reaction, but S, when it comes

will require only a simple reaction (Elithorn and Lawrence, 1955,

Marill , 1957, Halliday et al., 196o, Kerr et al., 1963, rg61rElithorn and
Barnett, t96T). The difficulty can be partly overcome by testing simple
reaction time to each class of signals separately (e.g. Davis, 1956), but
it is clear that conditions in this case are not always truly comparable
with those when both signals of a pair are given (M. Smirh, t967b). The
difficulty is reduced although not entirely eliminated when the subiect
is told in advance which signal will be S1e so that this too leads to a
simple reaction which can be used as an estimate of TDr, or if both S,
and 52 are each drawn from different classes of two or more signals.
Despite their difficulties, these experiments show clearly that the
delays which lengthen IR, are not eliminated by practice (Hick, 1948,
Davis, 1956, Slater-Hammel, 1958), are central rather than sensory in
origin because they still occur when one signal of a pair is visual and the
other auditory (Davis, 1957, 1959), and are not caused by the acnral
execution of the movement (Mr) in response to .S1 because they occur
when ML and Mz are made by different hands (e.g. Davis, 1956),
I7elford, 1959). Indeed they sometimes occur when S, has merely to
be observed and no response to it is required (Fraisse, t957, Davis,
t959, Elithorn, t96t, Koster and Bekker, ry67)., although they do not
always do so (e.9. Borger, 1963). Davis suggested that delays occur in
these latter cases when some established response to S, has to be
intribited or when there is difficrrlty in discriminating between S, and
52. Some evidence for the latter view is provided by Rubinstein (tg6+)
who found substantial delays when S, (to which no response had to be
made) was a large-field stimulus to one eye and S, (which was responded

to) was a similar stimulus to the other or when S, and S, were bursts of
noise to different ears, but no delays when S, was visual and S, auditory
or vice versa.
It seems reasonable to suppose that delays occur insofar as data from
S, are processed immediately even though no overt response is made.
Nickerson $967c) has shown that delays when no response is required
to S, are greater when it conveys information than when it can be
ignored: for example when .S, indicated to the subject which of tno keys

Single-channcl Opration

ro9

to press when .S2 arrived, delays were substantially greater than when
51 was neutral and the indication of which key to press came only with
,Sr. The complementary finding was made by Davis (lg6S) that no
delays occurred when, instead of S1 being given, the subject himself
pressed a 'trigger' key which led after a variable ^[ to the appearance of
52. This finding has, however, been both confirmed and challenged
(Koster and Bekker, 1967, Kornblum and Koster, 1967). Some variability of result is perhaps to be expected since it may be difficult for
the subiect to iguore feedback from his uiggering movement.
The evidence implies that although ody one signal, or as we shall see
later one group of signals, can be dealt with at a rer overlap is possible
in the sense that data from one signal can be dealt with by the translation mechanism while that from a subsequent signal is being received
and stored by the percepnral mechaniSmr and the response to a previous
signal is being executed by the effector mechanism. It is the translation
mechanism that seems to form the single channel which gives rise to the
delays in responcling to Sr.
If this is so, however, a difficulty arises with Eq. 4.r. Although it fits
the faas well, it assumes that the time for which the single channel is
occupied by data from S, is equal to the whole of IRr, md takes no
account of the fact that appreciable times are required for data to reach

the cortex from sense organs and for efferent impulses and muscular
contractions to make a response effective (Davis, r95T). A possible way
of avoiding this difficulty is to assume that some minimtrm feedback
from the responding action, indicating that it has begrrn, is necessary to
clear the decision mechanism. If this were so, the time taken for efferent
impulses to initiate a movement would automatically be included in the
decision time, together with the time required for afferent impulses from
the responding limb to rettrn to the brain. This last time would approxirnately balance the time taken by external signals to reach the brain from
a sense-organ, so that total reaction time would grve a close estimate of
the time required to make a decision and clear the mechanism in readiness for making another. The me relationships envisaged can be
expressed in the equation:
TRz

- TP' * TCr, * TE, * r/<1 + TP' * TC' *(/ TE'TPr)
- +

(4.2)

where TP is the time taken in the sense organ and afferent pathways,
TC is that taken by the central mechanisms, TE is that taken in the
efferent pathways and TK is the time required for the kinaesthetic or
other feedback from the responding limb to reach the brain. Since

rro

Fundammtals of Skill

TP, * TC. * TEL
- IR,
reduces to Eq. 4.r if TKL

and TP' * TC' * TE, TD* Eq. 4t.2
TP*
Some feedback of the kind envisaged almost certainly plays an

important part in checking whether errors have been made: in copylng
activities such as rapid qping, feedback data seem to be compared with
some trace of the input and any discrepancy alerts the subiect to stop
and discover where the error has occurred. Support for the view that
feedback is necessary to clear the central mechanisms to deal with
further input data comes for the 'motor stuttering' observed with
delayed visual feedback. A subiect writing a word when he cannot see
his hand direaly but only via a television screen where its movements
are shown after a delay of about .5 sec will often repeat a letter, suggesting that the 'orders' to write it go on being effective until visual feedback
confirming that it has been done are received (\f. Smith et al., 196o).
Similar results occur with delayed auditory feedback: for example Chase
et al. (l96la) found that subieas who had to repeat the sotrnd 'b' or tap
on a morse key in groups of three produced more sounds or taps than
they should when feedback was delayed by about .25 sec. Similarly
Yates (1965b) showed that delayed auditory feedback impaired the
accuracy of Morse code operators, usually by causing them to insert
extra dots or dashes.
A more positive indication that such feedback time is included in
decision time is perhaps contained in the finding by Fraisse (tgSil,
Davis (lgSg) and M. Smith (r g6lc) that delays found when no response

is required to S, are shorter than when a response is demanded presumably in these cases the release signal for the decision mechanism
comes from within the brain.

THE EFFECT OF A SIGNAL DURING THE MOVEMENT MADE IN
RESPONSE TO A PREVIOUS SIGNAL
Several experiments have shown that some lengthening of TR, may
also occur when S, comes shortly after the end of IR.. Vince's (r948a,
r95o) results suggested that if S, came during the initial r5o msec or
thereabouts of Mp dealing with the data from it was delayed until the
end of this period (tU(relford, r95z). More recent results (Sfelford, 1959)
where this point was examined in detail suggested that the delay was
trntil the end of M1. Probably the most plausible reason for the delay
was given by Hick who suggested that 'the attention may be switched
to that sensory field from which confirmation of the occrurence of the
response is expected. Or alternatively, to avoid the teleological concept

Single-channel Operation

rrr

of "expecting confirmation", we may suppose that the attention is
reflexly deflected by the inevitable stimulation of kinaesthetic or other
receptors by the response' (Hick, 1948, p. 43). If this is so, and if the
cenual mechanisms deal with data fed back from responses in the same
way as that from signals coming from the outside, we should elpect data
from the beginning of the response to 'capture' the decision mechanism
for a brief period gFb) - in other words the response would be manitored. Any fresh sigual from outside arriving after Tm had begrrn would
be dealt with only after it had ended. TFbrmight or might not coincide
with TMr: when it does so it may well be that M is 'tailored' to TF-b
rather than vice versa - we shall gtve other examples later of where this
seems to be true (p. tz7). The time taken bysuch monitoring presumably
depends on the information conveyed by the feedback and will therefore, like reaction time, depend upon such variables as the frequency of
the response as well as on its complexity.
The time relationships envisaged are expressed in the equation

TRz: TR, * TD, * TFh - I

(TR, <I

<(TRr* TFbr)) (+.1)

Delays in respondirg to 52 have also been fotrnd when it comes shortly
after the end of Mr, suggesting that the decision mechanism may
become occupied for a further brief period at the termination of M, rf

no .Sz has by then arrived. The time of occupation cannot be measured

with certainty but has been estimated for Vince's experiments at about
r5o msec (S7elford, r95z) and for the present writer's (Welfor4 1959)
at about 2oo msec - in each case a time similar to TFb. It seems reasonable to suppose that any 'high point' of kinaesthetic, tactile or other
stimulation arising during a response might captrue the decision
mechanism and that the end of a movement often provides such a 'high

point'. Alternatively, there may be some monitoring of the completion
of each response which enables the subiect to set hi-self in readiness

for the next. The time relationships involved are set out in the equation

TRz- IR'. * TDr* TMr* TFer- I

((IR'*rM)</<
(TRr * TM, * TFel)) (+.+)

where TFe is the time for which the decision mechanism is occtrpied by
data fed back from the end of a response.
If neither IRp TFbr, nor TFel varied from trial to trial the plot of
TR, against f would, accordirg to Eqs. 4.rr 4.3 and 4.4, assume I stwtooth form, starting high when .I was short, and falling at 45 degrees
as in Fig. z until the end of IRr; thereafter it would rise sharply to fall

rrz

Fundamentals of Skill

at 45 degrees until the end of Tfrr, and would rise again at the end of
M, to fall once more at 45 degrees to the end of TFe1. Variations in the
times concerned will, however, always obscure this picture and tend to
produce a gradual fall of TR, as .I increases up to a point some time
beyond the end of Mr.
Confirmation that monitoring of the kind envisaged in Eqs. 4.3 and
4.4 does in fact occur and that it takes time is contained in the results
of an experiment by Leonard (rgSl). His task could be presented in tvyo
ways. In the first a lever had to be moved to the corner of a uiangular
guide corresponding to that of a signal light in a triangular display. On
reaching this corner the light would go out and another one of the three
come on, and so on. The task in the second condition was the same,
except that with each light a second light came on which gave the
subiea foreknowledge of which light would be the next in the series.
Performance in the second condition was faster than in the fust, presumably because the advance signals enabled a decision to be made
about the direction of each movement during the preceding movement.
The important point in the present context is that the advance signals
also seemed to prevent attention to feedback from the responding
movements. Many subiects, to their surprise, found that with the second
condition in contrast to the first, they had little awareness of what they
were doing - the movements seemed 'automatic' and without conscious
control.
Why the signals in the display should have taken precedence over the
monitoring of responses is not easy to see, but Leonard's experiment is
not alone in posing this problem. If Eq. 4.1 holds, it seems clear that
whenever S, comes during IR, monitoring of M, must either be cut
out or deferred, otherwise TR, would be lengthened by a delay to the
end of TFbr or even TFelinstead of to the end of TRr. This is a point
which merits further investigation. Meanwhile we may note that performance in Leonard's experiment when advance signals were given
seems in many ways comparable with that of expert semi-skilled
industrial repetition workers who, over months and years at the same
iob, attain high speeds of working while at the same time losing awareness of their detailed actions.
Davis (rgS6) and Marill (tgSil in paired signal experiments srith
simple key-pressing responses, found no lengthening of TRrif S, came
after the end of TR1 - it looked, in fact, as if the responses made by their
subiects were not being monitored. This is plausible in that practice is
likely to increase the accuracy of actions so that they do not need to be
checked - in other words, the feedback will become redundant (Annett,

Singlc-channel Opration

rr3

t966a). The amount of practice needed before this state is reached is
likely to increase with the complexity of the response: for simple actions
such as those used by Davis and Marill, the practice possible within a
nonnal laboratory experiment could well be sufficient. With more
elaborate, graded actions, however, the practice required to do without
monitoring would probably be much longer, snd attained ody by
musicians, tlpists, industrial repetition workers, professional games
players and others who exercise a skill over a ntrmber of years.

SOME ANOMALOUS RESULTS

Although Eqs. 4.rr 4.3 and 4.4 seem to hold reasonably well for a very
substantial range of experimental data, there are two important t]rlpes
of case in which TR, tends to be shorter than these equations predict.
Grouping of signals and responses
When f is very short, reaction to & is sometimes delayed as Eq. 4.1
predicts, but is often hardly delayed at all. When this is so, however,
there are signs that Mrand Mzare in some way co-ordinated: in some
cases both are made simultaneously or M2begins before M. has finished;
sometimes the responses are in the wrong order suggesting that the
order of S, and S, has not been resolved: an interesting discussion of
temporal factors in the perception of simultanerty is given by Norrnan
(t967). More often temporal order seems to have been perceived
correctlS but as a rrnitary patteru. fn Vince's experiments, where S, in
effect cancelled &, the total result nith short .I was sometimes that no
response \iras made. fn some cases the shoft TRzis coupled with a TR,
which is longer than normal, suggesting that the subiea has as his
categories of response, not only M, and Mrbatalso (M, * Mr) which,
because it is rare, takes longer to produce (Hick and Welford, 1956).
In other c:nes both TRrand TR,. seem unusully short suggesting that,
when both signals have appeared, the subiea makes an undifferentiated,
and therefore rapid, response to both. The different types of grouping
effect have been illustrated by Halliday et al. (196o), by Kerr et al. (1963,
1965) and by Elithorn and Barnett (rg6il although their evidence has to
be taken with caution because they used only short /s (of up to roo msec)
and the subiects' whole strategT of performance may therefore have
been different from that in experiments where a wider range of Is were
presented.

The evidence suggests that when I is very short - up to about

rr4

Fmdamentals of Skill

roo msec - .S1 and S, may in some way be reated together as a single
group. Such grouping might arise if the subject deliberately waited a
brief interval after S, to see if any further signal arrived before committing himself to a response. This seems to have been done by one
group of Adams' (1962) subiects and by one of Borger's (lg6l) subiects,
and was clearly done by Sanders' (1964) subiects when instructed 'to
collect all perceptual data before any response is carried out' as opposed

to completing IRr 'without taking any notice of .Sr'. Alternatively
grouping might occur if both Sr and Sz arrived while the decision
mechanism was occupied with a previous signal, so that data from both
were waiting when the mechanism was cleared.
These e4planations srer however, insufficient to account for all cases
and it seems necessary to postulate in addition that the 'gate' which
prevents data from subsequent signals entering the decision mechanism
once it has started its computations, takes an appreciable time to close.
The t'tne appears to average about 8o msec: signs of grouping are very
frequent when / is less than this, rare when it is greater. The time is
probably, however, affected by a number of factors at present imperfectly understood: for example Kerr et al. (rg6f) found grouping to be
less frequent when 51 and M1 were on the non-dominant side and S,
and M, on the dominant than with the reverse order of sides, but that
among grouped responses those where TRrwas especially short tended
to follow the presentation of S, to the dominant side. Meanwhile it is
perhaps significant that the time during which grouping occurs is very
similar to that over which perception of a pattern exposed briefly (say
f,or 'r sec) can be prevented by a subsequent bright flash of light (e.g.
Cheatham, r95z).

The result of such grouping is that, if grouped and ungrouped
responses are not separated, the average TR, is shortened (e.g. SlaterHammel, 1958, Adams, t96z). In extreme cases it is actually shorter

when / is very short than when / is a little longer instead of falling
continuously with increase of / in accordance with Eq. 4.r (Elithorn and
Lawrence, 1955, Marill, 1957, Koster and Bekker, 196T).

It must be emphasised that grouping of this kind is not a process

confitted to the conditions we have been discusshg, but seems to be a
coulmon - indeed essential - feanrre of many high-speed performances.
Craik (lg+8) pointed out that if a musician dealt with each note, or a
typist with each letter, separately they would perform very much more
slowly than they in fact do. The rates acnrally attained imply that the
decision mechanism takes in data from whole musical phrases or words
or even groups of words as single units and issues 'orders' to the effector

Single-chatmcl

Opration

rr5

side in tenns of correspondingly complex series of actions. Vince Gg+g)
has demonsuated this point with experiments in which subieas had to
tap a Morse key in response to dots on a paper band moving past a slit

in a screen. She found the number of dots that could be reacted to
accurately was very much higher if they appqred in groups of two,
three or four than if they were presented singly.
Selective perception

A specid question about grouping arises with regard to certain aspects
of selective perception discussed in the previous chapter (pp. 98-ro3).
In a general way the single-channel model covers the facts of selective
perception in that when two or more strqlms of data each require
different actions, only one is dealt with at a time unless the data and
actions can be in some way co-ordinated. However f, as we have
suggested, signals queue until the decision mechanism is free, we should
expect rapid alternation between tasks according to the precise moments
at which signals in the different streams arrive, so that it should not be

possible to keep attention focused on one stream to the exclusion of
others. This can nevertheless be done in experiments such as those of
Cherry and Treisman surveyed in the previous chapter Gp. 99-ror) in
which subiects repeat ('shadow') a passage of prose played into one ear
while ignoring a different passage played into the other, and the question
arises of why there is not instead continual switching from one passage
to the other ? Admittedy, as we have seen, occasional words from the
passage to the neglected ear are produced, but these seem to be determined by their appropriateness in the context of the shadowed passage
or because of special familiarity - such as the subject's own name - and

not by their precise timing.
Broadbent (tgSZb) has suggested that a continuous flow of data
through one sensory channel might of itsef be enough to maintain the
attenrntion of data from others, and certainly he has shown (r95zb) that
when the individual words of trro messages are presented alternately,
selectivs fislsningto one is difficult.Vince's and subsequent experiments
suggest, however, that continuity would have to be such that no gap
occurred which was longer than about 8o msec - that is the time taken
for the 'gate' to close. Such a maximum seems impossibly short, so that
some additional reason for the maintenance of selectivity is required.
One possibility is to posnrlate a filter or switch prior to the decision
mechanism which enables data from all channels except one to the
excluded (Broadbent, 1958), but this has the disadvantage of making an

116

Fundamentals of Skill

ad hoc addition to the chain of mechanisms between input and output.
The alternative seems to be to question whether the time taken by the
gate to close is always as short as Vince's and subsequent experiments
suggest. Is it, perhaps, that with more connected data, especially with
verbal material in which coherent phrases last much longer than 8o msec,
the 'grouping period' can be lengthened and with it the gap that can be
left before attention is spontaneously switched elsewhere ?

Modification of response
Several snrdies concerned with the single-channel hSpothesis have
exanrined the time taken to modify an action (Vince, rg48b, Hick, 1949,
Poulton, r95o). The results appear to be consistent with Eqs. 4.r, 4.3
and 4.4, the equation applicable depending upon whether the modification depends on an amending signal given during IR, upon the beginning of Mror upon its results. fn some tracking experiments, however,
modifications are occasionally made in a very much shorter time than
these equations would predict - in about roo msec as opposed to 2oo3oo. It is perhaps reasonable to regard these as cases where an error has
been made by the central effector mechanism rather than by the decision
mechanism - the 'orders' given by the latter have been correct, but have
not been correctly carried out. Such modifications seem to impty a
comparison of feedback from the moving member with the decision
mechanism's 'orders', and correction without the issue of fresh'orders'.
If this is uue, modifications should take place in these cases without
visual observation of the error concerned.
Some evidence supporting this view is provided by Rabbitt (r966c)
who found in a serial choice-reaction task that errors cotild be detected
and corrected even when no indication that they had occurred was given

by the display, md that corrections were in most cases substantially
faster than the corresponding accurate responses. However, although
corections tend to be quicker than accurate responses (Rabbitt, t966a,
Rabbitt and Phillips, t967) they are often not greatly so, suggesting that

in these cases an error has occurred in the translation mechanism and
that some of the work of seleaing the response has therefore had to be
done again. It is understandable that in these cases the correcting
response should still be somewhat quicker: Rabbitt (r967b) cites an
unpublished snrdy by Burns who showed that errors in choice-response
tasks were not entirely random but were in part correct. They might,
for example, in a four-choice task be in the correct pair but the wrong
member of the pair. This is not the whole explanation because Rabbitt

Single-channel Operation

rr7

(t967b) has shown that responses to errors may be faster even if they are
quite different from correct responses. He used a choicereaction task in
which responses were given by keys under the several fingers of one
hand. Sfhenever an error was made the subject had to depress keys
trnder both thumbs. Rabbitt regarded this as an'anatomically awlcvard'
response, but from what was said in the last chapter (p. 78) about the
neurological control of the fingers, it might be regarded as relatively easy
and thus understandably fast: while each finger would have to be
depressed individually without moving the othersr the thtunbs could be
depressed by a relatively gross rotation of the hands. It should be
emphasised in passing that rapid correction of erors is by no means
universal: in some cases it takes substantially longer than making a
correct response (e.9. Adams and Chambers, tg6z). ![e may surmise
that in these cases there is not only recalculation of orders by the
uanslation mechanism, but some review of the strategT ofperformance.
The corrections discussed by Rabbitt do not seem to involve any
breach of the principle implied in Eqs. 4.rr 4.3 and 4.4. A clear breach
does, however, seem to be shown by the results of an experiment by
Vince on changing the speed of a response (Vince and Welford, t967).
These results suggest that an amending signal may sometimes get
through the closed 'gate'. Subieas were presented with vertical lines
rising r.5 in from a baseline on a paper band revolving on a kymograph
drun behind a screen in which was a vertical slit ro mm wide. The lines
were spaced so as to appear at irregtrlar intervals of z3 sec. One group
(A) was told that as soon as a blue line appeared they were to draw a
line of the same length in the slit and then return to the baseline, making

the whole movement smoothly and without hurrying. After some
practice lines they were told that occasionally a red line would appear,
in which case they should make the movement as rapidly as possible.
A second group (B) made their normal responses to the blue lines as
rapidly as possible and their occasional responses to the red lines at
leisure. A third group (C) were gtven the same instnraions as group A,
except that when a red line occrured they were to stop their movement
and pause before returning to baseline.
The results are set out in Fig. 4.3, in a way suggested by Bertelson
(r967a). They consist of all cases in which Sz arrived during the reaction
time to .S1 and separate responses were made to both. Cases were
excluded in which the response to the red line was wholly substinrted
for that to the blue - a kind of grouping effect which sometimes occurred
when the interval between S, and S, was less than roo msec. According
to the hypothesis, the points for each group should lie on a straight line

rr8

Fundamentals of Skill

sloping at 45 degrees. Those for group B are in very fair agreement with
prediction, and this was so for each of the subiects individually. The fact
that the observed TRrwere rather too long when S, came only a short

C.)
IJJ

a
a

c{

o

o-

UJ

=
F
z.

o
tr
o'g
?
UJ

G

0.3

0.2

INTERVAL BETWEEN ARRIVAL OF SA AND END OF'

0

REAcTToN TIME To sr tSec.l

Figure 4.3. Times taken to modiff the speed of a response. Results of
an experiment by Vince (Vince and $7elford, tg67).
O- O, Group A (slow to fast): three subiects.
O-O, Group B (fast to slow): three subiects.
O - - O, One subiect in group C (slow to stop)-whose performance
resembled that of group A.

O - - O, Two subiects in group C whose performances resembled

that of group B.
Each point is the mean TR, for a range of intenrals between the arrival
of Ss and the end IR,.. The ranges (from right to left) in msec were
v49, 5an9' roo-r49, r5o-r99, zoo-249 and zlo and over. The numbers
of readings contributing to each point for groups A and B vary from
7 to 38 (mean zo); those for the single subiect in group C from 2 to t4
(mean 8'z) and for the two subjects in group C from 7 to zr (mean r3.7).

time before the end of IR. may have been due to monitoring of Mr:
feedback from the beginning of Mrmay have become grouped with S,
and thus lengthened TRr. The results for group A showed no systematic

uend and this was again true for each subject individually. Two
members of group C behaved like those of group B: their shorter mean
reaction times are understandable in that the decision to stop a response
is probably simpler than one to change its speed. The third subject in
group C behaved like those in group A.

Single-chanrul Opration

rr9

The results suggest an important refinement of the single-channel
hlpothesis. The central processes following 51 were concerned with the
initiation of a phased pattem of muscular action, which had to be
modified in response to 52. Clearly, the extent of this modification was
very different for groups A and B. The latter, in order to slow a movement down, would have had to change the pattern of muscular innervation substantialln bringng antagonists into play in order to arrest the
rapid motion. For group A, however, the speeding up seems likely to
have involved merely an intensification of the pattern already in operation. Group C should at first sight all have behaved like Soup B, but it
is possible that the subiect whose performance was like that of group A
managed simply to discontinue his movements instead of actively
arresting them. If so, it seems reasonable to suggest that the decision
process protected by the 'gate' is the initiation of a fresh pattmt of action,
or to put it another wsy, the changing of relations betsteen input and,
output. Thus a sigRal to bring other muscles into play is excluded, while
siguals conveying instructions merely to increase or reduce the aaivity
of those already in action can get through.
Indications that the gate does not block potentially interfering signals
completely but merely attenuates them is in any qlse suggested by the
results of several experiments in which fRl has varied with /, sometimes
increasing and sometimes decreasing although the amotrnts have been
small. Further indications are contained in an experiment by Helson and
Steger $962) who fotrnd simple IR to be lengthened by the occrurence
of a second signal to which the subject did not have to respond. .S2 was
presented from ro to r8o msec after ,S1. The effect was maximum Gf/o,
i.e. about 25 msec) when .I was 4o-r2o msec. It was less with longer /s,
perhaps because in many cases IR, was by then over. This additional
evidence is, however, not strong for trro reasons. Firstly Koplin et al.
(1966) failed to duplicate Helson and Steger's results; secondly Gottsdanker et al. (lg6f) found some lengthening of TR trnder these conditions even when S, came '5 sec after Sr. In this case ^I was clearly too
long for there to have been any interaction of S, with the decision
relating to Sr, and the lengthening of TR. must imply a change of
strategy on the subiea's part. If so, a similar change might trnderlie the
lengthening of IR, in other cases as well.
More generallg the results reported by Vince and Welford are perhaps consistent with the finding that an irrelevant stimulus gwen iust
before a signal can quicken response, and that the quickening is greater
with strong than with weak stimuli (Iohn, 1964).

t2a.

Fundammtals of Skill

ALTERNATIVE THEORIES

Anyone who has attempted to test the single-charmel hypothesis will
know that the experiments required are difrcult to conduct and still
more difficult to interpret and present clearly. It is therefore understandable that despite almost overwhelming evidence, a number of
alternative e4planations have been offered. These have been discussed
by the present writer (rg1z, r959)rby Bertelson (1966) and by M. Smith
(t967a) so they will be mentioned only briefly here, except for one about
which evidence has recently accumulated. We may disctrss them under
four heads, showing in each case how far the theory accounts for the

writer's (lg5g) data.
(r) An early suggestion was that there is a refractory staterindependent
of TRr. and TMy following an event in some part of the chain of
mechanisms leading from signal to response. Various periods ranging

from'r to'5 sec have been proposed. None of these gives a good fit to
the data in that no one fixed interval will account for the observed
delays both when 51 comes during TRr and when it comes during TM,
(see Table 4.r(a)).
(D) Several authors have proposed that perception may be quantised

into samples about a third of a second long, and Broadbenr (lgS8)
suggested that a subiea begins a new sample when 51 arrives so that
when / is less than a third of a second the data from S, have to wait

until a new sample begins. This theory makes fairly close prediaions
of IR, in some cases but does not account for certain other features of
the data. For example it implies that when S, comes durin1 TRu TR,
for auy given /, or more generally ( TR, * /), will be independent of
TRr whereas Eq. 4.r imptes they will be correlated. It also suggests that
when 52 comes during TMr,, TRrand IR, $rill be positively correlated
since short TR, would be secured if fRl ended and 52 arrived before
the end of the first sample, but TRrwould be much longer if S, missed
the end of the first sample and had to wait to be dealt with until the end
of the second. Substantial correlations were in fact fotrnd by both the
present writer and by Borger (1963) when .I was /ess than TRrrbecoming

very low when .I was longer (see Table 4.r (b)). The theory also predias
that any factor affecting IRr will also affect TR if S, comes during IRr.
Sanders (1967) and Broadbent and Gregory Gg6il have tested this by
varying the degree of compatibility berween S, and its corresponding
response while holding the relationship between ,Sz and its response
constant. If intermittency is due to a division into fixed quanta, TR,
should be unaffeaed by the compatibility of .S1 and its response: if the

o
ct

S $E

3:'y I $.* H
'Ss s 'tsXcrr,rt

GI

o
P(

$a
+;

r$t fSHC,'E'
E+o#';..
ogeE€
sxs
*'s i
E sr
ilE s
s,s s

R3S
THR
h"Ps

SEb
S EE
.t qo
'sE*
q)

\-./

R) q)

+++

SAE
r{}{N;N

cDOv

Ek

E
b0
A
LT

Tb

€o
o
F{

k
k
ct
G!

v)

r{ r{
\o
l{OFa

6

+ll

+

t{H

ES

cD

(u

b0

d

.d

l-

$s.$

(u

E $.s
-Sa
I

S.E I
D

l\r

s
ss
!'dN
l tR6
t+s

pl.S s

E s$

U'

GI

6!

(rr)

tr-

t{
t{
H

f(r1

I

I

rF{

fE
t\a

$
H

cGI

+l
fn

a)

o\

r{
A
Fa
.tri
A

cn

O\N

N
6'b\qi
.tN
N

rnOrnr/l
rn

EH
u-o

Hl{

+l

€o
o

f,
+

SA
N.E

\o

rrr

(n

U

$
N
I

k
\OA
$s)
FrP

r'E

r,,l

$
I

.ts

tr

f{

cl

F{

a

d

rn

o

z

H

+

(.)

@

*

\rr

=

a
lt

fi8
89
a.g
cr€

.93 o

l,

d
cd

.Er
HB
6h
5H
o.q
l,

€o

d
lr,

-t ,;
la

.r1

H

o r-l

kk

as
loo
6H
gE

c
'aE
H9

.d
k

€a

t&

vE

)

€ss

8s*

S

a

a
tc,

€ss

^

h

ct,
F{

a;

cl

n ilfls

r{

$

+

€tr

t\

t$t
$E

E

t(

F
Fr

o

:R;
s$s

+o

tr

=
Ftt
)
3t: F trE

E&s

o

ll q

.Fa

BE8
Ed
.o.S H

cD
.ri

8+:
€-oo
O.tr
>r
.cg

-'

t.i

L.

E6
a

Fl

'E#
!.(u

'd.9
ssH $ H $ Ee
oatr
{98
E V€
H..i -Z^,8
HE Sa
fr
Yn-t
5i
50C)

(+{

o
k
o
k
k

o

Ek

€tr
GI

cl

a-,

clt

Ef
;=_;5
'EEtsEsIur
iB
pn-E

Eg

#

b9
aH
o

3 3.qn

g ob

&P g

;EE
5.9 e

\3^

SH

sgs
an\o o
Esa
B+F

E F
*
SggiftBii3gEE8} E{?a

I 2T

rzz

Fundarnmtals of Skill

single-channel hypothesis is correct the incompatibility which lengthens

IRr. should lengthen TR, by the same mean amotrnt. Both authors
found that this occrured. Broadbent and Gregory also fotrnd that when
one of the two possible signals which might constitute S, was given less
frequently than the other so that the reaction time to it was longer than
to the other, the delays to S, were also longer. Further confirmation is
given by Smith (1967b) who lengthened IR. by reducing the intensrty
of .S1 and fotrnd that TR, was correspondingly lengthened.
(c) Elithorn and Lawrence (lqSS) seem to imply the suggestion that
the results of experiments using pairs of responses made by differert
hands could be accounted for in terms of cortical or other cenual interaction. A somewhat similar suggestion appears to be made by Reynolds
(1964, t966). The fact that the cortical response to S1 in some sense
inhibits that to S, is not in doubt - it is indeed the foturdation of the
single-channel hypothesis. The question is whether the fust response
blocks the second at the output end of the decision mechanism or
whether it blocks entry of data at the input end. At present the latter
seems more likely on several grounds. If ML blocks the emergence of
My the lengthening of TR, should depend on TMrin such a way that
in place of Eqs. 4.r and 4.3 we could write
TRz

- IR' * TM' - I

(TD, < IR. * TM,

- I) (+.s)

This equation gives a very poor fit to the experimental data (see Table
4.1(c)). On an observational level, when grouping occurs one response
seems often to facilitate rather than inhibit the other. If an argument on

grounds of functional efrciency can be admitted, inhibition at the ouqpur
end implies a wasteful manner of operation by the brain with two or
more independent decision mechanisms each of which would presumably have less capacity than one which made use of the total resoruces
available.

(d) The most persistently canvassed alternative to the single-channel
hlpothesis has been the suggestion that the delays in responding to Sz
can be accounted for in terms of temporal uncertainty effects. Many
studies have shown that when a warning precedes a signal by an interval
(foreperiod) which varies from one trial to the next, reaction to the
signal is slower on those trials when the interval is very short - say
2oo msec or less - than when it is somewhat longer: it is generally
assumed, as we noted in the previous chapter, that the subject makes
some kind of preparation during the foreperiod but that the state of
preparedness cannot be held at optimum level for more than a fraction
of a secondr so that the subiect prepares for the mean or modal fore-

Single-channel Operatiut

r23

period and is less than fully prepared if the signal comes earlier. W'e may
note in passing that in signal-detection theory terms, the warning and
preparation seem to lower P so that response becomes readier but less
accnrate. Bertelson Q967b) who demonsuated this showed that changes
of reaction-time with foreperiod closely corresponded to changes in
ntrmbers of errors. He suggested that the warning signal may have a
direct facilitatory effect by, in a sense, adding to the signal.
It has been held that lack of preparedness can account for the delays
to TR, in the experiments we have been discussirg, although it seems
equally plausiblerprima facierthat the temporal uncertainty effects are,
at least in some cases, due to single-channel delays caused by the warning
capnrring the decision mechanism.
Attempts to distinguish betvveen these views consist of separating
temporal uncertainty and single-channel effects. The methods used so
far fall into five classes:
(i) Klemmer (lgS6), IGrlin (lgSg) and Drazin (1961) each controtled
single-chant'el effects while varying temporal uncertainty by presenting
a range of foreperiods which remained the same wtrile the minirmtm
foreperiod changed from one block of trials to another. Drazin, for
exlmple, compared ranges of z.v4.o sec \dth r.o-3.o, .5-2.5, .25-2.2s

and 'r25-z'\25 sec. In all cases, reaction times were a little longer at the
beginning of the range. The absolute lengthening - about zo msec differed little between the three higher rangs, all of which could be
regarded as clear of single-channel effects. Klemrner's and Karlin's
results were roughly similar. It looks, therefore, as if temporal uncertainty effects do exist apart from single-channel delays but that they are
much smaller.
(ii) Nickerson (1965b) used a somewhat similar plan but required
responses to both .S1 and S, while presenting different ranges of /:
'r-'5, '3-'Tr'5-'g and 'r-'9 sec. As /increased IR2 became shorter over
each of the ranges used, by about 60 msec in the .r-.5 sec range to about
30 msec in the '5-'9 range. Unfornrnately little weight can be attached
to his results as he took no account of possible effests of feedback
from My
(iii) Nickerson (r96fu) secured the independence of temporal uncertainty and single-channel effects by arransng that / in a succession of
trials varied in such a way that there was always at any instant an equal
momentary probability of Sz appearing. He varied this momentary
probability in different blocks of uials, and found that TR, increased
both with increase of / and with decrease in momentary probabifity,
implytng that both single channel and temporal uncertainty effects were

tz4

Fundamentals of Skill

occurring. Thornas Gg6il has calculated from various sets of data by
previous workers that expectancy in terms of the conditional probability
that a signal will arrive, given that it has not arrived before, is insufficient
to account for single-channel effects.
(iv) fn some experiments temporal uncertainty has been minimised by
keeping foreperiod length, or in double signal experiments /, the same
over a block of trials. Subjective temporal uncertainty is not wholly
excluded by this method since the subiect's ability to iudge time intervals
is not perfea, but after a little practice temporal uncertainty effects seem
clearly to be reduced (Reynolds, t966).
(v) Several experiments have compared delays in double signal conditions when responses are required to both S, and Sz with conditions
when response is made only to S, and 51 becomes in effect a second
warning. Delays in the latter condition have almost always been substantially less than in the former.
Methods (w) and (v) are both illustrated in an experiment by Kay
and l[eiss (1961) whose subieas made trials under several different
conditions after considerable practice. In all cases a trial began by
pressing a'ready'key. There followed after I r, 213 or 4 sec foreperiod
a click (Sr), and this was followed after an I of z5-rrooo msec by a
second click (Sr). Their results are shown in Fig. 4.4. When both foreperiod and f were constant over a block of trials and no response was
required to 51 (condition cc), the subiect could in effect begin to react
to S, as soon as he pressed the 'ready' key although occasional catch
trials in which S, was omitted would prevent him doing so completely
before S, arrived. Both temporal uncertainty and single-ctrannel effects
were thus excluded andr 8s expected, IR differed little with / up to
25o rnsec. The slight fall of IR with longer / perhaps indicates that
5oo rnsec or so were required to take full advantage of the warning given
by S,.. With an irregular foreperiod (condition zc) S, would convey more
information and it would be expected that its full benefits would take
longer to realise. It is therefore not strrprising that in condition zc TR
became longer with short values of .L The extra delays with irregular I
(conditions cv and zo) were clearly small and well within the range of
temporal rurcertainty effects found by Drazin (196r). The difference
between TR, in conditions VC and VV in which a response had to be
made to S. was also within this range. Delays in these latter conditions
were much greater than when no response was made to S, and it is
particularly important to note that delays were not abolished in condition VC when .[ was held constant: the delay here clearly cannot be
accounted for in terms of temporal uncertainty. Similar evidence has

Singlc- chanrul O per ation

125

()

lrl
(,
v'L

o
d,

o

o
(J
UJ
(.r)

o
F
UJ

0.1

=
tr
t-

o
F

()
ttt
(ts

0

0.2
0.5
INTERVAL BETWEEN FIRST AND SECOND SIGNALS (SEC )

r.0

Figrrre 4.4. Results of an experiment by Kay and !7eiss (196r).
The iesults are plotted as follows:

cc

Foreperiod I
Response to
Constant Constant S, only

\r
VC

Variable Variable 52 only
Variable Constant 51 and 52

;: ?r,Iffi ::t3lf. ::t*I

rhe data Jl"m a,#i:?",o*. n::ti."*x,Jillt;ve practised

subiects.

been provided from experiments by Borger (1963), Creamer (1961) and
Bertelson (r9 66, tg67a).
A further indication that the delays in responding to S, are not simply
due to the effects of temporal uncertainty is the substantial correlation
benreen (IR, + /) and TR* shown in Table 4.1. Since temporal uncertainty should have nothing to do with IR,., no correlation would be
expected (Table 4.r(d)).

The predictions of the single-channel hypothesis shown in Table

4.r(e) give a better overall fit than any of the alternative theories although
the times prediaed are a little too short both when S, comes druing TR,
and when it comes durin9 TMr. The discrepancy is more than enough
to cover the temporal uncertainty effects found by Kay and Weiss and
by Drazin. Some at least of the extra time was probably due to grouping
of Fb, with S, in a few cases when it came just before the end of TR,

r26

Fundamentals of Skill

t---l

A

#-rrl

to o -l

Hifoaol
--t

a

I

I

0.5

1.0

I

l- --l

l-

2.0 sEc

1.5
l_ _ _l

H

t--;l..*

B
t---l

t....1..'.fr"tl

q

0.5

1,0

I

l-s

2.0 s Ec

F---frvw-..t..1

%-.t

c
I--lk-l

t....1....to...t

0

0.5

1.0

1.5 sEc
Figure 4.5. Successive responses in continuous performance.
A. $7hen both TFb and TFe occrlr.
B. !7hen TFb occurs but not TFe.
C. \U7hen all TF is cut out.

and produced very long TRrs instead of the very short ones predicted
by Eq.4.T. Similar grouping may have occurred with TFel when S,
arrived close to the end of ML. The instances were too few to ueat
separatelp but had they been omitted from the calculations the mean
TRrs would certainly have been appreciably closer to those predicted.

THE SPEED OF CONTINUOUS PERFORMANCE
Craik (lg+8) assumed that the speed of a continuous performance such
as tracking was limited by the times, firstly to observe and decide upon
corrections for misalignments, and secondly to carry out the correcting
actions - in short by the sum of the TRs and TMs involved. He
observed, however, that correcting movements tended to run into one
another as subieas became more practised ffid, as we have noted,
subsequent work has shown that TR can overlap a previous TM. We
should, therefore, rather say that, so long as TM is shorter than IR, the
speed of a continuous performance will be limited by the decision times
involved plus the times required to carry out any essential monitoring
of responses - that is by the sum of IRs and essential IFs. This may
be either greater or less than (fR + fM) according to circumstances:

Single-channel Opration

r2T

(a) When actions have to be carried out meticulously and the display
is static, as for example when tracing carefully over a pattern, subiects
may well monitor the ends of movements as well as any earlier significant
points, as shown in Fig. 4.5(A)r so that TFe wifl have to be added to

(rR + TliI).

(D) When astion does not have to be so precise, or when the display

is changing so that misalignments are continually building up, TFe
is likely to be cut out, and the speed of perfonnance limited by
(fR + IFil as shown in Fig. 4.5(B). For tracking, taking TR as 3oo
and TFb as r5o msec, as in Vince's (r948a) experiments, yields a correction rate of r per 45o msec which is close to the half second suggested
by Craik, and in fact a better fit to Vince's data. This figure is also
consistent with the breakdown of high-speed tracking performance
found when the uack changes direaion more than about twice per sec
(\[elford, 1958, pp. 8696).
(c) \fhen the display changes very fast even TFb may fail to capture
the decision mechanism. In uacking tasks this would follow from
Eq. 4.r if misalignment built up so fast that there was always a substantial correstion waiting to be made before the end of the TR to the
previous observation. Speed of performance in this case would depend
on IR alone as shown in Fig. 4.5(C). Acctracy in these circumstances
would, however, tend to be low unless responses were very simple and
ungraded, since any error made in one movement could not be corrected
in the next, but only in the next but one. This t5pe ofperformance seerns
to have been attained in the writer's high-speed tracking e{periments
already mentioned: when the track was changing direaion three times
per sec, subjects maintained the corred.number of changes, but accuracy
was very poor compared with that attained with changes of two per sec

or less.

Limitation of the speed of performance by decision processes rather
than motor action is illustrated in several everyday activities, perhaps
most notably in speaking (Goldman-Eisler, 1956, 1958, tg6rrHenderson
et al., 1965). Although occasionally people think faster than they can
speak, more often they are unable to forrrulate the content of a statement
fast enough to maintain a constant rapid flow of significant words. They
may in consequence speak slowly and so adiust the speed of action to
that of cenual conuol. If not, they inuoduce redundant words or meaningless sounds (such as '.r'), make pauses or reduce ttre average information per word of their statement. Such effects tend to be more
pronounced preceding high-information words and in sentences involv-

ing difrcult constmctions or other complex cognitive aaivity, as would

r28

Fundammtals of Skill

be expected on the ground that these require relatively long times to
retrieve data from memory or to order words in a sequence.
The fact that both data from outside and feedback from a subiect's
own voice have to be processed, and that doing so takes time, is illustrated in an experiment by Broadbent $952a) in which subjeas had
to reply to a series of questions resembling military signals messages.
Occasionally a question would be asked while the subject was answering
a previous question. fn these cases errors tended to occur either in the
reply he was making or in replying to the message coming in at the time,
indicating that the decision mechanism was being overloaded by the
task of speaking and listening concurrently. Simultaneous uanslators
seem to acquire the ability to do this after long practice, but they
appear to operate under the conditions of Fig. 5(C), ignoring the
feedback from their own voices. In consequence their speaking voices
are often strange and they themselves report that they have little idea
of what they are saying or confidence that it is correct.

MEAsURING'urNrAt LoAD'
Vince's (1948a, r95o) and the present writer's (lg5g) experiments presented exlmples of tasks which were paced in the sense that the times at
which signals for action arrived were not under the subiea's control.
The load imposed on the operator varied from moment to moment and
the times when two signals came close together can be regarded as brid
periods of overload. In such a case the aoerage rate of response is not an
adequate measure of the demand made by the task: accotrnt must also
be taken of the maximum instantaneous rate required - in other words
of the way in which signals for astion are from time to time 'bunched'
together. A good example of the effects of bunching is given by Mackworth and Maclworth (lgS6) who used a task in which signals on a
number of moving belts appeared for limited times in windows, and
found that the number of signals missed was approximately linear with
the extent to which the signals presented in different windows overlapped in time. Putting the matter another wayrpacing reduces the extent
to which a subject can compensate for slowness at one instant by extra
speed at another, and thus increases the number of errors made at any
glven average rate of responding - as Bertels on et al. (1965) showed for
operating a letter-sorting rpsshine.
This type of task is analogous to many conveyor line and machinemindi"g jobs in industry. Often in these tasks the siguals for action
develop slowly and many decisions have to be made which do no issue

Single-clmnnel Operatbn

t29

in overt action. A detailed record of signals presented and of astions
taken, even if it could be obtained, would therefore grve an inadequate
picture of the load imposed on an operator. The single-channel model
suggests a way of measuring such loads: research is still in the explora-

tory stage, but two approaches appear promisi.g. Both are based
essentially on the simple assumption that dealing with the data from
each signal requiring decision takes time, and that if this is more than
the time available, responses will be delayed or omitted.
Attempts to measrue loaditg direaly

If the time / required to deal with each signal is an trnvarying quantity
and aU signals arrive at identical intervals 6 every signal can be responded

to so long as / does not exceed i. If it does, every alternate signal can
be responded to so long as / does not exceed twice i, and so on. The
relationship between response rate and signal frequency will take the
fonn shown in Fig. 4.6. Any variability in t or r will quickly lead to
I

ul

Q+,

E-l

oS
u,, E
U) lr,

=sZ
aG,
TU UJ
EO0

123

4

5

SIGNALS PRESENTED PER INTERVAL T

Figxrre 4.6. Theoredcal relationship bennreen response rate and signal
frequency when each sigpal takes a time , to deal with and has to be
responded to immediately it appears or be missed. Signals are asstrmed
to arrive at equal intervals of time.

smoothing of the 'saw tooth' pattern and its replacement by a curve
which is convex upward and aslmptotic to a rate at which the subiect's
whole time is taken up in responding and there are no gaps duing which
he is waiting for a signal. The smoothing will be quicker still if he can
not only respond to siguals immediately they arrive but can also respond
a little early or late. Such latinrde can arise either from the arrangements
of the displaywhich enable responses to be made at anytime during an
appreciable periodr or from the subiect being able to predict items or
hold them in rtrnning short-term memory, and for the present purpose
all these are equivalent.

r3o

Fund,amentals of Skill

Conrad (l96ob) has assruned that the number of responses made
under paced conditions will be the same as the number made within the
same time under unpaced conditions so that paced performance can be
predicted from the disuibution of unpaced performance times, but this
is clearly not always so. Brown (r9SZ) found that subjects in a plotting
task made more plots under paced conditions than would have been
prediaed from their unpaced performances: detailed sttrdy of the
results indicated, however, that their approach had become more hurried
and the extra speed was attained at some cost in accuracy. Conrad
himself (lgS6) in a study of telephone operators showed that the time
taken per call decreased linearly with increase in the log frequency of
calls coming in, although it is not clear how far this represented a true
speeding up and how far a reduction of ancillary activities and is therefore an illustration of Parkinson's Law. Evidence on the precise effects
of pacing on performance are surprisingly scanty.
The assumption that, even if r changes from unpaced to paced conditions, it is unaffected by degree of pacing yields reasonable results when
applied to data obtained in a series of e4periments by Conrad (r95r,
r95&rb). The subiect's task was to respond by pressing a key or ttrning
a knob each time one of a nunber of rotating pointers coincided with
one of several irregularly spaced marks on the edges of dials. The
number of signal soruces was varied by using 21 3 or 4 dials in different
trials and the signal frequency was independentty varied by changing
the speed at which the pointers rotated. An increase of either variable
led to an increasing proportion of signals failing to secure a response.
Conrad in a private comrntrnication has indicated that the rate of
responding was not limited by motor factors - subiects could respond
much faster when they made responses in a predetermined pattern
wittrout regard to the displap and although subjeas were allowed to use
both hands, the intervds benreen responses were such that they could
easily have made them with one hand alone. The limitation seems clearly
to have been in the speed at which the cenual mechanisms could deal
with the data from the signals and, perhaps, monitor the responses.
Crossman in a private communication has shown that Conrad's
results can be fairly well fitted by assuming that subieas dealt with data
at 4 bits Per sec and responded whenever there was a long enough gap
between one signal and the next in the series he used. The present
writer obtained close fits, shown in Figs , 4.T and 4.8 to Conrad's data,
assuming constant / for any grven number of pointers, random i and a
latinrde of either one or two items. The treaunent is the more plausible
in that the values of / for 21 3 and 4 dials (Fig. 4.7) grve a good fit when

Single- clwnnel O p

r ation

I3I

or
100

tr
80

.I[

UJ

F
f

=
=

U)

ur
a
z.
o
o-

(,
u.l

G

10

to'u*0.r1,9,*r* loo

tzo

Figrue 4.7. Conrad's (lgSr) data fitted assuming that sigrals arrive at
random intervals and that dealing with each takes an equal time r. It is
also assumed that if the subiect cannot deal with the signal at the instant
it arrives it can wait, but that data from not more than two signals at a
time can wait in this way - in other words the latitude (i) in responding
is two signals.
l: z dials, r : 'Jl sec.
II: 3 dials, r : 'Jp sec.
III: 4 dials, r _ '74 sec.

I
a

tr

(n
UJ

:)

=

60

=

(n

UJ

u)

z

o
ou)

t!

G,

60

80

100 r20 140 160
STGNALS/ tvrtnures

r80 200

Figrrre 4.8. I : Conrad's (tg14a) results fitted usi.g the same assumptions
: .74 sec. II: Conrad's (tgS+b) results fitted
using the sarne assumptions ercept that 1 : r. 4 dials, r : '83 sec.

as for Fig. 4.7.4 dials, r

t3z

Fundamentals of Skill

in the ratio r : r'58 : z required by Hick's Law for z, 3 and 4 choice
responses. Also group B in Fig. 4.8, who show a latitude of one instead
of trro items, was a rather poorer group of subjects than the others. The
method of calculation is described in the Appendix.
The model is obviously a gross over-simplification taking no account
of possible grouping effects or of subtleties in the results such as rhe faa
that at high signal frequencies many subjects neglected one dial, and
missed all siguals on it, for a minute or so at a time. Such a procedure
could raise the rate of responding because the loss of potential responses
to signals from the neglected dial would be more than offset by the
shortening of t as a result of reducing the number of possible responses
combined with the effective lengthening of i when signals on one of
the dials are neglected. The assumption that r and f remained constant
with signal speed would, however, then underestimate performance at
high signal speeds.

The interaction of two simultaneous tasks

It has long been known that if the attempt is made to perform two tasks
at the same time, the speed or accuracy of one or of the other or of both

is [kely to be lower than when the tasks are carried our separately.
Early e4periments in this field are those of Bornemann (tgqp) who
paired mental arithmetic with a task in which the subject had to 'dor
with a stylus through holes in a paper band passing over a drum, and of
Mowbray (r9 52, 1953, 1954) whose subiects were required to reporr on
data such as letters, digits, or prose passages presented either visually
or aurally. He found that subiects could not deal adequately with two
different streams of information, one presented to the eyes and one to
the ears, at the same time. The limitation could obviously not have been
sensory snd, since subiects did not have to report until afterwards, it was
not on the motor side. The effects seemed clearly to be the result of an
overloading of some central mechanism, snd it is reasonable to suppose
that the impairment was due to the single channel being caprured by
data from one task at a time, to the exclusion of data from the other.
Some evidence in favour of this view is Mowbray's finding that when
the two tasks were of unequal difficdry, it was the easier that tended to
suffer more. This is understandable if the more difficult task tends to
occupy the single channel for longer periods than the easy: capture of
the single channel by the more difficult task would mean the omission
of a relatively large block of data from the easier task, whereas capture
by the easier task would cause a relatively brief interruption of the more

TABLE 4.2: Some dual-task studies
Author
Bahrick et al. (rgS+)

Primary task, to which main attention

Secondary task

is normally gioen

Five-choice reaction task with either Experimenter read two numbers and

(a) repetitive or (b) random pattern of subiect said difference between them.
signals. Signals visual, responses by Two more numbers read as soon as
subject replied.
keys under fingers.

Results

Perforrnance of secondary task equal
for the two primary conditions early in
practice. Later, when repetitive pattern had been recognised, secondary
task performed better with (a) than

with (D).
Batrick & Shelly (tgS8)

Pressing appropriate one of 4 keys Pressing appropriate one of 4 keys Accuracy of atl fotrr primary tasks
under right hand in response to 4 rurder fingers of left hand in response equal when perforrred done. With
lights which appeared one at a time at to random digits from r to 5 presented secondary task added, accuracy decreased progressively from (a) to (d).
o.58 sec intervals. (a) repetitive pattern aurally.

of signals, (D) series with substantial

Secondary task performed equally well

with all four primary tasks.

sequential prediaability, (c) series with
some sequential predictability but less

than (D), (d) random series.
Broadbent (1956)

Answering messages presented aurally.
Each message separated from last by

interval of z sec.

Garvey & Taylor (t959)

Pressing a key or stamping the foot Secondary task impaired acanracy of
perfonrrance in primary, both when
when a brtzzer sounded.
brtz.z came during message and when it
came between messages. Practice reduced latter effect but not fonner.

Visual uacking with (a) aided, or (D) (i) detecting range and bearing of Normally performance at (a) better
targets on simulated radar scope, than at (D). Subjects then selected or
unaided, acceleration control.
(ii) subuacting one digit from another. trained so as to equate performances at

(a) and (A) without secondar5r task.

IThen secondary task added, perfotzrance at (c) again became better thar
ar (D).
Poulton (tgS8b)

Brown & Poulton (196r)

Responding whenever a pointer passed
a mark on the edge of either (c) z dials
or (D) 6 dials. Task similar to that of
Conrad (r95r, r954a, b).

Driving car in (a) residential and (D)
shopping areas.

Brown (1962, r965a)

Driving car.

Messages containing z-digit nunber
presented aurally every 5 sec. Subiect
said which number was the same in 3

Significandy more errors in secondary
task with 6 dran with z dials.

For subiects who were research

Errors in secondary task greater in
shopping areas than in residential.
Speed of &iving not affected by

out of ro consecrrtive messages.

workers or technicians and considered
'average' drivers, 8-diglt numbers presented aurally every 4 sec. One digit
changed between each group and next,
and subiect said which. For police subjects, regarded as 'advanced' &ivers,
summing gf,oups of 3 digits presented
atrrally and saying total, at intervals of
3 to 6 sec.

Either (i) digits heard at r.2S sec in- (i) a more sensitive indicator than (ii) of
tenrals and subiect reported whenever
sequence 'odd-even-odd', or (ii) digits
heard at 5 sec intenrals and subiect re-

ported after each group of ro which
digrt has been presented twice.

Brown (tg6Sb)

Broadbent & Heron (rg6z)

Schouten et al. (t962)

Driving car.

Listening to car radio.

Cancelling all examples of a particular As Brown (t965a) task (ii).

digit on sheets of random nu[tbers.
Pressing right and left pedals in response to 2rooo cps. and 25o cps. tones
respectivelg @) unpaced with new
signal presented after reaction to previous signal or (D) paced with signals
presented at fixed rate.

secondary task.

changes in the demands of driving.

Driving was slowed bV (ii) more hsn

bv (i).

Music slowed driving in hearry traffic
and reduced number of control movements in light traffic. No signfficant
effect of speech from radio.
Secondary task had greater effect on
performances of subjects aged 4S4o
years than of those aged t8-25 years.

(i) Puning washer and nut on bolt, Impairment of performance at primary

(ii) Inserting pin in hole with o'ro/, to or secondary task as total demand inSoo/o tolerance, (iii) Solving simple creased. Degree of impairrnent of
arittrmetical problems and writing secondary increased roughly in order

down answers, (iv) Traversing Porteus
Maze, (v) Spontaneous writing.

(i) to (v). Quality of writing deteriorated with speed of primary task:
characters became infantile and in-

forrration content diminished.
Kalsbeek (tg6S)

As Schouten et al. (b) with speeds of As Schouten et al. (v).
ro-r2o signals per min.

Step by step deterioration of q,.riting
with increased speed of primary task

resembled that seen in anoxia, fatigue,
etc. and regarded as shift to 'lower

level of organised behaviour'.
Kalsbeek (1964)

As Schouten et al. (a) and (&).

(i) removing rods from holes and placing in box, (ii) placing rods of different
sizes in corresponding holes, the rods
being aranged in order by size. (iii)
same as (ii) but rods in random order.
(i) intended to provide mooemenr task,

(i) and (ii) caused slower perfotmance,
(iii) dso tended to produce errors in
primary task.

(ii) movement plus positioning and
(iii) movement plus both positioning

and choice.

Tune (tg6+)

Predicthg, by pressing one of z keys, Listening to digits and reporting by Secondary task impaired accuracy of
which of 2 lights would come on. pressing a third key whenever r or 2 pred.iction. Interpreted as implying
breakdown of ability to retain sequence
Pattern in which they came on was heard.
of signals in short-term memory.

repetitive and subiects practised trntil

it was learnt before secondary task was
added.

Murdock (tg6Sb)

Dimond (1966)

Immediate recall in any order of list of Card sorting: (i) dealing on to single
zo words aurally presented once at rate pile, (ii) soning by colour (red/black),
(iii) soning by 4 suits.
of 6o words per min.
Pressing key in response to light appearing (a) at intervals of 6 sec or (D) at
irregular intervals of 4-ro sec.

Recall worse with (iD than with (i) and

with (iii) than \rith (ii), especidly of

words early in list.

Six-choice key-pressing task with After practice, when regularity of (a)

signals at irregular intervals averaging

I SeC.

recognised, secondary task performed
better with (o) than with (D).

Sfugle-chamel Operation

r33

difrcult task. Alternatively the subiect might try to maximise his overall
performance by concentrating on the easier task and doing it well, in
which case he would tend to do relatively better \rith the easy task.
This may have been the reason why Brown et al. (rg6S) found that a
tactile discrimination suffered more when paired with an easy visual
discrimination than with a more difrcult one - a result which is otherwise surprising in view of Mowbray's evidence.
From what has been said earlier in this chapter, it would be expected
that the main interference between two tasks would result from their
competing for the time of the uanslation mechanism. If sor interference will depend on the extent to which each task requires active
choice of responses and on the compatibility between signals and
responding actions. Results obtained by Noble et al. (tg6il and by
Trumbo et al. (1967) suggest this is the case. They paired tracking with
a task in which the subject had either (a) to try to predict a series of
numbers presented aurally at 3 sec intervals or (b) merely repeat them
or (c) generate random numbers in response to clicks at 3 sec interyals.
The first and third tasks clearly involved a more active choice of response
than the second, snd were shown to cause much more interference.
Tracking while performing the second task seemed little if at all
different from tracking with no extra task added.
A number of further sttrdies indicate the range of usefulness of what
is potentially an important means of assessing the load imposed by tasks
for which direct measurement is not possible, and of showing up
differences of loading which would otherwise be difrcult to detect.
Some of them have been briefly summarised in Table 4.2. These shrdies
indicate clearly that inctease in the load of either the primary or the
secondary task beyond a critical point can Sreatly impair performance
at one or both, and that additg a secondary task can show up differences
in the loads imposed by different primary asks which are unobservable
when they are performed alone. The effects of a secondary task on short
term memory fotrnd by Murdock (1965b) are at first strrprising since
during the period his subiects were taking in and reaining the material
no overt action was require{ and thus it might seem that there were no
translation processes with which the secondary task could interfere. V/e
shall, however, see in Chapter T that the translation mechanism is
often involved in short-term memory especially for the early items of a
list - in fact for iust those items Mtrrdock found to be most affested.
Perhaps the simplest example of the interference envisaged in these dual
tasks is the internrption of performance caused by sudden loud noises
which momentarily capnre the single channel to the exclusion of the

r34

Fundamsntals of Skill

task in progress at the time (e.g. l7oodhead, 1959, t964a., b, Sanders,

r96rb).

The dual-task technique is as yet in its infancy and a number of
problems still remain to be solved. We may mention three as especially
important.
(a) Methods of measilremenf. The primary task when given alone can be
conceived as not occupying the single-channel fullS but as leaving some
'spare capacity' into which the secondary task can, up to a point, be
fitted. Performance is impaired when this spare capacity is insufficient.
There is some question whether the primary task occupies the singlechannel continuously but not completely at any one instant, or whether
occupation is complete but intermittent so that spare capacity is in the
form of 'gaps' during which the single-channel is free. The experiments
we have surveyed earlier in this chapter seem to favour the latter view
although it is only fair to point out that if a subject is told to react as
quickly as Nssible he can be presumed to use the whole of his available
capacity in doing so. W-hether when he reacts more slowly he uses only
a proportion of his total capacity at any instant is a difficult question
which remains to be answered. It is, of course a variety of the old and
still unsolved problem of whether attention can be truly divided betrneen
two tasks, or whether it alternates rapidly berween one and the other
(S7oodworth and Schlosberg, rg14).
If the intermittency view is correct, as it clearly is in many cases, the
loads imposed by the primary and secondary tasks separately and
together can in princrple be calculated by methods similar to those tsed
to provide the basis of Figs . 4.T and 4.8 - the problem becomes essentially one of queueing by signals from the two tasks for use of the
translation mechan ism.
An empirical index of loarling in dual tasks had been proposed by
Michon (1964b, 1966, Michon and Van Doorne, ry67) on the basis of
experiments in which the secondary task has been the tapping of a foot
pedal at intervals of .5 to r.o sec which the subjea attempts to keep as
regular as possible. Michon noted that the interval between each tap
and the next seemed to be a sensitive indicator of interference by the
primary task, and suggested that tapping performance can be measured
in terms of the average difference of each interval from the one preceding. Adapting his equation we may write:

--L q:A
N-r x :t

Tappins performance __Y

(+.6)

Sfuglc-clmnwl Opration

r35

where /6 is each interval from the first to the (N
- r)th taken in nrn,
and f is the mean interval. Michon proposes that

.,.
- or?rloaorng
rnoex

?-,

-

Difference between tapping performances
with and without primary task

(+.1)
\ t-t t

Q) fhs stability of capacity.It has been tacitly assumed in most snrdies
using dual asks that the subiect's basic capacity rernains the same
whether or not a secondary task is added to a primary, and however
severe the demands of the secondary task may be. Brown's (tgSil and
Conrad's (1956) results on the effects of pacing do, however, raise the
question of whether some genuine increase of capacity occruls under
pressrue for speed. Certainly Kalsbeek and Ettema (lg6+) and Kdsbeek and Sykes (tg6il have shown that effort, as measured by the
incteased regularity of heart rate, increases with the load imposed by a
secondary task and produces an apparent increase of channel capacity,
but evidence is insufficient yet to decide whether this is a true increase
or merely represents a shift of criterion in the subiect's decision rnaking
- that is a shift of p in the signal-detection model. The problem has been
discussed by Kalsbeek and Ettema who distinguish the capacity a sub
iect is 'willing to spend'from the Eaximurn he can use in an emergency.
We shall consider the general problem of the effect of effort more fully

in Chapter 9.
(c) Ctordination of ?rintary and secandary tasks. Support for the view
that in perfonning dual tasks we normally alternate raprdly from one to
the other rather than expend parts of our total capacity simultaneously
on both comes from the faa that it seems impossible to carry out
actions, by say the two hands, which are tnrly simultaneous without
their being in some way co-ordinated and fitted into a common temporal
framework. For example, in the well-known children's party trick of
making tapping movements with one hand and circular motions with
the other, the ntrmber of taps seerns to be ineviably a multiple of the
number of circles. Such so-esdination is a type of grouping and seems
to enable a more elaborate performance to be carried out than would
otherwise be possible. For example, Kalsbeek (lg6+) noted that the
subieas in his double-task experiments tended to build up a rhythmic
pattern of perfonnance in which the two asks were regularly interdigitated, and that when this was achieved, the impairment produced
by combining the two asks was reduced, implyrng prestrmably that they
were no longer separate but had been combined into one more complex

ry6

Fundammtak of Skill

task. Further evidence that when the signals in one or both asks come
at regular intervals or are otherwise predictable, responses in the two
tasks tend to be co-ordinated and performance improves has been
provided for tracking by Adams and Chambers $962) and for serial
reaction times by Dimond (1966). Results obtained by Kalsbeek and
Sykes (tg6il suggest that co-ordination is fostered by uai"ing, but that
the scope for it is gteatest when the total loading is moderate and that
it may break down when the loading becomes very severe. Such coordination obviously complicates measurement of the load imposed by
the primary task in terms of the impairment of performance at the
secondary, although it remains possible to make such measurement in

principle.

V
Movement

Enough has been said in preceding chapters to indicate that movemen6
involve a complex co-ordination of various muscles brought into play
in a phased sequence, and thatthe exeantionofmovements is in importanlt
ways distinct from decisions to initiale them. Further evidence for such
a distinction is provided by the faa that there is commonly little or
no
correlation between reaction-time and speed of movement1".g. Henry,
r952t t96t, Henry and Rodgers, t96o, pierson, 196r, L. smith, 196;)
and factols which affest the one may hardly at all affect the other (..I.
singleton, t9s4, 1955, weiss, 1965, Brichcin, 1966, KimotsuH,
ry6;).
The psychological sttrdy of movement has been somewhat of a
cinderella, partly no doubt because it is less opeo to introspective
examination than, for example, perception; but largely because it is
amenable to much more clear-cut physiological shrdy than are other
processes traditionally ueated by e4perimental psychology. We shall

not attempt to outline physiologcal snrdies here as excellent summaries

are available elsewhere, notably those by Ruch (rgSl) and pailtard
(196o). Nor shall we be concerned \dth the anatomical factors which
determine the maximum mechanical forces that can be exerted in
various positions: these have again been reviewed elsewhere (e.g.
Darcus, 1954, Morgan et al,, 1963, Murrell, 1965). Instead we rtr"tt
consider the seflscrri-motor cuttrol of movement. Anatomical and physiological fastors undoubtedly play a part in such control, setting li-itr to
its operation and showing, for example, in the difference of speed and
accuracy of movements made $rith different limbs or in various limbpositions or direaions relative to the body (e.g. Searle and Taylor, 1948,
Begbie , r9s9, Hammerton and Tickner, 1966), but they seem to be of
secondary importance to the subtle interplay that takes place benreen

action and sensory feedback.

r37

r38

Fundammtals of Skill

SENSORY CONTROL

The control of movement follows a typical servo pattern in which action
is monitored and modified by sensory feedback of various kinds, the
relative importance of which vary with the precise tasks concerned. The
feedback may be visual via sight of the responding limb or of the obiea
manipulated: accuracy obviously suffers when movements are made
blind and is courmonly affected if visual feedback is in any way distorted. Our present knowledge of the effects of such distortion is largely
due to the work of K. Lf. and W'. M. Smith and their associates using
a technique in which the subiect manipulates obiects without being able
to see his hand directly: instead an image of the hand is picked up by a
television camera and shown to the subiect on a closed-circuit television
screen with various distortions or transpositions introduced into the
link between camera and screen (e.g. Smith and Smith, t962, K. Smith,
1966). Experiments by Gould (rg6S) and by Gould and Schaffer (rg65),
using a similar technique and manipulating the contrast to show up
particular parts of the field, have suggested that sight of the obiect
teing manipulated or of the tool being used is more important than
sight of the hand itself.
Evidence regarding the quantitative effects of visual feedback comes
from an e4periment by Chas e et al. (tg65) whose subiects attemPted to
hold a finger steady and were shown the finger-tremor rnagnified on an
oscilloscope. The extent of the low frequency components of the uemor
decreased markedly as the magnification increased from r to to times,
and thereafter remained steady when magnification was incteased from
ro to 4o times. The same authors found analogous results when visual
indication was replaced by a tone the frequency of which varied with
displacement of the finger. These findings are similar to those of a
number of uacking sfirdies summarised by Poulton (1966), in which
increase in the ratio of change in display to extent of movement by the
corresponding control improves perforrnance up to a point but not
thereafter. Seidenstein et al. (196o) have emphasised that the relevant
variable in calculating this ratio is the retinal image of the tracking error,
so that absolute rnagninrde and viewing distance are compensatory if
angtrlar displacement is held constant.

Other senses can play an important part in such feedback. For
example, Cha se et al. (r96rb) who set their subiects to taP a key with
constant amplinrde and pressure at a steady rate of three movements

per sec found that inability to see the finger produced no impairment of
Regularity of tapping decreased, however, f the soznd of

i.rfor-*.e.

Mwement

r39

tapping was masked by noise played through earphones, or rf pr*
pioceptiae signals were masked by vibrators placed on the wrist and
forearmr or rf touch was abolished by anaesthetising the finger. Provins
(r95lr r9i8) had previously observed that accuracy in applying a pressure was impaired by anaesthetising the skin over the part of the hand
used to apply the pressure, although the maximum rate of apping was

little affected. It is more diffic'ult to remov e kinacsttutic feedback, but
Laszlo (1966), using an infated cuff on the arm to produce nerye
blocking has shown that accuacy of tapping diminishes during the
period after cutaneous sensation has ceased and before muscrrlar power
has been lost, implying that accuracy is affected by loss of kinaesthetic
feedback. It is, of course, well known that patients suffering from t&es
dorsalis, which is marked by intermption of the kinaesthetic pathways,
show exaggerated movements which make adjusments ro the limbs
inaccuate and slow.

The origin of many sttrdies in this area has been the problem of
whether, when power assisted controls are fitted to cars, aircraft or
other machinery, there is any advantage to be gained by leaving some
'feel' in steering wheels and joysticks. Such feel will obviously increase
proprioceptive signals, snd it is therefore not surprising that the accuracy
of positioning a lever is improved by spring loading so that the fmce r*
quired increases with the extent of movement (Howland and Noble,
1953, Bahrick et al., r955a). There are indications that moderate pressures feld optimum results: ]enkins Gg+il obtained constant \ffeber

Fractions (6F/F) for incteases of force over pressures rangng from
ro to 5o lb but higher fractions below ro lb while pressures above 5o lb
would inuoduce complications owing to the amount of muscrrlar
effort required. Some improvement in the accruacy of simple movements has also been obtained by addi"g inertia to the conuol (Searle
and Taylor, 1948, Howland and Noble, 1953), and accuracy of both
simple and more complex movements has been improved by making
the moving conuol lever pult a pltrnger through a bath of oil thus
adding viscuts friction (Howland and Noble, rg13rBatuick et al.r r955b).
The suggestion has been made on the basis of such snrdies that
pressure cmtrols in which a control lever does not move at all but has
an effect according to how hard it is pushed, might be preferable to

moving controls and the results of several e4periments indicate that this
is so (e.9. Gibbs, 1954, North and Lomnicki, 196r, Burke and Gibbs,
1965). Other investigators have, however, found position better than
force as a basis of accurate control (e.g. ![eiss, rg10: presumably
which is better depends on the degree of force and the extent of

r4o

Fundamentals of Shill

movement or change of position. Optimum performance seems, indeed,
to be obtained with a combination of appreciable movement with springloading thus enabling kinaesthetic as well as other proprioceptive
signals to play a part. Briggs et al. (tgSil suggest that their results
bearing on this point, and those of Jenkins (1947) arc well fitted by an
equation proposed by Bahrick et al. (t955a):

rved:
Accuracy in terms of d.istance m(

dF

fr)

X a constant (5.1)

where F is the force required to produce a specified displacement of the
control and dF is the change of force associated with a change of displacement dD.

TIME REQUIRED FOR FEEDBACK TO BE EFFECTIVE
\ilToodworth (l8gg) in a classical monograph on voluntary movement,
divided the time taken by a movement into a period of initial impulse'
and a subsequent period of 'current control'. The former is a consequence of the fact that it takes time for sense data fed back from a
movement to become effective in modi$ing its course, partly because of
an inevitable reaction time and partly because of the delays discrrssed in
the previous chapter. Brief movements are thus ballistic in the sense
that they are initiated as a whole and have to nrn their course without
the possibility of modification. The most striking illusuation of this
known to the writer arose from a Siamese cat belonging to Mr K. F. H.
Murrell. It had the habit of sitting on the shoulder of anyone eating a
meal and of pawing at his hand as he raised a forkful of food to the
mouth. The result was that the food was deflected into the cat's mouth,
and the diner was quite incapable of preventing this happening unless
he moved his aun very slowly and deliberately. Crossman and Goodeve
(rg6l) have obtained the same type of result e4perimentally by having
subjeas make movements of a rotating knob to bring a pointer to a
target and adding aiding or opposing forces during the coruse of the
movement. The effects were to cause overshoots and undershoots respectively. After a little practice, subiects attempted to minimise these
by tensing the muscles being used - they were presumably increasing
the forces applied to both agonists and antagonists so that the disturbing forces would have proportionately smaller effects.
I[oodworth found that the accuracy of right-hand movements lasting
Iess than about .4 sec, and of left-hand lasting less than about .75 sec,
differed little whether or not the subiea closed his eyes while the hand

Mwement

r4t

was in motion, implylng that there was no ctrrent visual control.
Accrracy of movements lasting longer than these times improved as
duration increased with the eyes open, but did not do so when they
were closed. This frnding was confirmed by Vince (l94Sb) whose results
are shown in Fig.5.r. Analogous results were obtained by Provins

5to
E,

rl,

rU

o
F

z.

LIJ
()tr
oa \,

lJ

o-

0

0.4 0.6 0-8 1.0 1.2

TIME PER MOVEMENT (SEC.)

1.4

Figrrre 5.r. Relations between accuracy and duration of movements with
eyes open and with eyes closed. Data from an experiment by Vince
(r948b).

O : Eyes open.
O : Eyes closed as soon as movement began.
Each point is the mean of a series of readings taken from each of ten subiects. The movements were made by pu[ing a cord attached to a pointer
which moved downwards over the surface of a smoked drurn to a target
line r in away. The pointer was returned to its starting position by spriog
tension. Movements were made in time with a metronome.

(tgSil who fotrnd that the accurasy of pressrue with the side of the
hand was greater when it was exerted at the subiect's own preferred
speed - usually over a period of 23 sec - than when it was exerted as
fast as possible, but that the difference disappeared when the area of
skin at the locus of the pressure was anaesthetised.
Both !floodworth and Vince showed that in the course of rapid
movements there is a period of acceleration followed by an approdmately similar period of deceleration, md that these normally shade into

r42

Fundamentals of Skill

one another so as to produce a cenual zone of nearly trniform velocity.
If distance travelled is plotted against time the cluve seems to be
roughly a normal ogtve (Crossman and Goodeve, 1963). Peters and
Wenborn (1936) fotrnd that such patterns were remarkably constant for

different directions and extents of movements, snd that when movements made as fast as possible were compiued with those made at a
moderate speed, all phases changed to about the same proportional
extent. Woodworth's and Vince's slower movements showed the same
TYPI CAL RECORDS
MAXINIUM POSSIBLE

RATE

I20 STROKES PER MINUTE

ry{ \Il ffiil

A

60 STROKES PER MINUTE

\ t \

UT

20 STROKES PER MINUTE

I

\

I NG RETURN

MOVEMENT= 0'1 SEC

01234
SECONDS

EXPERIMET{T 3

B

0 0.2 0'4 0.6

0

SECONDS

Figrrre S.z. Examples of movements made in the experiment the results
of which are shown in Fig. 5.r.

initial period of rapid acceleration, but this was followed by a much
longer and less regular deceleration, suggesting that under conditions of
'current control' the movement was no longer a single entity but a
series of movements which ran into one another. Some of Vince's
results are shown in Fig. 5.2.
Looking at the matter from a slightly different standpoint, the
S-shaped roughly symmetrical curve with approximately equal times
taken over acceleration and deceleration, seems to be associated with
movements intended to be of a given extent, whereas the type showing

a prolonged final deceleration with one or more adiustments tends to

Mwement

43

occnr when the movement is aimed precisely at a target: the finrl
deceleration and adiustments represent a process of 'homi.g' on the
target rather than merely bringing the movement to a halt. Most of the
time taken by such accurate movements is thus spent over the last part
of the travel (Annett et al., 1958).
Fig. 5.2 illustrates the further point that within a series of rapid
movements, each too fast to incorporate any secondary adiustments,
there are nevertheless trends over groups, suggesting that adiustments
are made between one movement or group of movements and the next.

If so, it means that two servo-loops are operating. One is a short-terrr
loop which has to do with the immediate phasing of each individual
movement and if one can iudge from the fact that movements made at
maximum rate averaged about five per secr achieves about ten adiustments per sec - one to make each stroke and one to release it. Superimposed on this is a slower-aaing loop operating over periods of .5 sec

or so. The latter is presumably a visual loop, while the former is
kinaesthetic or proprioceptive and local to the effector mechanism - in
other words corrections made in terms of it do not require any new
decision. Similar implications follow from the obserrration that when
tracki4g a simple harmonic course at high speed subiects may make
rapid harmonic motions which rrer however, not 'tim+locked' to the
excursions of the target (Noble et al.r rgjj). Further evidence for such
a short-term loop comes from the rapid corrections for errors in some
tracking tasks which we noted in the previous chapter (p. r t6) and which
are far too fast to be visually controlled. We may also rote in this

connection that Mnrrell and Entrrisle (lg6o) and Crossman and
Goodeve (tg6l) observed a periodicity of acceleration and deceleration
of about ten per sec in detailed recordingp of movements aimed at a
target. It looks, in short, as if the translation mechanism feeds an
'order' into the effector mechanism, and that the latter carries out the
order by means of a servo mechanism with a loop which takes about
'r sec to traverse and which continues to act until sufficient accuracy
has been achieved.
Further evidence for the importance of time factors in the control of

movement comes from a consideration of how movements made at
murimun speed differ according to their extent and loading. Increasing
the amplinrde of a movement while keeping its tenninal acctrasy the
same makes relatively little difference to the time taken (Searle and
Taylor, t948), and if the ratio benreen amplinrde and terminal accrracy
is held constant, the times taken by movements of extents from about
2 to zo in are dmost identical (Fitts, 1954, Crossman t r957r Fitts and

r44

Fundammtals of Skill

Peterson, 1964, Crossrnan and Goodeve, $63). Similarly Fitts (lgS+)
forurd that 'dotting' witr a stylus back and forth between two targets

took about the same time whether the stylus weighed r oz or r lb.
Again El Temamy (1966), using a weight which slid back and forth on
a rod, found that the time taken for any grven degree of accuracy rose
by only about 30% as the weight was increased from rz1 to r2 lb. It
can be argued that the rise is likely to be small because the arm itself
is a large element in the total mass to be moved, but close examinations
by Taylor and Birmingham (lg+8) of uacking performance and by El
Temamy of his results made it clear that the force deployed rose considerably with heavier weights and larger amplinrdes. Had the same
forces been used with lighter weights or shorter amplinrdes the movements would have been considerably faster. ft seems, in etrea, that
with incteased amplitude or load the subjea incteases the force applied
while keeping the time roughly constant. This, of course, is what would
be expected if the limiting factor was the time required to process the
data on which control was based. Additional support for this general
view comes from the well-recognised fact that complex movements
involving a series of different directions or extents, and thus implying
more elaborate control, take substantially longer than simple linear
movements (Woodworth, 1899, Peters and Wenborn, 1936); also from
the finding that speed of arm movement has little to do with the strength
or effeaive rnass of the aun (Pierson, 196r).

Time has an important bearing upon the optimum sensitivity of
controls. Severd snrdies have shown that when a pointer has to be
moved to coincide with a target, performance varies with the extent of
the pointer movement produced by a given control movement, being
most rapid and accurate when the ratio between the two movements is
at some intermediate value between very high and very low (]enkins
and Connor, rg4grJenkins and olson, t952, Gibbs, rg14rNorth et al,,
1958, Hammerton, t96z).If small movements deflea the pointer a long
way it is difficult to make fine adjustments because the pointer will
move too far during the minimum time within which a movement can
be executed and corrected. On the other hand, if large control movements have to be made in order to produce a relatively small deflection
of the pointer, acqracy in terms of time and distance off target will
suffer because ofthe time taken to execute the movements required. This
latter is especially the case when the control is not a simple lever but,
say, a handwheel as used for steering a car. If a steering wheel has to
be ttrrned a relatively long way to produce a grven change in the car's
direction, it may be difficult to corner at speed. If, on the other hand

Mwentent

145

the steering is trnduly sensitive, it is diffictrlt to keep straight on an
open road.

RELATIONSHIPS BETWEEN SPEED AND ACCURACY

The relation benn een amplinrde, accuracy and the time taken to make
hand movements had been a subiect of disctrssion for some years before
Fitts (rgS+) suggested a formulation in information-theory terms which
connected all three together. Fitts proposed essentially that

Movementtime_@+blog(zA/V)

(5.2)

where lY is the width of the target within which the movement is
required to end, measrued pardlel to the direction of the movement; r{
is the amplinrde of the movement meastued from its starting point to
the centre of the target; and a and D are constants. The essentid point
of this formulation is that it makes movement time constant for any
given ratio betrreen amplinrde and target width.
Fitts recognised that the multiplication of A by z was arbitrary
although some such procedure appeared necessary in order to ensure
that the logarithm was always a positive quantity: had the fraction A/V

been used, 7 would have exceeded A in the case of movements made
from iust outside a target and this would have made log, (A/V) negative. There was also some conceptual plausability in that the subiect
could be thought of as having to 'choose' a movement which had the

possibility of either over- or under-shooting the target. Fitts also recognised that it would not always be accruate to take V as the measure of
the scatter of the shots: the subiea might, for example, concenuate his
shots in a narrower width, in which case the movement time would be
expected to be greater than that predicted by Eq. 5.2.
Fitts backed his formulation wittr four sets of experimental data. In
one experiment the subiect was required to 'dot' alternately on two
metal strips 6 in long using a metd-tipped stylus weighing r oz and
about the size of a pencil. The long axes of the strips were perpendictrlar

to the line of movement betneen them. The suips were mounted on a
board placed in such a position that the subiect's movements were from
side to side in front of him. Four widths of target strip were used:
2, rr.5 and .zS itat each of fotu distances benreen centres i 2t 4t 8, and
16 in.

Average tiu s per movement, plotted accorditg to Eq. 5.2 with A

taken at the distance between target centres, tnd V as the width of the
target strips, are shown in Fig. 5.3. All the points, except those at the

146

Fundamuttals of Skill

extreme lower end, lie close to a straight line and suggest that Fitts's
formulation was right in principle. Fig. 5.3 does, however, have three
unsatisfactory features which suggest that some modifications of detail
are required:

(o) The suaight line nrnning through most of the points cuts the
zeto information line below the origln making a in Eq.

S.z a negative
+

0.6

c,

0.5

rl,

a

z 0.4
F

UJ

rr,

o
=

G,

UJ

o-

,

UJ

=

o

x

F

0.1

3

45

7

LoG2 (ffi)

Figure 5.3. Times for reciprocal tapping with a r oz stylus plotted in
terms of Eq. 5.2. Data from an experiment by Fitts (lgj+). fne target
widths (w) areindicated asfollows x - 2 h, o : r i"; b : bin ana
+ : t in. The four amplitudes of (A) for each target width are, from
left to right, 2, 4, 8 and t6 in. Each point is based on a total of 6ry-2r669
movements obtained from 16 subiects.

quantity. Crossman GgSil has suggested that this difficulty can be
avoided by omitting to multiply A by 2 so that Eq. 5.2 becomes
Movement time

- a + blog(A/V)

(S.t)
The constant a works out at about .o5 sec, a figure close to that found
by Crossman himself in a similar experiment for the time spent on the
targets as opposed to the time spent moving beniveen them. ft seems
obvious at first sight that the time spent stationary on the tiugets
ought not to be included in the measure of movement time. It has been
found, however, in another similar experiment (Welford, 1958, pp. 99
and ro3) that results are, on the whole, more unifonn and plausible if
the times on target and benn een targets are added together than if the

t47
Mwetnent
latter measure only is taken. It seems as if there may be some corlpensatory tendency at work which makes it possible for time spent on
the targets to be used to shorten time spent moving at least to some
extent. Such a compensatory effect would be consistent with the view
that movement time is limited by the speed at which central processes
can control and monitor movement. These processes must obviotrsly to
some extent precede and outlast the movements concerned, and the time
thus required would be spent on the targets.
(b) Although the points in Fig. 53, except those at the extreme lower
end, lie fairly close to a straight line, the best fitting line through them
would curve gently upwards. The curve can be substantially lessened
by making a further modification to the equations by writing
Movement time

- K "r(ry)-

K

,rr(#* 'r) G.+)

This formulation makes movement time dependent upon a kind of
Weber Fraction in that the subiect is called upon to distinguish berween
the distances to the far and the near edges of the target. To put it
another way, he is called upon to choose a distance V oat of a tota
distance extending from his starting point to the far edge of the target.

The formulation also preserves the advantage which Fitts claimed for
the procedtue of multiplyrng A by 2, h that the logarithm can never
be negative, since in the extreme case when the movement begins at
the edge of the target A - lW.
G) The curve in Fig. 5.3 shows a distinct flattening at the lower end.
This is probably due, as Crossman has suggested, to some timiting
factor setting a minimunr time per movement however short or uncotrstrained. Snrdy by the present author of dotting benreen targets trsing
a pencil as a stylus suggests that this limiting factor affects the amoturt
of target used. When the targets are wide and the distance short, the
subiect uses very much less than the full target width. He is, in fact,
uansmitting more information than a calculation in tenns of Eq. 5.4
would assume because the effective lV is narrower. The narrowing of V
is to some extent reflected in a reduction of errors and if due allowance
is made for them Eq. 5.4 still holds reasonably well.
The method of correaing for errors has been dCIcribed by Crossman
(tg1il.It makes use of the faa that the information in a normal distribution is logro t@1, where o is the standard deviation of the distribution. Now \@) - 4.133 and a range of * half this, i.e. 2'06zo,
includes about 96% of a normal distribution. W'e can therefore argue
that if about 4% of shots fall outside the target, log, V is an accurate

r48

Fmdanwntals of Skill

rqpresentation of the information contained in the distribution of shots.
IUe can also argue that if the errors exceed 4% the ffictioe rarger width
is greater than Wrandif the errors are less than 4%the effective target
width is less than W. How much greater or less can be calctrlated from
tables of the normal distribution (e.g. Fisher and Yates, 1938, Table r).
For example, suppose W 2 in and the errors are f/o. Then the

-

effectivelV - 2 x 4'ry3/J'rJ2
- t.6o4in, since allbut to/oofanormal
distribution lie within a range of + 2.5T6 (i.e. * x S-rS2) of the mean.
It has been assumed so far that errors are distributed equally to both
0.8

0.7

(J

tlJ

a

F 0.4
z.
r!
t!
=

+

o
G,
rrJ

L

]rJ

0.2

=
0.1

0

23
LoG2 (

t# ) connecTED FoR ERRoRs

Figrrre 5.4. The same data as in Fig. 5.3 plotted in terms of Eq. 5.4
corrected for errors by Crossman's method. Target widths and amplitudes of movement are indicated as before.
sides of the target and that the mean is in the centre of the target so
that the mean amplitude of movement is correct. ff the mean falls to
one side or the other of the centre, the effective ^r{ will be different
from the distance A between target centres. Strialy speaking the
effective z{ should in any case be the geometric mean of the distances
between each shot and the next, although it is usually approximately
correct to take the distance betvveen the means of the two distributions
of shots. If the distributions can be assumed to be symmetrical it is not
necessary to record the positions of all the shots in order to calculate
the positions of the means: the mean of each disuibution can be inferred

Movemutt

t4g

from the proportions of shots falling on the near and far sides of the
target concerned.

Fitts's data are plotted in Fig. 5.4 using Eq. 5.4 and making appropriate adjustments to lV for errors. The data were unfortunately not
given which would have made it possible to calculate the effective As,
and these have been assumed to be correct. The results lie close to a
straight line which passes through the origin.
Further data on movement times
Crossman (tgSil has produced results which although obtained independently of Fitts's, are in striking agreement with them. He has also

shown that the length of the target strip as well as its width affea
movement time although to a smaller extent. The reason for this becomes clear when one studies records of dotting with a pencil betrn een
targets drawn on paper. The shots in each target fall roughly into an
ellipse with its long axis parallel to the line of movement: the scaner
perpendicular to the line of movement is very much smaller than that
parallel to it.
Closely similar results to those of Figs. 5.3 and 5.4 were obtained by
Fitts for a similar 'dotting' task but using a stylus weighing r lb instead
of r oz, although the slope was slightly steeper \ilith the heavier srylus.
Ifuight and Dagnall (1967) using a task in which the subiect aligned a
pointer \rith each of two targets alternately, obtained results which
closely followed Eq. 5.4 and were better fitted by this than by Eq. 5.2.
The slopes were a little steeper than in Fig. 5.4, but were of the same
order of magninrde. Addi"g inertia to the movement increased the
slope signifigantly although, as in Fitts's case, the increase was not
substantial.
Fitts and Peterson Gg6+) using amplinrdes of 3, 6 and t2 in and
target widths of l, '5, '25 and .rz1 in, found that Eq. 5.4 fitted well and
slightly better than Eq. 5.2 for single movements each carried out following a separate signal to move. The slope was somewhat less steep
than for continuous movements, perhaps because there were no tururound times at end points to be included. The time saved was understandably less than the reaction-time between the appearance of the
signal to move and the acnral beginning of the movement, since this
contained an element of temporal uncertainty absent from the continuous task. Fitts and Radford (1966) in a similar experiment showed
that although accuracy improved as the time taken to exec-ute a movement lengthened, it was little affected by extra time taken to preptrer

rjo

Fundamentals of Skill

and that rate of gain of information varied little whether instnrctions
were for speed or for accuracy.
Some indication of the detailed actions lying behind the logarithmic
relationship bern een speed, amplinrde and accuracy is given by Crossman and Goodeve (tg6f) who plotted the course of movements made in
continuous 'dotting' tasks like those of Fitts (rg5+), and with discrete
settings of a pointer made by rotating the wrist. They found, as we have
already noted, that the movements were not smooth but showed accelera*ion and deceleration with a periodicity of about ten cycles per sec. The
pattern was such as to suggest that each movement consisted of a series
of impulses which ran together to some extent but not completely. All
took about the same time, but they differed in velocity so that the first
covered approximately the first half of the distance to be covered, the
second the next quarter, the third the next eighth and so on. When
distance travelled was plotted against time taken, each impulse roughly
resembled a normal ogive like that of a movement made to cover a
given extent rather than aimed at a target. Crossman and Goodeve
point out that this implies that movement is controlled by a servosystem which initiates impulses having a velocity proportional to the
distance to be covered and which are corrected by feedback at intervals
of approximately 'r sec. They also prove that ifthe velocities are adiusted
in terms of the distance to go to the centre of the target aimed at, and
the subiect institutes a stopping procedure at the near edge of the
target, this system leads to Eq. 5.2. Ve may add that if instead velocities
are adiusted in terms of the distance to the far edge of the target and
stopping is again initiated at the near edge, the system leads to Eq. 5.4.
The authors suggest that this servo is within the motor system (effector mechanism) and largely independent of visual control since the
same pattern of successive impulses was obtained with the rotating
pointer task when a, sector-shaped screen was srranged to hide the
pointer from view as soon as it began to move, and the subiect was

required to rotate the knob so as to bring the pointer to rest as near as
possible to the far edge of the screen. Some visual control might well
have been involved, however, in checking whether or not the pointer
had emerged from behind the screen and could perhaps have affeaed
the adiustments made. In support of this suggestion we may note that
Crossman and Goodeve in a further pilot experiment found the number
of impulses to be fewer when the pointer disappeared from view halfway to the target than when it was hidden over the whole distance.
We have spoken of the subiect 'choosing' a target width out of a
longer extent, but this is obviously only a figurative description; some

Mooutunt

rir

more detailed process needs to be specified. Fitts (1954) and Fitts and
Peterson (1964) suggested that the subiect could be conceived as trying
to make a series of movements each of uniform extent but subiect to
random disttubances which may either add to or subtrast from the
extent aimed at. In criticism of this view it seems fair to reiterate two
points we have already noted. Firstly, the subiect's task was not, in the
experiments we have been discussingr to make movements of a given
extmt but ones which ended within specified limits. SecondlS the
actions when 'homing' upon a target are more complex than those involved in making a movement of a particular amplinrde or dtration.
Taking these two points together, it is plausible to think of the subiect
as having to consider all movements of an extent short of the far edge
of the target and to reiect all those which are short of the near edge.
He starts by reiecting a large class of those which are much too short
and then reiects finer and finer classes as the target is approached. His
behaviour would in a sense be analogous to that envisaged in the first
of the models discussed in relation to choice reactions (p. Tr), and i
each sub-decision took an approximately equal time it is trnderstandable that the first hafi the next quarter and so on of the total distance
would be traverrsed in approximately equal times as Crossman and
Goodeve found.
Ftrrther evidence on 'doning'times obtained by Welford et al. (1961)
indicate that these views are somewhat too simple, snd that Eqs. 5.2,
5.3 and 5.4, although on the right lines, need some elaboration. Before
discussing this work, however, we need briefly to review experiments in
which speed of movement has been snrdied in relation to accuracy,
using a different gpe of task.

Inserting pins into sockets

Fitts (rgS+) reported two further experiments. In one the subiect was
required to transfer discs with holes in their centres from a vertical
pin to another similar pin a given distance to the left. Four distances
betrneen pin centres were used with each of fotu sizes of hole. The
'target widths' were taken as the differences benreen the diameters of
the pins and of the holes. The other task was the transfer of metal pins
from one set of holes to another a given distance away. Four diameters
of pins were used with each of five distances. The holes were in each
qlse trnrice the diameter of the pins: the 'target width' was again taken
as the tolerance between the pins and the holes.
The results of these last two experiments trer unfornrnatelS not

t1z

Fmdammtals of Skill

suitable for detailed examination as the times taken to transport the
discs or pins to their 'targets' were not measrued separately from the

return ('transport empty') movements of the hand to pick up the next
disc or pin. We cannot assume these return movements to be of constant durationr or of a dtrration proportional only to distance. Evidence
that they may be comes from results by Annett et al. (lgS8), but Crossnum GgSil found that return movements tended to become slower
as the accruacy of the outward movements rose. There is considerable
support for the view that the speed of one movement in a cycle tends to
affect those of others (de Montpellier, 1935, Wehrkamp and Smith,
t952, Welford, 1958, p. ro5, Simon and Simon, 1959).
This diffiailty does not attach, however, to the pin-transfer experiment by Annett et al. (rg58). Their subiects, like Fitts's, were required
to transfer pins from holders to sockets. The pins were + in diameter
and different sets of sockets ranged in dianreter from fi through +?
and t* to t? rn, gving tolerances of #, *., * and r in respectively. The
distance betrreen centres of holders and sockets was in atl cases 8 in.
The holders were & in diameter. Movements were recorded by a film
taken at 48 frames per second and from this it was possible to determine
the intervd elapsing between the pin leaving the holder and its release
after having been placed in the socket ('transport loaded' plus 'position').
'Target width' was taken as the difference of diameter benreen pin and
socket.

The average times taken by the three subjects plotted in accordance
with Eq. 5.4 lay close to a straight line, but the times for the finer
tolerances tended to be too short. The line would have been straight
and have passed through the origin if the pins had been # in diameter

instead of t rn, and this fact led the present writer to consider whether
there were any reasons why the effectizte pin diameter might have been
a little less dran + in. Two factors appeared possible: firstly that the
ends of the pins were slightly rounded, and secondly that the subjeas
were apptying them to the holes at an angle so as to present a rotrnded
edge which would 'find' the hole more easily than would the flat end
of the pin. Enquiry from one of the authors revealed that both factors
were present: the ends of the pins were slightly chamfered since if this
was not done 'the subiects often ftrmbled when putting them into the
smallest holes'; and the film clearly showed ttre pins being applied to
the holes at an angle.
Fnrttrer work by Schouten et al. (see Westhoff, 1964) on placing pins
in sockets has shown that formulation in terms of tolerance is not quite
adequate: some account must be taken of absolute ph, or socket, dia-

Mwemilrt

r53

meter. Their subieas moved pins as fast as possible over a distance of
2J mm into a hole of r, 2, 4t 8 or 16 mm diameter. Different pins were
used to provide, for each hole, percentage tolerances of 5or25, r2'5 and
so on for a total of between eight steps with the smallest hole and eleven
1.2

I

2

4

1.0

o
9

-

o

(J
lrl

I
A

l6

vt

o

v

F-

zlJ.l

o

a

trl 0.5
=

o

=
G,
lrl

o- 0.4

l!

=

F

0
0

2

LOG2

A

4

6

I

t0

12

w

Figrrre 5.5. Times taken to place pins in holes. Data from experiments by
Schouten et ol. (see Vesthoff, tg64).
A - 25 mm in all cases. W' : the difference between the diameters of
pin and hole. The diameters of hole used are indicated on the right in
millimeues against the corresponding regression lines.

with the largest. Their results are shown in Fig. j.5. If tolerance hsd
been an adequate measure of V in Eqs. 5.2r 5.3 or 5.4 the regrcssion
lines for different diameters of hole should have been superimposed.
We shall consider reasons why they are not in relation to the results
obtained by Welford et al. (1963).
Deailed snrdy of accurary
The writer and his colleagues were concerned to overcome the difficulty
in Fitts's original experiment and in others since, that times are scored
for the task of landing within target limits rather than for recorded distributions of shots. Times were therefore taken for dotting from side to

rS4

Fundarumtals of Skill

side with a pencil betrnreen two targets drawn on paper so that the point
at which each shot landed was recorded. Continuous runs were made of
5o shots in each direaion with all combinations of (A + +W) - Jo,
r4z and 4oz nrm and W - 32, rr and 4 rnm. The range of values was
resuicted to these because the task was part of a battery of tests being
given to a large group of subjeas and the time available was limited.
The disuibutions of shots appeared to be rather too broad in the middle
for norrnality: they seemed to fall berween normal and reaangular. The
width of the distribution for each subject on each urget was therefore
taken as the distance betrreen the extreme shots, excluding occasional

wild deviants. Group distributions as opposed to those of individual
subjects would, of course, have been much nearer normal: so that this
procedure is not in conflict with that adopted in correcting for errors in

Fig. s.4.

ft was argued that, by analogy with Eq. 5.4rthe present results ought

to be fitted by the eqtution:
Movement time

- K ^r(#,* .r)

G.s)

where V' is the mean width of the two distributions, one at each end,
observed with any partictrlar combination of A and W and A' is the
distance between the centres of these disuibutions. The experiment
was designed to study age differences: the results for the forties, the
decade \pith the largest ntrmber of subieas, are shown plotted against
Eq. 5;.5 in Fig. 5l.6. It can be seen at once that, while the times for each
target width increased with amplinrde at rates similar to those found
by Fitts, *re times for the narrower targets were too high, as they were
with the results of Schouten et al. The results for each decade from the
twenties to the eighties showed the same pattern.
Two possible partial reasons for the difference between these and
Fins's results Brer fostly, that the meal targets he used may have tended
to emphasise acsuracy more than did ours drawn on paper: our subiects
tended to scatter their shots well outside the narrow targets and not to
use the full \ddth of the wide ones, whereas Fitts's subjects seemed to
have preserved acctracy relatively well on the narrow targets. Secondlp
and as against this, the point of the stylus used by Fitts must have had
an appreciable diameter and the electrical insulation between the targets
and the srurounding meal plate must have had an appreciable thickness,
so that all his targets might have been effectively wider by a small
constant amount than he claimed them to be. This extra would be of
little importance with the widest targets but would be a substantial
proportion of the narrowest.

Mwenunt

rjj

These reasons did not, however, appear to be sufficient to explain the
poor fit of Eq. 5.5 to oru results, &nd accordingly three other possible
e4planations were considered. The first of these was that movement
times depend more on aim than on accurary achieved, and thus more

+

a

0.5

tiIU
tr)

+

0.4

z
lr,
=

rU

o

o
=

E,

UJ

oUJ

= 0.1

F

0

0

I

^,2

LoG2 (f,,*o.s;

3

4

5

6

7

Figrrre 5.6. Times for reciprocal tapping benreen two targets plotted in
terlns of Eq. 5.4. Data from an experiment by S7elford a il. (1963).
O :32 IIIm targets, O : tI rnm and + - 4 rrrfn. The distance from
the centre o{ one target to the far edge of the other for each target width
Br€r from left to right, 5or t42 and 4oz rnm.
Each point is based on roo shots by each of 8r subiects. The subiects

were all aged bennreen 4o and 49.

on target presented than on scatter observed. If so, Eq. 5.5 is mis-

conceived and should be replaced by Eq. 5.4. This view, apart from its
innritive improbability, appears to be untenable because it would assume no relationship bennteen observed scatter and movement time for
any given target and amplinrde. There was, however, a clear tendency
for time to rise as the distribution of shots became narrower, as shown

in Fig. S.T.

A second suggestion arose from the fact that the results could be
made strikingly more linear by subuacting a small constant (c) of about

156

Fundamentals of Skill

3 rnm from the observed widths of the disuibutions of shots, thus
replacin1lZ' in Eq. 5.5. by (W' - c). The constant might perhaps be

attributed to tremor, the presence of which would broaden distributions
slightly and mean that, in order to attain a given level of accuracy, the
subject would have to aim at astghtly higher level. Such tremor might
arise from the fact that the time taken to traverse the servo mechanisms
controlling movement means that there is an inevitable period of about
.r sec during which operation is open-loop. Alternatively it might arise
from the fact that muscular action by a large member such as the arm
is not very finely graded and that movements are likely to be slighdy
disnrrbed by factors such as the pulse. Some support for this general
point of view comes from the fact that the constant required to make
the present results plot approximately on a suaight line is of the same
order of maguinrde as the scatter achieved when a series of shots all
aimed at the same point are made at a moderate rate, also Fitts's results
are appreciably better fitted than they are in Fig. 5.4 rf the same constant is deduaed from lY. The slope of regressions such as that in
Fig. S.T, when plotted with 3 mm deducted from all the observed
scaffers is approximately that for joining equal targets at different
distances in Fig. 5.6 - that is about ro bits per sec. We may also note
that when very accurate placing is required, subiects tend not to rely on
arm movements but to support the wrist, thus minimising tremor.
On the other hand this view has serious disadvantages which seem to
make it, like the first, untenable. The subtraction of a constant from lY'
of about 3 mm obviously makes nonsense of the data of Schouten et al.
for pin sizes of r or z rnm since in these cases the effective target size
would be a negative quantity. Perhaps more important, we shotild
expect tremor not to add a constant to lV' but rather to add to the
variance of the disuibution of shots in such a way that one should write
This, however, does not produce a good
not (tY' - c) but \/W,
fit to the results shown in Fig. 5.6.
A third possibility, free from the objeaions to the previous t\ilo,
arises from the fact that if in Fig. 5.6 we ioin the points for any one
rarget width at different amplinrdes of movement we obtain slopes of
about ro bits per sec as shown in Fig. 5.8(A), and if we ioin the points
for any one amplinrde at different target widths we obtain slopes of
about 6 bits per sec as shown in Fig. 5.8(B). This suggests that two
control processes or phases ought perhaps to be distinguished: a faster
distance-covering phase and a slower phase of 'homing' on to the target.
It seems reasonable to regard the former as similar in speed to that of a
ballistic movement of a given amplinrde and the latter as implyrng an

Movement

r5T

additional process of visual control. If so, the appropriate equation
would be of the type:
Movement time

- alog A' + D log #,

6.6)

where a and b arc the slope constants for amplitudes and targets respec-

tively. Since ballistic movements show substantial accuacy without
a

a

:

a
a

a

a

0.7
a

a

a
o
a
a

a

o
a
a

H 0'6

)/.,
/t't

9

Ot

z
lr,

o/

./
,/
,/ t'

t=
rj

I o.s

/.

G,

tl,

a

a'

a
a

ooo
O1

o
o

(L
UJ

=

a

l0
SCATTER ON TARGETS 1UU )

Figrrre 5.7. Relationship benn'een the time taken per shot and the width
of distributions of shors.
The results are those for targets r r mm wide set with centres 3g7 rrun
apart. Each point represents the mean of roo shots by one subiect. The
best fitting suaight line is for an incremental rate of gain of 6.3 bits per
sec.

current visual control we should probably write in place of Eq. 5.6:
Movement rime

- alog#r+ bWW,
- alog A' - blogV'r * Q - a) log W'o $.1)

when W'o is the scatter of shots for ballistic movements of amplinrde
A' , and lY' 6 is the scatter observed with any particular target wid*l
under consideration.
The part played by (D a)logP'o depends on whether or not lY'o
incteases with A'. This, as Woodworth (r8gg) has pointed out, is

0.9
+
0.8

0.7
0.6

+

:

0.5
d
lrt

3

F.0.4

z,r
=
I 0.3
o
ll,

,

tr o.z
llr

A

=
tr 0.1
0

L0G2 (

67

0.5)

0.9
+

0.7

0.6

ci

0.5

UJ

a
F

0.4

z.
IJJ
I,L' 0.3
=
o

G,

rrt

(L

IJ

B

tr 0.1
0

0

1.,23

LoG2 (fi,* o'5)

567

Figure 5.8. The same data as in Fig. 5.6.
A. Showing the relation of movement time to amplitude, keeping
acctrracy constant.
B. Showing the relation of movement time to accurocy, keeping amplitude constant.

Movement

r59

ffictrlt to measure direaly as it involves conuolling both movement
time and also vision, but if Eq. 5.7 is valid we can infer an answer from
the present data. S.rppose, for example, that Weber's Law held and that
W'o rose proportionately with A', it would then be possible to rewrite
Eq. 5.7 thus:
Movement rime

where C -

- aros# + bbsffi
- bbs#+ (D - a) logc

(s.8)

V'o/A' : the Weber Fraction. It is obvious that Eq. j.8
+

-()

Iu, o-z
F

z

tr,

F ,.t

+

o

=

fr o's
u,

t=r' o.c

o

0.4

- 0.2 - 0.t
o.lo4 toGrA'

0.1 0-2 0.3

- 0.177 Loc2*i

0.4 0.5

Figrrre S.g. The same data as in Fig. 5.6 plotted in terms of Eq. 5.7. The
regression line cuts the abscissa at (6 - a) log V'o. The value of this,
-'37, corresponds to a W'o of 3r mrl.

would imply that all the points in Fig. 5.6 should lie on the same suaight
line and that it therefore does not grve a good fit to the observed data.
If dternatively we assume that V'o is constant at all values of A' ,
the expression (b
- a) log V', in Eq. 5.7 becomes a constant, and the
data should fall on a straight line described by Eq. S.Trprovided suiable
values for a and b arc chosen. Values were accordingly calorlated from

16o

Fundarnuttals of Skill

the data of Fig. 5.6 omitting the leftmost point which was suspected of
being somewhat too high because of the lower limit to movement time
mentioned earlier. The resulting plot in Fig. S.g shows all the points
faling close to a straight line with the exception of the leftmost. It
thus indicates that Eq. 5.7 provides a good fit to the data provided the
accuracy of ballistic movements (lY') is independent of their e)fient.
Paillard (t968) has provided some evidence that this last point is true
by showing that the accuracy with which one hand could be moved
into alignment with the other did not depend on whether the distance
moved was long or short. fn other words it seems as if the accuracy of
ballistic movements depends on some absolute appreciation of end
position rather than of distance moved.
An estimate of V's can be obtained from Fig. 5.9, since the distance
from zero to where the regtession line cuts the abscissa (b log W ' s.
In the example shown V'oworked out at 3r rnm which is") certainly
of the right order of magninrde. The rates of gain of information implied
by the slope constants a and b arc respectively 9.6 bits per sec which is
close to the figure for Fitts's results, and 5.6 bits per sec which is close
to that found by Hick (t95za) and others for choice reaction times. It is
tempting to suppose that the former represents the capacity of the
effector mechanism and that the latter represents control involving the
translation mechanism.

In concluding this chapter it must be emphasised that although we
know a good deal about the mechanism responsible for the control of
simple individual or repeated movements, we are still almost totally
ignorant of the ways in which the more complex patterns of movement

normally involved in everyday activities are phased, timed and coordinated. Perhaps the best guess that can be hazarded at present as a
gurde for funrre research, is that they follow the same principles as those
which appear to govern the co-ordination of perceptual aaivities. These
we consider in the next chapter.

VI
Economy of Decision

Duing the second decade of the present cennrry the results were published of several independent researches all aimed at emphasising the
faa that incoming sense data are grouped and ordered. Thus as
regards vision, we do not nonnally perceive simply a mosaic of more
and less stimulated points, but coherent objects which have form and
strucnrre. All these researches dealt essentially with visual percqption
but the principles they enunciated obviously had sifficance for other
sensory modes, especially hearing. Most of the writers, snd particularly
those constinrting the Gestalt school, were mainly concerned to relate
conscious perception to the physical 'strucfilre' or 'patterning' of the
stimulus. Bartlett (rg3z), however, in his snrdies emphasised from the
start that perception was in a very real sense an activity of the whole
organism, shaped not only by the obiective stimuli but by the attinrdes,
interests, hopes, fears and experiences an individual brought to the
sinration; there is, he tuged, an'effort after nuaning' which embraces
far more than the dpamic interaction of various parts of a complex
pattern of sensory stimulation. Bartlett's work was ahead of its time and
anticipated in many important ways the modern treatment of perception

in information-theory and similar tems. We shall not here examine
these approaches, old and new, in detail, but shall instead try to outline
some of the factors which seem to be important in shaping perception,
and shall then look at some of their analogues and extensions in other
areas of human performance. The issues involved are at first sight
highly academic, yet they have important implications for the desigu of

visual displays and of many t5pes of machinery for practical use (see
e.g. Easterby, \967).

r6r

r62

Fundamm,tals of Skill

ECONOMY IN PERCEPTION

Hick, in a paper which has attracted less attention than it deserves,
noted that

'a curious feature of mental representations, which has a bearing
on certain coding problems, is the tendency towards economy . . .
Professor Max Born, in his S7aynflete Lectures, refers sceptically to
the notion that elegant simplicity in a scientific theory tells in favour
of its truth. In the literal sense of the proposition, he is obviously
right to be sceptical. But we are not looking merely for truth, but for
useful truth, by whatever standard we iudge utility; and elegance, in
this context, is apt to mean "fitness for ftrnction". A simple representation, if approximately true, is far more worth having than a
complicated one; and a similar valuation appears to operate at the
relatively primitive levels of perception. It has often been remarked
how powerful is the compulsion to try to make sense of things, even
when it is known to be impossible' (Hick, r95zb, p. Z2).
The process of organising and grouping incoming data may be thought
of as *re abstraction of constants from the total mass of data presented
in space and over time, together with the selecti,on of some data as
dominant and important while the rest are relegated to the background
and more or less neglected. Neurological studies have shown that the
cortex is constituted to analyse patterns into units such as lines, corners
and their orientations, achieving economy by this elementary form of
coding (Hubel and Wiesel , t962, 1968). More complex perception may
be thought of as involving the imposition of schemes or rules which
reduce the effeaive amount of data to be handled.
Obvious examples are to be seen in concept formation and in the
estimation of averages of, ssy, several positions of a pointer on a scale
or spots on a diagram (Bartlett and Macksvorth, r95o, Spencer, t96t,
1963, Edwards, 1963), and in both there is a link with topics covered in
previous chapters. Concept formation can be regarded as a type of
classffication in which several differing objeas are put into a single
category (see p. 94) on the basis of some common feature, either an
item or some relationship benueen items which is identical in them all.
The estimation of averages can be thought of as an aaivity involving
choice which may be handled in information-theory terms: the accuracy
with which the average is iudged can be expressed in terms of the distribution of the quantities which contribute to it, and thus can be
expressed in informational terms (see p. \47).

konomy of Daision

163

Principles of abstraction and economy seem to apply on a more or less
detailed scale in many types of perception. The fundamental approach
of many modern sttrdies on this subiea is epitomised in a tlpe of guessing game outlined by Attneave (rgS+). We may take as an illustration
an experiment used for some years in the experimental psychology
praaical class at Cambridge. The experimenter has a gnd in front of
hin containing a pattern as in Fig. 6.r. The subiect qmnot see the
experimenter's grid but has his own blank gdd and is required to guess
for each square in turn whether it \rill be black or white. The experimenter tells him the correct answer immediately after each guess and
the subiect fills it in on his grid thus gradually builcling up a copy of

II
II
rI
II
II
II
ABCD
Figure 6.r. Designs used in 'gtressing gaute'. The subiect is required to
guess, without seeing the design, whether each square is black or white,
begiruring at the top left corner and proceeding along each row in turn.
the experimenter's grid. If the pattern is random as in Fig. 6.r(A), the
subiect will on average rnake so%errors. If, however, the pattern is not
randomr 8s in Fig. 6.r(8) and (C), the subiect will quickly detect the
regularities and thereafter make no more errors. If the pattern changes
as in Fig. 6.r(D), he will make a few errors arotrnd the point of change
but none once its nature and direction have been recognised. The process is similar to that discussed in Chapter 3 G. ro4) whereby the subiect
sets up a running hlpothesis which is at each step confirmed and
extended or refuted and modified by the incoming data. It is not, of
course, suggested that anyone does in faA scan a pattern like this in
ordinary perception, but the game serves to emphasize that, once
regularities are detected, much ofthe rest ofthe pattern can be inferred in information-theory terrns, it becomes, redtmdant. Information is
concentrated at the points at which regularity changes, such as the
beginning of the slope in Fig. 6.r(D). More generally one can say it is
concentrated at boundarics, at angles and where crrrved lines change
their rate of curvature.
The main principles by which such redtrndancy is attained can be

roughly divided into seven tyPes:

r64

Fundamentals of Skill

Grouping

The Gestalt school, especially I7ertheimer, emphasised that one of the
fundamentals of perception was that discrete objects in close proximity
tend to be grouped. This is obviously true of objects close together in
space and is also true of successive events in time. Consider, for
example, Fig. 6.2. Practically everyone sees the spots in pairs separated

oo oo oo co oo oo
Figrrre 6.2.

by spaces. If we put r for the spots and o for the spaces, the
pattern could be presented as r ror ror ro . . . and the subiect may be
conceived as treatinB (cocling) each or r as a single unit. Such coding is
economical in the sense that, if the regularity in the sequence is recognised, once o or r r have occurred, the ensuing r r or o can be inferred

with certainty and are thus redundant in the sense that they convey
no further information.
Attneave (1954) has pointed out that certain other kinds of spatial
grouping can be economical in informational terms in the sense that the
approximate locations of the individual obieas becomes redundant.
Consider the groups of points in Fig. 6.3(A). Their approximare locations can be specified in terms of the relatively coarse grid shown in

aoo
oo
oo a

,O

oo '
,oo
o

o

A

B

o

ooo
o oo

!

oo 3
D OO
aa

C

Figure 6.3. Economy in specifying position brought about by spec$ing
the approximate positions of groups.

Ecmmy of Decision

r6j

Fig. 6.2(8) and all shown to lie within two of the 16 squares. Precise locations can then be specffied in terms of finer grids, shown in
Fig. 6.3(C), which are confined to these two squares.
Continuity of line
We have already noted that information is concentrated at points in a
pattern at which the directions of lines or contorus change. The same is
true at points where a line or contour is interrupted. It is thus understandable that a pattern such as that of Fig. 6.4 tends to be seen 8s

-

Figure 6.4.
plus /rrather than as _/ plus
-'the former arrangement requires the
specification of two lines ody, whereas the latter requires the specification of three.
Continuity is not confined to suaight lines but can also be perceived
in curves, especially when the curvature has some easily specifiable
characteristic such as a steady rate of change of direction as in an arc
of a circle, or a steady acceleration as in a spiral. The same is true of
several other types of constant change: for example few people observing the series

o oo ooo oooo ooooo

would be in much doubt about how it would continue.
Regularity and synmetry

Hochberg and McAlister (rgSf) and Attneave (rgS+) have both ernphasised that the Gestalt school's principles that regular and symmetrical patterns are easier to see than irregular, can be accounted for in
terms of economy of specification. An inegnlar polygon, of which the
nnmber of sides is not known beforehand, requires for its specification
either the lengths of all the sides plus all the angles except one, or dl
the angles and all the sides except one - in either case (zn
- r) items if
z is the number of sides. If, however, the figrrre is known to be symmetrical, only n items need be specified includi.g the angle of one of
the sides to the axis of symmetry. If dre figure is completely regular, specification of only one side and one angle will suffce. Similar

$6

Fwdarnmtals of Skill

considerations apply to figrrres with curved sides although the process
of specifying them is obviously more complex.
This kind of approach was extended by Hochberg and McAlister
and by Hochberg and Brooks (t96o) to the perception of ambiguous
line drawings which can be seen as either two- or three-dimensional.
For example, Hochberg and McAlister found that the percentage of

A

B

C

D

Figrrre 6.5. Various projections of a ctrbe, together with the percentage
of time each was seen as three-dimensional rather than two-dimensional,
and certain stimulus characteristics. Adapted from Hochberg and

McAlister (1953).
Percent three-dimensional
responses

A
B
C

D

98'7

99'3

5r'o
40'o

Numbers of elements required to
specify figures as two-dimensional
Line segments
Angles
t6
z6
t6
z6

13
rz

20
18

time that cubes similar to those shown in Fig. 6.5 were seen as threedimensional fell with the number of lines and angles required to
specify them as two-dimensional - that is it fell from A and B to C and
again to D in Fig. 6.5. If the patterns are seen as three-dimensional
there are in all cases rz lines and 24 angles, so that if the number of
lines and angles is taken as the criterion, three-dimensions should be

Economy of

Decision

fiT

preferred for A and B, two-dimensions for D, and C should be equivocal, as was indeed the case.
Hochberg and McAlister recognised ttrat scoring in terms of lines
and angles is cnrde and that other possibilities need to be considered.
Of these three may be mentioned which appear to have application far
beyond the present case:
(o) The economy of three-dimensional perception is much greater
than nro-dimensional in A, B and C if the number of diffrenl angles
is considered. In trnro-dimensions A consists of three pairs of overlapping parallelograms each with two different a'rgles - a total of six whereas when it is seen as a cube all angls are equal. In B and C two
of the pairs of parallelograms are similar, but there are still four different angles. In D dl angles are equal in both two and three dimensions.
(D) Although, if minor perspective effects are neglected, all foru
views are proiections of a cube, they are not all equally probable. A
real outline cube would only be seen as D by an observer who viewed
it in one precise alignment with one eye from a fixed position. Viewing
with one eye would also be necessary for C and movement would be
permissible only along a vertical axis: lateral movement would ttun the
pattern into one of tlpe A. Patterns of tlpe B would be obtained from
a real outline cube by fusing images from both eyes, so that monocular
viewing would not be required, but movement could still only occur
vertically without its becoming a pattern of tlpe A. None of these
limiations apply to patterns of qpe A: the size of the srnall parallelogram in the middle would differ according to precise orientation, bur
the same general pattern would remain with both vertical and horizontal movements and whether viewed with one eye or two.
G) The faa that D is a completely regular pattern in trnro dimensions
makes one wonder why it was seen in three-dimensions by Hochberg
and McAlister's subjects for as much as 4oo/o of the time. A likely
reason is that all subiects were shown all patterns and were thus looking for three-dimensional appearance. fn other words, their perception
depended not only on the immediate stimulus, but on expectations
brought from previous stages of the experiment.

Continuity of change
Considerations of probability mentioned in (D) emphasise that percep-

tion is concerned with stimuli which are not normally instantaneous
but continue in, and may change \rith, time. A specification which is
not economical in a static display may therefore become so if the

168

Fundamsntals of Skill

display changes with time. A well-known illustration is Wertheimer's
demonstration of grouping by 'common movement'. For example if
the spots rnarked with arrows in Fig. 6.6 moved in trnison they would
be grouped together as distinct from those which remained stationary,
even though in a static display they would be grouped in pairs.
The same principle can be conceived as trnderlpng the perception of
movement by obiects: to speciff a single obiect as moving from one
location to another is often more economical than speci$ing a number
of obiects at different locations. This is shown in the so-called d-phenomenon of apparent movement: a light at one point which goes out
iust as another comes on a little distance away produces a compelling

t r

tf

t

oo ao oo oo oo oo

Figrrre 6.6. If the spots indicated by arrows move in trnison they will
tend to be grouped together as distinct from those that remain stationary,
despite the fact that this is contrary to the grouping by proximity seen

in Fig. 6.2.
perception of one light changing position, however much the subiect
knows that two lights are involved. It should be noted that the timing
must resemble that of real movement: if the gap between the first light

going out and the next coming on is too short the two lights appear
distina and simultaneous, and if it is too long they appear successive.
Continuity of movement can grve coherence in cases where other
sensory data would not. An illusuation is contained in a class e:rperiment
conducted in Cambridge for some years past. Subjects view a coloured
'bar' which travels along a horizontal slit in a screen, passing behind
three black bars as it moves. When the coloured bar remains the same
throughout its iourney, praaically all subiects see a stngle bar moving
the whole distance. If the coloured bar changes width or colour as it
passes behind each black bar, most subiects still see only one bar
moving. It is only when both colour and width change simultaneously
at each black bar that a maiority regard the moving bar as no longer one
over the whole of its travel.
Perception of distance

We have already noted one example of economy resulting from perception in three dimensions with the proiections of cubes shown in
Fig. 6.5. The economy principle seems clearly to be involved also in
other features of depth and distance perception:
(a) Perspective gradients, such as that in Fig. 6,7, enable obieas of

Economy of

Decisiut,

t@

equal obieaive size but different retinal sizes at different distances, to
be perceived as the same size once the gradient has been recogRised
(Gibson, r95o). The extraction of the gradient from the data in the
observer's field of view thus increases the invariance of the obiects in the
field. For example, if Fig. 6.7 is specified in two dimensions it is necessary to assign different magniftdes to each row of bars and each distance benn een one bar and the next. If, however, a perspective gradient

t-------rr!--

r-----

III
-Z

-II

-II

Figrue 6.7, Perspestive gradient used in experiments by Vickers (1967).
is extracted so that the bars are seen as lying on a flat receding surface,
all can be specified as equal in size and distance apart.
The gradient appears to be extracted at some cost. If, for example, the
whole of Fig. 6.7 except the bottom two rows of bars is covered by a
card, the uncovered portion is not likdy to be perceived as threeclimensional. If now the card is slowly raised, to e4pose further rows,
the three-dimensional effect E'ilI suddenly appear. The same is true of
the example shown in Fig. 6.8 but the changeover from two to three
dimensions comes later. This can be regarded as due to the faa that the
amount of daa per row rendered invariant in Fig. 6.7 is greater than
in Fig. 6.8. In the former length, width and distances between bars

t7o

Fundamentals of Skill

are all made invariant, in the latter, only distance. Systematic studies
of this phenomenon using the patterns of Figrues 6.T and 6.8 and others
have been reported by Vickers (1967).
The gradients of Figs. 6.7 and 6.8 are for flat surfaces. Gibson
emphasised that these are not the only ones that can be extracted:
different gradients give impressions of ctuved surfacsr either concave
or convex, and the rate of change affects the apparent angle of slope

relative to the subiea's line of regard.
Two corollaries to this principle are now well known. First, if a

Figrrre 6.8. A simplified perspective gradient used by Vickers (1967).
gradient is extracted, obiects located along it are scaled for size accordingly, as is shown by some familiar visual illusions (e.g. Fig. 6.9).
Second, when a wrong gradient is extracted the sizes of obiects may
appear to be anomalous. The clearest example of this is in the Ames
room where the subiect looks through a peephole into a room of irregular

shape $/ith one far corner much farther away than the other. The
obieas in the room are so shaped, however, that all the wall panels
become similar in shape and all the floorboards the same width if the
room is seen as rectangulrr - the invariance in the pattern seen is
maximised by regarding the room as rectangular. When the whole

Econorry of Decision

r7r

room is seen in this way the sizes of isolated obieas appear very
different according to whether they are placed in the nearer or farther
corner: for example, of two people of equal height the one in the farther
corner appears shorter. ![e may note in passing that these phenomena
have little if anything to do with the fanriliarity of rectangtrlar roolns,
but wirh maximising the invariance of the data received by the observer:
it is, for example, possible to constnrct an obieaively reaangular room
which looks irregular in shape. The illusion of the irregular room which
appears rectangular has been shown to break down after prolonged
observation: this is likely because close scrutiny \rill reveal a ntrmber

FigUre 6.9. The two large rectangles are the same Size.

of details, such as the fine texnrres of painted surfaces, which are not
consistent with perception of the two corners as equidistant and thus
reduce the extent to which the rectangular view achieves invariance.
The inadequacy of familiarity as a basis for iudging size in relation to
distance has also been demonstrated in other contexts by Hochberg and

McAlister (lgSS) and by Gruber and pinns6tein (rg65).
(D) Gibson was at pains to emphasise that the visual scene which is
normally perceived is not the same as that observed in a single fixation
of the eye, but represents the integration of many different glances. A
good illusuation of this is obtained by viewing Fig. 6.10 in a stereoscope. At first glance no depth effect will be seen: it comes only gradually as first one then another pair of circles are fixated and firsed. It
is, in short, achieved when the gradients of retirwl disparity and
binocular convergence have been extracted. Once this has been done,

t72

Fundam,entals of Skill

the depth effect is virnmlly immediate - if one looks away from the
stereoscope and then back again, the depth effea is seen at once.

(r) Gibson also emphasised that the observer is not normally static
but moves, and in doing so produces different changes in the relative
positions of objects at different distances. Maximising invariance in these
circumstances requires the integration of much more data than when the

observer is stationary, and although the economy achieved is greater,
the final complexity of what is seen is greater also. The same essential

principles apply to the perception of solid obiects: as the observer
moves their shapes undergo a series of transformations which can be
reduced to invariance by specifying the observer's movement.

Figure 6.10. If the right-hand circles are viewed by the right eye and the
left-hand circles by the left eye in a stereoscope, the outer iircles will

appear nearer and the middle circles farther away than the centre circles.

(d,) These several points jointly lead to the rather surprising conclusion that the visual scene as we perceive it has little to do with the
actual data received by the eyes at any moment. It is rather a kind of
frameatork in which the kaleidoscope of individual glances is reduced
to relative invariance and stability, enabling each new datum to be
fitted into a wider context, and providing the observer with a means of
orientation and a continuity of space arotrnd him.

Constancies

The foregoing discussion of distance perception has obvious implications for size and shape constancies. Once a perceptual framework has
been built up, perception of size as invariant with distance and of shape
with orientation are economical. These constancies are destroyed only
in cases in which the gradients indicating distance cannot be extracted.
Thus Thouless (lgll) found that a circular disc viewed tilted at an
angle was seen as such except when it was placed against a black velvet
background and viewed monocularly with the head held in a rest.
The fact that objeas retain the same apparent degree of whiteness or blackness - when seen in different illuminations can also be brought
within the economy principle. In a manner somewhat analogous to that

Economy of

Decisim

ry3

of grouping in Fig. 63 it is economical to speci$ the brightrcss of
different parts of the field and to specify the whitutess of obiects within
these parts rather than speciS all the different amounts of light reflected by the various obiects in the whole field. Despite sratemenrs in
some of the standard textbooks and the findings of classical workers
such as Katz (lglS), it is not generally recognised that the visual
mechanism usnally has fully adequate data for dividing the field up
into areas of different brightness. The amount of light reflected by a
white object in shadow is often far less than that reflected by a black
obiect in good illumination within the same field of view - in other
words, differences between black and white cover a relatively small part
of the total range of light intensities present in any perceived scene.

Familiarity
The factors making for economy in perception that we have considered
hitherto have been concerned with the physical characteristics of the
incoming data and not \rith the subiect's past experience. It must,
however, be recognised that past experience can upon occasion play an
important part in determining iust how incoming data are coded. The
categories of identification discussed in Chapters 2 and 3 are like
templates which the subiect tries to fit to incoming data. Normally he
is successful and the process is so rapid that he is unaware of its happeDhg, but when objects are uncourmon, or seen from unusull angles or
under adverse conditions, the process may take much longer and the
search for a 'fit' may be much more conscious.

The imposition of a template from past experience has much the
same effect as grouping enabling a complex of data to be apprehended
as a unitary whole. Typically the fit does not have to be precise but has

to fall within acceptable limits of approximation, and the template itsef
seems constandy to be modified as the result of funher experience. For
example, the category 'modern car' will include a range of shapes, sizes
and colours, snd will change substantially during the lifetime of any
individual. It was such facts that Bartlett (tglz) expressed when he
referred to the 'fitting' process as 'schematic' and the templates as
'schemata'.

Bartlett noted that complex line drawings to which a match could be
made in this way were regarded by his subjects as 'simpler' than less
obieaively complex drawings that could not be readily linked to common objeas. Such observations would almost certainly be confirmed
quantitatively by Attneave's tlpe of guessing game, in the sense that

r74

Fundamentals of Skill

fewer errors would be made if a familiar shape was recogRised than if
all the subiea had to go on were simple continuities and regularities.
One can perhaps see an analogy to Bartlett's results in Michotte's
(tg46, rg(4,) experiments on the perception of causality in sequences of
events: the subiea imposes a pattern, couched in terms of one event
causing another, which gives coherence to several separate events and
enables them to be treated as one. For example, if one obiea such as a
ball moves up to another, which then immediately moves away, the
first is seen as pushing the second. The successful application of such
a pattern is likely to depend, as Michotte insisted, upon the obiective
sequence conforming to certain time limits: the causal sequence would
not be seen if there was a substantial time interval betrnreen the movements of the two balls.
Bartlett was clear that the schemata actually imposed showed some
effects of social conventions and of the subiects' individual interests,
presumably implying that the likelihood of any one schema being
applied depended to some extent on its availability at the time of the
experiment. The possibility of deliberately shaping perception by inducing bias towards partictrlar schemata was illustrated in the classical

experiment of Carmichael et al. (tglz) who showed, for example,
that the reproduaion of two circles ioined horizontally by a short line
and exposed briefly differed substantially according to whether the
subject was told beforehand that he was going to see a pair of eyeglasses

or a dumbell.

Two results run through Bartlett's and much preceding and subsequent work, whether concerned with effects of familiarity or of the
obiective structure of the stimulus.
(o) Much of the detail which is apparently perceived is in fact inferred. It seems as if a schema which is imposed on incoming data
brings a considerable amount of detail with it and this detail is incorporated into the resulti.g perception.
(b) Details which cannot be fitted into the schema are either ignored
or become what Bartlett described as 'dominant details' which are specially noticed. Woodworth (see Woodworth and Schlosberg, r95D made
some play of the same idea under the title of 'schema with correctior',
suggesting that a subject makes an approximate fit to the data with
some familiar category which, if reasonably adequate, suppresses the
perception of deviant details. If, however, the lack of fit is substantial
in one or two respects, the deviant features are specified separately.
Attneave has argued that such a procedtrre could still be much more
economical than no categorisation. We might add that it could often be

Econonry of

Decision

rys
more economical to accept a fit \dth a yickly found schema and to
speciff deviant details than to make an extensive search of the material
stored in memory for a more precise category.
Turning this second result the other way round, several sttrdies have
shown that the difficulty of detecting a gtven deviation from a standard
pattern rises as the deviation becomes smaller in relation to the size or
complexity of the pattern (e.9. Hillix, t96, van de Geer and Levelt,
1963, Sengstake, 1965). Many other studies have shown that once a
familiar scherna has been applied it may be difficult to change: for
example it is often very difficult to reorder the letters in a word to
produce an anagram. The difficulty of such reordering rises substantially as the transitions from one letter or pair of letters to another in
the words presented become more fanriliar (Beilin and Horn, 196z) or
as those in the solution words become less familiar (Tresselt and

Mayzner, 1965).

Problems of formulation
Economy in perception is a concept which nrns through a number of
somewhat different conceptual models of the processes involved. We
have spoken here of economy of specification. In its crudest statement
this may refer to the number of lina, angles or other features required
to define a pattern. More fundamentally it means that several details
of a pattern can be subsumed trnder some more embracing rule or
schema which makes it possible to infer or predict one part of the
whole pattern from another. In these terms there are obvious links
with information-theory, although the precise manner of linking is open
to a good deal of discrrssion (see Attneave, 1959, Garner, 1962, Staniland, 1966, Evans, ry67). Broadly speaking it can be claimed, on the
one hand, that imposition of a rule or schema makes much of the detailed data redtrndant and that perception maximises redtrndancy or
invariance. At the same time we can say that the ntrmber of separately
variable items in the data is reduced, so that inforrration is minimised.
In other words, the co-variation benneen different items has been
recognised and used to determine the way in which the data are ordered.
In either case the overall mathematical formulation describes a complex process within the subiect which needs to be spelt out in detail
if a full understanding is to be obtained. From this point of view we
may tentatively suggest that economy can usefully be considered in
terms of the amount of data that has to be specifud in ordn to anizte at

tT6

Fmdamentals of Skill

a sufrcient definition of the pattern We can regard the process of definition as one of classification or choice analogous to those disctrssed in
Chapter 3, involving one or more decisions of the qpes discussed in
Chapter 2. \Mhat is 'sufficient' will depend as emphasised in Chapter 3
on the degree of precision with which the classificatiou must be made.
This method of stating the process covers a number of results not
easy to account for in terms of other formulations:

(") Consider the pattern shown in Fig. 6.rr(A). Subjects tend not to
group by pro*imity as they do with Fig. 6.2, which would yield alternate groups X/ and /X, but by similarity so as to produce alternating
A

B

x

lr

XX

ll xx

XX

X

-x-l lr---xl l-i--i-l lr---xl l'-

Figrrre 6. r r.

groups of / / and XX. There is no advantage of one arrangement over
the other in the simple information-theory terms used for Fig. 6.2,
but there is some advantage for alternate groups of lines and crosses in

RAN

DOM

CONSTRAINE D

Figure 6.r,2. Examples of bar-diagrams used by Fitts et al. (t956).

that they can be defined in terms of whether or not they come above
dre line in Fig.6.rr(8). In other words, the subiect can group simply
in terms of the tops of the lines, and virnrally neglect the rest of the
pattern.
(D) Fitts et al. (lgS6) required their subjeas to identiff bar figures
of the grpe shown in Fig. 6.12 from among sets of eight similar figures.
The bars could be of eight different heights and in different trials the
figures were construaed either with random bar-heights or with the
constraint that all eight heights were represented once but in random

Econwny of Decision

ryT

order. It was expected that because there were 88 possible random
figures but only 8 ! constrained, that the latter would be identified more
quickly. The reverse, however, was found to be true, and this result
has been confirmed by Anderson and Leonard (1958). Fitts et al.
noted that the constrained fiSrres appeared more similar to one another
than did the random, and if one considers the subject's task in detail
one can see that this is likely to be so. If he compares only the first
bars of the standard figrre and of one to be distinguished as similar or
different he canr oD average, decide trat it is different seven times out
of eight. On the eighth occasion, when the first bars are the same he
will have to compare the second bars, and again he will be able to
decide seven times out of eight. The same will be tme if he has to go
to a third or subsequent bar. \fith constrained figures he will again be
able to decide seven times out of eight by inspeaing the first bars ody,
but if he has to go to further bars he will only be able to decide six
times out of seven for the second, five times out of six for the third and
so on. Alternatively if he inspects the figrrre row by row from the top
downwards he will in many cases be able to identiff a random figure
from its having two bars of the same height longer than any others, but
he will never have this cue with the constrained figures. fn short, whichever method he adopts he \d[, with the constrained figtres, have on
average to inspect more of the pattern than is necessary with the random
figures.

(c) Attneave(tg1l) required subjects to learn to associate boys'names

with sets of eight irregular polygons. Each set was constructed by making minor variations in a basic design. He found that learning was
better when all the variations were in the same corners than when any
of the corners might be altered. It is perhaps obvious that one design
is likely to be distinguished from another more easily if all the variations
are concenuated at a few points since the remainder of the figrrre can
then be neglected.
(d) However, even if it is strialy possible to ignore some parts of a
pattern, subiects may find it difrcult to do so. For instance Bricker
(rgSSb), whose subieas responded as quickly as possible with a dif:
ferent nonsense syllable to each of the eight patterrrs shown in Fig. 6.13,
found that reaction was quicker when only the rightmost three items
in each row were exposed than when all five were shown. The patterns
could be unamblguously identified by means of either the first three or
the last three items, so that there was no need to observe them all.
Nevertheless subieas seemed to have done so, at least to some extmt.
The main advantage in having all five items exposed is that if one is

rl8

Fundamentals of Skill

unreliable, the correct response can still be inferred from the others so
that accuracy does not suffer. This beneficial effect of redundancy was
shown by Bricker to ocqlr, as it has been by many others in other
conteKts.

Perception in everyday life is, of course, much more complex than

it is with the single obiects or groups of

obiects used in laboratory experiments.
I

2

3

4

5

6

1

o

o

o
o

oo
oo

o

o

ooo
o

o

o

o

o

o

ooo
o

o
o

This is so not only in the sense that many
and varied obiects are dealt with more or
less simultaneously but, more importantrthat
the co-ordination and integration of data
seems to take place on several different
scales at the same time. A clear example is
that when reading we simultaneously integrate letters (or graphemes) into words, words
into sentences, sentences 'into paragraphs
and perhaps also become aware of even
broader features such as the style of writing.

Similarly in looking at a painting we may
be aware not only ofthe various obiects depicted, but of details such as brush-work,
and of broader aspects ,such as the overall
composition of the picture. The upper limit

to the integrative process at any one

moment seems to be set by the extent to
which there has been time and opporttrnity
to observe enough detail to produce a larger
Figure 6.13. patterns of framework within which it can be ordered.
Iights used by Bricker The lower limit seems to depend on how
(r955b).
far the action to which the perception is
directed demands attention to detail. It is
well known, for example, that when a book is being read rapidly, misprints may pass unnoticed that would almost certainly have been recognised when reading more slowly. Attneave (lgS+) glves the illusuation
that when looking at a furry animal we do not normally observe each
separate hair, but see all the hairs together as a'furry texture', or again
that looking at a monled surface we speciS merely the overall texnre
and nor each individual variation. This, he points out, is equivalent
8

o

o

o

o

o

to adopting a relatively coarse grid in the 'guessing game' and so treating

together coherent areas which contain given proportions and distributions of black and white (or whatever colours are involved), as having a

konomy of Decisiut,

t7g

particular texnlre as opposed to other areas \ilith different proportions
or distributions. The setting of this lower limit taks us beyond perception in the immediate sense to the interplay of perception and action
to which we now turnr.

RECODING AND TRANSTATION

The coditg of input data in the manner we have been discussing is
tlpically succeeded by one or more cenual stages of what may be
broadly termed recoding - that is to say the coding initidly imposed on
the data is uanslated into a different code. These processes are obvious

in uanslating from one language to another, or in putting into words
material presented visually, and they clearly turderlie tests of mental
funaion such as digit-symbol substiffiion.
Recoding may result in substantial further economies in dealing
with incoming data. Oldfield (rgS+) grves the following example:
IOOIOOOI IOI I IOOOI I IOOIOOOI I IOOIOOOI I
The series can be broken up into units of roo or orl, and if we recode
roo as A and or r as B it can be uanslated into the very much shorter
seria
AABBABAABAAB

This could be further shortened by recoding AAB as X and BAB as Y
to produce

xYx)(

The economy resul 'tB from each recodirg is achieved at the cost of
having the 'key' to it stored in memory and available for dgcsding the
message when it is required to put it back into its original form: if this
key is not retained most of the origrnal information is lost.
Recoding can make material easier to handle in several ways. For
example, Miller (lgS6) notes that whereas a subiect can repeat back
only about nine random binary digts if he deals with them as such, he
can reproduce many more if he groups them in threes and converts into
rrr

: T. In this way the series
IOIOOOIOOIIIOOIIIO

becomes

5o47t6

In this case the total information in the message is unchanged by the
recoding but the number of items ('chunls' as Miller terms them) is
greatly reduced. An alternative, although less efrcient, method of reducing the ntrmber of items is to retain the first and then the lengths of

r8o

Fmdamentals of Skill

run of each type afterwards. Miller's sequence would in these terms
become

rrr I3r2323r
Miller found thatthe method he described required considerablepractice
before its full effect was obtained: in other words the benefits of recoding
again depended on the establishment of a 'key' in memory and having
it readily available for use. Such recocling also implies grouping and
can be hindered when conditions make this difficult. For instance

Klemlner (1964) found that practice in transforming binary digits to
octal did not substantially improve the span of perception for binary
digits exposed for very brief intervals. Again several experirnents
have shown that patterns are less accurately perceived if different portions are shown successively in time than if they are shown all at once
(e.g. Harcum and Friedman, 1963, I(Ieene, 1965), snd that words are
less easily read if shown letter by letter instead of in larger units
(Newman, 1966). Recoding also takes an appreciable time and may
therefore not be achieved when material is presented at fast rates
(Pollack and Johnson, 1965).
Inuospectively such recoding seems to be a kind of response to the
material and may therefore be only a special case of the recoding that
takes place in the uanslation from perception to action - the difference
being simply whether or not the response is overt. To consider the
transition from perception to action as a process of recoding may at
first sight seem surprishg, yet a moment's reflection must make it
clear that there is a difference between the neural aaivity involved in

the identification of an obiect and that required to initiate action in
response to it: in subjective terms there is a profound difference between
knowing what has happened and deciding what to do about it.
Problems of recoding berween perception and action have in recent
years assumed considerable practical significance in the design of controls for vehicles and machine tools and of consoles for the monitoring
and control of automatic plant, strd it is research aimed at clarifying the
principles involved in these that probably provides our most systematic
present knowledge of recoding operatiolls.

Compatibility' - the relationships betreen displays and controls

Many experiments have shown that performance of a task is affected
not only by perceptual requirements and actions involved but by the
relationships between them, and have attempted to relate performance
to the nafi,ue of the intervening steps needed to bridge the gap between

Ecanomy of

Decisiort

r8r

perception and action. Such steps are at a minimum in the situation
used by Leonard (tgSg) for the choicereaction experiment mentioned
in Chapter 3 (p. 8f) in which the subiect held his fingers lightly on the
armatures of a set of relays: the signals were vibrations by one or other
of the armatures, and the subiect responded by pressing the same armature. Relationships benreen perception and action are similarly suaightforward when moving an obiect by hand from one position to another,
in that the direction and extent of the movements of the hand are
directly related to the perceived positions and changes in position of the
obiect. Baker (196o), for example, found that uacking was much better
with a stylus that could be used to trace the target direaly on a cathoderay nrbe than with a ioystick which moved a spot on the firbe.
One of the clearest exarnples of complicating these relationships is
given by Garvey and I(nowles (1954) who used the displays and conuol
panels shown in Fig. 6.14. In system A each signal was the lighting of
one of the roo bulbs, and subjeas responded by pressing the button
immediately below the light concerned. The same was true of system B
except that each signal consisted of two lights, one in each column, rnd
responses were made by pressing the correspondi.g t\tro buttons, thus
again Sving roo possibilities. The completion of each response brought
on the next signal so that the subiea set his own pace of work: he was
told to work as fast as possible without making errors. Systems C and D
were similar to A and B respectively except that the lights and buttons
were on separate panels. Performance with C and D took about twice
as long as with A and B, prestrmably because with C and D the subiect
had first to identify the position of the light or lights on the display
panel and then search for the coffesponding positions on the conuol
panels, whereas with A and B one search sufficed for both. We cirn
perhaps assume that the same economy accotrnted for the flattening of
the slope relating reaction-time to degree of choice with highly compatible arrangements which were found by Crossman (tgS6) and Griew
(t958b) aod mentioned in Chapter 3 (p. 8z).
With system E the row and column of the display panel in which the
light appeared had to be indicated by pressing the appropriate buttons
in the left- and right-hand columns of the control panel: there was
thus a recoding required berween display and control. A rcssding in
the opposite direction was involved in F. These systems yielded considerably slower performance than any of the others, and this seems to
have been due essentially to the need for recoding and not to any
feanue of the display or control as such. This is not to say that display
and control feanrres had no effects: systems B, D and E in which two

o

oooooooooo

H

z

(

a
o

O o
aO

O

a
a
a
o
o

oo
oa
ao

ooooo oo o oo

o

-D

o ooaa

ao
oo

ao

9O (JCt lrJlL(D ---r

O o

o

E

oo
oo
O.

oc

oo
oo
oo
oo

o
O

o
o
o
a
a
o

LL

u
=l

F

(n

a

(f

z.

o
c)

ooooo oo o o o
ooooooo o
ooooo oo o o o
ooooo oo o o oo
o oooooo o o
o oooo oo o o o
oo

F Gl G) t+tf)(I)t\(DCl:<

oooooooooo

trl

=

UJ

o
o oooooo
o oooooo o o o
o oooooo o o o
oooooo o o

t-

oooooooooo

(n

<lrl()OuJr,,l,---,

u)

e $l<rtSrnOf-OO: p

-(\l

oooooooooo

(r) -ttft(^ol.-@Cl)g

o
,-1.

=
F
UJ

H

z
o

oooooooooo

ocoooooooo

< ,Er (Jour.L(r--)

{tD qlOLr.ru-(Df,-;

< O()t3[dtl()I--

{O<.rOuJu_(9I_r

$
r/t
o\
F'

u)

a

d

o
o
B

o
A
)r

E

ooooo oooqo ol
ooooo ooooo
F ooooo ooooo
(o
ooooo ooOOO to
ro ooooo ooooo rr)
-l ooooo ooooo rf
(vl ooooo ooooo
(\ ooooo ooooo
ooooo ooooo
I

c)

z

ooo

ooa
ooo
ooo
ooo
ooo
ooo
ooo
ooo
oao

CD

o ooooo ooooo o

o

(J

oooo
cooo
oooo o
oooo o
oooo o
oooo o
oaoa O
oaooo
ooooo
ooooa

M

€E
C)

=

UJ

1-

a
a

G,

h
o
k

G'

()
>-t

rO

1,
o
U'

)
d

a

r.<

iF- (\

(Y) <f

ul, (J) f-

@ Ol sz

oaoooooDoaororoDoD
oDOOaOOtOODO.ooD

H

z

o

F

()

u
=l
F
0
a

o

tr

c,
O.
Fra

o
k

.F,

tr

o
u

€c
CU

o, oD ot ooDoroa

o

zo

co

F

oo

>1
(Il

oa oa o, oDoaoro
oo oD oa ocooao ot qa
OD OD

Fa

OODO Oa OO OO OD or oa

cD
.<

OOOOD ooorooo or OD
o, o, oo oa oo oD oD oa oa
OD
oaoro oroooao
oo or
oD

oD
ODODO ot oD
oa oD OD ODOOOO' OD ca
oD ot o, oo OD

q

u
=l

F

a
a

a

+
l.{

\C'

o
*{
A
J

h0

.H

fr{

r8z

Eonomy of Decision

r83

buttons had to be pressed for each response were all a little slower than
the correspondi.g systems A, C and F which required only one, and
the difference might well have been gf,eater if it had not been partly
offset by the difficulty of discriminating one light from another in the
ro X ro displays used in systerns A, C and E. Evidence of such difficutty has been provided by Crarvey and Mitnick (lgSS) who found that
performance with system C could be speeded up by marking the display and control panels with lines dividing the ro x ro matricc into

o
o

o

o o

o
o
o

o

o
o

o

o

o

o

Sa

Se

Ra

Rs

scALE |

o
Sc

6!N' I

Rc

Figure 6.15. Display and control panels used by Fitts and Seeger (1953).
lfith the display panel SA the sienal was one of the 8 lights. lfith-Sn
each signal consisted of either one light or two adiacent lights. Strirh SC
each signal consisted of either one light alone or one in each pair. All
display panels thus provided 8 possible signals. !7ith control panels RA
and RB the subiect responded by moving a stylus from the centre to one
of the 8 indicated positions. Tfith RC he had two styli, one on each bar,
and moved one or both as indicated by the signal.
smaller groups. The differences betrn een the several systems appeared
to be resistant to at least moderate amounts of practice: although per-

formance improved markedly over rr5oo reactions, the differences
betnreen the systems remained substantial. Whether they would have
disappeared after much longer practice is, of course, another matter.

Similar effects of recoding have been shown by Fitts and Seeger
(lgSf) who compared all possible combinations of the display and
conuol panels shown in Fig. 6.rj. They found that reactions were
fastest and most accurate for each display when it was paired with its

r84

Fundamentals of Skill

corresponding control panel - that is when the recoding benreen display and control was at a minimum. It was these authors who coined

the term 'compatibility' to denote the degree of directness benreen
display and control.
Looking at these experiments in the perspective of others, the recodirgs normally required berween perception and action can be seen
to fall into two main classes : spatial transpositions and symbolic translations. Included in the former are mirror reversals and the effects of
various other arrangements of display and control which require some
mental reorientation of the one in order to relate it to the other. An
example, amongst the many that could be cited, is a serial reaction-time
experiment by Kay (rgSS) who presented his subjects with a box containing a row of tz lights and a second box with a corresponding row
of tz Morse keys under three conditions: (a) with each key immediately below its corresponding light, (D) the same but with the box of
lights 3 ft away across a table and (c) the same as (b) but with the box
of lights reversed end to end so that the leftmost light corresponded to
the rightmost key and so on. The signal lights appeared in random
order, each being brought on by correct response to the one before.
The mean times to complete 30 responses under the three conditions
were (a) 24'5 sec, (D) 4o'8 and (r) 9T'4 and the errors made were o,
4.o and 9'o.
Increasesof timererrorsror both have been found also when horizontal
movements have to be made to match distances indicated on a vertical
as opposed to a horizontal display (Szafran, see Welford, 1958, pp.
14z-l.46) and where view of a target is distorted by viewing through
prisms (Kalil and Freedman, r966a, b). The most extensive series of
studies in this area has been made, as mentioned in the previous chapter,
by K. fI. and'W'. M. Smith and their colleagues. They observed writing
and other activities by subjeas whose hands were obsctrred from direct
view but could be seen via a television screen on which the image was
rotated by varying amounts (e.9. Smith et al., 1956, Smith and Smith,
1962, Gould and Smith, 1963). Performance at these tasks improves
rapidly with practice, and the adiustment once achieved seems to be
general in the sense that when it has been acquired for actions by one
limb, it transfers to other limbs (e.9. Bray, r9z8).
The general concept of spatial transposition as a type of recoding
which takes 'me and central capacity to achieve is of obvious application to certain spatial tests of intelligence which demand that shapes
should be rotated 'in the head' to make them fit into other shapes. fr
also applies to the difficulty experienced when writing letters and nurn-

konorry of Decision

r85

bers in reverse: for example Brown (see Welfor4 r95rr p. 66) found
that the average ttme taken by subjeas to write the ten figrres r-o in
a normal rnanner was 9'4 sec but rose to 33.o sec when they were written
with each digit reversed. The difference of time required by the same
subiects to trace over normal and reversed figures was very small. The
reason for the difference was presrunably that when uniting reversed
figrres subieas had to transpose the orders that would normally be
given to gurde the hand, whereas when tracing they had merely to

T HIRD ANGLE

FIRST ANGLE

n

B

A

n

A

B

Figure 6J6. Illusuation of first and third angle orthographic proiections
for machine drawings as studied by Spencer (1965).

follow straightforwardly the pattern presented to them. The reverced
writing showed a substantial practice effea - the mean time for a second
attempt at the figures r-o was only 22.2 sec - indicating that once the

transposition had been rnade it remained to some extent available for
use and did not have to be rebuilt from sctatch.
A further problem in the field of spatial transposition has been sudied
by Spencer (1965) who compared the comprehension of two standard
fornrs of engineering drawings shown in Fig. 6.16. Untrained subiects
read the 'third angle' drawing more quickly and acctrately than the
'first angle': it is obviotrs that the 'thfud angle' is the more straightforward in that sides adiacent between A and C and between A and B

186

Fundanuntals of Skill

in the actual article remain adiacent in the drawing. Again the difficulty
can be reduced \dth practice - experienced draughtsmen performed
equally well with both qpes of drawing. The cost of the recoding
required is, however, indicated by the fact that all subiects fourd perspective or isometric drawings easier than either first or third angle
proiections.
Comparisons of relationships involving symbolic translations between
display and control \rith more compatible arrangements have been
made by Knowles et al.(rgSf) and by Fitts and Deininger (1954). The
former compared systems C and E of Fig. 6.14 with systems in which
the display was replaced either bya window showing a letter and ntunber
(e.g. A-r ot E-7)r or by a letter and nurrber spoken over a loudspeaker,
to indicate the response required. Their results are given in Table 6.r

TABLE 6.r Effects of swbolic translation between display and coilffol.
Rerults obtaircd by l(nos)les et al. (196l). Each figure is the tnea, time per
response in sec based on roo readings from each of nine subiects.
Type of control

Type

of

display

panel

IOXIO

Letter and

Letter and

matrix as C
in Fig. 6.t4

number shown
in window

number heard
from loudspeaker

IO X IO matnx aS
C in Fig. 6.t4
Double column
as E in Fig.

6.t4

2'o7g

2'234

r'668

2'7oo

2'179

r'55o

which shows that although, as Garvey and Knowles (1964) found, the
matrix response panel (C) was better than the double column panel (E)
with the matrix display, the double coltrmn panel was relatively better
with the window and loudspeaker displays: the double column panel
is the more compatible with'figure-letter' presentation. The superiority
of presentation by loudspeaker over other methods was presumably due

to its having enabled *re subjects to look directly at the control panel

all the time instead of uansferring their gaze to it only after the signal
had appeared.
Fitts and Deininger (1954) compared the display and control shown in
Fig. 6.t7 with an affangement in which the same control was used but
the display was replaced by a window in which figures could be shown
indicating clock positions - for example r2.oo, 4.3o, 9.oo P corresponding to the different directions on the control panel. The average times

Economy of

Decision

$T

taken to move from the centre of the control panel to the end of the
approPriate 'spoke' and the errors made are shown in the first two
rows of Table 6.2rfrom which it is evident that the symbotic translation
impaired both speed and accuracy: taking both into account, it about
halved the rate of gain of information. The remaining four rows
of
Table 6.2 show the effects of increasing the complexity of the spatial

DISPLAY

OF LICHTS

ooo
o
o

o
o o

RESPONSE

PANE L

Iig*S 6.17. Display and control panels used by Fitts and Deininger
(1965). Each signal was one of the 8 lights. fhisubiect responded-by
moving a stylus from the centre of the iorrtrol panel to &e ena of onl

of the arms.

rule relating display to control by either mirror-image reversal so that
if the display signals 3.oo the subiea moves to what would normally be
the 9.oo positionr or by complete randomisation. It can be seen that the
effect of these changes is much greater for the normally comFatible,
spatial display than for the symbolic, so that when relationships .r.

completely random the slmbolic display yields better performance.

Kay Gg54' see also Welford, 1958, r96zc, d) has shown that adcli'g
one recoding to another may impair performance to a greater extent
than would be accounted for by assuming that the effeas of the rwo
recodings were simply additive. The apparafts alreadydescribed (p. r84)
was used but with the additional complication of an index card bearing

r88

Fundanmtals of Skill

TABLE 6.2 Effects of symbotic tarulation between display and coritrolResults obtained by Fitts and Deininger (rgS+). Eachfigure is the tnean time
per response in sec or percentage of errors based ort t z8 readings from each of

tm subjects.

Type of display

Correspondence
between display
and control
positions

Circular as

in Fig. 6.t7

Clock times
shown in

Difference

window

Straightforward Times
Errors
Mirror reversal Times
Errors
Times
Random
Errors

'387

r'9
'54t

'67s

*'288

5'O

+ 3'r

'777

7'2

4'4

'885

I.TII

r o'o

I 5.1

*'136
* z'8
-.226

- 5.r

the numbers r-r2 (one for each lighQ in random order. The layouts for
two of the three conditions used are shown in Fig. 6.18. Subjects were
told to think of the lights as being numbered r-r2 in order from left to
right, and when any one came on to find the corresponding nunrber on
POSITION Z ( DISPLAY)

POSIIION A(OISPLAY)

ooo o oo ooo 6oo.

LIGHTS

LI 6HTS

INDEX CARD

3 FT.

3 FT.

INDEX
CARO

POSITION A(CONTROL)

POSITION Z (CONTROL)

KEYS

Figure 6.18. Layout of appararus used in Kay's (1954) experiment coIDbining spatial transposition with symbolic translation.
the card and press the key in line with this number. The task was performed with the index card in three different positions: (l) immediately
above the keys as on the left of Fig. 6.18, (z) halfway betriveen the lights
and keys and (3) immediately under the lights as on the right of Fig. 6.18.
The instruitions applied equally to all three conditions yet they
differed widely in difficulry. The average times for a run of zo responses

Economy of

Decision

r89

by the same subiects as in Kay's (r95S) experiment were (r) 65.9, e)
lo8'3 and G) tgl'r sec with average errors of 2.4r 6.9 and 2r.g respectively. Comparing these tasks with the purely spatial transpositions
involved in Kay's other experiment, we can regard condition (r) as
replacing the alignment required across the 3 ft gap in condition (b)
by a symbolic uanslation using the numbered card. Condition (3) required the same symbolic uanslation and also aligument across the gap,
and thus combined the difficulties involved in both conditions (l) and
(b). If these difficulties were simply additiver we should expect the
differences between the times and errors for conditions (a) and
$) to be
the sums of the differences between (a) and (b) and between (a) and (r).
In fact the differences between (a) and (l) far exceed these sums:
allowing for the fact that 30 instead of zo responses were made in
conditions (a) and (b), simple addition would predict a time for condition $) of 69'6 sec instead of the rg1.r sec observed, and 5.1 errors
instead of the 2r'g observed. The combined difficulties of conditions
(t) and (b) seem to have had a quite disproportionate effect on the
speed and accuracy of performance.

Relationships between controls and their effects

The princlples of compatibility benn een display and conuol dso apply

to relationships berween conuols and their effects, and both are obviously facets of the same problem. In the former the subject is conceived
as translating from what he perceives on the display to an appropriate
responding action, in the latter he has to relate action to its observed
or expected effects on a display. The evidence in this area has been
reviewed by Mitchell and Vince (lg5l), Mtrrell (tgST, 1965) and
Loveless (tg6z) so that only the general principles involved will be
outlined here. They fall broadly into three classes:
r. Arbitrary rules. Conventions such as that electrical switches move
down (ot up) for 'on' lead to an obvious economy of decision and
freedom from uncertainty and possible confusion in so far as they
apply generally over a wide range of situations. Several shrdies have
shown that many such conventions, or 'population stereogpes', become
deeply ingrained, such as that a knob needs to be nuned clocknise to
increase the intensity of a sound or light (Bradley, 1959). The conventions ffe, however, arbiuary and differ from one country to another for example the on and off positions of electric switches; or from one
context to another - for instance a water-tap is turned anti-clockwise to
increase flow. Difficulties have been noted in practical sinrations where

r9o

Fwdamentals of Skill

different conventions are inevitably mixed, such as in the engine-roouts
of ttubo-electric ships where both steam and elecuical controls are
present.

z. Seemingly'natural' linhages. Many expectations about the effects of
controls appear to assume a simple linkage benneen perception and
action similar to that of the ordinary co-ordination benreen hand and
-{>

-{>

t

l

ffi

tIB

A

(#

t
c

-4r

I

x
#E#Fc
T

-*

D

t

.-u

Figrire 6.19. Expected relationships between the movements of controls
and displays. Movements of the levers and knobs are a(pected to produce
corresponding movements of the pointers in the directions indicated by
the arrows.
eye, and several affangements of display and control seem to be easier

to operate if they are deliberately conceived by the subiea in these
rerms (Abbey , 1964). Examples of such simple linkages are that moving
a lever in a given direction is expected to move the pointer on a linear
scale in the same direction, as shown in Fig. 6.tg (A and B). Other
expectations seem to assume a simple mechanism connecting control and
display. Thus where the pointer on a linear scale is controlled by a
knob, the tendency is to assume that the pointer moves in the same
direction as the part of the knob nearest the scale, as if the two were

geared together by a rack and pinion (Fig. 6.tg C and D). Various
expectations are not always consistent with one another. A conflict
leading to confusion may occur, for example, with pointers moving on

Econonry of

Decision

r9r

circular dials. If the clochnise movement of a contro[ing knob sinrated
below the dial moves the pointer clochrise, the top of the knob and
the pointer will move in the same direction while the pointer is near
the top of the scale, but in opposite directions when the pointer is
near the bottom (Fig. 6.tg E and F). The seriousness of such conflicts
seelns to be affected by a number of factors as yet imperfealy understood. For example, Thylen (1966) has noted that clockwise rotation
of a knob placed above a linear scale as in Fig. 6.tg (G), is expected to
move the pointer to the right from a starting position at either end of
the scale but to the left from a starting position in the middle: presumably the expectation that the pointer will move as if geared to the knob
does not extend to the ends of the scale.
Expectations of these kinds dmost certainly contribute to the difficulties of operating velocity-controls where a movement of the control
lever or handwheel produces a proportional ctrange not in the position
of the pointer, but in its rate of change of position. To move ttre pointer
from one position to another it is necessary first to move the control in
one direction to start the pointer and then in the other direction to
stop it. There is still a seemingly natural linkage if the inirial movement of the control is 'compatible with the direction in which the
pointer moves, but it is not as direct as with a straightforward positional
control.
Mitchell and Vince (lgSt) noted that more intelligent subieas
were less affected by 'unnanrral' linkages. The effects of these can also

be reduced by training and tend to be less with uacking tasks where
the subject is constantly making movements, so that each can be made
with reference to the last, than when only occasional adiustments are
required. They do not, however, entirely disappear as is shown by the
faa that conftrsion from unexpeaed relationships may occasionally
occur even with well practised subieas, especially under conditions
of suess (Taylor and Garvey, 1959).
3. 'Mental models.' The linkages berneen single displays and controls
shade into more elaborate conceptual models, akin to the spatial frameworks of everyday perception (p. rT2), which enable the various parts
of a machine or industrial plant to be related together and conceived as a
unified whole. Like the perceptual frarneworks already disctrssed, the
models need not be stricdy accurate in order to be useful. The model
conceived by the operator of an industrial plant is often crtrde and
grossly inaccurate, but it enables him to co-ordinate the individual
items of his task so that they appqr less arbitrary than they otherwise
would. It seems clear that these models represent a recoding of the data

rg2

Fundamentals of Skill

provided by the plant and by observation of the effects of conffols
which is more economical than a set of rules-of-thumb in that fewer
separate instrustions have to be carried in the operator's memory. In
some industrial plants deliberate attempts are made to show the essential relationships between different parts by means of 'mimic diagrams'
such as circuit diagrams connecting meters and switches on an electrical control panel (for other exaurples see Welford, r9fua, pp. 7t 9
and z5).
The precise ways in which these 'mental models' are built, used and
maintained is not at present well understood, and the pioneering sttrdies

in this area of Crossman (196o) and of Beishon and Crawley (lg6S)
deserve to be extended.

HIGHER UNITS OF PERFORMANCE

The grouping and co-ordination of data in perception has its analogy
on the motor side in the building up of sequences of actions which
tend to become coherent 'higher units' of performance. The classical
studies are those of Bryan and Harter (t8gg) who found that as Morse
operators became more skilled they tended to pass from dealing with
single letters as trnits to syllables, words or even phrases. They suggested
that these units constinrted a 'hierarchy of habits' the levels in which
could be conceived in terms of size of unit. Much the same ideas were
put forward regarding tlpewriting in the classical monograph by Book
(r9o8). Questions of how the higher units are related to the lower and

how they come to be formed out of them were not pursued in detail,
and the application the authors made of the hierarchy princrple was
timited to the tasks they were sttrdying. The concept does, however,
seem to be of very much wider application and, indeed, to denote an

important general princlple of performance.
Even a seemingly very simple response to a signal usually requires a
complex interplay between perception and action. For example, we
noted that the superiority of Garvey and I(nowles's systems A and B
(Fig. 6.14) in which signal lights and their corresponding response
buttons were immediately adjacent was due to the fact that only one
visual inspection was required to find both light and button. In other
arrangements where lights and buttons were on different panels, the
subiect had to search the display panel to find the signal and then to
search the control panel for the appropriate button, and thus had a
rwofold visual task. fn everyday life this compound nature of performance is clearly marked, ild we may indeed argue that the simple

Economy of Decision

t93

signal-response unit is an absuaction seldom if ever found except in
reflexes and some very simple laboratory tasks.
Let us consider by way of example looking up a telephone number
and dixlling it. Looking up the number will require a series of actions,
turning over pages, running the finger down the columns, and so on.
It will involve a constant interplay betrreen receptor and effector functions, each turn of a page being made in response to information on the
page open at the time in relation to the information sought. The various
actions are all in one sense disctete signal-response units but are bound

together by the aim of finding the number required. fn this sense, they
are all included in the receptor activity aimed at obtaining information

for a larger unit of perfonnance. When this information has been
obained it is uanslated into a series of nrrns of the telephone dial which
are then made. These are again in one sense discrete sigqal-response

units each requiring the observation of a number and turning the dial.
Again, however, they are in another sense all included in the effector
activity of a larger unit of perforuunce. 'We can thus ttrink of the whole
operation as a single unit of perforrnance incorporating many smaller
units, in which both the larger and the smaller trnits are similar signalresponse units in the sense that in each, whether larger or smaller,
information is gathered and used to direct action to a specific end.
The hierarchical ordering of units of performance is perhaps better
illustrated in some industrial skills. If, for example, \ile went into a
workshop where a uran was using a lathe and internrpted his activity
at a particular instant of timer sre should find a detailed muscular action
in progress - say a trristing of the wrist to nun a handwheel on the
tool carriage. The action would, however, be only one of a series required to move the tool over the surface of the work. This again would
be only one part of the cycle of operations required to machine the
article concerned, and the article might be only one of several needed
for the job of construction on which the man was engaged. The action,
the series of actions, the cycle of operations, and the iob of constnrction
are all in a sense units of performance of a task. The larger units at each
level embrace the smaller, organisirg, co-ordinating, 'stegring' and

indeed 'driving' or motivating those which lie below. Were we to ask
the man on the lathe what he was doing we should be asking an ambiguous question because an answer in terms of any of the units would be
correct. The acnral unit he chose to grve as his answer might be expected
to depend on the level at which the outcome of his actions was least
certain. LInits higher in the hierarchy would be 'taken for granted',
those lower would have become more or less 'automatic'. If this view is

rg4

Fundammtals of Skill

correct, $re should expect the level at which awareness is centred to
rise as the operator becomes more expert and masters larger and larger
units, but that it might fall again if conditions of work, fatigue or other
factors made the performance of smaller units sufficiently difficult for
their outcome to be it .ppreciable doubt.
Where conditions require or permit virtually exact repetition of a
unit many times, performance tends to become stereot)rped in the
course of practice, and the whole cycle can be run off very much as a
chain response with each member acting as the cue for the one that
follows. Even in this case, however, the trnit seems to behave as a whole
rather than as a simple chain because it is often impossible, and almost
always difficult, if a cycle is interrupted, to begin it again in the middle
without some rehearsal of the parts already completed.
!7here conditions are more fluid, and in early stages of practice,
performance appears to be more variable. The precise sequence of subunits may differ from one performance to another in much the same
way as the precise form of an action varies with detailed circumstances.
Often the results of each sub-unit indicate what should be done subsequently, as when in looking up a telephone nurrber the names appearing
on each page of the directory opened indicate which way the Pages
should next be turned. Sometimes, however, the subiect will have no
alternative but to remember at each stage what he has done and what
still remains to be done to complete the task. Thus in dialling a telephone
nunber it is necessary to remember as each figtue is dialled what remains to be done because the dial itself gives no indication of what has
been completed. This last point focuses attention upon an important
factor implicit in the whole idea of higher units of perforlrrance and
the integration of data over time, namely short-terrn rilemory rctaining
data early in a series trntil they can be combined with later, holding
data while the decision mechanism is 'busy' and keeping a tally of
what has been done in a complex task. To this we ttun in the next
chapter.

Experimental snrdies of the hierarchical organisation of perfonnance
since those of the early pioneers have been remarkably few. We may,
however, mention three which are especially relevant and which link
our present approach with previous discussions:
(") Craik (rg+il noted that when a subiect is required to turn a handwheel at a constant rate to keep a pointer in line with a target, his intermittent corrections are superimposed on a steady rate of turning. In
other words the subject has extracted the steady rate of nrrning as a
constant in a manner analogous to the exUaction of constants in percep-

Econorcy of Decision

r95

tion: he ntrns at a rate which roughly matches the rate required and
makes periodical adiustments to this.

(D) Several studies have shown that the speed of one part of a cycle
of actions affects that of other parts so that a uniform tempo is imposed
on the whole (e.9. de Montpellier, 1945, Wehrkamp and Smith, 1952,
Simon and Simon r 1959, Simon, 196o). Perhaps the clearest illustration
of this is in an experiment by N. Welford (see Welford, 1958, pp. ro3ro5) whose subjects tapped continuously from side to side between
two targets of either r or z in diaureter set either r ot z ft apart. As
expected from the results surveyed in the previous chapter, petrrnunce
was slower when the targets were r in diameter than when they were
2 in. The interesting result in the present context was that obtained
when subiects tapped alternately benreen one r in and one z tn target.
It might have been expected that movements in the two directions
would have taken different times, those from the smaller to the larger
target taking substantially less time than those from the larger to the
smaller. fnstead, movements in both directions took about the same
intermediate length of time.
k) Control of details by broader aspects of perfonnance is further
illusuated in a series of experiments by Pew (1966) whose subiects
attempted to keep the spot on a cathode-ray nrbe in a cenual position
by operating two keys. If, say, the left-hand key was pressed, the spot
accelerated to the left. Pressing the right-hand key would cause it to
decelerate, reverse direction and accelerate to the right. Pressing the
left-hand key again would cause the target to slow down, reverse and
move off to the left once more, and so on. Accrrrate performance
demanded a rapid alternation benreen the nro keys. Pew noted that the
suategies used in dealing with this task could be divided into three types.
The simplest, but least effective, \ryas to observe the effect of pressing
each key before pressing the other: this resulted in a series of large overshoots. A second strategy was to press the keys alternately in rapid
succession, making corrections by leaving a longer interval when the
spot was seen to wander appreciably to one side or the other. This
strategy is similar to that employed by Vince's (1948b) subiects making
series of movements at rapid rates (p. r43). The third strategy again
involved pressing the keys alternately in rapid succession but adiusting
the intervals beureen pressings so that the spot remained approximately
centred the whole time: the subject essentially imposed a pattern of
timing upon his performance which ordered the individual actions so as
to maintain a uniform overall result. It should be emphasised that both
the second and third suategies required each key to be pressed before

ry6

Fundammtals of Skill

the full effects of pressing the previous key could be observed so that
performance was essentially anticrpatory and ballistic. In this respect it
resembles a number of skilled industrial tasks snrdied by Crossman
(196o) and by Beishon (tg67, Beishon and Crawley, 1965) which involved substantial time lags before controlling actions had their full
effects. In these cases the operator may have to take a series of actions
before he can observe any overall effects. l7hether one regards his
performance as open- or closed-loop depends on the scale on which it
is viewed. Each individual action may be closed-loop in the sense that
its immediate course can be observed, but open-loop in the sense that
its ultimate effects may not appear until long after it has been completed.
One further point about the rurits of performance we have been discussirg needs to be stressed. !7e have spoken of them as signal-response

trnits and this is correct in that both perceptual and effector funaions
are involved. In another sense, however, such a description is misleading. The functional trnit of performance does not typically consist
merely of percepnral processes leading to motor responses, but of
attempts by the organism to bring about modifications in the sinration
in which it finds itself. To put this in signal-response terms, we should
have to say that the unit of performance extends from a signal to a
modified signal and that response or action is merely a link benreen
these t\ilo. This way of looking at performance has two important consequences. First, it places the main emphasis on perception and decision
and thus makes the essential matrix of behaviour cognitive. Secondly
since actions merely bridge the gap between one perceptual sinration
and another, they can vary substantiatly without the functional unit of
performance having to be regarded as different: the central mechanisms
are capable of producing a range of aaions the details of which are
matched to the precise requirements of the occasion so that the same
end may be achieved in several slightly d,ifferent ways. Perhaps the most
striking evidence of this is the way that compensation can be made for
deficiencies resulting from fatigue, rge or iniury, by adiustments of the
method or manner of performance so as to shift the load away from
capacities which are impaired to those which remain intact (e.9.
Welford, 1958).

VII
Short-term Retention
It is well known that after a severe blow on the head which has produced
temporary unconsciousness, the patient's memory for events prior to
the blow is disturbed. At first he may be unable to remember anything
that happened during a substantial time before the accident, except
perhaps in a fragmentary or disordered manner. This period gradually
shrinks, the more distant memories usually renuning before the more
recent. There remains, however, a short period of a few seconds or
minutes which is pennanently lost. Similar retrograde armcsia effects
have been forurd using other agents producing violent assaults on the
brain, such as electro-convulsive therapy (ECT) (for a review see
Glickman, 196r).
Facts such as these have led to the view that learning is a two-stage
process. The material being learnt is conceived as held for a few seconds

in a short-term 'store' consisting of some kind of brain actiztity - selfregenerating circuits of neurones analogous to the dynamic memory
stores of some early computers have been suggested (Hebb , 1949). This
shofi-term retention is regarded as providing an oppornrnity for a more
endtrring memory trace to be built up in the form of either a submicroscopic change of structure, or a stable biochemical change, in
partiorlar brain cells. The more enduring uace is assumed to be weak
at first and therefore liable to be distorted or rendered unavailable by
neural noise resultir g from cerebral assaults, but to become stronger
with time. There has been considerable controversy as to whether shortand long-term retention are stages of a single process or whether they
imply two separate memory stores in the brain. Some evidence favouring the latter view is given by clinical snrdis which show that the one
may be severely impaired while the other is little affected (e.g. Symonds,
1966). Further evidence is provided by Baddeley Q966a, b) who showed
that the tlpes of error differ in the two cases: in short-term memory
they tend to be due to confusion benreen acoustically similar words
but not benreen words similar in meaning, whereas in long-term
memory the opposite tendencies appear.
r97

r98

Fundamentals of Skill

In practice, however, it is difficult to separate the two mechanisms
since they conrmonly seem to work closely together. Many of the very
substantial number of snrdies which have been made of short-term
memory, especially during the last ro years, suggest that long-term as
well as short-term stores have been participating. At the same time
short-term retention appears to play an important part in the process
of learning for long-term retention. !7e shall nevertheless for convenience treat the two separately, dealing with short-term retention in
this chapter and leaving questions of, learning and long-term retention
to Chapter 9.
BASIC FACTS OF IMMEDIATE MEMORY

Early work on short-term memory has been surlmarised by Blankenship (tgf8) and more recent studies by Posner (t963, r967a) and
Peterson (1966a). The basic facts are well established and can be
broadly sunmarised under four heads:
(o) Perhaps the most striking fact about short-term memory is ttre
limited amount of material that can be retained at any one time. The
'immediate memory span' varies according to the ctiteria adopted in
meastrring it, but Jacobs (rS87) who reported the first results in this
field, found that the maximum number of random digits heard once
that could be repeated back with complete acquacy was, on average,
ro for a group of subjeas in their late 'teens. Cardozo and Leopold
(lg6f) found the maoimum number of random digits repeated back
correctly evety time in a series of trials to be about 6. Averages for the
number coffect 50 o/o of times or for more subtle measures are ustrally
between T and 8. Corresponding figures for random letters of the

alphabet are about one less in each case.
(b) The span is reduced if the subiect shifts his attention to other
material during the period between presentation and recall. The magninrde of the effect is well shown in the results of an experiment by
Brown (1958). The subiects viewed pairs of consonants on a paper
strip which passed behind a small window at a rate of r pair per .8 sec.

The ntrmber of pairs varied from r to 4 in different trials. In one
condition the last pair was followed by 5 pairs of digits, all at the same

time intervals: the subieas read out the letters and digrts as they
appeared and immediately afterwards wrote down as many of the
letters as they could remember. A control condition was similar except
that instead of the digrts there was blank interval of the same length.
It can be seen from Table 7. r that the letter span in this control condi-

Short-tqm Retqttion

r99

TABLE 7.t Interference aith short-term ffiemory by material presmted
durtng the period of retmtiur. Data from Brown Q1SS). The figures are the
moan runnbers of letters recalled, based ut 9 trials by each of to xrtjects
Number of letters presented:
Experimental condition : letters presented,
then 5 pairs of digits
Control condition: letters presented,
then blank interval of 4.7 sec

2

68

r'95

4
2'72

2.45 2.Ol

Not

3'98

5.6r S.z3

tried

tion was benveen 5 and 6, while in the experimental condition it was
benreer z and 3: presumably the digits had interfered in some way
with the retention or recovery of the letters.
Such interference can be substantial from even a single item as, for
example, when a subject is required to say 'o' before lsgxlling a string
of digits (Conrad, r96oc, Dallett, 1960 or to dial o before a telephone
number (Conrad, 1958). The effect of such a single item is negligible
upon a string of 4 digits butbecomes appreciablewith 6 or 8(Mortenson

and Loess, 1964).
Interference may arise not only from extra items during the period of
retention but from the presentation of additional items to be recalled
(Norman, ry66)., strd it tends to be greater when the original and interpolated items are closely similar than when they are very different.
Thus interpolated letters which are phonetically similar to those being
retained produce greater effects than letters which are phonetically dissimilar (S7ickelgren, r966a, c, Dale, ry64)., strd learning a list of letters
interferes less with the retention of digits than does either learning
further digits or even merely reciting digts (Sanders, r96ra).
Loss of retention may also result if the subiect has to perform certain
kinds of recoding operation upon the data being retained and increases
as such recoding becomes more radical. For example, Posner and
Rossman (t965) showed that errors increased from a condition when
each pair of a string of 8 digits had to be repeated bachrards, through
conditions where members of each pair had to be added together, to a
condition in which each pair had to be identified as 'high' or 'low'
(above or below 50) and odd or even. Posner and Rossman argued that
the loss of retention depended on the extent to which the information
in the original pairs (about 6'6 bits each) had to be reduced in the final
answer, that is from not at all in the first condition, to reduaions to
3'8 bits for the sums and to z bits for the classification. Posner and
Konick (1966) in further experiments have indicated that the effects of
such recoding and of interfering items are independent.

zoo

Fundamentals of Skill

Similar interfering effects to those observed for digits and letters have
been shown for simple mooements by Boswell and Bilodeau (t964).
Their subjects moved a lever a few inches from left to right and z8 sec
later attempted to reproduce the same movement. The average correlation between the two movements was reduced from .84 to .72 by

requiring the subiect to pick up a pencil from the floor benreen the
first and the second, instead of waiting passively by the apparatus.
G) Short-term retention can be greatly improved by rehearsing the
material between presentation and recall. Brown (rgS8) in a furttrer
experiment using the same general technique as that already outlined,
found that leaving an interval of z-5 sec betvveen the last of the leuers
and the first of the digits substantially improved recall of the lemers
after presentation of the digits. ft was clear from his subjects' remarks
that they were going over the letters in this interval and that doing so
reduced the disruptive effect of the digits. Similarly Sanders (r96ra)
who told his subjects to rehearse during an interval after hearing 8 digits
found that retention of the digits after learning or reciting further digits
or letters was better if the rehearsal had lasted 4o sec than if it had
only lasted 12 sec - the longer period of rehearsal had increased the
resistance to interference.
The reasons for these rehearsal effects are not, however, entirely clear.

To some extent rehearsal may serve to keep the memory traces from
decaying but this cannot account for the increased resistance to interference from intervening aaivity. Brown reported that many of his
subjects made remarks which implied that they were somehow recoding
the material during rehearsal or were applying mnemonic devices such
as forming associations - he mentions one subject who associated the
Ietters ND with the words 'National Debt'. Such recoding has been

shown to improve retention (Schaub and Lindley, 1964). It seems
possible, on the other hand, that the active response to the material
imFlied in rehearsal tends to uansfer the material to the long-term
memory store where it would be much more resistant to interference
effects. There is considerable evidence that learning and long-term
retention are enhanced by active response to material (e.g. Gates, t9r1,
Belbin, 1958) and indeed it seems reasonable to suppose, as suggested
in Fig. r.3 (p. r9), that entrance to the long-term store is via a decision
about the material by the*translation mechanism which would also
tend to result in action. We shall consider this point again later and deal
with it more fully in Chapter 9. Meanwhile direct evidence about the
role of active response in short-term memory comes from experiments
by Murray (1965, 1966) who found that letters were retained better

Short-term Retention

2ot

if subiects spoke them aloud, or even mouthed them silently, as they
were presented instead of merely reading them silently. Such vocalisation dso improved resistance to interference. The complementary finding has been made by Turvey (tg6il that repeate d presmtation of the
stimuli under conditions trnfavourable to any kind of rehearsal produces
little or no improvement of retention.
While rehearsal generally aids retention, it seems that it may occasionally impair it. Heron (t962) foturd that 8-digit numbers were

dialled on a telephone dial less accurately after they had been rehearsed

vocally or written down than if they were dialled immediately after
presentation. To some extent this may merely indicate that reproduction is bemer when immediate than when delayed, even when rehearsal
takes place, or that rehearssl, while it consolidates retention may introduce and also consolidate errors. To some extent it may have been due
to the fact that Heron's subjeas rehearsed and recalled by different

methods: the spoken or written rehearsal may perhaps have coded the
material in a form unsuitable for use subsequently when fialling.
Somewhat similar indications are contained in the results of experiments
by Wickelgren (lg6S) and Margrain (tg67).
(d,) The immediate memory span can be improved to some extent
by practice. For example, Martin and Fernberger Ggzg) found that the
span for digits presented at a rate of r per sec rose by about 36% during
the course of practice involving two trials a day for 5o days. Pollack
et al. (lgSg) found even larger rises - up to tooo/o- in digrt span
between the first and eighth blocks of 15 trials, but their task differed
from the maiority in testing a running ffiemory span: subjects heard
groups of digits up to 40 in length and wrote down 'as many of the
last numbers in the group' as they could remember. Martin and Fernberger emphasise that such practice effects seem not to imply a true
inctease of capacity to hold data in short-term memory, but rather
increased skill in putting it in a form easy to retain. They reported
'grouping and organising' of the digits - in other words a form of recodirg. Pollack et a/. found much less improvement when digts were
presented at a rate of 4 per sec than at rates of 2 per sec or less, suggesting the fast rate did not glve enough time for recodirg to take place.
The problems of accounting for these several effects seem to pose
three main questions: firsdp what sets the limit to the short-term
memory span; secondly, in what units should the span be measured;
and thirdly what is the nahrre of the short-term store and where is it
located within the central mechanisms? !7e shall in the rest of this
chapter consider these questions in turn, and having done so will look

2o2

Fundamcntals of Skill

briefly at some of the ways in which short-term memory seems to
play an important part in higher mental functions such as intelligence
test performance, problem-solving, thinking and the ordering of complex skilled activity.

SOURCES OF LIMITATION

In snrdies of long-term memory it has become customary to think of
forgetting as due in part to the decay of memory traces over time and
in part to interference by other material either during the period of
retention or at the time of recall. Much of the snrdy of short-term
memory has been concerned with the question of how far the limited
span can be explained by the same principles. Fotrr main types of
theory have been proposed:

r. Decay with time
A theory in these terms was put forward by Broadbent (r9 STb) who
suggested that a subject can hold data for a limited time but can extend
retention beyond the critical period by means of rehearsal which recirculates the material through the central mechanisms and back to the
input. The length of span was thus held to depend on the amount of

material that could be recirculated before the traces had decayed beyond
repair.
Evidence which appears at first sight to favour this view is contained
in results obtained by Conrad (rgSil whose subjects listened to 8-digit
numbers at a rate of either 30 or 90 digits per min, and had to write
them down immediately at the same rates as soon as the last digit had
been presented. The speed of writing was paced by a series of clicks. Conrad fotrnd that the faster rate was superior, yielding 4r o/o correct recall
as opposed to 3zo/o for the slower rate. Similar results were obtained by
Fraser (lgS8) and Posner Q964a) who compared rates of 4o with rzo
per min and of 3o with 96 respectively. A parallel finding with nonverbal material is that of Eriksen and Johnson (t964) who found that
the acflracy with which subiects reported whether a tone had been
sounded while they were reading a novel declined with the length of
time elapsing berween presentation of the tone and the posing of the
question about whether it had occurred. On the motor side Adams and
Dijkstra (1966) found that the accuracy with which subjects reproduced
the movement of a lever fell as the interval bennreen the original movement and the attempt to reproduce it rose from 5 to 80 sec.

Short-tqm Retention

2a3

On reflection, however, this evidence is unconvincing. One would

expect that, if '.ne alone was important, halving the rate of presentation
and reproduction would halve the span. This it clearly does not do.
What is mor€r tt number of snrdies have found that memory span falls
rather than rises with increased rate of presentation. Bergstrom (lgoZ)
obtained fewer errors in word and digit spans when items were presented at half sec than at r sec intervals, and fewer still at 2 sec interyals.
Similar findirgs were obtained by McReynolds and Acker (1959) for

nonsense-syllables presented at rates of z-tz per sec. In both these
cases recall was an approximately linear ftrnstion of the logarithm of the
interval berween items. Again Pollack et al. (lgSg) found that the run-

ning memory span improved as the rate of presentation fell from 4
through z and r to '5 items per sec. Similar indications in a non-verbal
context come from an experiment by Fraisse (tg+z) who fotrnd that
erors made in tapping out rhythms which had been heard once rose
both with the nurnber of sounds in the pattern and with the speed at
which it was presented. It is, of course, true that extremely rapid rates
of presentation might give rise to problems of grouping and failtrre to
distinguish number and order discussed in Chapter 4, but the rates in

ail these experiments were far too slow for this explanation to be valid.
Looking at the problems from a diferent anglel Murdock (196o)
found the number of words learnt in a single trial to rise linearly with

the product of the number of items and the time per item - that is with

the total time taken to present the list. Again Mnrdock (r965a) fotrnd
that retention of pairs of words from a list depended upon the length
of time for which they were presented and not upon the total number
presented or upon whether the time was concentrated in a single trial
or distributed over a number of trials. Similar findings have been
reported by Waugh (rg6il who found that probability of recall was a
linear function of the time for which material was presented. It is
interesting to compare these results with those of Wallace (lgS6) who
found that probability of correct identification was a ftrnaion of the total
time for which material was viewed, more or less regardless of whether

the time was continuous or split up into several shorter periods (see
p. g7).
One difference of Conrad's (tgSil and Fraser's (lgS8) experiments

from others is that they controlled the rate of recall whereas the others
did not. We may perhaps suspect that the clicks which Conrad and
Fraser used to pace recall may have acted as a sotuce of interference
which would have been more severe at slower speeds in that the longer
the time between each click and the next, the less easy it would have

2o4

Fundamentals of Skill

been for the subjea to acquire a rhythm which would have enabled
him to ignore the clicks. Conrad and Hille (tgSil have produced evidence in line with this view by showing that errors in the immediate
reproduction of strings of 8 digits were less with urpaced than with
paced recall, and less with a fast pace of recall (9o per min) than with
a slow $o per min). Rate of presentation had little effect, although what
it had was in the direction ofthe faster rate being superior. More recently
Waugh and Norman (lg6S) have shown clearly that recall of individual
digts from lists of 16 read at rates of either 4 or r per sec was greatly
affected by the number of digits intervening betrreen presentation and
recall, but very little by the time interval concerned. Similar indications
appear in a snrdy by Conrad and Hull (1966).
Evidence regarding the effects of time on short-term memory span
are still not quite unequivocal, but it can be argued that it is virnrally
impossible to be sure that time is important as such: the longer the
time between presentation and recall, the more chance there is of random disnrrbances and disuaaions interfering with retention. As against
this, slower rates of presentation give greater oppornrnity for recoding
and rehearsal which might more than offset any effects of interference.
Some indication in favour of this view is given by Corballis (1966) who
found that retention was better at slow speeds when the material was
also exposed (visually) for long times, but that with short exposures
faster speeds were better - the relatively long gaps betrn een presentations with short exposures at slow speeds would have given the opportunity for distraction of the subject's attention. Time cannot on present
evidence be wholly excluded as a cause of limitation of the short-term
memory span, but it is clearly an insffiient cause, and may well prove
also to be unnecessary.

2. Interference by retrieval
It is an obvious possibility that iust as retention is impaired when
another task has to be done between presentation and recall, so the
acnral recall of items may impair the retention of others waiting to be
recalled. Evidence in favour of this view has been provided by Anderson
(196o) whose subjects listened to three groups of four digts and then
either immediately or at intervals ranging up to 30 sec were told which
group or groups to write down. She found that recall was most accurate
when only one group was called for and least when all three had to be
reproduced. Similar results have been reported by Howe (t965, t966)
using three groups of three consonants all of which had to be repro-

Slwrt-term ktmtion
q)

L

I

\J
v,
q)

:\
c)

,s
\

rr{

\)

\
e
L

(nNr,,|

N

rn
o
H
fr1

lt

Y,

fr
,-i
\t
r^
0\

(n t{

++

8 tEE
a -'n
ETE
goa

2,-5

rqHE HX

l.{
I

3
o
L

l-a

+

I

I

a
G'
t
IL,

o

*,
$
\l

f)

EEE

sR

ao\(n
(n

oE.Ef!-l

rno
1{

f- t{

ft{

B9

N-+
t\o

HES
U)gcl

il9

$f-

a

a

g

#,8HU'

c\l

aS

q)

Ua

L
q)

"ttr

o

G'
rL

s

*o'

tt

o

r\a

q3

q
q)
L

\)

"t
LI

GI
E

o

tr

t\a

.S

L
q)

llt

\!
sG'R

b
l.rr
q)

-,
q)

G'
t

o

.s
FI

.SL
O ,l.r
el(J
^tP

Bl$ ggi;
i EEr
€E; :BI
Ef

Lrq
?\)L
.SD d

6f

(l

i$
q)
L

st
s8

\io'o

NB

;s
r-S
as

rqI
FlP
!qq

il*

$

O\ !f,

2ci6

Fundammtak of Skill

duced immediately following presentation of the last group: the groups
were recalled in various orders and accruacy tended to increase the
earlier the group came in order of recall.
The interpretation of these results is not entirely straightforward,
however, because of earlier findings by Brown (rgS+) whose subiects
were presented with a series of arrowheads together with nro-digit
numbers, and had to reproduce either the one or the other or both which was indicated after they had been presented. Recall of the numbers was substantially poorer if the arrowheads were to be recalled
subsequently than if only the numbers were required. Similar results
were obtained using series of digits and letters. Both sets of results are
shown in Table 7.2. These results seem clearly to imply that recall of
some items rnay be impaired by the need to retain others. The magnltude of the effect of prior recall will therefore very much depend on
whether or not this effect is taken into account - as can be seen by
comparing columns E and F of Table 7.2. An analogous result has been
reported by Crowder Gg6il who found that performance at a fourchoice serial-reaction task was impaired by the need to retain strings
of 3 or 5 words during the period of the task.
We may remark in passing that this principle can also explain the
fact that material presented before items to be retained may impair
recall (Pillsbury and Sylvester, r 94o, Murdock, 196r) although it does
not always do so (Brown, 1958): its effect seems to depend on whether
it can be ignored or can be dismissed before the material to be retained
is presented. Thus Pollack et al. (rgSg) in experiments on rtrnning
memory fotrnd that length of span fell as the number of items increased
when the number to be presented before recall was demanded was not
known beforetrand, so that all had to be retained. V/hen, however, the
number to be presented was known in advance, it had no effect: subjeas could iguore all except the last few items. Again Murdock (l 9$)
and Conrad (196o) have both shown that although items become confused within lists, there is relatively little interference from previous
lists - presumably the mernory store is effectively cleared once retention
is no longer required.
What appears at first sight to be clearer evidence that the recall of
some items impairs the retention of others is given by Inglis and Caird

(rg6f) who used Broadbent's (tg14, 195il technique of presenting
simultaneous pairs of digits, one of each pair to each ear. When asked
to recall, subjects tended spontaneously to reproduce all the digits from
one ear before any from the other. The 'earful' reproduced first tended
to be strikingly more accurate than those reproduced second, and the

Short-tqm Retentinn

2oT

obvious implication seems to be that recall of the fust had impaired
retention of the second. However, ftrrther studies by Inglis and Ankus
(l g6S) showed that this was onlypartly true. They comparedthe sinration
studied by Inglis and Caird in which subjects had been free to choose
which ear's digits to reproduce first, with conditions in which they were
told the order either before or after the digits had been presented. Their
results were extensive, covering from r to 6 pairs of digits and each
decade of age from the 'teens to the 'sixties. A sample is set out in
Table 7.3, showing &at the difference berween the series reproduced
TABLE 7 -3 The effects of order of recall and of whether ornot this order was
knoam beforehand on the accuracy of reyoducing two series of digits, one
Presented to each ear. Two sa,mples of results obtained W Ingtis and Ankus
(1955). The figures are the mean numbers of digits recalled, based in each
cclse on twenty subjects, tm m"En and ten women, who perforuted
4triats in
afiich they were free to choose which'earful' to relrroduce first, 8 in which they
zuere told the order for reproduction beforehand and 8 more in which they
were told after presuttation

A

B

Subject
free to

Order of

choose

order of
recall

recall
given
beforehand

CD
Effect of
Order of not knowrecall ing recall
given order

aftefwards

beforehand

i.e. C-B
Subjects aged 3r-4o
3 digits to each ear
Series recalled first
Series recalled second

Difference
Subjects aged 2r-jo
+ digits to each ear
Series recalled first
Series recalled second

Difference

-

r'79

2.6r

z.06

'90

1,42

.89

I.5I
- l'o7

2.26

3'37

2'47

T'O4
1.22

r-2(J^

I.IO
- r'37

-2'r7

.64

-'55

-.r2

-'9O

-.ro

first and second was gteatest when the order was known beforehand
and least when it was signified afterwards. In other words, the disadvantage of not knowing the order of recall until after presentation was

2o8

Fundamentak of Skill

greater for the first series than for the second. The implication seems
clear that part of the difference when order of recall is left to the subiect
or distated in advance is due to the first series being more thoroughly
leamt - the subject employs a different strategyin the process ofacquisition. This point was made by Kay and Poulton (tgsl) in a pioneering
snrdy and has emerged to varying degrees in subsequent work. The
results of three studies are suurmarised in Table 7.4. The essential
points to note are that the differences in column A are consistently more
negative than those in column B, and that the figure in column C for
the upper of each pair of rows is more negative than that for the lower.
Furttrer evidence that dictating the order of recall before presentation

affects suatery is perhaps contained in the rather surprising result
shown in Table Tg that performance was poorest when the subiect
was left free to choose his own order of recall.
Taken together, these results suggest that, although recall of some
items may sometimes impair the retention of others substantidly, the
effect is often small and certainly not enough to explain the limitation
of the immediate memory span. Further evidence suppor ilrg the same
conclusion has been provided by Murdock (1963) and by Tulving and

Arbuckle (1966).

3. Interference dnring presentation
Both Anderson (196o) and Howe (1965, t966) who, as already mentioned, presented three groups of items to their subjeas and required
recall of one or more, found that groups presented last were recalled
best. If, as we have seen, time held in store is not important as such,
this must mean that the presentation of later items interfered with the
retention of earlier. Confirmatory evidence comes from Mackworth
$964a) who showed that accuracy of recall fell linearly with length of
list, and from experiments in which a series of paired items has been
presented and then the subject has been given one member of one pair
and asked to recall the other: accuracy of recall falls with increase in
the ntunber of pairs presented between the pair concerned and its
recall (Murdock r963a, tg66a). Similar results have been obtained using
single items: at the end of a string of items one is re-presented and the
subject is asked to recall the one which followed it (Waugh and Norman 1965). This technique has been elegantly employed by Norlnan
(rg66) who showed that rate of forsrtting depended essentially upon the
nunrber of items intervening betvyeen presentation and recall and was
little affeaed by speed of presentation, sense modality, tlpe of item or

Short-tqm Retmtim

2og

TABLE 7.4 Further effects of order ol recall and of whether or not this wcts
hnown beforeha,nd. The fi.gures are in each case the percmtage of ituns correct

A

B

C

Order of

Order of

recall
given
beforehand

recall
given
afterwards

Effect of
not knowing
recall order
beforehand

i.e. B-A

Kay €, Poultan (rg1r). Two
series, each of 4 pairs of arrows
given in each trial. Each of 3z
subiects had zo trials.

First series
Recalled before second
Recalled after second
Difference
Second series
Recalled before first
Recalled after fir-st
Difference

39

36
3o

3

-6
3o

3r

-I

45

36
39

-t7

+3

-26
-6

32

*r

Brosm QgS4). Including some
of same data as in Table 7.2.
Numbers
Recalled before arrowheads
Recalled after arrowheads
Difference
Arrowheads
Recalled before numbers
Recalled after numbers
Difference

6z

z6

r9

24

20

2

*r

-7
-4

Rabbitt (r 96 z). Series of 5
cards each bearing coloured

letter. Based on r oo trials in
each condition by each of 18
subjects.
Colours
Recalled before letters
Recalled after letters
Difference

Letters
Recalled before colours
Recalled after colours
Difference

8t

79

8

72

l6

+4

- 15

-3

83

7o

49

6z

-34

-8

- 13

+13

zro

Fundamentals of Skill

length of list, although all the faaors affected the starting level from
which forgetting began and can thus be regarded as having influenced
acquisition of the material.
Several attempts have been rnade to treat the results of this tlpe of
experiment in terms of signal-detection theory: subieas have been
asked to recall or to recognise an item as having appeared or not appeared

in a series, and to rate the confidence with which they made their
judgments. The results thus obtained have been plotted to produce the
tlpical ROC curves of the signal-detection theory model (Fig. 2.3rp. 35).
If this can be done it implies that traces are not wholly destroyed but
are in some way progressively impaired or rendered unavailable (Murdock,

r965c, t966a, Norman and WickelBren, rg65r ltrickelgren, r966d).

It is an obvious fi.rst guess that they are rendered 'noisy' as postulated
by Brown (1959). Support for this view is contained in the faa that
short-term memory may show 'reminiscence effects' in that recall may
be better a few seconds after presentation than it is immediately the
noise has died down during the interval (Crawford et al.r t966, Peterson,

t966b). One would also expect a fall of d'with forgetting. Wickelgren
(r967a), who presented two items and asked whether the second had
followed immediately on the first in the original list, indeed found that
log d' fell approximately linearly with the number of items intervening
between presentation and recall. Mrudock (r966a) however, who
presented one item from an original list and asked for recall of the itern
immediately before or after, found that although recall of early items
was less frequent, it tended to be accurate, implying a change ofB rather
than of d' . The conflict of evidence is more apparent than real because
the two authors were asking different questions. Wickelgren was asking
what was the probability of obtaining the correct response at each serial
position, whereas Mtrdock was concerned with the likelihood of each
item being attributed to its correct serial position. The results can be
reconciled if it is asstuned that the serial order of the early items became
uncertain so that more items were competing for recall or recognition
at the early positions. In terms of Fig. 2.9 (p. 4) d' would fall if each
item was viewed against the noise produced by all the competing items
together, but could remain the same if the item was viewed against its
own background noise only. In the latter case, however, a rise of P would
be required for accuracy to be maintained (p. 4T). In both cases, by
analogy with experiments showing that reaction-time rises with degree
of discrimination and of choice, it is understandable that speed of
response should be slower for items earlier in the series (Murdock,
rg66a, Morin et al., t967).

Short-term Retention

ztr

Interference effects are particularly severe when the acquisition

of new items coincides with the recall of items already in store, as
occurs in a certain type of running-memory task. The original experiment on this kind of task seems to have been that by Kay (1953, see
also Welford, 1958, pp. 248-z5o). The subject sat facing a row of
rz light bulbs with a Morse key direaly under each numbered r-rz
in order. For the fust part ofthe task one of these lights was put on
by the experimenter snd, by pressing the key immediately under it
the subject put it out and another orr. Pressing the key under this
second light made it go out and a third appear, and so on through
the whole rz in random order. This simple task caused no difficulty.
The task was then changed so that the lights came on autornatically in
random order at r'5 sec intervals, and the subject had to press the key
under each light which had just gute cff: in other words he had to work
'one back' in the series. The task was then changed again and the subject
required to work two back, then three back, then four back. The 'oneback' task was fotrnd to be easy: subiects simply moved their hands to
the key where the light was on, waited until it went out, pressed the
key and then moved to the new position of the light. On the 'two-back'
task subjects would sometimes mark the position of the intermediate
light with a fiqger of the hand they were not using, thus making the
task somewhat similar to the 'one-back', but beyond this they were
compelled, if they were to make a perfect score, to carry the intermediate light positions in some form of rtrnning short-term memory. This
they found extremely difrcult to do: every time they 'took a ntrmber
from store' in order to make a response they seemed to forget the other
numbers still in store. It seemed impossible to take one item from store
and leave others: taking one item appeared to clear the whole store.
Many subjeas aftempted to avoid this difficulty by placing a finger on
the key under the light which was on, watching several changes and
then pressing the corresponding keys while ignoring the lights which
appeared while they were doing so. This involved abandoning any
attempt to deal with more than about half the signals, but seemed often
to be the only alternative to complete inaaivity. The results are shown
in Table 7.5, which gives both the total correct responses rnade and
also scores which take account of the simplification of the task we have
iust mentioned. These latter scores gave credit for a correct response
only when the subiea both pressed the correct key and subsequently
made some response to *re light which was on at the time. Only one
subject made a near-perfect score on the 'four-back' task. His procedrue
was, whenever he nnade a response, to shout aloud the other numbers to

2rz

Fundamentals of Skill

TABLE 7.5 Results of an experiment on serial short-tetm retention by Kay
(tg[l). The figures are the percenta.ges of correct reslutses out of a total of
16 by each of 58 subjects
r' 5 sec betwem signals
4 sec between
t-back z-back j-back 4-back signals 4-back

All correct responses 95
Correct responses fol-

6l

47

35

6z

57

23

13

35

lowed by response to

light on

at the time

95

be retained including the new one for the present position of the light.
In this way he seemed to have recirculated the data and thus avoided
losing it. Several other subjects seemed to be doing something of the
same kind in a less striking manner, and most verbalised the positions
of the lights as they appeared. It is relevant to note that after the main

series the 'four-back' task was repeated but with the lights at 4 sec
intervals, and in this condition scores were much higher: presumably
the longer intervals made it easier to rehearse and recirculate the
material.
Additional experiments with Kay's apparatus carried out by Kirchner
(tgS8) have shown that this kind of running-memory task is especially
diffictrlt for older subjects. Further studies showing the beneficial
effects of longer time per item have been made using more complex but
analogous tasks (Mackworth, 1959, Mackworth and Mackworth, 1959).
Similar experiments using a simple running memory task for digrts
have also confirmed the effect of presentation rate and the faa that
spans measured in this way are substantially shorter than conventional
digrt spans (Pollack and Johnson, r963b).
The tendency by Kay's subjects to try to reduce the more difficult
tasks to simpler ones appeared also in uacking experiments by Griew
(t958a) and Poulton (1963). In both cases subjeas had to follow an
inegular wavy ink line on a paper strip. The course could be seen some
distance ahead of the tracking point, but the portion immediately ahead
was hidden by a mask. The subiect had thus to retain the pattern of the
track for the time it was passing under the mask if he was to follow
accurately. The amount of pattern to be retained was varied in ditrerent
uials by varying either the width of the mask or the speed of the track.

Short-term Retmtion

2tj

In all cases subieas tended to track earlier than they should. The
amount by which they did so was related to the time taken by the
uack to pass under the mask, being small when the time was short and
increasing very rapidly as it lengthened. The effecr was to reduce
markedly the number of swings of the uack that had to be retained - to
a maximum of about r in Griew's case and r. 5l in Poulton's.
Vince (lg+g) and especially Poulton (rgS+) have reported several
other serial reaction tasks in which manual responses were required to
visual displays and subjects were either instructed to react to items one
or more back in the series, or were forced to do so for brief periods
because the items were presented too fast for each to be dealt with
before the next arrived. Perforrnance was usually adequate when the
subject was responding to items ody one back in the series, but tended
to break down when reaction was required to items fiuther back. Performance in these last experiments has obvious similarities to looking
ahead when readirg and to copying behind dnring dictation (e.g.
Hogan, 196r). Poulton (r9S8a) has pointed out that srtrdies of reading
or dictation are difficult to link up with other work as the material used
has usually been connected prose, but it seems clear that subiects,
especially when well practised, deal with the material in relatively large
units, listening and responding to phrases or sentences rather than to
single words. The size of the group dealt with can perhaps in principle
be regarded as providing some measrue of short-term memory span.

4. Limited capacity of the store
Probably the most frequently asstuned reason for the limited span of
immediate memory is that the short-term store has a limited capacity
and that interference from subsequently presented items arises because
they push out those already in the store. If this is true, however, it is
clear that they do not do so on a strictly rotational basis of each new
item pushing out the oldest already in store. Several snrdies have shown
that when the number of items presented exceeds the immediate
memory sPo, the beginning and end of the series are recalled more
acclrrately than the middle (Fraisse, r944t Deese and Kaufman, 1957,
Murdock, 196z). IJad there been a strict rotation in the store, the last
few items should have been retained and the early ones forgotten.
It mayrhoweverrbe true that drere is a limited capacitystore operating
as part of a more complex system. The writer was alerted to this possibility some years ago by noticing in a class experiment that when
recall was allowed in any order subiects tended to produce the last few

2t4

Fundamentals of Skill

items first, and to do so correctly. After a short pause they usually
produced the first item followed by some of the intervening ones, and
these tended to show the distortions found by Bartlett (rglz) to be
typical of long-term memory.
This impression has been confirmed by the results of two experiments
by Kassum (tg6il in which he presented subjects with lists of ro
monosyllabic English words. In the first he compared the effects of
1.0

0.8

J
J

() 0.6
ut
(E

u-

o
J

6 0.4
cl
o
G.
(L

0,2

0

678910
POSITIOI.I IN LIST

Figure T.r. Relationships between position in which a word was presented

in a list and the probability of its recall. Data from Kassum (1961).
O : Recall required in order of presentation.
O : Recall permitted in any order.
Each point is based on ro trials by each of 4 subiects.

requiring subiects to recall the items in order of presentation with those
of leaving them free to recall in any order. The results are shown in
Fig. T.r.It is clear that with both types of recall the ends of the list are
better reproduced than the middle. The fact that the terminal items are
reproduced better when recall is in any order may either be a result
of interference by recall which would place these items at a disadvantage
in serial order reproductionr or may represent, as suggested earlier, a

different strategy of performance under the two conditions. If sor

however, it is relevant to ask why it should have been adopted: is there

Short-term Retention

zts

for example some advantage in recalling the terminal items first? There
does seem to have been some advantage in freedom to recall in any order
because the spans in this condition were on average about half an item
longer than with serial recall - 6.3o as opposed to 5.67 items.
Kassum's second experiment was designed to snrdy the recall of
terminal items in greater detail. Again two conditions were compared:
in one subjeas were simply asked to recall in any order, in the other
they were asked to recall the last few words first, before attempt'ng to
recall the others. With the second tlpe of insurrction, subiects tended
to produce what Kassum termed a 'final block' of items which were
correctly recalled with no omissions and no intrusions. Final blocks also
occurred with the first tlpe of insu:uction but less frequently as shown
in Table 7.6. A substantial maiority of these blocks, especially under the
second tJpe of instruction, were gtven either in order of presentation or
TABLE 7.6 Results of an experiment W Kassum Qg6il on the immediate
recall of lists of r o words

Experimental group

r5 subjects

First
Second
ro trials ro uials
Recall Last few
in any items to
order be recalled

Control group
9 subiects

First Second
ro trials ro trials
Recall in any order

first
Average number of words

per uial recalled correctly

in final block
Number of subjects producing final blocks
Percentage of trials in

6'3

6.o

6.5

3'5

4.6

3'7

rz/15

t5 i'15

8/e

42

83

34

47
24

68

79

r3
r9

r5

Average number of words

6'5

which final blocks
appeared

Percentage of final blocks
recalled in
Order of presentation
Exact reverse order
Other orders

29

6

216

Fwtdammtals of Skill

in exact reverse order. Concentration on reproducing the terminal items
first seems again to have been beneficial, lengthening the span by an
average of about half an item. There thus again seems to have been
some special feanrre attaching to the last few items presented in each
Iist. Additional evidence for this is provided by a scrutiny of words
wrongly recalled which were intrusions from [sts previously presented:
three quarters of these words $6 out of 48) came from words recalled
outside frnal blocks, although these were only about a third of the total
words recalled. The inuusions thus came from words presented relatively near the beginnings of previous lists. Craik (rg6il has noted the
same tendency.
ft seems fair to suggest with Waugh and Norman (rg6S) that the
last few items of a list are held in a short-term store of limited capacity
which works on a rotational basis so that later items tend to push out
earlier, especially under running-memory conditions. Earlier items ffiay,
however, be preserved by gening through into the long-term store.
This would be more true of the first than of subsequent items: if, as

suggested in Fig. r.3 (p. 19) the long-term store lies beyond the
translation mechanism, the first item is likely to capture the translation mechanism and block the entry of subsequent items in the manner
suggested in Chapter 4. The mechanism may be cleared at intervals,
however, dtrring a long list or one which is presented slowly, and this
may be the reason why memory span lengthens slightly with the number
of items presented (e.9. Binet and Henri, 1894, Seibel et dl., t965),
although the number of items that can be recalled from the extreme
end of a list is independent of the length of the list (!7augh 196o).
Such a model seems to be consistent with a number of other results:
(o) IUhen items are recalled in order of presentation the first items
will have to be recovered from the long-term store and may tend to
disrupt those being held in the short-term store: serial reproduction
although understandably 'natural' (Corballis, rg6il since most everyday materials are sequential, is to this extent essentially inefficient.
Evidence on this point is given by Shepard and Sheenan (tg6S) whose
subjeas reproduced strings of 8 digits of which either the first or the
last 4 were a familiar sequence and could be regarded as a single item
held in long-term memory, ffid the remainder were random and probably within the capacity of the short-term store. They found that performance was strikingly poorer when the familiar sequence came before
the random digits than when it came after. In the latter case the shortterm store could be cleared before the items were taken from the long-

term store.

Short-term

Retention

2rT

(D) The middle items of a list will obviously tend to be the least well
remembered: they are liable to be expelled from the short-term store
by later items, while they may be blocked by earlier items from entering

the long-term store.
(c) The preservation of early items is likely to be favoured by rehearsal
since this tends to transfer material to long-term store. The converse
of this may be the reason why Gibson & Raffel (1936), who snrdied
short-term retention of series of desigm, found that initial items were
not well remembered but that retention rose steadily from the begrnning
of the series to the end: designs are obviously less easy to rehearse than
verbal material.
,.Looking back over the four theories we have discussed, the present
position seems to be that time in store is of little importance as such
although it may be very important in the oppornrnity it gtves for, on
the one hand rehearsal and recoding, snd on the other for disrupting
influences such as shifts of attention. Interference resulting from recall
and from the presentation of new material to which attention has to be
paid are both influential in limiting the immediate memory spo, the
latter probably more than the former. Theories of limited capacity ue
at present somewhat speculative but, especially in the form put forward
here in which both short- and long-term stores are involved, seem able
to account for a range of facts otherwise difficult to explarn.

MEASURING THE CAPACITY OF THE STORE

The question was raised by Miller (1956) of whether the limitation of'
short-term memory was in terms of information in the informationtheory sense: if the span for digits was betrreen l and 8 this capacity
would be benreen T nd 8 tiures log, ro - i.e. about 25 bits. The question and the answers given to it preceded most of the research outlined
in the previous section and assumed a simple limited-capacity theory:
they do therefore not at present link fully with other results. We shall
discuss here what has been done and then suggest tentative links with
the work we have dready surveyed.
Miller concluded after snrdying the evidence then available that the
capacity of the store should not be measured in terms of information,
but that rather a maximum number of items, or as he termed them
'chunks', could be stored. His treatment was important because it
called attention to the fact that short-term memory deals not onty with
single letters or digits but with coded items such as words or syllables
as its units. We have noted in the previous chapter (p. rTil that the

zr8

Fundammtals of Skill

immediate memory span can often be substantially increased by recoding
items such as binary digits into larger nnits such as decimal digits, so
that the number of items to be stored is reduced. However, the evidence

on which Miller based his conclusion was derived from studies in
which items from different sizes of set drawn from larger sets were presented and these, as subsequent sflrdies have shown, ffisy be misleading.
We have already seen in Chapter 3 that it seems to be no easier to deal
with a selection of letters than with the whole alphabet, although performance is slightly better when the subset is a very familiar one such
as ABC. The same is true of digits. fn both cases the subject appqrs to
behave as if he was all the time dealing with the whole alphabet or
whole set of ro digits.
Crossman (1961) attempted to overcome this difficulty by using
familiar sets of different sizes as shown in the first column of Table 7.7.
It can be argued that {.s.d. and N.S.E.![. constinrte subsets of letters
which are sufficiently familiar for them to be treated as separate from
the whole alphabet. Items from each set were presented verbally in
random order at a rate of about r per sec and recall was required in the
order ofpresentation. Ifthe immediate memory span depended straightforwardly on the information contained in the items concerned, the
information per item multiplied by the number of items in the span
should be constant. It can be seen from column 5 of Table T.T that
TABLE 7.7. Item, and order information'in mentory spans for different size
sets according to Eq. 7.r . Data from Crossman (t 96 t). Each figure is based on
5 trials by each of 6 subjects

I

Set

Black/White

d.s.d.
N.S.E.!tr.
Days
Digits

Months
Alphabet
States of USA
Playing cards
Dates of year

2345

6

7

Number Infor- Average Item
Total
Order
of alter- mation length inforinforinfornatives per item of span mation mation mation
(bits)
(bits)
(bits)
in set (bits)
r'oo
r'58

9'r

9,7

t 8.8

27'9

3

6'9

ro'9

I I.9

22.8

4

2'OO

6.8

t3'6

r r'8

7

z'8r

5'7

r 6'o

8.6

25'4
24'6

ro

3'32

7'3

24'2

I3.I

5'4

t9'4

7'7

27'l

z6

3'58
4'7o

6.8

3r'9

r r.8

49

5'6t

4.1

23'o

4'9

43'7
27'9

3.6
2.8

27'5

z

rz

Sz

5'7o,

3'4

19'4

365

8'5 r

2.9

24'7

37'3

23'O

Shmt-terrn Retention

2rg

this is not so: the spans for the smaller sets are shorter than would be
expected.

Casting arotrnd for a reason why there should be this discrepecy,
Crossman was suuck by the large number of cases in which errors
occurred because the correct items were produced but in the wrong
order, and suggested that in a typical short-term memory experiment
the retention of items and of the order in which they were given are
separate aspects of the information to be stored. He proposed that the
order information for a span of m items can be expressed as the logarithm of the number of possible permutations of these items, that is
m(m
- t)(m - z) o . . r, expressed as m!, and that rf n is the size of
set from which the items are drawDr we can express the capacity of the
store by the equation:

mlogn * logm! - consrant

(7.t)

The final coltrmn of Table T.T indicates that this is roughly so for
Crossman's data except with digits and letters where the figures are
much too high. One possible reason is that these sets are more familiar
than any of the others. There is a good deal of evidence to show that
immediate memory is better with familiar than with relatively unfamiliar
material. Thus Dale and Gregory (1966) obtained more acctuare recall
of lists of high-frequency than of lower-frequency words, and Postman
et al. (rg6+) fotrnd that running-memory spans were longer for more
familiar words, at least early in practice. Again Korn and Lindley (lg6l)
found that the recall of consonants improved both with their frequency
in written English, and with their frequency of presenrarion in the
panicular experiment, and Kasstrm (tg6il fotrnd that with English
speaking subjeas the number out of ro French monosytlabic words
recalled accurately was only about I of the ntrmber of English monosyllables similarly recalled. The more familiar items are perhaps more
easily coded on presentation or may in some way be more readily
'available' at recall.
An alternative explanation of why digits and letters yielded longer
spans than were expected in Crossman's experiment is that these particular items lent themselves to some recoding. For example, suppose
subiects took the digits in pairs so that a span of T would consist of three
pairs each drawn from a set of roo, plus a single digit drawn from a set
of lo, instead of seven individual items each drawn from a set of ro.

In this case Eq. 7.r would yield
3 log, roo

* log, ro + log, 4l : 27.8

2zo

Fundamentals of Skill

which is comparable with the figure for other sets. The recodirg does
not reduce the information in the items, but markedly reduces dre
order-information from 1og, 7! to logr4!. In the same way it might be
argued that strings of random letters would be likely to throw up
occasional pronounceable syllables which could be treated as single
units and so reduce order-information.
Crossman's formulation has been attacked on two main grounds. One
is that span is affected not only by the factors he considered but by the
discriminability of the items from one another (Conrad, r964a). For
example, several experiments have shown that letters are less well retained when drawn from an acoustically similar set such as BCDGPTV
than from a more readily distir.guishable set such as FHLMQSX
(Dale, 1964, Conrad and Hull, 1964, Conrad et dl., 1965). Similar
difficulty with acoustically confusable words has also been demonstrated
(Conrad, 1963, Baddeley, r966a). This criticism is fair in the sense
that information analysis, at least in the simple form of Eq. T.Trassumes
discriminability to be held constant. Sfhile, however, it calls attention
to an additional complicating factor it does not invalidate Crossman's
treatment as far as it goes. 'W'e may note that Sfickelgren (r965b) found
acoustic similarity to increase errors of order but not to result in wrong
items being recalled: indeed Conrad (rg6S) suggested that acoustic
similarity is capable of accounting for all the errors of order observed
in letter spans, although McNicol (1967) has subsequently calculated
that this is an exaggeration.
The second criticism of Eq. 7.r is that it counts the order-information
twice in the sense that order-information is already contained in mlogn.
For example, the rorooorooo possible different spans formed by I decimal digits will include all possible permutations of any Tt ot in other
words any 7-digt number can be specified as one out of ro7 possibilites.
For this criticism to be valid it is necessary to assume either that as
each item arrives, its serial position is automatically preserved, as it
would be if the store were conceived as analogous to a tube in which
items were packed one after the other with no possibility of their
exchanging positions afterwards, or else the subject must be able to
observe the whole of the items simultaneously in order to code them
with maximum efficiency. The many errors of order which occur seem
to rule out the first possibility. The second might be true with visual
presentation of the items all at once but with serial presentation it
could not occur unless the items were already stored for inspection. In
this case it would assume the storage we are seeking to explain. Similar
arguments apply to the possible saving of order-information when two

Short-term Retmtion

z2r

or more items in a span are the same: with serial presentation the
subiect cannot know in advance which these items will be and so cannot
neglect to record their order. The fact that identical items could be
interchanged might, however, save some errors during the process of
recall.
Crossman proposed that order is not autornatically preserved so that
inforrration has to be recorded not only about each item but also about
the serial position in which it comes. He suggested that the shorr-term
memory store is capable of holding only a limited total of such item

and order information together, so that the two are in a sense interchangeable. This would tend to level up the spans for different size
sets since, although the small sets would take less 'space' per item in the
store and would thus produce longer spans, any lengfiening would
rapidly increase the space required for order-information.
Two other approaches to the storage of data about order deserve to be
mentioned. First, $Tickelgren (t964, t967b) presented subjects with
strings of digits at r sec intervals with instructions to rehearse silently
in groups of rr 21 31 4 or 5. Recall was found to be best after rehearsal
in groups of 3, the superiority being more in the accuracy of order and
position of items than in the recall of the items as such. The effectiveness of the grouping was shown by the faa that most errors of position
were either within the correct group or benueen sirnilar positions in
different groups. \[ickelgren suggested that items tended to be 'tagged'
with labels indicating 'beginning', 'middle' or 'end' in a hierarchical
fashionr so that for a list of, say, 9 items this taggng would be in two
stages indicating first within which third of the total list the item lry,
and then in which position of the third concerned. This approach can
be brought within the broad informational scheme because it bears a
striking resemblance to the second of the serial classification models
proposed in connection with choice-reactions in Chapter 3 (p. T3).
In this model the subiea is assumed to divide the material into groups
and to inspect the material in each group seriallyr so that on average he
finds what he wants within a group in (n + t)/z inspections when n is
the number of items in the group. His optimum strategy is to divide
the material into thirds, find the required third and then divide this
into thirds, and so on. Thus, suppose that to find an item for recall he
had to search among a string of 9, he would take 2 inspections to
decide whether it was in the first, middle or last group of 3 and then a
further 2 inspeaions to decide which member it was of the group
chosen, making a total of 4. Tf, however, he grouped in twos plus one
single item, it would on average take approximately 3 inspeaions to

222

Fundamentals of Skill

decide to which of the resulting 5 groups the item belonged, and then a

firrther r.5 inspections with 4 of these groups -that is with S/g of the
items - making an average total of 433. If alternatively he grouped into
4 x 5, it would take an average of approximately r'5 inspeaions to
identify the correct group, followed in +/g cases by z'5 inspeaions and

in /g cases by j inspections to identify the precise member of the
S

Broup, making an average total of 4'28.
0.20

0.t
a

ln
G,

o
(r

g.
LU

LL

o
0

o
zI
E,
o
(L
o
0. 05
E,
(L

t
0
0

5

l0

CONFUSION UNI TS

l5

Figure 7.2. Relationship between probability of correct recall and the
confusability of the serial position in which an item was presented, with
other positions. Confusability is plotted in terms of Crossman's Confusion Function (Eq. 2.7). Data obtained by Ryan (rg6il plotted by
McNicol (1967).
O - First 5 items in list of 9 with confusability calculated from beginning of list.

O - Last 4 items with confusabilty calculated from end of list.
Each point is based on zo trials by each of zo subjects.
The second alternative to Crossman's approach is that of McNicol
(tg6il who separated order- and item-information by presenting strings
of digits aurally and then showing the digits visually in cardinal order,
so that the subject was in no doubt about u;kichitems had been presented

and had merely to recall their order. He noted that accuracy of positioning the first few items in the string could be accounted for by
assuming that each item was tagged with a 'quantity' linearly related
to its serial position, and that the subject discriminated benreen these

Short-terrn Retmtion

223

quantities according to Crossman's Confusion Function Model (Eq.2.7,

p. 2g). Thus the first item was beuer discriminated from the second

than the second was from the third, and so on as one would expect if
the subiect was discriminating betrreen r : 2, 2: 3 and so on. The total
errors of each position could be accotrnted for in terms of the totd
discriminability of that position from all the others. McNicol fotrnd
that considering discrimination in this way did not account for the
errors of position in the last few items of a string, but that these were
well fitted if a similar process was envisaged as taking place with the
end of the string as its base, discriminating the difference of each tag
from the magninrde of the tag attached to the last item in the string.
Data ueated in this way are shown in Fig. 7.2, and support the
idea that discrimination of order takes place from either end of the
suing, whichever is nearer to the item concerned.
It/hether or not any of these models are fotrnd to be futly acceptable,
Crossman's insistence that a more subtle measure of storage than discrete items is required, and the various indications that order of items
has to be considered, are both important contributions to the theory of
short-term memory. Their validity depends, h the first case, on whether
size of set from which items are drawn does in fact affect immediate
memory spe, and on the other, whether various devices for saving
order-information are effective. W'e shdl consider the evidence about
both these questions in tunr.

Effea of size of set
Crossman's data provide good pri*o facie evidence that span does
diminish as the size of set from which items are drawn increases,
although some doubt might perhaps be cast on his results by arguing
that playrng cards and dates of the year represent two items each ntrmber and suit or day and month. If so, the spans for these items
should be regarded as trrice those gtven in TableT.T. The same would
apply to some of the states in the USA such as New Hampshire and
South Dakota. On this t)?e of reckoning all the observed spans would
lie betrn een a little more than 5 and a little less than 7 items orcept for
digits and for Blacv$fhite. The latter might perhaps be recoded more
easily than other types of item by methods mentioned in the previous
chapter (p. r7g).
Further studies on the effects of size of set are not always easy to
interpret. We have already noted that selections of letters or digits are

224

Fundamentals of Skill

unlikely to grve very different results from the whole alphabet or set of
ro digits, and that studies using such selections do not constinrte evidence against the view that span is related to set size (Pollack, r953b,
Pollack et al., t959, Pollack and Johnson, ry63b, Conrad and Hull,
1964, Baddeley et al., T965, Conrad et al., 1965r Ifoodhead, 1966). In
two cases longer spans have been obtained with larger sets. However, itr
one of these the effect was small (Crannell and Parrish, t95T)., and in
the other, where it was large, the task was of the running type (Woodhead, ry66): subjeas had to report digits 3-back in a series drawn from
either 2, 6 or all ro digits. Accuracy of recall in this type of task tends,
as we have seen, to be poor, and it is possible that the larger sets
might have conferred an advantage by making the different members of
the series more discriminable from one another.
V/arrington et al. (1966) report that in an experiment in which the
items were inclined and curved lines, memory span decreased markedly
as the number of different types of line increased. All the lines in a
given presentation were, however, shown together and the presentations
with fewer tJpes of line tended to be easier to group and to show more
symmetry. The effect may therefore be one of perceptual coding rather

than of memory span.
Probably the best authenticated difference of span associated with
size of set is the longer span obtained with digrts than with letters
(e.9. Jacobs, 1887, Cramell and Parrish, rg11rWamington et al., t966),
although the evidence is not quite unequivocal: for example Cardozo
and Leopold (1963) found that the span for digits was the same as that
for letters when the items were presented visually all at once, although
it was longer with auditory presentation: in the latter case the spans
contained equal item information.
Further evidence indicating that memory span is affected by set size
is given by Crannell and Parrish (rgSil who found that the spans for

random 3-letter words were shorter than for single letters, and by
Lloyd et Al. (196o) who found that errors tended to be greater when
random words were drawn from large than from small sets.
A different method of varying set size was used by Harrison Gg6l)
who presented strings of 4 items consisting of the words 'Black' or
'White' with each word varying in one or more of certain other ways:
they might be presented to either the left or the right ear, in either a
man's or a woman's voice, and with the voice either clear or distorted,
so that each item contained eithet 2t 3 or 4 binary attributes giving a
set of 4, 8 or 16 possibilities respectively. Recall became substantially
poorer as the number of attributes per item rose. Harrison suggested

Short-terrn Retention

2zs

that the different dimensions behaved as separate, additive items within
- a view which as far as it goes is consistent \dth
Crossman's. We may note that adding dimensions did not extend
capacity for retention as it does for discrimination (p. +l): the different
dimensions appear all to use the same store.

a fixed overall capacity

Relationships between successive items
Several lines of snrdy have shown clearly that memory span is related
to the probability with which one item follows another. Thus Conrad
et al. (rg6S) and Baddeley et al. (1965) have shown that accuracy in the

recall of suings of lemers rises as the uansition between each letter and
the next approximates to that fotrnd in normal printed English. Again,
several experiments have found that the memory span for words increases as the sequence changes from strialy random and comes to
approximate to that of normal English (Miller and Selfridge, r95o,
Marks and ]ack, rg52rDeese and Kaufman, 1957, Richardson and Voss,
196o). The span is still larger if conneaed passages of prose are used
and reprodustion of ideas is required rather than literal recall (Zangwill,
1956, Deese and Kaufman, 1957). In line with these results is Lawson's
(1961) finding that subieas look further ahead in readi.g as the
sequence of words becomes less random and more like normal prose.
Lachrnan and Tuttle (lg6S) have attempted to localise these sequential effects as benreen accruacy of ptrce1t'ion, stwage and recall. They
fotrnd that the poorer recall of more random material and the better
recall of that which approximated to normal English still occurred when
subieas had to read the material aloud as it was presented, thus providing proof that it had been accruately perceioed.In a second experiment
they excluded the possibility that high approximations to English are easier
to reconstruct at recallrby comparing recall with recognition of individual words presented one at a time amongst other words: the material
which more closely approximated to English was both better recognised
and better recalled. The authors therefore concluded that the effect lay
in the storage processes. The distinction between the retention and
recall phases is, however, extremely difficult to draw. Given that a
trace has to be laid down, preserved and then recovered, the better the
trace has been preserved, the easier it is likely to be to distinguish from
other traces and thus the easier to recover accurately for recall: at the
same time any difficulty inherent in the process of recovery will be
accentuated by any deficiency in the traces to be recovered.
If so, we may appropriately consider effects of the probability with

226

Fundamentals of Skill

which one item follows another in the light of the 'cerebral dictionary'
concept put forward by Treisman and discussed in Chapter 3 (p. lo3):
if each item partially activates others which are similar to it or associated
with it in some wsy, recall will be facilitated if these similarities and
associations are in line with the material presented, and hindered if they
are not. The same explanation can be advanced to account for the confusability of some items such as those which are acoustically similar.
It can also account for a number of other results which are otherwise
difficult to understand:
For example, Broadbent and Gregory (r g6+) notecl that lists of alternate digits and letters were less easily recalled than lists consisting of
all digits or all letters, but that the difference became less as the rate of
presentation became slower. This finding resembles that of Bertelson
(p. 8o) who found that reaction times for repeated responses rose as
the interval betrreen signals increased: we may think of the item or
response as facilitating for a brief period the production of those like
it or in the same class and as acting as 'noise' to others in different
classes. The effect of alternation between classes of item has been
fnrther shown by Warrington et al. (1966) who fotrnd that although
sequences of letters were less well recalled than sequences of diglts they
were better recalled than mixed sequences of letters and digits. Recall
in the latter case fell as the number of transitions from letters to digits
or vice versa rose. In line with these results are those of \ilTickelgren
(lg66b) who found that pairs of digits presented once and followed by a
task desigued to interfere with retention of them, were recognised more
accurately if they were both the same or in ascending or descending
sequence (e.g. 55 or 43). Also in line are the findings of Cohen (r963)that
when recall of a list of words is permitted in any order those which are
in some way related tend to cluster together in recall: the production of

one item tends to facilitate the production of others related to it.
Ftuther support for this last point is provided by Murdock and vom
Saal Gg6il who found that the items in lists of words were better
retained if they all came from the same category, although the order in
which they were recalled might be less accurate.
Looking over the evidence we have surveyed in this section it seems
clear that both size of set from which items are drawn and the relation
between one item and another affect memory span in the kind of ways
that Crossman's approach requires. Qualitatively, therefore, the idea
is well justified that information rather than items are stored and that
item- and order-information need to be distinguished. It is also clear
that Eq. T.r can be made to give a fair account of Crossman's data -

Short-term Retmtion

z2T

which is by far the most complete yet available for this purpose - but
that the alternative formulation proposed by McNicol to accotrnt for
order calls its precise terms into question. Diffictrlty also arises for
Eq. 7.t from the fact that the increase of span when recall is allowed in
any order instead of being required in strict sequence seerns to be far
less than would be expected if no order-information was being retained.

This last ctifficulty might, perhaps, be overcome by assuming that subieas inevitably record order to some extent, as Kassum's results suggest.
A possible synthesis with the ideas put forward at the end of the
preceding section of this chapter is to posnrlate a short-term store
deating with the last few items presented and preserving order relatively
efficientlS provided there is no interference by subsequently presented
material or recall of earlier material. This, as suggested, might work in
coniunction with a long-term store where order is less well preserved.
Decision as to whether or not these ideas are correct must wait upon
future work. Further progress seems to demand a much more detailed
examination of items recalled and of the order of recall with different
types of material and under different conditions of presentation and
recall.

LOCATING THE STORE
Broadbent (e.g. r954r rgs7a) originated a series of experiments in which
pairs of digits were played simultaneously, one to each ear, and found a
marked tendency for subiects to recall all the digrts from one ear then
all from the other rather than in pairs as presented, thus digrts presented:

Right ear 7

2

5

r

8

3

Left ear

would be recalled 725183 rather than 7l:816. On the basis of these
experiments he posnrlated that short-term memory storage is located
between the sense organs and the central data-processing mechanisms.
He assumed that each ear is associated with a separate store and that
when recallingt subiect takes all the material from one store before
"
any from the other. He recognised, however, that part of the material,
say from one ear, might not be held in such a peripheral store but could
be passed direct to the central mechanisms which would thus participate

in short-term memory.
The view that the short-term store holds unprocessed, 'raw' sensory
data as opposed to the processed data retained in long-term memory has

2zB

Fundamsttals of Skill

an obvious elegance and is at first sight supported by several lines of
evidence. On close scrutiny, however, these appear less convincing. We
shall consider three lines upon which discussion has centred:
(") It is a matter of courmon experience that a remark may be heard
in conversation but that at the time the listener is unable to understand
what has been said. He does, nevertheless, retain the sounds of the
words he has heard and may come to reco$use their meaning later. The
retention of the sounds in these circumstances ffisy, however, imply
merely that the data are retained in this form of coding only so long as a
more efficient form cannot be achieved. Some support for this view is
contained in Clark's (tg6S) finding that simple nonsense designs presented once were recognised equally well whether or not the subject
named them, whereas complex designs were better retained when
named - the capacity of the store is enough to cope with the raw data
of simple designs but not of complex. The latter require a more efficient
coding if they are to be retained.
(b) Murdock (1966b) showed that the retention of words presented
visually follows a somewhat different pattern from that of words presented aurally, and argued that this implies different stores with different
characteristics for eye and ear, since if both merely fed into one central
store, retention should be the same in both cases. McGhi e et al. (t965)
in a somewhat similar experiment had found digits presented visually
to be much less well retained than those presented aurally and suggested that subjects tend to recode visual material in auditory form by
some kind of rehearsal because it is more difficult to retain in visual
form. This indeed seems reasonable in the sense that visual patterns of
digits, and certainly of words, could be regarded as more complex and
so as requiring more storage 'space' than their auditory equivalents. If
so, the difference between visually and aurally presented material may
result from the recoding of the former. Murdock encouraged such recoding by requiring his subjects to repeat the material as it was presented
and found that visual presentation produced better recall of early items
in a list, as might be expected if the process of recoding had tended to
transfer the material to long-term store. The importance of such recoding was confirmed by further experiments in which the material was
not repeated as it arrived: in these visual presentation yielded poorer
results than aural at all positions in the list except the last where recall in
both cases was near perfect (Murdock, tg6l).
(r) Several experiments in which a substantial amount of material is
flashed on a screen all at once for a fraction of a second have shown that
although the subject cannot reproduce it all, he can nevertheless produce

Short-tqm Retention

229

any one part of it if this part is indicated to him within a half second or
so after the presentation (Sperli''g, t96o, Averbach and Sperting, tg6r,
Klemmer, t96t, Mackrrorth, t964a, Eriksen and Lappin, ry67). This
obviously implies that the whole of the rnaterial has been stored briefly,
probably in the form of some visual after-effect, but that the effect
ceases before there is time to read off more than a part of the information it contains. Further evidence in favour of such a system is given by
Michon (tg6+) who found that accuracy of reproduction was similar
whether the material was presented as a whole or in groups, provided
the total time over which presentation took place remained the same.
This very brief visual storage is far too ephemeral to account for the
short-term retention we have been discrrssing hitherto. Sperling (lg6f)
suggested, however, that the process of reading off recoded the material
in auditory form and that this was more durable. Strong evidence that

auditory recoding of visual material tends to occur automatically is
contained in the finding that items which are acoustically similar - such
as the letters B and V - are often confused even though they are presented visually (Conrad, r96zb, t963), and that the frequency of such
confusions increases ifthe letters are deliberately mouthed or spoken as

they are presented (Murray, 965, tg66). Items which sound alike are
also more easily confused than those which merely look alike when
written, such as the words THROUGH and COUGH (Baddel.y,
1966b). These results do not necessarily imply, however, that the material is retained in a peripheral auditory store, but merely that auditory
coding is preferred to visual, perhaps as we have already noted because it

is more efficientr or perhaps because with material such as digits or
letters the auditory form is more generic. For example, THREE, 3 and
III are all pronounced 'three', and F, f and fi are all pronounced 'eff'.
Fnrthermore, some effors in short-term retention are due to confusion
of items similar in meaning (Baddel.y, r964b, r966b, Dale and Gregory,
t966), implying that storage is cenual rather than periphersl, although
it is hard to say how far this indicates a central short-term store, and
how far it results from material having got through to the long-term
store: we mentioned at the beginning of this chapter that Baddeley
(t966a) found similarity of meaning to be a greater cause of error than
similarity of sound in long-term memory.
Arguments in favour of a cenual short-terrn store have also been
based on the results of experiments usirg Broadbent's technique of
presenting pairs of items simultaneously to the two ears. Moray (196o)
noted that when subjects recalled ear by ear many errors were due to
items from the wrong ear being substinrted, and argued that such errors

23o

Fundarumtals of Skill

could not occur if there was a separate store attached to each ear and
subieas rqproduced all the material from one before any from the other.
The alternative is either to assume that part of the material can be taken
from one store then part from the other alternately, or more simply
that the material from both ears is stored cenually and is distinguishable
by being 'tagged' with the ear from which it has come: some such
tagging would be necessary if the material presented simultaneously to
the mvo ears was not to be fused into a single combined sound. Such
experiments as Kay's (tg6f) (p. zrt) which indicate that it is very

difficult to take one item from store while leaving others tends to

argue against the first view and, by implication, in favour of the second.
More direct evidence has been provided by Gray and Wedderburn
(tg6o) who presented three pairs of items to the two ears thus:

ear CYC- T
ear r
LO-

Right

STYLE

Left

3

and found that recall by class of material - that is 'CYCLOSTYLE
rT3' was as good as ear-by-ear recall. Broadbent and Gregory Gg6+)
using the same method with digits and letters or digits and names of
colours alternated between ears also found that recall by class was as
good as ear-by-ear. Yntema and Trask (rg6f) using digits and words
found recall better when made class-by-class than ear-by-ear, which
seems an unlikely result if subjects had been switching rapidly betrreen
stores for the different ears. Again Emmerick et al. (rg6S) who presented

three pairs of words, obtained more accurate recall in the pattern

RLRLRL than RRRLLL when the former made a meaningful sentence
even when the words to the two ears were spoken in different voices.
These effects seem to depend upon the association between items in
successive pairs and not to occur within individual pairs. Thus Bartz
et al. (tg6il who presented separate halves of words thus
Right ear

AB-

EI-

Left ear

LE

THER STRAINT

Right ear

FOOT- HAIR- MOON-

Left ear

BALL

CON-

or

CUT

GLOV/

found recall tended to occur ear-by-ear.
As regards vision, Sampson and Spong (l96la, b) and Sampson

Short-tqm Retmtion
23r
(tg6+) have shown that when digits are presented simultaneously to the
two eyes recall in pairs rather than eye-by-eye is the rule. It is perhaps
understandable that this should be so and that the two eyes should be
regarded as feeding into a common store since they work so closely
together, yet the close integration of data from the trro ears in achieving
auditory localisation iugues that vision and hearing are not very different
as regards the combination of data from the two paired sense organs,
and that principles of storage might reasonably be the same for both.

As benreen vision and hearing, Broadbent and Gregory (1961) found
that digits presented alternately to eye and ear were recalled better
sense-by-sense than in acnral order of arrival. We may suppose that
they were inevitably tagged with the sense from which they had come
so that this method of recall does not necessarily favour a theory of
separate stores for eye and ear but can be accounted for in terms of
recall being easier class-by-class as fotrnd by Yntema and Trask.
Sperling (tg6il in a revision of his former (1963) theory has suggested
that material is rapidly read from a peripheral store capable of holding
it for up to about '5; sec into a central store capable of holdi.g it for a
somewhat longer period - long enough for it to grve rise to overr or
covert motor acts of rehearsal. The transfer of material to this central
store is asstrmed to be very rapid: Sperling suggests about 3 letters in
50 msec - say about r4o bits per sec. The faa that such transfer does
not necessarily i*ply auditory recoding appears to follow from results
obtained by Parks (lg6S), who exposed designs on a band which moved
horizontally past a fixed narrow vertical slit at a rate such that the whole
design was seen in '25 to '5 sec. Under these conditions the whole
desigu was seen simultaneously in the vicinity of the slit. This result
implies some short-term storage of the early parts of the design in
visual form trntil the later parts have arrived. One obvious possible
explattation was that the storage was on the retina and that the eye
tracked in the direction of the movement of the band, so projecting the
slit on to different retinal areas. Parks ruled this out, however, by showing that the same results could be obtained for two designs presented
simultaneously on bands moving in opposite directions. He concluded
that the storage must be central. The extent of uansfer from the periphery to such a central store would depend on both peripheral and
central factors: on the peripheral signals lasting long enough and not
being desuoyed by the masking effects of signals occtrrring immediately
afterwards, and on the cenmal store not being overloaded by too much
data (Lawrence and La Berge, 1956, Eriksen and Stefri, 1964, Steff
and Eriksen, 1965).

z3z

Fundamentals of Skill

This model bears a striking resemblance to that proposed to account

for single-channel effects in Chapter 4, where it was suggested that

there is a short-term store located between the percepnral and translation mechanisms, and that data from signals can be stored there until
the translation mechanism is ready to use them as a basis for action
(p. ro9, see also Fig. r.3, p. r9). We have already seen in Chapter 3
(p. 86) that the percepnral mechanism appears to operate at a rate
which is of the right order to fit in with the rate of uansfer from peri-

pheral to central storage proposed by Sperling.
We may therefore sum up the evidence regarding the location of the
short-term store by saying that it seems necesssry, and probably sufficient, to posnrlate three processes:
(i) Very brief peripheral storage holding data over periods of less

than about r sec.
(ii) Cenual short-term storage holding data over periods of several
seconds. Data thus held have been processed in the sense that they have
passed through the percepnral mechanisffir but they have not passed
through the translation mechanisrr - they have been perceived but
not responded to. Data in this store are extremely vulnerable to interference from other material coming after or to any shift in the subiect's
attention. The results obtained by Kay (rgSf) and others (p. zrz)
suggest that this store tends to be emptied completely if any item is
taken from it.
(iii) Recirculation of the data from the short-term store through the
translation mechanism not only prolongs the period of retention by
'rewriting' the traces in the short-term store itselt but also tends to
pass material to long-term storage. It is understandable that the digits,
letters, syllables and words that have been cornmonly used in shortterm memory experiments have tended to be recirculated and stored
with an auditory coding, but other codings are also possible.
SHORT-TERM RETENTION AS A FACTOR IN COMPLEX PERFORMANCE

]acobs (1887), who measured digit and letter spans for the guls in a
North London school, found that average spans increased from the
lower to the higher forms in the school and tended to be higherumon!
the abler pupils in each form. On the basis of these restrlts, which are
set out in Table 7.8, he suggested: "'span of apprehension" should be
an important factor in determining mental grasp, and its determination
one of the tests of mental capacity'. Digit span has indeed been incor-

Shut-term Retuttion

43

porated into a number of intelligence tests such as the Wechsler Adult
Intelligence Scale. We may suspect, however, that it enters into many
other intellecnral tasks besides. IUe mentioned in the previous chapter
TABr-Y 7 .8 Digit and letter spans attatned by London schoolg:irls. Datafrom
Jacobs (1887). Figures are the aaerages of the highest spcm attaincd by each

indioidual in tuto sets of trtals,

Digits

Class

Letters

Top t of Bottom $ Top * of Bottom $ Mean of

class of class class of class Means

vr.

ro'5

Upper V.

9'r

9'o

9'8
7'9

9'r

8.8

8.6

8'r

8'2

8.r

8'S

v.
I-ower V.R.
Lower V.

Upper IV. R.
Upper IV.
IV.R.

8.r
8'2
7'8

9.O

8'o

8'r

8'r

9'o

8,2

8.o

8.4

8'+

8.o

8'4

7'5

8'r

8.4
8.6

7'8

7'4

Z'6
6.6
6.7

7'2

6.5
6.9

7.6

7'r

6'5
7'5

7'3

7'5

7'O

6.3

6'4

7'l

6.4

8.5

6'7

5'4
6'4

7'4

IV.

8'o

Lower IV.R.
Lower IV.
Upper III.

8.o

7'5
7'4
7'8

7.o

III.
II.

6.8

r.

4'9

7'4

7.r

6.5
6.8

7.O

Mean of means

8,2

7'7

7'3

7'r

6'o

9'2
8.r

7'5
7'O

6.4

6'r
7't

the example of skilled performance in which a complex sequence of
actions is undertaken to achieve a given end and short-term memory is
required to keep a tally on what has been done and what remains to be
done. We may here outline briefly four tlpes of sinration in which
different kinds of tally are required.

r. Keeping track of several variables
Yntema and Mueser (196o) presented subiects with a number of srnall
doors behind each of which could be written a symbol, for example a
shape, i.e. circle, squarer triangle or heart; or a numburi.e. r, 213 or 4;
or a letter denoting a colour, i.e. R(ed), Y(ellow), G(reen) or B(lue).
Each set of symbols was associated with a particular door. The subiect

234

Fundammtals of Skill

heard a series of messages indicating the symbol to be written behind
each door: sometimes these were the same as the symbol already there
in which case the subject left things as they were, sometimes they were
different in which case the subject erased the existing symbol and substituted the new one. At intervals questions were asked as to what
symbol was present behind a particular door and the subjea had to
answer without opening the door concerned to inspect. He had therefore
to retain the present states of the symbols behind all the doors at the
same time. Further similar experiments using a slightly modified apparatus were carried out by the same authors (r 96z). The task was found
to be relatively easy when only one in four messages informed the subiect
that one of the series had changed and when the sets associated with the
various doors were all different. Accurary fell, however, if every message
modified one of the sets or if all the sets were the same - for example if
all were numbers. Difficulty increased as the number of sets rose from
2 to 8 but did not do so to any significant extent as the number of
possible symbols in any one set rose from z to zl. These results are in
line with the results of more orthodox short-term memory experiments
in showing that retention becomes poorer as the number of items to be
retained increases and as they become less readily discriminated from
one another, but is often relatively little affected by increase in the size
of set from which the items are drawn.
A similar fall of accuracy with increase in the number of variables to
be monitored was found by Mackworth and Mackworth (1958, 1959)
when subjects had to keep track of a number of simultaneously developing situations. Their subjects, unlike those of Yntema and Mueser,
were able to observe the present states of all the variables continuously,
so that it is understandable that they were able to cope with a rather
large number. The interesting point is that errors were nevertheless
made: it seems that even when a whole display is on view at once, the
fact that only one part of it can be scrutinised at any one instant imposes
a substantial load on short-term memory if all parts of it are to be kept
under review. Several studies of similar type have shown that, as in
Kay's (lqSf) task, performance improves as the length of time between
messages increases, suggesting that the time for which data have to be
held in store is less important than the opportunities for rehearsal given
by a slower pace of presentation (Monty et al., t965, rg67a, b, Glucksberg et al., 1967, Taub et al., tg67).
Somewhat similar principles seem to apply to a concept-formation
task snrdied by Restle and Emmerick (t966) in which subieas were
shown a series of simple designs and required to discover the salient

Short-tenn futmtion

45

features by which they should be classified. To do this, if they were nor

to proceed by blind uial and error, they had to retain da12 about examples presented previously and about the correctness of their responses

to them. The authors found that this task was much more dimcdt if,
instead of each example coming from a single set of classesr successive
items came from 2 to 6 different sets in nrrn. These latter conditions
clearly increased the memory load since with, say, two sets the subiect
would have had to retain twice as much data in total in order to retain
as much data about any one set as he would if he were dealing with one
set only.

2. Problem-solving
As mentioned in Chapter t, short-term retention seems to play a crucial

part in certain gpes of problem-solving in which subiects have to
gather data and then hold it while gathering ftrrther data. It seems as if
performance is often limited by their forgetting the earlier da1x while
acquiring the later. Results have so far come from e4periments designed
to study differences between adults of different ages, and the decline
of short-term memory with age has made them clearer than they would
otherwise have been. Ife may briefly consider two examples. The first,
an electrical problem experiment in which subiects had to relate terminals on small boxes to points on circuit diagrams by taking ,edings
on a meter, has already been described in Chapter r (p. 22). The pre-

ferred method of tackling this problem seemed to U. that of aking

readings until one pair of terminals had been identified, then t"ki"t
ftrrther readings to identiSr a third, snd so on. At each stage the readin;
not immediately required were usually forgotten, so that they had to b.
taken again when required for the idenffication of another terminal
later. This strategy, although seemingly wasteful, had the advantage of
keeping the load on short-term memory to a minimum. Simil.r r.r.ilt,
were obtained in closely analogous tasks not couched in terms of electrical circuits by Clay (see Chapter r p. 22, also Welford, 1958, pp.
2o5-2o9) who also noted that accuracy improved when subiectt **t.
down the readings they took instead of trying to hold them in memory.
Clay G954, 1957, see also Welford, 1958, pp. 2rr-2r9) dso found

much the same kind of difrculty due to short-term memory with a
different kind of problem. In this subjects placed counters on chequerboards-3 x 314 x 415 x 5 or6 x 6-so as to add simultaneously
to marginal totals specified for both columns and rows. The 5 x 5
board, with counters in place, is shown in Fig . 7.3. The difficulty of

Fundamentals of Skill

46

this task seemed to lie mainly in the making of corrections for errors
when all, or almost all, the counters had been placed upon the board.
Subieas would make one or two rearrangements of the counters and
then seemingly become confused. Such confusions seemed to arise
from a failure to hold enough data in mind to effect the required correction. For example, to correct the error in Fig. 73the subiect would
have had to note that the second column added to 12 instead of 13 and
then find which of the other columns added to one too many. Having

3 3 2 t4
t3
2 3 3
2

I

t2

3

I

3 3
t4 r3 t2 lt
I

-l

to

I

lo
+l

Figure 7.3. 5 x 5 chequer board used by Clay (1954, rg57). The numbers enclosed by circles represent counters placed by a subiect in the
attempt to make them add to the correct totals on both rows and columns
simultaneously. The rows are correct, but the second column from the
left adds to T2 instead of 13 and the rightmost column adds to rr instead
of ro. To correct this error two pairs of counters marked with small
crosses need to be exhanged between the coltunns.

found this he would have to discover which counters to interchange
between the columns in order to correct the error without inuoducing
errors into the rows. Many subieas tended to lose essential data during
this process: for example they would forget the column first discovered
to be wrong while searching for the second. It was interesting to see, in
confirmation of this, that many subjects indicated incorrect totals by
devices such as moving a counter slightly out of place, for the acknowledged pu{pose of helping them to remember which totals needed
correction. Closely analogous results have been obtained with a more
complex problem by Jerome (1962) who again found that performance
fell with age from young adulthood onwards.

Short-tqm Retention

4T

It seemed fairly clear that, when solving problems such as these,
subieas more or less gradually obtained some kind of grasp of the way
in which the various pieces of data fitted together. What this means
in precise terms is not yet possible to say confidently. It does not appear
to depend wholly on building up a (structure' in long-term retention
because, at least in its earlier stages, the overall 'picttrre' is plastic and
subiea to modification by subsequent data. It seems rather to be akin
to the concepnral frameworks discussed in the previous chapter in connection with the co-ordination of aaion and the maintenance of orientation, and we may note that the second of these tends to be lost in
certain clinical disorders of memory.

3. Thinking
Perhaps the main interest of problem-solving is the lead it gives to the
snrdy of certain processes involved in thinking. Bartlett (r95o, 1958)
has described a number of experiments by himself and by Adiseshiah
(rgSr) and Jeeves GgSil which treat ttrinking as a series of discrete
stages which are either interpolated betrnreen a starting point to a known
conclusion or extrapolated towards a conclusion as yet unknow:n. Anyone who has been required in his snrdent days to undertake exercises in

inuospection will recognise the plausibility of such stages. Subieas
set to think about a problem and to inuospect while doing so commonly
report a series of images. These do not seem to represent the true
processes of thought: rather they are the stations bennreen one process
and the next. It is tempting to suggest that they are a kind of storage
berweel computations which, as the so-cdled Vurtzhtrg School rccognised in the early years of the present centuryr tre outside the scope of
inuospection. Obviously litde weight can be laid on such evidence, but
it is attractive and provocative to conceive of thinking as akin to the
operation of a computer going through a series of stages in each of
which data are taken either from the sense organs or from a memory
store to be combined with other data in some kind of computation, md
the result is then stored temporarily to be used later with other data

in a further computation, and so on.

If this is so, limitations in the range and capacity of thought could lie
either as traditionally assumed in the computational process, or in the
storage phase (Posner, t965). Effects of storage could be due either to

the limited amount of data that can be held at any one time or to
particular conditions, such as disuactions of attention, which are inimical to short-term retention. The former is suggested by Garner's (1962)

48

Fundamentals of Skill

finding that correlations between two dimensions or feanues of a stimulus can be readily perceived and used to simpli$, perception, whereas
correlations benreen three dimensions are much less easily recoguised
and used. The possibility that thinking may be impaired by conditions
which tend to disrupt short-term retention suggest that close scrutiny
would be worth while of detailed procedures during experiments on
tasks such as concept formation. Is it, for instance, that the requirement
to make a judgment about each example from which a general principle
has to be extracted tends to preclude the storage which is essential if
one example is to be compared with another? Again we may well ask
how far performance at certain tests, such as the pattern completions of
Raven's Matrices Test, are limited by difficulties of observing one part

of a pattern and then holding it in memory while looking away to
another part to see if it fits ?
The computations benreen each stage of storage and the next can
perhaps be regarded as recodings comparable with those discussed in
the previous chapter. Interpolation is analogous to the taking of action
in a skilled performance in which the end to be achieved is determined
in advance, but some flexibility of means is allowed. Extrapolation can
be thought of as the application of a coding or schema to data which
enables further data to be inferred in the same way as detail is often
inferred in perception. The range of possible extrapolations from any
given set of data will be constrained to a great extent by expectations
based on familiarity. One facet of originality is, perhaps, to set aside the
most probable extrapolation and to follow up a less familiar line.
4. Prograrnming of action

The analogies berween certain types of human performance and computer operation draws attention to the fact that short-term retention is
required not only of data but also of programmes of action. Such prograrnmes have been discussed under the term 'plans' by Miller et al.
(t96o) and the computer analogy has been well emphasised by Simon
and Kotovsky (r 963). Experimental studies are sparse but Broadbent
and Heron (t962) have shown that speed and accuracy of cancelling
digits in lists of random numbers tend to be poorer, especially for older
subjeas, when the digit to be cancelled changes frequently so that the
subiect has to keep revising his programme, than when it remains the
same throughout a list. Simon and Kotovsky point out that relatively
simple programmes couched in terms of computer operations can not
only enable subieas to carry out routines but can be the basis of

Shut-term Retention

239

extrapolation from a series and of the inductive recognition of regularity.
They show that the type and complexity ofprogramme required for such
tasks is related to the difficulty of the Thurstone Letter-Series Completion Problems in which subjects have to continue series ranging from

simple sequences such as cdcdcd . . . through sequences such as
abmcdmefmghm . . . to more complex examples such as rscdstdenref

Short-term prograrnmes raise a number of problems which are as
yet largely unrecognised: what is the maximum length and complexity
of prograurme - for example how elaborate a pattern can a woman knit
without gving continuous attention to the task? In what units should
the length of a prograrnme be measured? Almost certainly the means
of their retention are different from those of items in short-term memory
spans: the short-terrn programme seems to be more robust than the
memory span and much less liable to disruption by intervening activity
and shifts of attention, although still mlnerable to some extent. It is
tempting to draw analogies between, on dre one hand the short-term
retention of data and various clinical deficiencies of memory, and on the

other hand between short-term programmes and certain types of
aphasia, apraxia and disuastability, and thus to link them to neluological mectranisms. Doing so would, however, take us beyond the scope
of our present discussion. Suffice to say there seems to be a prima facic
case for distingoishing between the retention of data and of programmes:
iust what the difference is remains to be seen.

VIII
Effects of Loading

S7e have already in Chapter 4 discussed some of the effects of brief

overloading of the subject's capacity. In the present chapter we shall
consider the effects of longer-continued overloading and also of underloading, bringing together three areas of study which have been
prominent in experimental psychology at different times, namely fatigue,
stress and vigilance.It is impossible to separate these entirely from one
another, but we shall for convenience deal with them in ttun.
NEUROMUSCULAR FATIGUE

The snrdy of fatigue stands at one of the uaditional meeting-points
between physiology and psychology. Until the early years of the present
cennrry research centred mainly round the decrements of performance
that occur in the course of prolonged muscular work, and the division
of labour berween the two disciplines could be simply conceived: phy-

siolory studied the neuromuscular lmechanism itself, psychology the
accompanying subjective feelings of discomfort and exhaustion. The

shift of interest in the psychological field during the past fifty years
to* a predominant concern with conscious processes to an almost
exclusive concentration on observable behaviour and measurable performance, served in its early stages to strengthen rather than diminish
psychological interest in fatiguel here, it seemed, were striking changes
of performance that could be related closely to definable conditions.
Soon, however, it became apparent that neither in its physiological nor
in its psychological aspect is fatigue a simple phenomenon. The neuromuscular mechanism involved is complex, the changes that might be
regarded as fatigue effects are manifold. Behavioural changes, though
clear in outline, vary in important detail from one set of circumstances to
another. The situation became even more confused when attempts
were made to see analogies to neuromuscular fatigue ir p*ely mental
operations carried on for long periods.
249^

Effects of Loading

24r

These difficulties have led some to wish to abandon the term'fatigue'.
Yet there is need for a term to cover those changes of performance that
take place over a period of time during which some part of the mechanism, whether sensory, central or muscular, becomes chronically overloaded. If fatigue is defined in this wsy, the task of the investigator
ceases to be one of finding a single entity that can be labelled 'fatigue' :
instead he has, first, to snrdy the detailed changes of per{oruunce
brought about by such overloading and, secondly, to seek to explain
them in terms of changes of function in any of the many bodily mechanisms concerned.

Summaries in English of work on fatigue have been provided by
several authors (e.9. Viteles , 1932, Bardey and Chute, 1947, Floyd and
Welford, 1953, Simonson, 1965) so that the subject will be d,iscussed
relatively briefly here, first re-examining the classical tJpe of experiment
on neuromuscular fatigue and then going on to the much more complex
problem of so-called mental fatigue.
Classical experiments on neuromuscular fatigue have used an appsrtnrs known as an ergograph. Typically an arm and hand are strapped in
position in such a way that only one finger is free to move. ![ith this
finger the subiect is required to depress a lever that can be loaded to
different extents by means of a spring or weight. The lever is depressed
and released in time with a metronome or other timing device g1ing
regularly repeated signals.
If the loadirg is light or the rate slow or both, the task can be con-

tinued indefinitely, but if the load is substantial and the rate more
frequent the depressions of the lever will, after a time, begin 1s diminisfo
in amplinrde and evenhrally fall to zero. A tlpical ergograph record is
shown in Fig. 8.r. The decline of perforrnance begins sooner and proceeds more rapidly as the load or the frequency increase. If, when
performance has ceased, the subiect is allowed to rest for a period and
then tries again, recovery will have occurred: after a short rest it may
be partial, in the sense that amplinrde may not be reattained in full
and will fall again to zero quickly. Itrith longer rests, recovery will be
complete.

The whole pattern of decrement and recovery may be envisaged as
that of a system having a limited capacity for continuous operation and
some reserve that can be used to deal with temporary overloads. If the
overload is continued, as when the weight or spring tension is relatively
heavy and the lifting frequent, the reserve becomes extrausted - slowly
if the overload is s[ght, more rapidly as it becomes greater. Rest allows
the reserve to be re-established. It must be emphasised, however, that

z4z

Fundamentals of Skill

such a concepnral model, though a convenient aide mdmoire., is not
necessarily true in a literal sense: the same pattern would hold rf, for
instance, exercise progressively built up inhibitory or toxic substances

that were gradually dissipated and if work output depended on the

Figrrre 8.r. Section of ergograph record. Each vertical stroke was made
by a flexion of the right forefinger against a lever weighted in such a

way that the force required to move the point of contact with the finger
was approximately 14 lb. The time marker below shows r sec interyals.

balance between accumulation and dispersion. Nor does such a model
specify what part of the neuromuscular system is limiting performance:
limitations at various points could produce the same paftern of overload

and recovery effects.
A considerable volume of physiological research has aimed at finding

the locus of fatigue under various conditions in, firstly the muscles
themselves, secondly the myoneural junctions and thirdly the central
mechanisms supplying the muscles.
Peripheral limitations

Experiments with nerve-muscle preparations have shown that, with
repeated stimulation of the nerve, the muscle may after a time cease to
contract. Although nerve cells can lose sensitivity and reactivity with
long-repeated rapid stimulation and can thus show a fatigue effect, it is
clear that the contraction ceases long before failure of the nerve. It
also occurs before the muscle itself becomes incapable of contraction,
as is shown by the fact that, when blocking has occurred, the muscle
can be made to contract again if direct electrical stimulation is applied
to it. The site of the blocking thus appears to be the myoneural iunction.
Even direct electrical stimulation of the muscle may eventually fail

Effects of l^oading

243

to secure contraction. Although this is sometimes attributable to local
conditions developing at the point of contact ofthe electrode, it is usually
taken to imply that the muscle itself can show a fatigue effect. Convincing evidence that in the intact htrman subiect the muscle can be the
locus of fatigue has been given by Merton (1956), who showed condi-

tions in which, when a muscle had been fatigued by repeated rapid
contractions to the point of complete voluntary inactivity, it could not
be made to contract again by direct electrical stimulation. Ftrrther
evidence is gtven by Lippold et al. (lg6o), who showed that the electrical activity in muscles exerting a static force against a spring or
weight rose with time and that the rise was steeper the heavier the load.

This finding appears to mean that changes in the muscles rendered
contraction less powerful for any given level of efferent stimulation, so
that neural activity had to be increased to maintain the required output
of muscular power. Both Merton and Lippold et a/. forrnd that cutting
off the blood supply to the fatigued muscle or muscles prevented recovery during a period of relaxation. It therefore seems clear that the
lowered performance of the muscles was either due to lack of oxygen
or other supplies or to accumulation of 'fatigue products', which would
have been corrected by blood flow. If so, it is trnderstandable that fatigue
effects are found to be greater with static contraction than with rhythmic,
since blood flow is likely to be greater during the latter than the former.
Direct evidence about the effects of sustained contraction on blood
flow in muscles is equivocal (see Hemingpvay, r953rp. T4), but we can
perhaps expect that the force of the conuaction would be critical in
determining the extent to which blood flow is restricted, so that results
would depend on the precise conditions of different experiments.
The fact that maintenance of blood-sugar level is important for
athletic endurance (Edwards et al., 1934; Douglas and Koch, r95r) is
usually taken as further evidence that failing fuel supply to muscles is
a direct cause of fatigue. The argument is not conclusive, however, since
lowered blood-sugar can affect a wide range of bodily mechanislns,
neural and central as well as musqilar.

Cenual limitations

In his classical studies of the scratch reflex in dogs, Sherrington (tgo6)
found that after a period of repeated elicitation the reflex ceased,
though the same muscles would still respond to a different reflex.
Clearly there could be no question of myoneural iunctions or other
effector parts of the mechanism constinrting the limiting factor; the

244

Fundamentals of Skill

limitation was presumably sensory or central. The cessation came later
when the reflex was elicited by stimulating various slightly different
points on the skin rather than by repeated stimulation of the same point,
but it came eventually nevertheless, indicating central rather than
sensory origin. Sherrington's results have been directly confirmed by
Lloyd (tg4z), who found that action potentials in the efferent nerve of
the reflex diminished with repeated stimulation of the afferent nerve.
\Ve may note in passing that Sherrington found the fatigue effects ro
be highly specific: if a reflex had ceased after repeated elicitation by
stimuli applied to one point on the skin, it could be restored immediately
by shifting to a different point. There is no question here of regarding
fatiguer os has sometimes been done, as due to the general circulation
in the blood stream of 'fatigue products' resulting from exercise and of
these affecting the central nervous system.
The importance of cenual faaors in neuromuscular fatigue has been
stressed by Reid (1928). Using an ergograph with human subjects he
showed that, although voluntary contraction had ceased, the muscle
could still be made to contract either by direct electrical stimulation or
by electrical stimulation of the efferent nerve trunk. The essential locus
of fatigue here seems clearly to have been central. Reid further showed,
however, that recovery was much less after a period during which
voluntary effort ceased but the muscle concerned was direaly stimulated electrically and thus kept contracted, than after a period in which
the muscle was rested. This result implies that, although the fatigue
was central, local conditions in the muscle nevertheless influenced it,
presumably by means of afferent impulses. If this conclusion is corr€ct,
the question is reopened in a new way of how physical or chemical
changes in muscle resulting from exercise influence fatigue and of why
continuous contraction causes fatigue effects more quickly than does
rhythmic: although conditions in the muscles may not limit performance directly, they cause signals to the spinal cord and brain that
affect conditions there. The subiective counterparts (if any) of these
afferent impulses are not certainly known, though it seems reasonable

to link them with the feelings of discomfort and pain that mount
rapidly as severe muscular contraction continues. If so, it means that
fatigue decrements in muscular performance will depend to some
extent on a subject's sensitivity to and toleration of pain.
It must be emphasised that Reid's work did not i-ply that the
limitation of performance due to fatigue was invariably central. His
results showed that when a series of exceptionally rapid voltrntary contractions had fallen to zero direct stimulation of the muscle or the

Effects of Loading

245

nelve trunk did not restore conuaction or did so only partly. It seems
dear that with rapid or intense contractions a tnrly peripheral muscular
fatigue can be produced, but that with conditions nearer to those of
everyday life, central limitations are more likely.
Two further points about neuromuscular fatigue indicate the importance of central faaors.

(o) Motioation. Figrre 8.1 shows a number of minor variations typical
of ergographic records; the rate of decline is not smooth, but is interrupted by temporary partial recoveries. To some extent such variation
is to be expected from the random fluctuations of function to be found
in almost any complex biological mechanism; but subjectively some, 8t
least, correspond to periods of special effort, and it is easy to show that
they can be produced by urging the subject to 'try harder'.
The effect of incentives upon fatigue effects has been vividly illustrated in an experiment by Schwab (rg53), who required subiecrs to
hang on a horizontal bar. He found that with instnrctions to hold on 'as
long as possible' the average length of time before lening go was less
than r min. lf/ith strong urglng the time was raised to rather over
r min. With the reward promised of a $S bill for bettering their previous records, subiects managed to tang on for an average of nearly

2 min.
What appears to be a different kind of motivational effect was shown
by Ash (t9t4), who found drat, after performance on an ergograph had
ceased, it could be made to begrn again not only by lightening the
weight, but also merely by leading the subiect to be[eve that the weight
had been lightened. Again Jarrard (196o) showed that subjects who
uansferred from one weight to another which was in fact the same but
because it was smaller in size appeared heavier, made fewer lifts before

reaching 'exhaustion' than did subjects who had the same size weight
throughout, while those who uansferred to a larger size weight which
appeared lighter made more lifts. He fotrnd that they adjusted their
muscular tension according to what they believed the heaviness of the
weight to be so that those who transferred to smaller weights were in a
sense exerting themselves unduly. His results are complementary to the
well-known fact that athletes pace their performance in races from the
begrnning accrcrding to the distance to be covered, thus adopting a
strategy which prevents premattue onset of fatigue due to over-exertion
in the early stages (e.9. Ward, r95o).
These results are important, as showing that fatigue effects are to
some extent under voluntary control in the sense, perhaps, that the

246

Fundamentals of Skill

subiect sets levels of effort he is willing to make and of discomfort he is
prepared to bear. Such variation is not strialy attributable to the
system that fatigues, but must be taken into account when measuring

fatigue effects.

(b) Recruitment. It is easy to observe in ergograph sturdies that when a
load is placed upon a small group of muscles other muscles spontaneously become active and that, as fatigue proceeds, the activity in other
muscles spreads until almost the whole body is involved: the subiect
may tense his legs and grit his teeth in the effort to depress one middle
finger so strapped in position that these activities cannot possibly help.
The classical demonstration of this phenomenon is by Ash (r9r4). A
more recent demonstration of the same kind of effect has been given
by Lundenrold (1958), using electromyographic recordings in a typewriting task. In everyday life this recruitment would be adaptive in
the sense that other muscles could take the load off the group that was
becoming fatigued; the fact that the phenomenon occurs in ergograph
experiments although it is not adaptive in thern implies that it is largely
involuntary. The exma muscles become active in a specific order
(Seyffarth, r94o, Denny-Brown, 1949), presumably along lines of ftrnctionalorneural proximity. Onemayperhaps assume the relatively simple
neurological model that focal aaivity concerned with one muscle or
group of muscles tends to spread to surrounding areas, the amount of
spread becoming greater the longer the focal activity continues or the
more intense it becomes. Lippold et al. (196o) showed that as other
muscles become active, electrical activity in the fatigued muscle may
diminish, presumably implying that activity in areas srurounding a
focal area may continue after it has ceased in the focal area itself. It
may be noted in this context that excessively exercised muscles tend to
go into contraction more readily than rested ones, sometimes showing
spasm, as in writer's or telegraphist's cramps. This appears to imply
that a focal area, when it has recovered from acute fatigue, ffioy be left
hypersensitive - a condition found in some sftrdies not concerned with
fatigue to occur in nerve tissue subjected to prolonged stimulation.
It is in line with this general picture that any voluntary aaivity by
an irrelevant muscle during a task will tend to increase the involuntary
activity by the same muscle subsequently: for example Yensen (lg6S)
found that when one arm was supporting a heavy weight, a single
voluntary contraction by the other arm tended to increase the subsequent involuntary aaivity in it - the voluntary activity seemed to have
facilitated the involuntary. The picnrre Dsy, however, be complicated

Effects of Loding

247

by general activation effects produced by anxiety, and also by individual
differences of approach to the task (Benson and Gedye, ry62): we shall
return to these factors later in this chapter and again in Chapter ro.
To sum up the position reached in the snrdy ofneuromuscular fatigue,
it appears that in the intact organism changes in the muscles brought
about by prolonged or repeated contraction can, according to circumstances, have one of rwo limiting effects. Either the muscles themselves
become temporarily incapable of further conuaction or the condition
of the muscles produces afferent stimuli and these in turn affect the
central mechanisms and lead to the cessation of efferent impulses. Vhich

effest occurs first depends on factors at present not entirely clear. It
Esy, however, tentatively be suggested that peripheral limitations are

likely to result from intense aaivity and central limitations from
activity which is less intense but more prolonged.
MENTAL FATIGUE

If the term 'mental fatigue' is to have a meaning in line with that of
neruomuscular fatigue, it must denote the impairment of some brain
mechanism as a result of long-continued use. The impairment must be
reversible in the sense that it disappears with rest, and may take the
form of lowered sensitivity or responsiveness or capacity. The last of
these may show as a reduction in either the amount of information that
can be handled at any one instant, and thus in reduced 'mental power',
or in the amount that cirn be handled in a given period and so in slowness of perception, choice and so on.
Such a definition enables a distinction to be made benn een mental
fatigue and several other cenual changersuch as adaptationrhabituation,
satiation, inhilifion and monotony or boredom, all of which lead to
decrement of performance with time. Adaptation implies a loss of sensitivity or discriminating power over one paft of a range of possible
stimulus values, but simultaneous gains over another: there is not so
much a lowering of sensitivity as a shift in the point of maximum
sensitivity. Habituation denotes a learnt ignoring of stimuli. Satiation is
a state in which action ceases because the need or appetite that gave
rise to it has been satisfied. Inhibition, although often loosely used,
strictly means the reduction of one process by the activity of some
opposing process. Boredom or monotony refers to a state in which the
organism is rurderloaded, not overloaded: it seems as if a certain
throughput of information is necessary to mainain full efficiencS and
tlpically boring situations seem to be those in which attention is required

248

Fundamentals of Skill

but little information is conveyed; the classical bore compels his hearer
to listen to conversation that is insignificant in content.
The distinction between fatigue and these states is easy to make
formally, but is often difficult to draw in practice. For exlmple, even
neuromuscular fatigue can in a sense be viewed as a central inhibitory
state brought about by afferent impulses from the muscles, although it
should be noted that Sherrington was able to distinguish betneen patterns of decrement in a reflex due to what he termed fatigue as opposed

to what he identified as inhibition. The greatest difficulty arises in
distinguishing fatigue from monotony or boredom: many tasks used in
snrdies of fatigue are repetitious and thus liable to become monotonous,
whereas some used for snrdying monotony require actions or decisions
to be repeated frequently enough for them to be a possible cause of
fatigue. It is indeed reasonable to suppose that some tasks are both
fatiguing and boring: some parts of the subjects' central mechanisms
may be overloaded even though the overall throughput of informarion
is low. Usually, however, it seems fairly clear which of the two effects
limits performance sooner.
We shall here survey work that has been regarded, probably correctly,
as studying fatigue, leaving till later studies of conditions producing
monotony and boredom.
Phenomena of mental fatigue

We may note at once two resemblances betneen mental and neuromusctrlar fatigue which, although penpheral to any discrrssion of the
natlrre of fatigue, have an important bearing on methods of measurement. Firstly, if given the oppornrnity, subjects tend to distribute their
efforts over a working spell so as to minimise fatigue effects, adjusting
their pace to the expected length of spell right from the beginning,
working fast if the spell is to be short, more slowly if it is to be long
(Bills and Brown, 1929, I(nreger, 1937, Barmack, 1939, Katz, 1949,
Forrest, 1958, Saufley and Bilodeau, 1963). This means that fatigue
effects are likely to show more if the subjea works under pressure for
speed for an unknown period than if he works at his own pace or for
a time known in advance. Secondly, fatigue effects tend to be specific to
the performances that produce them, leaving other performances little
if at all affested. As a consequence, tests in which a subiect is taken off
his main task and put for a brief period on to another to assess his
state of fatigue are often unsuccessful: little change of performance at
the tests occurs until such extreme states of exhaustion have been

Effects of Loading

249

reached that no test is needed to supplement common observation. In
consequence, the most sensitive indications of fatigue are usually ob-

tained by a detailed study of the fatiguing performance itself. There
have, however, been a number of researches reported in which fatigue
tests have proved successful. S7e shall in discussing fatigue effects
consider both types of evidence.
Bartlett's Ferrier Lecflre to the Royal Society (rg+f) was a landmark
in the development of ideas about fatigue and indeed about htrman
performance generally. In particular it showed that the phenomena of
fatigue are more varied and more complex than is often supposed. On
the basis of this lecnrre and of work done since, we may recognise four
main types of change that can come about in mental fatigue.

(o) Smsory or perceptual changes. The classical snrdies of visual fatigue
have been strmmarised and discussed by Bartley and Chute (tg+il and
by Weston (1953). They have, for the most part, been concerned with
the possible fatigue of eye mtrscles and its relation to feelings of eye
strain under conditions of low illunination, glare or close attention to
detail. It is doubtftrl if the effects are to be wholly, or even mainly,
attributed to the eye muscles: some probably result from frowning and
general muscular tension built up as a result of concentrated efforts to
see trnder difrcult conditions. Be this as it may, it seems clear that there
are some more strictly sensory or percepnral effects. For instance,
Berger and Mahneke (1954) found that visual acuity (cancelling Landolt
rings) and critical flicker frequency (CFF) both fell when tests were
made continuously for 55 min, but rose again after 5 min rest. Examples

of their results are shown in Fig. 8.2. Haider and Dixon (1961) in
experiments in which they made continuous recordings ofthe threshold
for detecting the 'difference of intensity between two spots of light,
found that thresholds rose substantially during the course of a t4
min session, mainly between the second and tenth minutes. Sddanha
(1955, r95il fotrnd that repeated seuings made on the vernier scale of
a cal[per gauge for t hr or more became less regular, snd thus less
accurate, with time, but that accuracy retruned'after + hr or so of rest.
We may note in passing that the rehrrn was greater after a period of
rest than after a similar period spent cancelling Landolt Rings - the
fatigue effect was such that a change of work was not as good as a rest.
The motor components in all these tasks were trivial, and thus it is clear
that there was some loss of fine differentiation either spatially, as with
visual acuity or vernier settings, or temporally as with CFF.
It has also been widely reported that CFF is lowered by mental work

z5o

Fmdarnentals of Skill

such as calculating or reading. The evidence has been summarised by
Simonson and Brozek (tgSz) and by Grandjean and Perret (196r).
Falls of CFF have not always been found when they have been sought,
but enough positive results have been obtained to suggest clearly that
the effects are real. It is very plausible that they should be so if fatigue
0.95
DECR.

23olo
0.90
41

=
L)
l

0.85

DECR.: 1l olo

40

(,
tJ.,

G,

i

; 0'80

lt 39
tro

=

;

1r)

F

U'
UJ

G,

38

0.75

= 'e

c

r=
n

327 - 254 m
37
0.70
0

20

60

40

TIME (MIN.)

AB

0

20

40
TIME ( MIN. )

60

Figure 8.2. Examples of fatigue effects obtained by Berger and Mahneke
(tgS+) with repeated measurements of visual functions.
A. Visual acuity. For each determination a Landolt ring was brought
progressively closer to the subject until the direction of the gap ctuld
be recognised: 90 determinations were made continuously over a period
of 55 min and 20 more after an interval of 5 min. Retinal illumination was
held constant at 3'5 lux. Each point is the mean of ro determinations by
one subiect.

B. CFF. Flicker was produced by a rotating disc illuminating a diffirsing glass surface i r2o determinations were made continuously over a
period of 55 min and further ones after an interval of romin. Theflickering
target subtended r'r5o at the subiect's eye.Each point is the mean of
ro determinations by one subject.
implies some temporary central impairment since CFF has been fotrnd
to decrease following brain iniury and under the influence of depressanr
drugs, and to fall progressively with age during adulthood. The evidence
suggests that CFF falls under conditions in which the signal-to-noise
ratio in the brain is lowered so that there is a fall of d' in the signaldetection model when detecting the difference betvyeen 'on' and 'off'.

Effects of Inadfug

2Sr

It is, however, possible that some at least of the effect might be due to
a change in the criteria of iudgment and thus of P.
As regards the site of the effect, Grandiean and Perret have shown
that it is not due to a change of pupil size. They did, however, in line
with Berger and Mahneke's results show that CFF fell as the time over
which the test light was exposed increased from ro to 28 sec. This fall
tended to mask the effects of fatigue caused by having repeatedly to
rearrange sets of seven random digts into their normal order. The
observed fatigue effect thus depended to a considerable extent on the
way in which CFF was measrued, and this may explain some of the
failures to find effects in the past. Evidence that the effect is not purely
visual in origin is contained in work cited by Simonson and Brozek
which found a substantial fall of CFF following calculations carried
out while blindfolded. Further evidence is contained in the finding
by Davis (lgSS) of a similar lowering of the critical rate at which an
intermittent auditory signal ceases to 'flutter' and is heard as a continuous sound. Davis tested both CFF and critical flutter rate following
r and 2 hr work at multiplying trro-digit numbers mentally: the
numbers were exposed visually and answers were written down. Both
measures declined, with the flutter rate proving somewhat more sensitive than CFF.
(b) Slouting of puforyrutnce. One of the most frequently observed fatigue
effects is the slowing of sensory-motor performance. It is often suggested that this may be due to muscular fatigue, but there is no doubt
that central factors are often, and probably rnainlp involved. An indication of this is contained in the results of an experiment by Singleton
(tgSf) who used a serial choice-reaction task. The apparanrs is shown
in Fig. 8.3. Subieas sat in a chair and pushed a ioystick from a centrd
position along slots in four directions in response to lights at the ends
of the cross on the display shown in the top left corner of the Figrre.
Upon the subiect's reaching the end of the correct slot the light went
out, and on his return to the centre another came on, trntil 6+ responses
had been completed. He was told to work as fast as possible. Three
variations of the task were presented; they were, in ascending order of
difficulty, (i) 'Direct' with the ioystick having to be pushed away when
the top light came on, to the left for the left light, and so on; (ii) 'r8o"
with the ioystick pushed away in response to the bottom light, to the
left for the right light, and so on; and (iii) '27oo' with the ioystick
pushed away in response to the left light, to the left for the bottom
light, and so on. The times per response gradually lengthened during

2Sz

Fundamentals of Skill

each run, but, as can be seen from Fig. 8.4, the lengthening was
much more in time spent at the centre, that is in deciding which way to
move, than in the acnral execution of movements. Moreover, this
lengthening increased with the difficulty of the condition, implying that
the fatigue effect became greater as the demands of the central task
rose.

Figure 8.3. Four-choice serial reaction apparatus used by Singleton

(r953).

Slowness may cause several complications when the subiect cannot,
as he could in Singleton's experiment, work at his own speed, but is
externally paced. The classical example of such a task used in the snrdy
of fatigue is the Cambridge Cockpit, which was the basis of experiments
by Craik, Drew and Davis (Davis, 1948). Subjects were tested for z hr
spells in a simulated aircraft cockpit under blind flying conditions and
had to deal with a series of maneuvres. Although in a general sense
subieas could control the timing of these, the complications of the task

were such that, once begun, many of the actions required in the
rnaneuvre were in effect paced by the apparatus. Under such conditions
subjects can react to slowing in one of two ways.

(i) Some of the actions required may be omitted. Drew, whose

experiments are cited by Davis, fotrnd this tendency to be characteristic

Effects of

Loading

zS3

of some of his subiects during the latter part of their z hr spells. He also
noted that most subfeas tended to pay less and less attention to the
more penpheral parts of the task as the spell continued, giving their
o.7

0.7

1 Boo

OIRECT

L)

lr,
(,
lrJ

0.6

0.6

0.5

0.5

l(

0.4

- 0.4
C)

UJ
LD

f o.s

0.3

=

=

F

F

0.2

0.2

(a)

0.1

0

0
START

(b)

0.1

TRIAL 6

END

TRIAL 6

START

END

o.7
zToo

0.6

a

0.5
a
L)

u,

-,.-X-)<---X--X-

t,

t!,

=

x

0.4
0.3

F

0,2
0.1

( c)

0

START

TRIAL 6

END

Figure 8.4. Results obtained by Singleton (1953) in three conditions of a
serial reaction task.
The results shown are for the sLnh trid of 64 reactions under each of
the 3 conditiorrs. Each point is the mean for 8 reactions by ro subiects.
The dots and solid lines are for times between the end of one movement

and the beginning of the next, with the ioystick in a central position,
and are essentially reaction times. The crosses and broken lines are for
the movement times from centre to end of slot and back.

main attention increasingly to the conuols in constant use. For example,
the fuel indicator had to be reset every ro min, but came to be more
and more often neglected, as shown in Fig. 8.5. We may note that
Btrsill (rgS8) found a similar tendency for peripheral items to be

254

Fmdamentals of Skill

neglected under conditions of high temperature in a task in which
subjects had to uack a moving target and respond at the same time to
signal lights at various distances from the target centre. Such omissions
may perhaps be regarded as spontaneous attempts to simptr& the task
somewhat like those discussed in Chapter T (p. zrt).

(ii) If, alternatively, the subjea tries to complete all the actions required in the time available, he will have to hurry and may not have
60

A

3ao
c)

x

Lr.l

(D

3
(,
tr
o
$-

z.
U.J

(J

u. 20
t l,J

o-

345

7

SUCCESSIVE PERIOOS

Figure 8.5. Percentage of subjects, in experiments with the Cambridge
Cockpit, who omitted to reset the fuel indicator during successive periods
of a z lu test. Based on results from r4o subiects. (After Davis, 1948.)

time to make his decisions and iudgments accurately. In tasks like
that of the Cambridge Cockpit, any slowing that results in a longer time
being needed to observe the various instruments accurately rnay be
felt as a 'stickiness of attention'. Performance under these conditions
will tend to suffer a disruption that builds up in a vicious crcle: the
longer time taken to observe an insuument means that the resulting
error tends to be larger before correcting action can be taken; when
action is eventually taken it may, in order to make up time, be poorly
controlled and require subsidiary corrections; these take further time
and mean that subsequent correcting actions have to be larger again.
The result, as Davis showed for many of his subieas, is that the onset

Eflects of Loading

255

of fatigue may lead to marked overaaivity, often coupled with signs of
anxiety. These in ttrn may direct attention from the task to worrying
about whether it is being performed adequately and thus lead to still

Figure 8.6. Records of two different types of response in attempting to
move a pointer on to a target by means of a velocity conuol. The records
are plots against the time (from left to right) of the movements of the
control. The upper record shows a normal skilful response with a movement of the control to get the pointer going and a second movement to
stop it. The lower record shows a disorganised response made in the
attempt to achieve the same result. Data from Davis (1949).

further slowing and disruption. An example of this kind of disruption
is shown in Fig. 8.6, taken from the records of another experiment by
Davis (tg+g), in which subjeas had to bring a pointer from one position to another by means of a velocity control.

(r) Inegularity of timing. Long-continued performance tends to become
not only slower but also less regular. Some of the most striking illusuations of these tendencies have been from industrial tasks in which,
during a strift, there has been a moderate rise in the me:m time required
to perforrn each cycle of an operation, accompanied by a much gf,eater
rise in the standard deviation (for a review see Murrell, 1965). To some
extent irregt larity may be more apparent than real: the distributions of
times for individual cycles of repetitive tasks tend to be skew, with a
tail of long times and with a variance increasing with the mean; any
overall slowing will therefore increase the variance and the number of
what seem to be trnusually long times. It has, however, been suggested
by Bills (1931) that irregrrlarity is due rather to occasional 'blocking'.
That is to ssy, every now and then a short gap appears in an otherwise
rapid performance, and the frequency of such gaps incteases when the
task is continued for a long time. On this view the greater part of the
disuibution might be only a little affected, but there should be a marked
increase in the size of the tail.
One source of such blocking in paced tasks is easy to understand.

256

Fundamentals of Skill

For example, Vince (1949) showed that subjects required to respond by
pressing a key to signals at regular intervals kept pace up to a certain
rate, but at higher rates gradually fell behind until eventually they
stopped 'to make a fresh start'. Bills was, however, primarily considering
unpaced tasks, such as alternate addition and subtraction of 3 from a
list of digits, colour naming, substinrdng letters for digits according to
a code or giving opposites of words. He found that the frequency of
times for individual items exceeding twice the subject's average time
tended to increase during 7 min of work. His results are not very convincing and could probably have been due to simple slowing, but much
clearer evidence has been obtained by Bertelson and loffe (lg6l) using
a serial choice-reaction task. The subiects in this were required to press
one of four keys in response to the figures r to 4 shown in random order
on a 'Nixie' tube. Each response brought on the next figure, so that the
task was continuous but unpaced. Samples of reaction times were scored
at the second minute of work and at the end of 3o min. No change was
found in the averages of the shortest or the median reaction times, but
there was a marked increase in the average of the longest. The percentage of 'blocks' (defined as reaction times longer than trrice the mean,
excluclirg responses in which errors were made) rose rapidly during
the first 5 min of work and slowly thereafter. More important, there
was a clear tendency for reaction time and errors to rise during the
responses immediately before a block and to fall immediately after, as
shown in Fig. 8.7. The results are consistent with the view that some
kind of fatigue effect builds up gradually over a series of responses and
is dissipated by the block. It should be noted that on this view there
are two fatigue effects involved: a short-term effect dissipated at each
block and a longer-term effect causing a rise in the frequency of blocks.
The longer-term effect could perhaps be regarded as due to recovery
during a block being not quite complete, so that the time taken to build
up to the next block is shorter than it would otherwise have been.
The cause of such blocking is not clear. Perhaps the most obvious
suggestion is that some part of the sensory-motor mechanism becomes
momentarily inoperative, although saying this does little more than
restate the phenomenon. Broadbent (tgS8) has suggested that the
mechanism which selects information relevant to the task in hand from
the whole mass of data impirrging on the organism, becomes temporarily
ineffective and allows irrelevant signals to gain attention. This would
neady account for the increase of distractability often observed in
fatigued subjects, although there is an alternative explanation available
in the likelihood that stimuli arising from hard seats, awkward posnres

Effects of

Loading

2ST

or tensed muscles would become more insistent srith time. In other
words, distraction could be due not to any failure of central mechanisms
directly involved in perforumnce, but to increased competition from
ancillary stimuli which might capture the cenual mechanisrns and so
produce intermittency effects like those discussed in Chapter 4.
These irregularities of timing have all been observed in the acnral

(J
ul

a
r!

=
UI

(n

z

o
(L

\,

(n
uJ

(r
llj
(,

(n

z. G.
r! o
G.

()

G, G,

llJ u,

(L

-21 -20

-8 -7 -6 -5 -4 -3 -2-l B +l +2 +3 r4 +5+6+7+8

+20 +21

STEPS FROM BLOCK

Figrrre 8.7. Response times and errors made before and after a'block',

in a serial reaction experiment by Bertelson and loffe (1963). The graph
was constructed by taking for each of z8 subiects the last r r blocks
observed (and not immediately associated with an ertor) during 3o min
work and plotting the mean reaction times and errors for the first to
eighth reacrions before and after the block. Thetwentiethandtwenty-first
reactions bdore and after are also shown as an indication of reactions
well clear of a block.
performance which has caused the fatigue. How far they are shown in
other tasks undertaken imnrediately afterwards has not been fully explored although Rey and Rey (rg6f) found rate of tapping less regular
after 45 min of work at a cancellation task, with reaction time becoming
longer and CFF lower at the same time. Again Takakuwa (undated)
found that the accruacy of aiming at a target over a period of one
minute deteriorated after a nunrber of tasks regarded as fatiguing, snd
that this seemed to be a better test of fatigue than CFF.

(d) Disorganisatim of performance. Bartlett (rg+l), with the preliminary
results of the Canrbridge Cockpit experiments in mind, suggested that
fatigued subiects may sometimes perform correct actions, but in the
wrongorder.In other words, the co-ordination of their performance, the
ordering of individual actions into'larger units', has broken down. This

258

Fundamentals of Skill

line of thought has not been followed up to the extent it deserves,
probably because it is far from easy to sttrdy the kinds of complex performance in which such breakdowns might show. It does, however,
tally well with the mild confusion, inability for sustained thought and
impairment of judgment often observed in states of fatigue. Three
possible explanations seem likely to repay further rsearch.
(i) Failure to place items in context The tendency for a word or
phrase repeated over and over many times to 'lose meaning' is well
known and has been demonstrated also under experimental conditions
(e.9. Bassett and !7arne, r9r9, Lambert and Jakobovits, 196o). The
mechanism of this phenomenon is obscure but it seems to i*ply some
loss of connection benreen the word or phrase concerned and its
normal associations.
(ii) Impairment of routine. When a situation or task in encountered
repeatedly we tend spontaneously, as discussed in Chapter 6, to build
routines that enable us to treat several actions together as a single
ordered 'unit', instead of having to make individual decisions about
each one. The building of such units depends, however, upon the
ability to carry out the individual actions accurately enough for one to
follow another without the flow having to be interrupted in order to
make corrections. Any change due to fatigue or any other factor that
impairs acnracy will tend to break up these routines and make it
necessary once more to deal with the task piecemeal.
(iii) Disturbance of short-term retention. We have noted in previous
chapters that implicit in the concept of organised performance and the
integration of actions or information is some form of short-term retention that holds earlier items to be combined with later ones and keeps
a tally of what has been done and what remains to be done. Such shortterm retention has been shown to be very liable to disruption by strifts
of attention, especially after brain iniury, in old age and in other conditions in which some organic impairment can be presumed. There is
some evidence of a similar breakdown with fatigue from tests on civilian
aircrews. fn one experiment (Welford, Brown and Gabb, r95o) radio
officers were tested with a type of electrical problem similar to that of
Fig. r.4 (p. zz) before and after flights from the United Kingdom to
Africa, India, Australia or the Far East. For each problem subfects were
given a box with six terminals on the top, a circuit diagram and a resistance meter and had to find out which terminal on the box corresponded
to each on the diagram by taking readings on the meter. Subjects took
many more readings if they were meeting the task for the first time
after a flight than before, ond it was clear that while taking one reading

Effects of

Loding

zS9

they were tendi.g to forget others already obtained and so were having
to take some readings several times. The results are shown on the left

of Fig. 8.8 (the results on the right of this figure are discussed in
Chapter 9). This experiment was on crews of unpressurised aircraft,
and the results might have been partly due to chronic anoxia. This
question does not arise, however, with an experiment by Kay (1953) on
radio officers, stewards and stewardesses of pressurised aircraft on the
Atlantic and Australian routes from the United Kingdom. This experi60

c,.
llJ
irr 40

=
z.

o
]lJ

x

-----

r.o20

rlD

(D

z

6

ut

e
I

o
2

3

4

PROBLEMS

Figrue 8.8. Numbers of readings taken on a meter by radio officers of
civilian aircraft in solving electrical problems (!7elford, Brown and Gabb,
rg50).

The upper curve shows results of tz subjects who solved their first
two problems immediately on renun from a flight and their last two
after at least 8 days' stand-dowrr. The lower curve shows results for

rz subjects w'hose first two problems were after stand-down and whose
last two rvere after flights.

ment which involved a running short-term memory task has already
been described in Chapter 7 (p. ztr). Yotrnger subjects were, on avertBe,
consistendy poorer after flights. Surprisingly, the older were not so,
but scored consistently less than the younger both before and after
flights. It looks as if the effects of fatigue and age did not stunmate,
perhaps because the lower achievement of the older subieas meant
that they were exhausting themselves less. A rather similar pattern of
results, dft older subjects achieving less but showing less fatigue
effect, was obtained by Botrninick and Shock (tgSz) with an adding task.
Two further points from experiments on fatigue may be mentioned
more briefly:

2fu

Fundamentals of Skill

(e) Temporary improoemmt of performance. Somewhat surprisiogly, the
first sigrr of oncoming fatigue is sometimes an improvement of performance, in the sense that it becomes m,ore active and achievement
rises: deterioration does not set in until later. For example, in another
experiment by Welford, Brown and Gabb (lgSo) on the same radio
officers as those who took part in the electrical problem experiment,
performance at a plotting task was better after a trip with a relatively
easy schedule than it was before going out. Performance was, however,
poorer after a trip with a more arduous schedule than it was before
flight, and the rise after an easy trip occurred only with the relatively
easy ploning task: it did not occur with the more difficult electrical
problems. Nor did it appear with the plotting task among stewards for
whom physical demands during flight were rather more arduous than
for radio officers and who were probably of less intellectual capacity.
In short, the temporary improvement of performance seemed to be
confined to narrow limits, which depended on the degree of fatigue, the
difficulty of the task and the grade of the subjects. It was perhaps the
analogy on a longer time scale of the 'warm-up' effect seen in many
laboratory experiments : performance at atask such as uacking improves

during the first few seconds or minutes of work even though it had
been previously well learnt and well practised (e.g. Ammons, 1947,

Irion, ry66).

(f) Reduction of fatigue effects by familiarity. Fig. 8.8 also shows that
the differences benreen the performances of those tested before and
after flight were much greater for the first than for the second problem.
Similar results have appeared in several other experiments and imply
that fatigue effects are greater for novel than for familiar tasks: some
capacities called into play to deal with new material seem especially
susceptible to fatigue. What these capacities are we cannot at present say.

Explanatory models of mental fatigue

The attempt to posnrlate more fundamental explanations of mental
fatigue than those so far considered is made difficult by the fact that
many other influences, such as local muscular fatigue or, as with the
aircrews, anoxia or loss of sleep, ffisy affect performance as well as any
true mental fatigue. Taking the evidence as a whole, however, two main
hypotheses about the nanre of mental fatigue seem reasonable.
{o) Local neural impairment. The maditional assumption is that some

Effects of

Loading

z6r

group of nerve cells concerned with the performance that fatigues, or
with some essential link in it, becomes insensitive or unresponsive
throtrgh continued activity. Such a view explains well the similarity of
some fatigue effects, such as fall of CFF, to those of brain iniury. It
can account for slowing of perforurance by asstrming that some stage
in the sensory-motor chain requires a stronger stimulus to operate it
and that a given level of stimulation can be integrated over time. Blocking is accounted for by assuming that the breakdown rnay be of only
short duration. Loss of short-term retention would result if the selfmaintaining neuronal circuits, on which it must almost certainly depend,
became insensitive, so allowing the memory traces to decay.
Rest, on this view, would have its effea by allowing for recovery of
the nerve cells involved. Overactivity and improved performance in the

early stages of fatigue could be accotrnted for either by analory with
the recruiment effect fonnd in nenromuscular fatigue or by assuming
that, faced with incipient failure, the subiect makes compensatory increases of effort that may, for a time, more than offset losses due to
fatigue. Such compensatory efforts are perhaps seen in the results obtained by Bills and Shapin (lg16), who found that deterioration of performance could be posqponed by pacing the task a little faster than the
comfortable rate instead of allowing the subiect to workat hisownspeed.

(b) Inqease of 'nanral noise'. AII alternative view can be based on one
put forward by C,rauford (196r) as a possible explanation of fatigue in
car drivers. He suggested that such fatigue might result from the
accrrmulated after-effects of stresses and annoyances from other road
users during prolonged spells of driving. If so, it could be regarded as
due not to under-functioning of particular brain mechanisms but to
overactivity in the brain or parts of it.
We shall discuss the mechanism of such neural overactivity in the

next section of this chapter. Meanwhile we can argue that, on this view,
fatigue results when netrral activity, either general or local, rises beyond
some optimum level. Such a theory would account for sensory fatigue
directly in terms of the bhuring effects of netral noise. It would accotrnt
for slowing of performance by arguing that the effective strength of
stimuli from one part of the sensory-motor mechanism to another
should be meastued in terms of signal-to-noise ratio aud that a rise in
the noise level thus implies a need for more powerful signals in order to
maintain the ratio required. Any randomrress from moment to moment
in the level of neural aaivity would mean brief periods of specially
intense noise, which might accotrnt for blockiag or, at a lower levei,

z6z

Fundamentals of Skill

brief periods of facilitation which might allow unwanted signals to
cause distraction. Loss of short-term retention would be assumed to
result from disnubance rather than decay of the memory traces. Rest
would allow excess activity or sensitivity to die away. Overactivity and
improved performance during the early stages of fatigue would be
accounted for directly by facilitatory effects of mild rises in activity.
This view could also account well for the irritability and diffictrlty of
relaxing or sleeping often observed after a long period of taxing mental
rvork.

Present data on fatigue seem to be equivocal in the sense that they
are consistent, grven plausible supplementary hypotheses, with either
view. Distinction berween the two models appears to need a substantial
amount of new research carried out by a combination of physiological
and psychological methods. The two models may well both operate in
different circumstances, and it seems reasonable to suggest that new
research should aim at establishing the conditions in which each applies
rather than iustifying one theory to the exclusion of the other.
STRESS, EFFORT AND AROUSAL

It has been recognised for at least three-quarters of a cennrry that one
of the important variables of human behaviour is the intensity with
which action is carried out. A man may work either more or less hard
at a iob, he may be moved either more or less strongly by an emotion,
his motives may be either more or less powerful. The methods of
assessing intensity have, for the most psft, been subjective using techniques such as rating scales. For the past 40 ye:rs or sor however,
research especially in the field of emotion has attempted more objective
measurement, linking intensity either to the activation of the skeletal
muscles in terms of muscle-tone or to autonomic activity as indicated,
for example, by skin conductance. !ilork along these lines favoured a
concept of intensity in terms of generalised actioation of the organism,
with the direction of aaivity being determined by features of the immediate cognitive situation or of past experience. Such a view accorded
well with ideas that had grown up in some other fields, for instance
that hunger, thirst and other appetites lead essentially to a generalised
activation of the organism which is given direction by knowledge and
experience.
This broad line of approach received encouragement from findingp
that activity and mental effort increase the tone of muscles not directly

involved in the task (e.9. Freeman, t933, Eason and $7hite, t96t,

Effects of Loding

263

Eason and Branks, 1963), ond ftrrther that irrelevant rnusctrlar activity
may improve performance. For example, Bills (1927) found that adding
and reading were performed faster, and Andreassi (lg6S) obtained
more accurate perception of figrrres exposed for 20 - 6o msec, if the
subject at the same time squeezed a hand dynamometer. Again Boder
(tgfS) found that the rate of tapping with one hand was increased when
a bar was gpsped with the other hand. It is interesting to note that the
increase of speed was due more to a saving oftime at the top and bottom
of the movement than to the movements themselves becoming quicker.
Early work in this field has been reviewed by Courts (tg4z). It is easy
to see in terms of the evidence discussed in relation to the single-channel
hlpothesis in Chaptet 4, how such unrelated activity might disuact the
subiect's attention and lead to poffier performance, but difrcult to see
how it could lead to improvement unless the irrelevant activity produced
some general facilitating spread of effect.
The main impenrs of this approach has undoubtedly come, however,
from two faas: firstly the increase in the dominant frequency of EEG
rhythms from deep sleep, through relaxed waking states to states of

concentrated attention; and secondly the neural mechanism provided
by the pathways benueen sensory input, the brain stem and the cortex,
by which the brain stem produces more or less general activation of the
cortex in association with autonomic aaivity. The extent to which
effo6 muscular tension, EEG rhythms and autonomic activity tie up
together varies in different snrdies but seems clear enough to indicate a
real connection (e.9. pinnss, 196r).
It was originally assumed that because the reticular formation in the
brain stem, which is intimately concerned with arousal, receives fibres
from the main sensory input channels, activation or arousal (the words
tend to be used as synonyms) was directly due to raw sensory input, so
that level of arousal was a direct function of the level of stimulation
impinging on the organism. It is, however, clear from subsequent sttrdies
and from coulmon observation that such an assumption is untenable.
If it were true, we should be unable to sleep in a room overlooking a
noisy street, and nuny people who are used tosuch noise sleepwell. The
stimulation that causes arousal may reasonably be identified as that to
which the organism has not habinrated, that is to ssy, stimuli conveying
information of significance to the organism. This is not quite the same
thing as saying that arousal is a function of presented information in the
information theory sense, because information thus defined deals with
obiective signals without taking the state of the organism into account.
Admittedly such a view, if it is not to be circular, requires us to speci$r

264

Fundo.,rnentals

of Sktll

what is 'of significance' to the subject using criteria other than arousal
effects, but it seems to be possible in principle to do this by taking
account of the relationships between present events and the subiect's
previous experience and funrre aims. In these terms, stimuli which
cause arousal will be those which are novel, very intense or signals of
danger because it is to these that habinration least easily occurs (e.9.
Berllme, t96o, Berlyne et al., 1963, Grim and White, 1965). More
generally one can suggest that zuousal level is raised by any task which
is in some way challenging or demands an effort, and by anything which
acts as an incentive.
There has been some discussion as to how far arousal effects are
general or are confined at different times to different brain mechanisms.
It seems now clear that both very general and more restricted effects
ocsur. We must also remember that the effects we are considering do
not only arise from the arousal system: any spread of effea from aaivity
in the brain and any continuing after-discharges from previous activity
will have something of the same effect. For our present pupose the
important point to note is that all such aaivation, whether widespread
in the brain or not, is more general and diftrse in both space and time
than the detailed signals involved in the perception of specific obiects
and in the taking of particular actions. We are thus essentially discussing
tonic background activity to the more precise signals which carry
information in the brain.

Activation and performance
One of the main difficulties in interpreting the nature and function of
arousal has been that its effects on performance are varied. In particular,
several sflrdies in which performance and physiological indications of
activation such as muscle-tone or skin-resistance have been measured
together have shown that the relation between them is not linear: performance improves with activation from a low level up to an optimum
and then, as activation increases further, performance deteriorates (e.9.
Freeman, 1933, Hebb, 1955, DuEr, 1957, Stennett, rg17rMalmo, 1959).
Snrdies in which only a rise or only a fall of performance has been
shown with increasing activation may perhaps be regarded as having
explored only one part of the whole range.

The mechanism linking rise of aaivation to improved performance

seems to be fairly generally agreed. Freeuurn (tgll) suggested that
activation increasing from a low level might produce nerve impulses
which would subliminally stimulate cells in the cortex and thus render

Effects of Loading

265

them more readily fired by weak siguals than they would otherwise be:

in this way the organism would become at once more sensitive and
more responsive. The same suggestion has been repeated by subsequent
writers, especially by Hebb (1955). The fall in perforuunce at high levels
of activation is recognised as being more difrctrlt to explain. It has been
variously attributed to competition between incompatible habits called
forth by tmduly high levels of facilitation, to fatigue due to overstimulation, to some kind of hormonal exhaustion following excessive
cell activity or to the activity of inhibitory centres which are presumed
to have a higher threshold than centres which produce positive action
so that they come into play only when high activation levels are reached.
None of these theories appears to have been much more than an ad hoc
speculation and the last has the disadvantage of being unduly flexible:
if both facilitation and inhiliiion are posftlated with different thresholds
aLnost any behaviour can be explained.
The simplest way of accounting for the fall in performance at high
activation levels is probably to assume that, if the stream of impulses
impinging on the cortex becomes very intense, the cells there are not
only rendered more sensitive but are acttrally fired. When this happens
the cortex becomes 'noisy' so that signals coming from outside or passing
from one part of the brain to another tend to be blurred. Furtheunore
cells which would otherwise be ready to carry signals are unavailable
because they are refractory from having iust fired, so that the channelcapacity of the brain is reduced. Once this point has been reached, my
ftrrther increase of aaivation tends to impair performance rather than
improve it. The essential feature of this model is that it accounts for
both the rise and the subsequent fall of per{onnance accompanying
progressive rise of activation, without posnrlating more than one type of
process: both follow suaightforwardly from the increase of activity in

the system (Welford, r962a).
The effects of activation on performance would, in terms of this
model, depend on the balance between facilitation on the one hand and
netrral noise and loss of channel-capacity on the other. At the extrems
of both activation and inactivation performance will be poor, but the
type of inadequacy will differ in the two cases. Vhen activation is
trnduly low the system will be inert and signals are likely to be lost either
in the perceprual system or at some later point in the chain tsading to
response. With very high levels of activation the system will be overreactive, action \ilill be confused and capacity will be insufficient for
high-level judgments or precisely graded responses which \pill tend
instead to be undifferentiatd, massive and tense. The measurement of

266

Fundamentals of Skill

such effects obviously cannot be made by merely taking single scores
of overall achievement and relating them to measures of skin resistance,
galvanic skin response, muscle toner.heart rate, suppression of sinus
arrythmia or other indicators of activation. Enough data must be recorded from a performance to see not only what was achieved but also
the detailed manner in which this achievement was attained.
It may be remarked in passing that if the breakdown of high-grade
iudgment under stress can be attributed to a lowering of channelcapacityr alt important new approach is opened up to the problem of
defining 'levels' of mental activity. It is often suggested that in srates of
fatigue and under stress the 'highest levels' are impaired first: what is
meant by 'highest' has not, however, been clear. Attempts have sometimes been made to define it in terms of the newest acquired
ability in the development of either the individual or the race. On the
present view the 'highest' tasks would be those demanding the handling
of most data, or the making of the finest discriminations or which relied
the most heavily on short-term retention and through this on absuaction
and thinking. 'Levels' would in this way be defuredquantitativelyrather
than qualitatively, and we should be freed from any diffic'ulty of trying
to idenury different levels with different parts of the brain.
The great maiority of evidence so far gathered is not detailed enough
to favour one theory of activation effects rather than another, but the
model proposed here does not seem to be contradicted, and in a number
of cases it either accounts for facts not otherwise covered satisfactorily
or provides a more precise lead than others for further research:

(") Speed and accuracy. The most direct evidence in favour of the
present model is probably that of Freeman (rg33). Electric shocks of
various measured strengths were delivered to the subject's right middle
finger, and both the latency and the extent of the reflex movements were
recorded. The shocks were given in pairs and the subiea had to respond
with the left hand to indicate whether the second member of the pair
was equal to or stronger than the first: the reaction-time and accuracy
of this discrimination were recorded. The experiment was performed
under three conditions: (i) 'normal' with the subject sitting in a chair,
(ii) with the subiect 'relaxed' by methods advocated by ]acobson Ggzg)
and (iii) with 'tension' induced by a somewhat complex arrangemenr
which made it necessary to tense the trunk muscles in order to maintain
balance. The results are shown in Table 8.r from which it can be seen
that with increase of tension from the 'relaxed' through the 'normal' to
the 'tension' condition the latency of the reaction to shock and the

Effects of

Loading

z6T

TABLE 8.t Effects of relaxation and of induced muscular tension on reactions. Data front Freqnan (tg3.j.). Each entry is the mean of zo readings

from each of 9 subjects

Relaxed Normal
Muscle-tonus in quadriceps
(arbitrary units)
Reaction time to shocks delivered
to right middle finger (msec)
Extent of finger withdrawal to
these shocks (mm)

Time taken to discriminate
whether second shock of pair
was equal to or stronger than

first (msec)

Accuracy of this discrimination

(%)

$Iith induced
muscle
tension

-29.r

* r8.z

+64.9

4ol

z1g

247

ro.3

28.o

27.5

6l+

442

355

65

s7

6g

discrimination reastion-time shortened and the finger movements became greater, all of which would be expected from the facilitation due
to increased activation. The accuracy of discrimination rose from the
'relaxed' to the 'normal' condition but fell again to the 'tension' condition: in short, the increased tension made performance faster but at the
expense of accuracy. This is what would be expected if it produced
both facilitation of astion and noise in the systems involved in disctiminating signals and selecting which responses to make. If alterna-

tively accuracy was maintained, reaction time would lengthen. This
result occurred in a later experiment by Freeman (lg+o) in which
reaction time was found first to become shorter and then longer again
with progressive fall in skin resistance. The shortening of reaction time
was confirmed by Andreassi (1966) but not the lengthening, possibly
because his subjects did not attain as high levels of tension as did
Freeman's.

(b) Sutsory thresholds. Increases of arousal from very low levels obviously improve sensory discrimination, but the effects of smaller changes

on sensory and percepnral functions seem to have been very little
shrdied and the results of snrdies that have been made are conflicting.
It would be reasonable to expect that thresholds would be lowered as
arousal rose, and Symons and Mackay $962) have produced what is
perhaps evidence in favour of this by finding that auditory thresholds

268

Fundammtals of Skill

were lower when subjects were doing mental arithmetic than when they
were relaxed. Yet Cohen and Lindley (rgf6) fotrnd that the threshold
for deteaing vibration on the skin was raised by increased muscletonus. Taking the model we have outlined in coniunction \rith the
signal-detection model outlined in Chapter z (p.3r) we can see why
results are tikely to be conflicting. Any mild increase of sensitivity
brought about by arousal would probably increase both noise and signal

x
tr-

o

u-,

I

f
J

t

s
zrU

t
I

I
,
t

I

()

,

t

I

,

I
t

z.

,
,

u-

o

I

t

,

F
J

6
co

o

d.
o-

xc

x

Figrrre 8.9. Hypothetical effect of arousal on the curves of the signaldetection model shown in Fig. 2.4 (p.38). Increased arousal is assumed
to expand both the 'Noise Alone' and the 'signal plus Noise' curves in
such a way that d' remains relatively little affected.

in roughly the same proportion, since the increased likelihood of cells
being fired would raise the rate of random firing as well as of response
to signals from outside. Any externd noise would similarly be increased
in suict proportion to the signal. The result would be to expand the
cnrves in Fig. 2.4 (p.38) as shown in Fig. 8.9. Such expansion since
it affects both the means and *re variances proportionally would leave
d' relatively limle affected. If, however, the cutoff point remained in
the same position there would be a substantial fall in fi leading to a
higher rate of signal deteaion but also to a larger proporti,on of falsepositives. fn short, increased arousal would not greatly affect capacity for

discrimination but would lower the criterion of iudgment. It seems

reasonable to suppose that, since the criterion represents the threshold

Effects of Loading

269

of some later stage in the chain of cenual processes, it should remain
the same and might even be lowered by increased arousal, if so it
would accennrate the process we have described. If, on the other hand,
the subiect took steps to keep his false positive rate steady by raising
his criterion-level his threshold would be little affeaed by mildly
increased arousal.
Very high degrees of arousal seem likely to raise the level of neural
noise more than that of the signal, thus leading to a fall of d' in addition
to the change of p dready discussed.

(r) Results of stress. From what has been said we should expect that
stressing agents such as dangers, annoyances and incentives which
increase arousal would all in mild degrees improve performance and
in more severe degrees impair it. We cannot here attempt a full discussion of suess effects but will mention a few points briefly. The revival
of interest in activation and arousal seems to have come largely from the
suggestion by Lindsley (lgSt) that the excited emotions of fear and
anger can be thought of as states of heightened arousal. Such a view
provides, in terms of the model proposed herer 8 solution of the longstanding controversy about whether emotions 'organise' or 'disorganise'
performance. Mild degrees of emotion, producing moderate degrees of
arousal, are likely to be beneficial and therefore 'organising': more
severe degrees will lead to disruption of perforuumce and thus be 'disorganising'. This view tallies well with the observations made by Mira
(tg4) of fear reactions to civilian bombing during the Spanish Civil
'W'ar. Mild degrees of fear, he suggested, led to prudence and selfrestraint, concentration and caution, and tended to improve achievement. More severe degrees led to anxiety and darm accompanied by
breakdown of high-grade skills, disorganised aaivity and uemors. Very
severe degrees resulted in panic marked by uncontrolled activity and
loss of memory afterwards of what had happened, or in extreme cases
led to collapse into a kind of tense shrpor. The more severe states of
fear and anger resemble those of animals from which the cerebracortex has been removed and led Darrow (1935) to describe these emotions as states of 'relative firnctional decortication'. This is effectively
what the model proposed here posnrlates.
It should be emphasised, however, that the concept of arousal and
activation applies also on a much more subtle scale. We have already
noted that smaller but siguificant changes of skin resistance, heart rate
and other indices of arousal can be observed with tasks such as mental
arithmetic. Perhaps the most elegant example of such changes is in

2To

Fundamentals of Skill

car drivhg, studies of which have shown the level to fluctuate continually: every minor event such as passing or being passed, the approach
to traffic lights or sight of a car turning in from a side road lowers the
driver's skin resistance, which rises again during clear nrns (Hulbert,
1957, Michaels, 1960, Taylor, \964).

(d) Interacting factors. Relating stress effects to objective conditions is
not always easy. For example loud noise might reasonably be expected
to be arousing and has, indeed, been shown to be capable of counteract-

ing some of the lowering of arousal which results from lack of sleep
(Wilkinson, 1965). Yet it often fails to have an effect, seemingly because
of the ease with which habituation can occur to at least some noises.
The beneficial effects on performance that might be expected to follow
from moderate noise levels are also liable to be offset by the raising of
B that may occur with noise (see p.47) and by the disuacting effects of
noise discussed in Chapter 4 (p. r33).

A different type of complication is illustrated by the finding that

tension and arousal levels are often higher before an exacting task is
begun than they are once it has startedr or that they decline as performance proceeds (Freeman, $33rDeane, 196r, Zimny, r965rPugh et al.,
1966) or the task is learnt (Freedman et al., 1966): preparation for the
task and the initial stages of performance may be more exacting than
later stages when initial uncertainties have been resolved and some
learning has taken place. A further anticipatory effect has been shown
by Forresr (rgS8) who found that muscular tension increased with the
length of task to be completed.

(r) Effects of dfficulty of task Freeman (1938) suggested that the optimal degree of muscle-tension, marking the uansition between an improvement and an impairment of performance, becomes lower as the
difficulty of the task increases. This view, although not very convincingly
supported by Freeman's own data, provides a plausible explanation of
the classical work by Yerkes and Dodson (19o8) on the learning by
dancing mice to discriminate between different brightnesses. The mice
had to learn to go to the brighter of nro boxes and received an electric
shock for entering the wrong box. The number of trials taken to learn
when the two brightnesses were very different and the task therefore
easy, fell progressively as the strength of shock was increased. For more
difficult discriminations the trials fell as the suength of shock increased
up to a point, but thereafter rose. The optimum shock was weaker for
a very difficult discrimination than for one of moderate difficulty. These

Effects of Loading

27r

results have been repeated subsequently with both animals (e.9. Broadhurst, 1957, Denenberg and Karas, 1960) md, using rather different
tasks and incentives, with human subjects (e.9. Castaneda and Lipsitt,
1959, Tecce and Testa, 1965, Elliott, 1965). Similar implications follow
from the work of Ray (tg6S) who has shownthat increasingpressure for
speed leads to progressively poorer performance at solving relatively
difFcult problems. Again Stabler and Dyal (tg6f) fonnd that reaction
time was longer among highly arxious, than among less anxious subiects early in practice, but that the positions were reversed later when
the task can be presumed to have become easier.
These results are susceptible of two explanations in the terms we have
been using and both may be correct. The first explanation assumes that
level of arousal rises with suength of incentive: a high level of incentive
which resulted in the brain becoming somewhat noisy would have little
effect on the performance of an easy task requiring relatively little
channel capacity, but would have a much greater effect as the required
capacity increased. Secondly it might be argued that the tasks dlemselves induce a degree of arousal which rises with their difficulty and
that this is added to the arousal produced by the incentive. If the optimum arousal level was the same for all degrees of dffidty, the addition
due to incentive that would produce this optimum would fall as the
task became more difficult.

(f) Relationship with fatiguc. ff activation and arousal outlast the circumstances grving rise to them, they will be liable to build up progressively during performance unless time is allowed for them to die away.

If so, they can be reasonably equated with the activity posnrlated in
the second of the two theories proposed to account for mental fatigue.
The main activation in fatigue would presumably result from effects
spreading from the fatiguing activities themselves, but arousal due to
annoyances or to discomforts resulting from prolonged sining or standing in awkward posnres might well add to it. Such activation and
arousal would, in the Yerkes-Dodson principle, account for greater
fatigue effects with more difficult tasks requiring greater channelcapacity (cf. Singleton, 1953) and for the fact that fatigue may attack
the higher co-ordination of performance while leaving more elementary
details unimpaired (Bartlett, 1943).

272

Furdamentals of Skill

Effects of long-continued stress
Animals that have been used in studies of 'experimental neurosis' have
tended to show fear and other stress symptoms for long periods, sometimes for years, after trials have ceased: in shortrthe stressin theexperimental episodes has had a lasting effect (for a review see Liddell, r9M).
Long-lasting effects can perhaps also be observed in some human beings
engaged in hazardous or over-exacting work who seem to 'lose their
nerye' after a time. For example, some of the airline pilots who about
20 years ago were being compulsorily retired at the age of 45 would
guardedly admit that they did not resist retirement because, over the
years, they had been losing the will to fly. The possible dangers outlined in briefings before flight seemed to loom larger and larger, and
the responsibility of the plane with its load of passengers began to weigh
heavily upon them. Again Hauty et al. (rg6S) have shown that symptoms
of suess among air-uaffic controllers increase with age and length of
time on the iob, especially the latter.
The traditional explanation of these states is that stress reactions have
become conditioned: that is to ssy, they have become firmly associated

with certain sinrations, and occur in these whether they are really
iustified or not. Clinical studies of the effects of prolonged stress on
human beings suggest a chronic state of heightened autonomic aaivity
to which the subiect's other bodily mechanisms adapt to some extent
but not completely (e.9. Selye, r95o). The same appears to be true
of laboratory animals kept under stressful conditions (e.9. Thiessen
and Rodgers, 196r, Thiessen et al., t96z).
To some extent at least, these effects can be attributed to overdevelopment of the adrenals in stressful conditions and thus to a raising
of the subiect's level of adrenaline, but it seems likely that some neural
mechanism must also be involved. We may perhaps argue that trains of
impulses passing synaptic iunctions make these more readily passed on
subsequent occasions and that, although resistance may be restored
with disuse, it is more readily lowered again on a subsequent occasion
(see Eccles, 1953). This princrple has been of the greatest importance
in the development of some modern theories of learning (e.9. Hebb,
r94g). It is held that the material to be learnt produces a pattern of
neural activity and this in turn gives rise to a pattern of lowered synaptic
resistances which constinrte an enduring memory trace. It follows from
this princlple that the prolonged use of particular pathways, such as
those involved in general activation and arousal, will tend to facilitate
their aaivity. ft also follows that continued unpatterued neural aaivity,

Effects of Loding

zl3

such as might occtrr with chronic over-activation due to stress, would
tend to produce widespread random lowering of s5maptic resistances in
the brain. If this occurred to a moderate degree it might merely render
more sensitive and responsive a brain that would otherwise have been
relatively inert. If it occurred to a more substantial degree as a result of
prolonged severe stress, it could bhu patterns of activity in the brain by,

in a sense, breaking down the norrnal 'insulation' of one cell from
another. In either case the effects on behaviour would resemble those
of chronic heightened arousal which in severe cases would reach the
level at which it impaired performance.

VIGILANCE

The extensive experimental work on the maintenance of vigilance that
has been done since the Second lforld War arose from the fact that
when radar and other watctrkeepers look for infrequent signals they
tend to miss them when they come if they have been on watch for an
hour or so. Research in this area has gained attention not only because
of its interest to the armed servicsr but for its obvious implications for
inspection and monitoring work in industry and because it raises important theoretical issues.
The initial experiments on synthetic radar displays and other laboratory tasks made it clear that the proportion of signals detected fell
sharply over a period of + hr or so when the signals were faint, brief and
ffiequent. With synthetic radar displays showing signals at unpredictable intervals, it could be argued that the subiect, in order to be sue of
seeing them when they ocsrured, would have to make a rapid and continuous series of checks. These could be fatiguing, and the fall in performance might therefore be a fatigue effect. Other laboratory tasks
have, however, presented the subiect with clearly obsenrable siguals at
discrete intervals and required him to detect occasional signals that are
slightly different from the maionty. The intervals have then been too
long (r to 5 sec) to make a fatigue theory tenable: some other factor is
clearly involved. The problem appears to be one of rmdqload ruther
than ooerload.
Of the 'pure' laboratory tasks probably the best known is the original
Clock Test used by Machrorth (r95o). The subiect sat facing a circular
dial on which a Gin pointer moved in steps of tz' of arc once per sec.
Occasionally the pointer made a iump of z4', and the subiea's task was
to report these by pressing a key. The double movements occurred
tn elve times in zo min (r in zoo movements on average) at irregular

274

Fundammtals of Skill

intervals. Other tasks have involved the use of a wide variety of signals,
such as faint spots of light (e.9. r$(/ilkinson, 196l) or small changes of
sound (e.9. Mackworth, r95o), or occasional features of a regularly
presented series of signals, such as the detection of 'odd, even, odd' in a
series of digits (e.g. Bakan, rg12). The decrements seem to occur with
both visual and auditory signals and may be shown not only by signals
being missed altogether, but by slower responses to those detected
(Wallis and Samuel, 196r, McCormack, tg6z, Surwillo and Quilter,
1964, Buck, t966). Failures are not the result of any peripheral factor
such as omitting to look at the source of signals, since Mackworth et al.
(rg6+) have shown that visual signals are often not reported even though
they are being fixated at the time: some central failure appeus to be
involved. The fall of performance seems to follow a different pattern
from that of simple, monotonously repetitive tasks such as adding,
simple assembly or bean-sorting (Baker and Ware, ry66). It is not,
however, due simply to lack of activity on the subiect's part; thus,
$flhinenburg et al. (rgS6), when they repeated the Clock Test requiring
the subject to respond to every jump, making different responses to
small and large, fotrnd that the number of occasional large jumps
detected still fell with time on watch. Nor is it due to lack of readiness
for the signalsr 8s Sfilkinson (196r) showed that decrements still
occurred when warnings of signals were grven or even when the subject,
himself determined the times at which they appeared: the subiea, needless to say, did not know whether or not any partictrlar signal would
require a response, but he had every opportunity to be ready if it

did.
Dectements tend to be less when signals are relatively strong (Mack-

worth, r95o) or of long duration (see Broadbent, 1958) or when, in a
series of signals some of which are 'wanted' and some not, the propor-

tion of wanted signals is raised (Colquhoun, t96r, 1966) and these
factors appear to interact in the sense that signals can be detected at a
lower physical intensity when they are more frequent (Martz, t966,
tg67). If the rate is raised very high performance Dsy, of course, fall
again due to the subiect not being able to deal with all the signals
presented in the time available: there is thus an optimum rate of
presentation which avoids both overload and underload (Poulton, 196o).
Several experiments have explored the possibility of raising detection
rate by combining the vigilance task with another - in other words
raising the wanted signal rate by giving siguals from two different tasks.
The results, however, are equivocal (S7allis and Samuel , tg6trAntrobus
and Singer, 1964), probably because any improvement due to the extra

Effects of Inading

2TS

signals from the second task may be offset by the subiea having to
divide his attention between the two tasks.
Mackworth showed that performance could be temporarily restored
by a telephone message during the 'watch' and later research has confirmed that periodical changes of aaivity (Bevan et al., 196T), brief
rest pauses (Bergum and Lehr, ry62) or the prsence of others in the
room (Fraser, 1953, Bergum and Lehr, t963a, Williams et al., 1965)
which may be presumed to introduce momentary disuactions, serve to
maintain performance. Mackworth further showed that decline in performance could be prevented by telling the subiect, each time after a
sigual had appeared, whether or not he had detected it correctly - in
other words by gling him 'knowledge of results' of his performance.
Subsequent work has shown that detection is improved even if knowledge of results is incomplete (McCormack et al.r tg63rMcCormack and
McElheran, 1963, Wilkinson, 1964) or false (Loeb and Schmidt, 1963),
and that detection rate in a primary vigilance task rises if knowledge of
results is given about a secondary task performed at the same time - the
effect of the knowledge of results seems to spread to both tasks (Baker,

r96r).

Theories of vigilance

Theories to account for the declines observed in vigilance have been
critically discussed by Broadbent (lgS8) and by Broadbent and Gregory
(r963b), and we shall not attempt to go over the grotrnd again here.
Broadly speaking the most plausible view seems to be that originally
proposed by Deese (tgSS) that decline of vigilance represents a fall in
the level of arousal or activation. Several other theories have, however,
been advanced and we shall consider four of them briefly before rettuning to the arousal theory:

(") Motiuation. Mackworth compared the vigilance sinradon with that
of a conditiorring experiment in which the'reinforcing agent' or 'reward'
has been discontinued so that the response is inhibited or 'extinguished'.
He argued that most of the watch for signals was unrewarded so that the
decline of perforurance should be attributed to inhibition which could
be disinhibited by brief interruptions and dissipated by longer rests.
This theory in its simple form is little more than an ad hoc guess: it
does, however, gain some support if the extinction situation is spelled
out a little by saying that lack of reinforcement leads to decrease of
motivation. The decline of vigilance with time has been reduced or

216

Fundammtals of Skill

abolished with better grade subiects who might be regarded as more
highly motivated (Kappauf and Powe, $59), or when army trainees
performed their task in the presence of an officer (Bergum and Lehr,
1963a)r or when subiects received money rewards for good detection

(Sipowicz et al., t96z), although this last effect may be short-lived
(Bergum and Lehr, 1964). !7e have already noted that it can be prevented by knowledge of results - a factor known on other grounds to
act as an incentive (Helmstadter and Ellis, 1952, Gibbs and Brown,
1956, Spielberger et al., 1966).

An explanation in terms of motivation poses the further question,
however, of how motives affect behaviorlr. \U(re have already seen that
one of their effects appears to be an increase of activation, so that a
motivation theory of vigilance might be regarded as a sub-class of an
aaivation or arousal theory - anything which increases motivation will
tend to offset any fall of aaivation during a prolonged watch.
(D) Expectancy. Several experiments have shown that the level of performance at vigilance tasks and the extent of the decline with time are
to some extent a function of the rate of signals expected by the subiect
on the basis of instructions or previous experience (Broadbent, 1958,
Colquhoun and Baddeley, I 964, 1967). The same idea has been applied
to individual signals by postulating that the likelihood of their being
missed depends on the predictability of the moment at which they will
arrive (IGppauf and Powe, 1959, Baker, 1959), although this latter
result is contnoversial (Boulter and Adams, 1963). Furttrer evidence
for expectancy effects is the 'end spurt' of improved detection when a
watch is known to be nearing its finish (Bergum and Lehr, r963b).
Expeaancy has also been proposed as a reason for some of the errors
made in industry and in the driving of transport, for example a uain
driver who is not used to being stopped at a particular signal may not
notice when, exceptionally, it is against him (Davis, 1958, t966, Buck,
1963). Expectancy might perhaps also explain the finding by Hardesty
et al. (tg6f) that the improved performance in a monitoring task produced by knowledge of results persisted for a substantial time after
knowledge had been withdrawrl.
This theory again, however, leads to the further question of how
expectancy exerts is effects, and from what has already been said earlier
in this chapter it seems reasonable to regard it as raising or lowering
activation level - the rises or falls of activation may anticipate the onset
of the task to which they refer. If so the expectancy theory becomes, as
Deese (tgSS) recognised, another special case of the activation theory.

Effects of Load,ing

z7T

G) Blocking. The fact that signals are missed less frequently when they
are of relatively long dtration led Broadbent (rgS8) to suggest that
they are missed because they come and go again during brief lapses of

attention which inctease in frequency as a watch proceeds. Such lapses
have considerable innritive support and some fiuttrer evidence of their
occlurence has been put forward by ]oh'rs1on et al. (t966). From our
present point of view the crucial question is why such lapses occur.
We shall consider four possibilities:
(i) The mechanism responsible for percephral seleaivity fatigues if
attention is concentrated for a long time upon one class of stimuli, so
that it temporarily breaks down allowing other stimuli to gain ad.mission.
(ii) Discomforts due, for example, to sining for a long time in one
position build up to a point at which they are strongenoughtoovercome
the bias against them produced by the selection mechanism. When they
gain attention they caPttrre the central mechanisms in the manner
outlined in Chapter 4 so that there are brief periods during which anention to other incoming signals is delayed, and dtring these delays they
may be lost.
The weaknesses of both these hlpotheses is that they glve no reason
why the decline of vigilance should be reduced or prevented by faaors
such as increase in signal frequency or knowledge of results.
(iii) It has been suggested by Hebb (lgSS) that arogsal level tends
to be self-adiusting with the organism tending to seek an opdmtrm.
Since novel stimuli tend to be more arousing than long continued]ones
(see Berlyne, 196o), a fall of arousal during a prolonged watch would
increase the tendency to seek a change of s 'mulation. Such a view is
questionable, however, following work by ]ohnston et al. (rg6) who
found that in a vigilance task in which some stimuli were repetitions of
a signal that had appeared before and others were new, it was only the
latter that were detected less well as the watch proceeded.
(iv) Activation level shows not only broad changes over relatively
long periods, but also moment-to-moment fluctuations dtging which it
might well fall occasionally to a level at which the system became so

insensitive that incoming signals were temporarily blocked. These
blocks would obviously become more frequent as the general level of
activation declined. This seems to be the most plausible of the views we
have outlined. ft does, of course, once more make the blocking theory
a particrrlar case of the activation theory.

(d) Changes of catoff in signal-detection. Several exami"ations have been
made of vigilance decrements in terms of the sigud-detecrion mode

278

Fundamentals of Skill

described in Chapter 2 (Broadbent and Gregory, r963b, Loeb and
Binford, 1963., Taylor, 1965, Binford and Loeb, t966, Colquhoun,
196l). All these found that as a watch proceeded the number of both
detections and false positives fell in such a way as to indicate a rise of
B but little if any change of d' - in short the vigilance decrement was
due to fewer signals passing the criterion and not to any true change of
ability to detect them. This result is clearly consistent with a fall of
activation level as indicated by Fig. 8.9: the distributions of both 'noise
alone' and 'signal plus noise' are moved to the left in such a way as to
leave the signal-to-noise ratio nearly the same but the cutoff point
apparently raised.
This model can also account for a further result obtained by Broadbent and Gregory, who found that when the subiect had to rate the
cerrainty with which he detected a signal on a five-pcint scale from
'Sure', through 'Uncertain' to 'Sure not' the change of P for the most

stringent criterion was greater than for the least, which was little
affected. The curves of Fig. 8.9 are redrawn with two criteria instead of
one in Fig. 8.ro. It can be seen that when the curves change from one
position to the other, B changes more for the upper criterion than for
the lower.

The picnrre is not always as clear cut, however, as Fig. 8.ro would
suggesr. For example Chinn and Alluisi (lg6+) who studied effects of
differenr rypes of knowledge of results on vigilance, obtained evidence
which suggests that the cutoff point may be moved and that d' may
also be affected. Informing subjects of the occasions when they responded in the absence of a signal reduced the number of false positives
and increased the number of signals missed as would be expected from a
change of cutoff brought about by 'punishing' false positives. On the
other hand, information about correct responses or missed signals
reduced both the number of signals missed and the number of false
positives implying an improvement of detectability - that is an increase

of d'. It seems fair to regard these results as complications due in the
first case to factors affecting the setting of the cutoff as discussed in
Chapter z, and in the second to the improvement of performance
brought about by detailed knowledge of results which we shall consider

in Chapter 9.
What appears at first sight to be evidence clean contrary to the model
proposed in Fig. 8.ro is provided by a series of studies which have
shown substantial falls of d' with time during a watch (Mackworth and
Taylor, 1963, Mackworth, r964br 1965). The task used in these experiments, however, was that of deteaing momentary interruptions of the

Effects of Loading

z7g

movement of a hand revolving round a dial and there was thus a risk of
missing signals if the watch was not completely continuous. It was
thus likely to have been fatiguing and in this respect differed from the
tasks of the experiments mentioned in the previous paragraph: in these
the points of time at which the signals might come were resrricted so
that the watch could be intermittent. Mackworth (1964c) compared
her results with those of a ntrmber of other experiments on fatigue

including the visual acuity data of Berger and Mahneke shown in

,

,

lr

,

o

,
,
,

)=

TI,

,
,
,

t
t

t
,

zlrl

t
t
I

o

t
,
,
,
,
t

z-

,
,
I
t
t

lJ-

o
F

,

t
t
,

-)
cO

o

d.
A

LESS

STRINCEM STRINCENT

x

CRTTERIA

Figure 8.ro. The same curves as in Fig. 8.9 but with two different ctiteria.
As the curves move to the left from the positions indicated by solid
lines, 0 for the more suingent criterion changes from about r.z to about
3'3 - a difference of 2'r. B for the less stringent criterion changes from

about'2 to '4- a difference of only .2. d'changes from r.4 to r.3.

Fig. 8.2, and data from a tracking task, and found that in all of them
log d' declined linearly with the square root of time on the task. These
results of Mackworth's are especially interesting as pointing to one of the
very few means at present known of distinguishing fatigue decrements
from those due to loss of vigilance: fatigue appears to result mainly in a
fafl of d'implying a true impairment of function, while loss of vigilance
leads mainly to a rise of P indicating a condition in which there is little
change of basic discriminatory power but ferver signals pass the criterion
of iudgment.

28o

Fundamentals of Skill

(r) Actiaation and ArousAl. We now return to the activation theory itself
and look at the evidence which bears directly on it. Suggestions in
favour of such a theory are given by some of the results already noted,
such as the prevention of a fall with time in detection rate by either
stronger signals or the presence of others in the room or an occasional
telephone message or knowledge of results, all of which involve some
increase in the amount of stimulation from outside impingrng on the
subiea. Again Adams et al. (1961) found that reaction times to changes
in a display lengthened during a 3 hr watch when the changes had
simply to be reported, but did not lengthen when subjects had to report
more detail about the nature of the change which had taken place.
Increased sensory input does not, however, always abolish decrements
of performance, presumably because habituation can occur to input
which is steady, regular or otherwise capable of being excluded. For
example Kirk and Hecht (lg6f) found that the proportion of visual
signals detected was higher when subjects were listening to a noise of
randomly varying intensity than with a steady noise of the same average
intensity. Complications may also occur when the conditions are
extremely unarousing as in so-called sensory deprivation experiments.
For example Smith et al. (tg6il noted that their subiects who were
given a vigilance test, in which faint irregular signals had to be detected,

after 72 hr of living under dark, quiet conditions, performed better
than did controls who were tested after living trnder more normal conditions. Why this was so is not clear: it may have been an indication
that underarousal is itself arousing so that the organism tends to main-

tain an optimum state as suggested by Hebb (1g55). Perhaps more
simply, it may have been that the subjects had become adapted to the
uneventful nature of their envirorunent so that any event became more
noticeable. If so we must postulate an adaptation which takes longer
than the 23 hr which is the usual duration of a vigilance test.
More positive evidence for an activation theory is contained in the
fall of skin conductance dtrring the course of a watch noted by Nishioka
et al. (196o), Dardano (1962) and Eason et al. (tg6S) although correspondence is not always as close as one might wish and some other
measures of activation show little change: for example Eason et al.
fonnd heart-rate to be trnchanged. Possibly, as Nishioka et al. suggest,
these rather fickle measures are sometimes disnubed by other irrelevant factors or by compensatory efforts to restore a performance recognised as failing. Such compensation seems to be the reason why loss of
sleep, which in some circumstances markedly increases vigilance decrements, does not always do so - the subiect's efforts to keep awake pre-

Effects of l^oad,fug

z8r

vent the lowering of arousal level (Wilkinson, 1965, Williams et al.,
1965). Compensatory tendencies may also be the reason why Micko's
(1966) subiects tended to pay more attention as time passed to iokes
proiected on to a screen during an auditory vigilance task. The same
explanation may also resolve the conflict of evidence about the effects of
increased body temperature on vigilance: Bell et al. (tg6+) found that
vigilance deteriorated as body temperanue rose, Fox et al. (lg6f) found
it did not but noted that their subiects tended to become restless and
irritable with increased body temperatune. Some vigrlance tasks may
also be fatiguing: if so, and if fatigue tends to increase activation, we
should have the rather curious case of one potential source of poorer
performance tending to compensate for another.
Perhaps the most striking direa evidence in favour of the activation
theory is Mackworth's (lgSo) finding that no appreciable decrement of
performance occurred during a z hr watch in subjects who had been
given ro mg of benzedrine shortly before the experimental session
began: the drug, which has a known stimulating effect on the arousal
mechanisrrlr seems to have prevented any fall of vigilance. Complementary evidence is contained in the greater decrement found after
taking r mg of hyoscine hydrobromide (Colqutroun, ry62). The effects
of benzedrine further serve to distinguish vigilance dectements from
impairments due to fatigue since Davis (rg+8), whose experiments with
the Cambridge Cockpit have already been mentioned (p. 252) fotrnd
that the drug had no consistent effect on the deterioration of performance in his task over a z br period.
The most difrcult data for an activation theory to accommodate are
from experiments in which a decline of vigilance has occurred even
though the task has been such as to keep the subiea more or less con-

tinuously active (e.9. Whittenburg et al., 1956, Adams and Boulter,
rg6zrAlluisi and Hall, rg63rWiener et al.r t964). These asks obviously
produced a stream of kinaesthetic and other signals which might reasonably have been expected to maintain arousal if sensory input alone was
sufficient to do this. As we have seen, it is not: habinration seems often
to occtu readily and we may suppose there can be habinration to the
feedback from repeated activity iust as there is to repeated external
stimulation. If so, however, and activation falls with time, how is the
aaivity required by the task maintained? It seems fair to suggest from
evidence of fatigue effects that relatively simple, repetitious, actions
can be well maintained at a lower level of neural funaion, and thus
at a lower level of arousal or activation, than higher grade and more
complex iudgments requiring greater channel capacity.

zLz

Fmdamentals of Skill

'FATIGUE' AT woRK

Much of the original interest in fatigue was directed at the elimination
of accidents and poor productivity in industry, and a number of distinguished field strdies have been done in the pursuit of this interest.
From the theoretical point of view indusuial tasks have the advantage
of enabling performance to be snrdied over much longer periods than
are normally possible in the laboratory. It is clear, however, that what is
commonly regarded as industrial fatigue contains many diverse elements
and that, iust as in the laboratory, closely similar'phenomenaof slowing,
blocking and disorganisation of performance can have widely different
causes. !7e shall in concluclirg this chapter look briefly at work in this
area in order to illustrate the subtle interplay of fatigue, vigrlance and
other factors which make their snrdy at once challenging and rewarding.

Output measures at different times during a shift
Presurrrptive evidence of some kind of fatigue effect during an industrial
shift is contained in the classical reports of the Industrial Fatigue Research Board (later reuamed Industrial Health Research Board). These
reports showed not only that shorter working sffis led to higher hourly
oulput (Osborne, r9r9, Vernon, r9r9), but also that a net reduction in
working hours could sometimes lead to a net rise in total ouqput

(Vernon, rgzoa., b). ft seems as if industrial workers tend to anticipate
fatigue and to distribute their efforts over a working shift in much the
same way as do long-distance runners and cyclists and subiects in some
experimental tasks: they work faster throughout if the shift is short.
It was also found in many operations that the rate of production rose
a little at the beginning of the sffi, presumably owing to some 'warmup' effect, fell towards the end of the morning, recovered somewhat
during the lunch break and fell again during the afternoon. The falls
could be largely prevented by inuoducing brief rest-pauses of 2 to r5
min shortly before the fall would otherwise have begun (Vernon and
Bedford, t924, Wyatt and Fraserr 1925, \[/yatt, rgz7). The early work
has been strmmarised by Chambers (1961) and other writers on indus-

uial psychology.

The interpretation of these findings is far from easy. I7hen work
involves heavy musctrlar effort, or is carried out in hot and humid
conditions, heat stress may slow work down or make pauses necessary.
This, however, cannot be the complete explanation of the changes
found, because many of the iobs were light and done under reasonably

Effects of Loadirrs

283

easy envirorunental conditiorrs. The fact that 'spurts' were sometimes
found to occur during a shift, or even right at the end, need not weigh

against a fatigue hlpothesis in view of what has been found about
motivation effects in laboratory snrdies of fatigue. Some effects may
arise from physiological rhythms over the z4 hr and from physiological
changes related to time from last meal; in this connection we may note
that Wyatt and Fraser Ggz1) found breaks more effective if refreshments were taken. Again, however, such factors are not a sufficient
explanation of the changes in production obseryed.
Apart from the suggestion of a tnre fatigue effect, two main alternative explanations have been offered.

(") Monotony. Snrdies by Vernon $924) and by \[yatt and Langdon
(tglz) showed that output per hotrr could be raised by switching from
one iob to another, md Wyatt and Fraser (1928, also \[yatt et dl.,
t929) showed that switchi'rg also reduced irregularity in the speed of
individual cycles in a repetitive operation. These findiogp have been
taken to support the hypothesis that the falls of production towards

the end of a sffi are due to monotony, although in view of what is now
known of the specificity of fatigue effects, they could still reasonably be
attributed to fatigue. There is, indeed, some support for this view in
the finding by Wyatt and Fraser (1928) that the benefits of switching
were reduced if the iobs concerned were closely similar.
More telling evidence in favour of a monotony theory is that obtained
by Wyatt and Langdon Gglil, who compared ouqput figures with
answers to questionnaire items about the incidence of boredom and
dso showed that output could be raised by 'music while you work'.
Wyatt and his co-workers have further shown that dectements of performance tend to be greater among workers of higher intelligence who,
it might reasonably be presumedr are less 'absorbed' by their work.
It seems fair to say that present evidence does not enable us to distinguish clearly betrreen fatigue and monotony as explanations of these
findings; it would be necessary to do much more detailed snrdies based
on the factors fotrnd to be significant in laboratory experiments.

(b) Ancillary actizritres. Dudley (tgSS) has suggested that the variations

in output over a shift may not be due to fatigue or monotony but to the
way in which the iob is organised. fn his snrdies of repetition work he
found the usual oulput curves showing a rise at the be#nning of the
shift and a fall towards the end. However, when he made detailed
studies of the work done, he found that the times for individual work

284

Fundamentals of Skill

cycles did not vary. In short, the initial rise and later fall were due to
production work being interspersed at the begiru,ing and end of the
shift with other activities, such as getting out and putting away tools.
These sttrdies emphasise the importance of looking at iobs in detail,
but they do not necessarily preclude an explanation in terms of fatigue.
!7e have seen in the experiment by Singleton (1951) that there was little,
if any, change in the speed of acttral movements made during a working
spell: fatigue effects showed in lengthsning of the times betrreen one
movement and the next. Although the time scale in Singleton's study
was much shorter than in Dudley's, it seems possible that fatigue effects
did not affect rycle times in the operations Dudley studied, but tended
to make the men turn briefly to other activities as a means of taking a
rest. Such rests might not be taken consciously: the whole process was
more likely to be unwitting. Certainly Dudley's explanation seems
hardly adequate to account for the substantial effects of brief rest
pauses foturd by Vernon, Wyatt and their colleagues.

Performance of long-distance lorry drivers
Several researches on long-distance lorry drivers in the United States
have been summarised by McFarland and Moseley (lgS+) and discussed

together with some other results by Crauford (196r). Lorry drivers
tested with various sensory-motor tasks after long periods of driving
have shown changes of performance similar to those fotrnd in laboratory
e4periments on fatigue. Comparable changes have also been fotrnd in
sttrdies of driving. Lauer and Suhr (tgS8) have shown that ill effects
can be greatly reduced by frequent rests. On the other hand, accidents
and near accidents have been found to be more frequent at the beginning of a trip, or after only a few hours of driving, than at the end of a
long haul: in one study, McFarland and Moseley observed zz out of 48
near accidents to occur during the first z hr of driving and only four in
the last z br of a g hr haul.
It is difficult to say how far the effects of prolonged driving are to be
attributed to true fatigue effects and how far to monotony and loss of
vigilance. 'We have already mentioned that Crawford (t96r) suggested
that fatigue may result from stress and annoyance caused by other
road users. On the other hand, certain 'hypnotic' effects of longdistance driving which have been observed indicate that monotony
may be important. It is perhaps reasonable to suggest that driving is
both fatiguing and sometimes, especially on long American roads, morxF

Effects of

Loading

285

tonous, and that the observed effects are a balance between or combination of the two.

Long-term indusuial fatigue
About the long-term tiredness, lassinrde and lack of enthusiasm often
termed 'chronic fatigue' there is little systematic knowledge. Subiectively, such states seem to arise from long-continued overload, leadi"g
to mild chronic overtenseness, which in severe forms can lead to inability to concentrate or make decisions, irritability and feelings of
futility and seem to be effects of chronic stress. Perhaps these effects
should be regarded as examples of the second t]rye of mental fatigue
effect we have outlined earlier. It may be that, just as in discussing the
work of Bills (1931) and of Bertelson and Joffe (lg6f) it was necessary
to posnrlate short-term and long-term effects operating over seconds
and minutes respectivelp so there may be still longer-term effects
extending over, say, a week or a year and dissipated by rests at weekends and annual holidays.
On the other hand, some of these effects appear to be coupled with
low morale and thus with various factors of social and indusuial organisation, such as type of leadership, seei4g 'results' for one's work and
avoidance of hold-ups in the flow of production. Factors of this kind
could reasonably link long-term morale to the same kind of mechanisms
that we have considered in relation to monotony and boredom. Hold-ups
in production, leading to waiting and idleness, are an obvious case in
point. Seeing results of work has a clear affinity to the 'knowledge of
results' which, as we have seen, tends to maintain vigilance. The same

is perhaps true of leadership: it is often claimed that morale is better
with a 'democratic' grpe of leadership, meaning that the worker, though
taking orders from management, is also able to exert an effect on
management. Looking at such leadership in cybernetic terms, we can
say that the servo loop from management to worker and back is completed iust as it is by knowledge of the results of performance. We shall
discuss this idea further in Chapter ro.

IX
Acquisition of Skill

It is not possible without artificiality to separate problems of acquiring

skill from those of human learning in general and thus from a vast
literanrre extending back to the flrn of the century. No attempt will
be made to survey the whole of this work here: the task would be
altogether too great and has in any case been largely done in a number
of text-books (e.g. McGeoch and Irion, r91z, Woodworth and Schlosberg, 1954, Bilodeau, 1966). Instead, some attention will be given to
particular topics which are especially pertinent to the subiects discussed
in previous chapters.
Experience, and therefore learning, are cumulative in the sense that
each new sinration is inevitably dealt with in terms of previous experience, and each new experience modifies what is carried to subsequent
situations. The modification seems to take the form of seleaing, qualiffing and reordering the schematised material held in memory, snd to

affect all the main links in the chain of central processes shown in
Fig. r.3 (p. r9). As regards perception and translation, this means that
material becomes more thoroughly coded or recoded as outlined in
Chapter 6. On the effector side we seem to learn fine control and tem-

poral ordering of action which can also be conceived as the achievement
of a more thorough coding. The durability of what is learnt seems to
vary somewhat between the different links. For reasons at present
unknown, it appears to be much firmer and more resistant to interference from subsequent activity on the motor than on the perceptual
side (e.g. Fleishman and Parker, 196z).
The kind of progress made as learning proceeds has been discussed
in the case of uacking by Slack (rgSl) who showed that mastery began
as a general acquaintance with the task and apparatus, and that this
was followed by adjustment of the velocrty of individual movements so
that the longer were made at higher velocities than the shorter and
later by recognition of the regularities of the track which enabled its
cogrse to be at least partly predicted. In the same wsy, the material
286

AcryLrition of Skill

z8T

brought from previous experience is at various levels of generality,
including not only specific details of perforrlrance but broader strategies
ancl techniques.

Past experience usually assists with the new situation, improving the
speed or accuracy of dealing with it. Occasionallg however, the previous experience which is applied is irrelevant or bears only a superficial
resemblance to what is required, so that it hinders rather than helps
mastery of a new task - in short the 'tansfer effect' from previous
learning is negative rather than positive.

SOURCES OF LIMITATION IN LEARNING AND RECALL

For learning to take place and for the subiea to furnish evidence that

it has, a number of stages must be gone through:

(l) The material to be learnt must be perceived and comprehended
and any responding actions selected.
(z) If the view outlined at ttre beginning of Chapter T is correct, the
material must be held in some kind of temporary short-term storage,
perhaps in the form of self-regenerating circuits of netual aaivity in the
brain, until there has been time for more permanent registration to take
place (Hebb, 1949).
6) Some kind of dtrable trace must be established which is capable
of remaining relatively uu i-Faired by subsequent activities of the
organism and of standing up to gross assaults on the brain which may
severely disrupt or temporarily suspend its activity. Such a trace seems
to require either an alteration in the mictostrucnrre or a stable biochemical change in some cells of the brain.
(+) The traces have to endure until the time of recdl. They rnay,
however, undergo some changes during this period, either because of
inherent instability or because they are overlaid or partly disrupted by
subsequent learning.
6) There must be recognition of a further situation which demands
the reuse of the material as modified by any changes during srage 4.

(6) The material must be recovered from among other material

stored in memory.
Stages 5 and 6 may at first sight appear to be the same, but should be
distinguished from one another: we can often recognise what material
is required without being able to recdl it, as when we cannot remember
a name although it is 'on the tip of the tongue'. This indicates a failure
at stage 6. On the other hand, we show a failure at stage 5 if we can
recall material without being able to see its relevance to a new sinration.

Fundamentals of Skill
288
(il Finally the recalled material has to be used in such a way as to

produce an overt, communicable response. Any one such response may
incorporate many items of recalled material together with data from
the sinration present at the time.
It is important to bear in mind that although normdly only the final
result at stage 7 is observed, failures at this stage may be due to breakdown at any of the previous stages.
Let us consider the main stages in ttrrn:

Understanding the task

Many apparent failures to learn are due to failure to comprehend what
has to be learnt, rather than to any diffic'ulty of registering or holding
the material in memory (King, 1948, Seymour, r954a). The practical
importance of this has been strikingly illusuated in studies of the invisible mending of woollen cloth, traditionally regarded as a very difficult
operation taking many months to learn (Belbin et al., 1957). It was
found that the main difficulty lay neither in the dexterity required to
manipulate dre needle, nor mainly in the visual acuity needed to see the
weaves, but in understanding the way the weaves \rere constructed and
thus the correct sequence of 'unders' and 'overs'. Trainees were accord-

ingly set to construct waves out of elastic thread on frames and to mend
specially produced large weave cloth before beginning work on cloth
of standard size weave. This method sf ffaining drastically reduced the
time required to learn as compared with the more usual method of
watching a skilled workerr or with a method based on the principles of
the'Training \Vithin Industry' scheme. The new method also enabled
middle-aged women to learn a iob uaditionally regarded as possible
only for school-leavers (Belbin, 1958).
The need to secure comprehension means that it is often beneficial to
teach certain points about a task before practice begins. Most discussion
has been about verbal formulation of the task, but other methods can
also be effective: for example Chaney and Teel Gg6il found diagrams
of faults to be an effeaive aid to the uaining of industrial inspectors,
and Posner (t967b) noted that the reproduction of movements which
had been visually guided was much better than those which had originally been made blind and thus learnt by kinaesthetic cues only. Even
if such 'pre-training' has little effect on perforrnance of the task concerned, it may improve the extent to which the skill transfers to other
tasks (Neumann, r9fu). Two cautions about pre-training have, however,
to be kept in mind. Firstly, although verbal or other formulation may

Acquisition of Skill

z8g

help a trainee to understand a task, the fonnulation will inevitably be
in terms of some 'concepual model' or method of codi$ing the dau
and correspondi.g actions, oDd the efrciency of the pre-training will
therefore very much depend on whether the'model'is appropriate and
easy to understand. For example, McAllister (tgSf) found when training subiects to position a lever that describing positions in angular
degrees was less effective than description in terms of a clock face or
directions related to the subiect's own body. Secondly, the well practised subiea will abandon the verbal or other intermediary and associate
incoming data with action direaly. For example, Fleishman and Rich
(tg6f) noted how visual cus gave way to kinaesthetic with practice on
a tracking task, and \[est (tg6il observed trat allowing tlpists to see
what they were doing early in practice was hehful, but that vision later
gave way to reliance on taaile and kinaesthetic cues. Pre-uaining
methods need to take care not to make the subiect dependent upon the
extra cues provided in the early stages of uaining and thus to hinder the
changeover to more direct relations between input and output at a later
stage.

It is important to recognise that a uainee may use cues in his learning
other than those which the instnraor intended. For example, London
post-office sorters are, or used to be, trained with packs of cards, each
card bearing a name, street number and name of district, which had to
be sorted into correct disuict numbers. Belbin Gg6+) found that
uainees seemed, in some wsy, to use not only the district name but the
whole of the data on the cards so that performance fell as soon as they
began to sort different packs or acnral mail. There is some conflia of
evidence as to whether flexibility, in the sense of ability to tackle a

range of similar although not identical tasks, is favoued by uainirg
with one example throughout or with several different examples (Adams,
1954, Callantine and Varren, 1955, Duncan, 1958). The conflict is
perhaps due to the fact that although variation benreen one task and
another prevents too close attachment to particular detailed methods, it
also tends to impair the subiect's grasp of essential principles. Optimum results may require a compromise: for instance Morrisett and
Hovland (rgS9) found that subiects uained with 64 trials at each of
three problems did better when uansferred to a new problem than
those trained with rgz trials at one problem or eight uials at each of
z4 problems.
Several snrdies have considered the question of whether information
glven about a task should concentrate on general principles, or whether
it should detail niles of procedure. Broadly speaking the findings suggst

a
v'

2go

Fundammtals of Skill

that, with very complex tasks, instruction in principles yields better
results ttran laying down a detailed drill, while with simpler tasks the
drill is at least equally effective (for a summary see Clay, 1964). The
reason is probably that a complex task commonly involves a number of
alternative sequences of actions each appropriate to particular varieties
of the circumstances under which the task is carried out. Any attempt

to reduce this to a drill will mean that several different drills will have
to be learnt, together with rules for applying them. In such a case,
general princrples, even if more difrcult to master than any one detailed
drill, may on balance effect a substantial saving.
A furttrer implication of the need for comprehension is that the speed
at which material is presented should be limited. If presentation proceeds too fast, I subject will be unable to deal with all the incoming
material and will thus be left with gaps in his knowledge which, in the
case of rote material, will lead direaly to errors of omission and in the
case of meaningful material will destroy the coherence of his trnderstanding. Ftrrther, h lecnrres and demonstrations it is not easy for a
trainee to go back and refer again to a point which may assume a new
significance in the light of later information. For these reasons there are
in some cases advantages in self-instruction methods rather than
direct personal teachirg. Much of the recent success of 'teaching
machines' is probably due to the subiect being free to take insuuction
from them at his own pace. Hulicka et al, (tg6il have shown that time
allowed for presentation is much more important than time allowed for
recall.
These various requirements can only be met if there is precise infor-

mation about what has to be learnt, and this in ftrn presupposes a
careful and thorough analysis of the task to decide what should be
perceived, what actions have to be taken and in which ways the latter
are conditional upon the former. Such analysis is not always easy to
make because a skilled performer seldom knows precisely how he
achieves his results, snd his actions may be so rapid that they are difficult to obserye. S7e may presume that in the course of practice his
actions have become accurate enough to be made ballistically without
the monitoring of sensory feedback, and that in consequence he has
lost awareness of his actions in the same manner as Leonard's (1953)
subiects mentioned in Chapter 4 (p. r rz).
Subtle variations of movement seem often to appear as qualitative
differences to the subiect who uses terms such as 'keeping the ball
down' or 'follow-through'. These seem to be reflections of orders by
the translation mechanism rather than observation of his own actions,

Acquisit;on of Skill

2gr

and it seems clear that the main analysis of a task needs to be at the
level of these orders rather than in terms of detailed movements.

The short-term link
If it is correct to assume that material has to be stored temporarily in
short-term memory while long-term traces are being formed, the very
small capacity of short-term memory we have seen in Chapter T implies
a severe limitation upon the amount of material that can be handled at

any one moment during learning. This view provides an approach to
two questions each of which has been the subiect of many experimental
snrdies with somewhat conficting results:

(a) Shmld a tra'irue attempt to master a complex task as a whole, or shmld
it be sPlit up so that he can learn one ?art at a timc? Where the whole task
is a closely co-ordinated aaivity such as aiming a rifle or simulated
flyrng of an aircraft, the evidence suggests that it is better to tackle the
task as a whole. Any attempt to divide it up tends to destroy the proper
co-ordination of action and subordination of individual actions to the
requirements of the whole (McGuigan and MacCaslin, rg11tBriggs and
I7aters, 1958, Crossman, 1959, I(napp, 1963, Naylor and Briggs, t963),
and this outvyeighs any advantage there might be in mastering different
portions of the task separately.
Wherer otr the other hand, the task involves a series of component
actions which have to be performed in the correct order but each is
largely independent of the others, there seem to be advantages in
practising the different components separately. An example is capstanlathe operation studied by Seymour (tgS+b, 1955, 1956). The components of the cycle of operation - loadirg the collet, operating the crossslide, operating the turret and urloading the collet - were practised
separately, then in pairs, then in threes and finally all together. The
advantages of this 'progressive-part' method of learning were small in
&e laboratory, but more substantial in the field (Seymour, 1959). The
splitting-up of the task in this case enables each portion to be mastered
quickly without overloading the learning mechanisms.

The same plan of learning separate portions of a whole task and
combining them together was found successful by Singleton (1959) in
uaining shoe-machinists. Belbin (tg6+) in her uaining of post-office
sorters found them very slow to master the full range of destinations if
all were given together at the beginning, but that if trainees learnt one
group of destinations then switched to another, they tended to forget

292

Fundamentals of Skill

the first group while learning the second. She accordinglyused amodi-

fication of the 'progressive-part' method, which proved successful,
whereby sorters first used a restricted range of destinations then added
more, a few at a time, until all were included. In this way the amount of
new material to be learnt at any one time was kept small, while learning
already established was maintained.
A similar method to Belbin's has been shown by Postman and
Goggin (1966) to yield quicker learning of a list of nonsense syllables
than either learning the list as a whole or learning each part separately
before combining them. It seems possible in this case, and perhaps in
the others discussed here, that some of the advantage ofthe'progressivepart' method of learning l.y, not in the avoidance of overload dtrring
learning but in the fact that dre early parts were learnt much more
thoroughly than the later. There is a good deal of evidence from other
work that the extent to which an item has been learnt can serve as a
distinguishing cue when it is being recalled, so that confusion between
items is less likely to occur when they have been learnt to different
extents (see McGeoch and Irion, pp. 4r1-4rl).

\
a.

(b) Wlrat is the optimum lmgth of training session? Is it easier ro learn a
new task by praaising it continuously until mastered, or is it better to
divide the time into short periods interspersed with rest or other
activity? At first sight, continuous practice ought to yield quicker
learning because it would ensure that traces were consolidated without
the chance of disruption by other activity which might interfere between
one practice session and the next. Yet many laboratory shrdies favotu
the spacing of practice with frequent brief rest-pauses. The superiority

of continuous or spaced practice clearly depends upon a balance

between several factors.

Continuous practice seems to facilitate mastery of complex, meaning-

ful material and the establishment of co-ordinated rhythmic aaivity.
Such comprehension or co-ordination means that individual items of
data or action are grouped together into larger units: in this way less

has to be recalled by rote because more is recovered by inference from

one item to another. Also, continuous practice seems to be preferred by
older trainees (Belbin, 1964) - a frnding consistent with other experimental results which show that short-term memory is more liable to
interference from other activity during retention among older than
among younger adults (e.9. Kirchner, 1958, Inglis, 1965).

Two main tJpes of reason has been advanced to account for the

cases in which spaced practice is superior. Firstly, if material is indeed

Acquisition of Skill

293

hdd in some kind of dynamic short-term store while more permanent
traces are bei.g established, 'consolidation' of the trace will oudast the
actual presentation of the material to the subject and will continue
during part of the gap between one practice session and the next.

Spaced practice could thus be more efficient than continuous if only

the actual duration of the sessions is counted and &e time between
sessions is ignored. This is what most sturdies of spaced practice have
done: when the time benveen sessions is included, continuous practice
is usually more efficient (Tsao, 1948).
The second type of advantage suggested for spaced practice is that
the pauses allow certain after-effects of previous uaining uials to die
away and thus reduce any adverse influences they might have on subsequent trials. One example is the very short-term mental fatigue discussed
in Chapter 8 (pp. 225-227). Anotherpossible adverse effect is interference
caused by the cross-connections bennreen items in the'netrral-dictionary'
mentioned in Chapter 3 (p. ro3). These would tend, unless time was
allowed for their effects to subside, progressively to activate other
erroneous responses and so blur the traces of the wanted items. Lfnderwood (1961) has concluded that spacing of practice reduces confirsion
between responses rather than between signals. Such a findi.g is perhaps consistent with the evidence outlined in Chapter 3 Gp. 8z-9o)
showing that choice of response is usually a very much slower
process than identification of signal, and therefore likely to exert a
longer after-effect.
Both factors imply two points about the effectiveness of spacing practice. Firstly it will depend very much on what is done dtuing the times
benn een practice periods. If they are spent in rehearsal of the rnaterial
these times effectively form a part of the total practice time and learning will benefit accordingly, unless the task is fatiguing in which case
the continued practice may depress subsequent performance (Adauls,
1955, Ifuapp, 1963, Rosenquist, 1965). If on the other hand times between practice periods are spent on another task, learning or later recall
of the first task may be impaired, the degree of impairment depending
on the degree of similarity between the two tasks. For example, Sanders
(l96ta) found that learning a list of letters interfered less with the recall
of a list of digits than did learning or even merely reading over, another
list of digrts. Again it seems as if similaiisy of response is more important than similarity of signal (Postman et al., 1965, Friedman and
Reynolds, 1967).
The second implication is that very brief pauses benreen practice sessions should be as effective as longer ones since the kinds of

294

Fmdammtals of Skill

perseverative or after-effects envisaged are likely to last only for a short

time - a few seconds at most. Experimental studies have indeed confirmed that rest pauses of about one minute are almost as effective as
those of a day (Lorge, r93o). The! effectiveness of very brief pauses
may in some cases be the reason why continuous practice appears to
be as efficient as spaced: when a subiect can perform a task at his
own speed, he commonly takes many short breaks of a few seconds
which are seldom recorded when his performance is snrdied, but may
be very important in allowing learning to consolidate and in avoiding
mental fatigue.
A further, as yet unanswered, question is raised by the problem of
optimum length of training session. It is clear that there are severe
limits to the rate at which material can be learnt when considered on a
time scale of seconds and minutes, but are there any additional limttations operating over periods of hotrs, days or even longer times ? Common experience suggests that there may be, but the question does not
seem to have been posed in a scientific context. An indication drat it
might be worth asking is contained in the finding by Henshaw and
Holman (1933) in an industrial snrdy, that 8o min training per day at a
chain assembly task yielded as rapid improvement as 16o min.

Passing material to long-term storage
Classical snrdies of memory by Bartlett (rglz) showed very clearly that,
in perceiving material to be remembered, subiects interpreted what they
saw or heard, shortening, simpli$ing, re-arrangrng and sometimes
elaborating it, and that it was this interpreted version of the material
that was retained. His results imply that memory normally does not
store the acttral data provided by ttre subiect's sense organs, but his
reactians to them, and that if we want to represent the flow of data in
the organism we need to put the memory store beyond the'translation'
stage as shown in Fig. r.3.

It follows that anything which ensures that data are passed though

the translation mechanism and not blocked at the previous stage will
tend to improve learning. The classical work on this point is that of
Gates Ggril who found that subieas learnt verbal material by rote
much more thorougtrly in a given time if they alternated readi.g with
attempts to recite, than if they merely looked at the material for the
whole time available.

It was at first thought that this effect was due to the sheer actions
involved in recitation, but recent sturdies by Von Wright (tgSlb) and

Acquisition of Skill

zgs

by Belbin (1958, 1964, Belbin and Downs, 1964) have shown that it
depends on the subiect making actioe choices instead of passively observ-

ing the data presented. \[e shall discuss Von Wright's work later.
Belbin's experiments used a card-sorting task. fn one of these which
we rnay take as an example, subiects learned to 'post' nunbered cards
into colotred slots : 2c.-29 went into one slot, 3o-3g into another and
so on. Some subjects were glven a list of associations betrn een colours
and numbers to memorise, others posted special cards each bearing a
number and the appropriate colotrr trntil they felt confident they had
learnt the required association. The latter tended to learn better. The
extra cues can, however, iD some cases impair mastery: a similar method
used in the uaining of post-office sorters to place cards with different

destinations into appropriate pigeon-holes did not prove superior to
straightforward memorising: it appeared that the uainees relied on the
colour code to the exclusion of active attempts to associate destinations
with their corresponding holes.
A different method of seorring that active choices are made is illustrated in an experiment by Chown et al. (t967) whose subiects learnt to
associate 20 village names with their appropriate counties by a pro-

grammed-instruction method. The prograulme consisted of books
with six 'frames' on each page, and subiects'moved from frame to franre
using a cardboard mask which exposed only one frame at a time. All
subiects learnt five villages in one county (Buckinghamshire) first,
then half the subiects were presented with 'discovery' frames listirg
five villages and headed by a statement such as 'T$fO of the following
villages are in SUSSEX, the others you know as they are in BUCKS.':
the subiects had thus to infer the county to which the villages they had
not seen before belonged. The other half were presented with further
frames listing five villages all in one county to be learnt in the normal
way. It was found that uainees aged zo-34 learnt about equally well
by both methods, but the scores of those aged 35-49 were clearly

higher with the 'discovery' method and about equal to those of
younger trainees.

Further evidence that active choices rather than motor actions are an
important factor in learning is provided by a nunrber of experiments
which have shown that 'mental practice' in which the subiect performs
a task in imagination, can often be substituted for a substantial amount
of practice involving full perfornrance with little if any loss of effectiveness (e.9. Hillix and Marx, 196o, Ifuapp, 1963, LXich, 1961).
Two further points follow from the seeming fact that data are processed and coded before being placed in long-term memory:

296

Fmdammtals of Skill

(o) The predotninance of initial experience. Zangwill (t937t t939), who
rnade a number of sturdies following Bartlett's work, fotrnd that when
subjects were required to reproduce designs orpassages of proseseveral
times on different occasions and were then asked to recoguse the orignal
from among other similar versions, they tended to choose a version
based on their own first reproduction rather than the original itself.
The same tendency for initial experience to determine subsequent
reaction has been fotrnd in several subsequent studies. For example
Bilodeau et al. (lg6+) who required subjeas to make movements
against a stop and then to repeat them several times with the stop
rernoved, found that the movements made on later trials tended to
resemble the first ones made after the stop was removed. Again Welford,
Brown and Crabb (tgSo) who, as already mentioned in Chapter 8
(p. 258), tested groups of aircrew with a series of problems both before
and after flight, found that the style of performance adopted by the
subjeas when the problems were first encountered carried over to the
later occasion. Thus, subjects tested first when they were tired after a
flight adopted somewhat inefrcient methods which they used again
when they were tested a second time with similar problems after several
days' stand-down. In the same way subiects tested first when they were
fresh carried the more efficient methods they adopted over to a second

test when they were tired after a flight. The effect can be seen in
Fig. 8.8 (p. zSil where the numbers of readings required to solve the
problem are less throughout for the group who started fresh than for
the group who started tired. This tlpe of result was obtained with four
different sub-groups in trnro different experiments and seems, therefore,
not to have been due to any accident of selection having produced
groups of unequal ability.
Why the first experience should tend to have this predominant effea
is not clear, but it is presumably bound up with the cumulative nanue
of learning. When a subject meets an entirely new problem he has to
constnrct his solution of it from past e4perience dealing with different
problems. Once he has done this, however, he has an outline method
ready prepared for application to any similar problem on a subsequent
occasion. Even if this method is not the best possible it will often be
more efficient to use it than to work out a better method dc nooo.
(D) Pqsistutt *rors and their prevention. One of the less fortunate consequences of the predominant influence of initial performance is that
errors in the early stages of learning tend to become ingrained. Indeed
Kay (rgSr) could argue &at one of the maior problems of learning is

Acyisiti* of Skill

2gT

concerned with gening rid of - unlearning - these initial errors. Kay's
task was simple enough: learning to press a series of five Morse keys in
the correct order by trying them again and again until all five could be
gone through trrice without error. Yet errors made in the first two or
three trials tended to persist for many subsequent trials, in spite of the
'Goeu'

mm

START

2mm

Figrrre 9.r. Von $lrright's 'moving tnaze' (from Von Vright rgsTa).
The pattern shown on the left was drawn on a white paper Uana and
appeared, moving from above downwards, in the slit of the screen shown
on the right. The subfect had to hold a stylus in the slit and trace over the
path of the rrraze' learning to move left or right as each diamond came
into view so as not to cross the double lines drawn on one side or other
of the upper half of the diamond. The double lines were sometimes on
one side, sometimes on the other, as shown at the extreme left of the
Figure. The subiect could not see them at the moment when he had to
decide which way to move. His only way of being sure not to make
errors was, therefore, to remember the positions of the double lines trom
previous trials.

fact that the correct key at each point in the series had always to be
found before passing on to the next. Even when subiects evennrally
avoided these persistent errors they were sometimes observed to move
their hand towards a wrong key and stop iust in time saying 'No zol

that one'.
Kay's work was followed up by Von \trright (tgSla) whose subjeas
learnt a pencil maze of a tJpe shown in Fig. g.!. The first time through

298

Fundamentals of Skill

the maze they tended to adopt some kind of systematic procedure such
as always going to one side or alternately right and left. The second
tirne through they obviously tried to apply what they had learnt in the
first trial, making some correct responses and some errors. The pattern
of errors made on the second uial tended to persist, so that the positions
of errors on subsequent uials correlated with those on the second trial.
These findings of Kay and Von I(rright will evoke echoes in the experience of those concerned with teaching athletics, games and skills such
as typewriting: errors need to be corrected immediately or they tend to
become ingrained.
Von Wright (tgSlb), following a number of earlier workers, argued

A

B

C

Figrrre g.z. Three different versions of Von lTright's 'moving maze' used
dtrring the first four training trials. From the fifth uial onwards all subiects used version A. The mean results per subiect with the three versions
were:

A
B
C

,.Ix'1i'::?L::"
23.25
ro.45
r8.8o

"l:[f;t;r*
7630
r5.8o

S2.tS

The subiects were 6o undergraduates randomly divided into three
groups of zo - one group for each version of the maze.
that if errors could be prevented in the first few uials, mastery of the
task should be very much quicker. He accordingly prepared three different versions of the same maze, as shown in Fig. g.z. Of these A was
the same as the original type of maze. B was the same as A except that
the bars were moved down to a position which enabled the subject to
see them just before he reached the choice-point and thus to make the
correct choice every time. Type C omitted the incorrect pattrs altogether,
so that the subject had merely to follow the correct path through the
maze. Three groups of subiects were used: one (A) simply made trials
with maze A until they went through twice running without error.
Groups (B) and (C) had their first four trials with mazes B and C
respectively and then transferred to maze A on which they made furttrer

Acryisitian of Skill

2gg

uials until they too went through nrice running without error. The
results are shown with Fig. 9.2: groups (B) and (C) learnt much more
quickly and made fewer errors than group (A). The superioriry of
group (B) to group (C) presumably reflects the faa that the former had
to observe the bars and made an active decision at each choice point,
whereas the latter could follow passively through.
Holding and Macrae (1964), who gtve a valuable list of references to

early work in this field, obtained comparable results using the very
simple task of sliding an aluminium sleeve a distance of 4 in along a
steel rod by means of a knob. Each subiect was grven an initial test to
determine the accuracy with which he made the required movement
without training and was then given one of several different kinds of
training before a final test. Performance improved less between initial
and frnal tests if the subiea held the knob while it was passively moved
over the required distance than if such guidance was alternated with
unguided attempts or lq instead, the subject made free movements
trntil he reached a stop set at the correct distance. Subsequent work by
the same authors has shown that such guidance dtrring uaining is especially beneficial when tracking movements have to be made with a relationship berreen display and conuol which is not fully compatible,
presumably because it prevents the errors that would otherwise result
from the incompatibility (Macrae and Holding g65, Holding and
Macrae, t966). They have also shown that although gurdance in which
the hand is passively moved is no better than normal practice when
learning to track a simple repetitive course, it is superior for learning a
more complex course, in this case probably not only because it prevents
errors but because it forces the subiect to move more actively than he
otherwise would (Macrae and Holding 1966).
An interesting analogue to the foregoing methods, especially Von
I7right's, has been described by Annett (1966b) for training perceprtral
discrimination. Subieas were trained to discriminate whether small gaps
in circles presented for brief periods were on the left or right by
presenting the one qpe of circle on a green and the other on a red background. Complete circles with each of the two colours of background
were also included so that the subject had to discriminate between 'left
gaP', 'right gap' and 'no gap'. The use of this:aid greatly increased the
accuracy with which rings on a neutral background were discriminated
after the training had been completed. Presumably the colours served to
concentrate the subjea's attention on the appropriate parts of the rings
while they were being exposed and thus helped him to observe what
the gaps looked like.

3oo

Fundammtals of Skill

Principles in this area are not yet fully understood and there is room
for further research. For example Belbin (t964) in her snrdy of postoffice uainees found tendencies for sorting errors to become ingrained

and attempted, seemingly successfully, to reduce them by requiring
subieas, if they were doubdul where to place a card, to put it in a
'doubtflul' box rather than hazard a guess. Yet Belbin et al. (tg6+) found
that it made no difference to subjects required to associate lists of village
names with corresponding counties, whether they guessed or not, and
that they had no special difficulty in learning a fresh list of cotmties for
the same villages on being told the original list was incorrect. The
authors interpreted their findings as implying that the guess made with
a low degree of confidence faut de mieux is less likely to become ingrained than the committed decision or choice.
Whatever the present unanswered questions and conflicts of evidence,
it seems clear from the work reviewed here that relatively little learning
occurs if the subject is a passive spectator or even a passive performer,
but that he must be involved in active decisions and choices about
what he is doing, and it is these that he will retain whether they are
right or wrong. Indeed it is arguable that even in those qlses where a
subiect appear.s to learn passivelp he does so because he occasionally
engages in active 'mental' practice.

Retrieval from memory store

The obvious confusions, omissions and substinrtions that occur when
remembering imply that memory traces are liable to be disrupted or
overlaid by other traces established subsequently. There has in the past
been some controversy as to whether subsequent learning destroys
memory traces already laid down or merely makes their retrieval when
required more diffictrlt. Probably both processes occtu (Melton and
Irwin, r94o). Explanation of the gradual return of memories following
retrograde amnesia poses a similar problem: how far is the rettrrn due
to a piecing together of fragments of memory that rernain, md how
far is it due to the trauma not having desuoyed the traces but having
somehow rendered them uravailable? Again both processes probably
operate. By the same tokens we can perhaps assume that re-use of
memories as, for example, when a task is repeated, or a recall or recognition test is grven, will tend both to strengthen the traces concerned
and to make them more readily available. This will apply to errors as
well as to correct responsesr so that the effect of repetition may be beneficial or the reverse. For example a recognition test in which the original

Acquisition of Skill

3or
material is re-presented may facilitate later recall while a recall test in
which errors are rnade may hamper later recognition (Belbinr r95o,
Estes and Da Poliro, 196l).
The main problem of retrieval can perhaps be conceived as the discriminability of uaces from one another. It is thus obvious that corrfusion is likely to be greater benpeen very similar than benreen widely
different items. For example, Posm,an and Goggrn (rg6+) fotrnd that
learning a list of ro consonant-vowel-consonant nonsense syllables in
which each of four consonants appeared five times was slower than
learning a list in which each of zo consonants appeared once only.
Again recall is easier when the possible items belong to a relatively
small set, and the commonly found superiority of recognition to recall
seems to be due essentially to this. For example, Davi s et al. (t96rb)
presented subiects $rith lists of 15 nro-digit numbers between ro and
99
and then required either recall of the originals or their recognition
from among a list of 3o, 6o or all 9o numbers. They found that recognition was superior to recall only $rith the shorter lists in which, of
course, range of choice was restricted. Recognition from among the
whole 90 possible numbers was not superior to recall. Similar results
have been obtained by others since (e.g. Dale, t967, Slamecka, 1965).
Many of the faas about the availability of material in recall follow
from the cross-connections between items in the 'cerebral dictionary'
postulated by Treisman and discussed in Chapter (p. ro3). Avail3
ability is a function of the extent to which cross-connections converge
on an item and thus on the likelihood of its being to some extent preactivated. Thus nonsense syllables which evoke many associations are
more readily recalled than those which evoke less (e.g. Posman and
Goggrn, 1960 and such'association value'is more important than either
'meaningfulness' or pronounceability (Lindley, 1963). Again Murdock

(l96ob) noted that rate of learning a list of unconnected words was
proportional to the logarithm of the frequency of usage of the words in
nounal English. Obvious illustrations of the same general principle are
that rote learning of words is assisted by deliberate attempts to form
associations (Eagle, D67), and that when pairs of words are presented
and the subiea has to recall one member of the pair when given the
other, recall is easier if the words of each pair are in some way related
(Robinson and Loess, 1961). Less obvious, but clearly emphasising the
role of association rather than fauriliariry is the finding by Baddeley
(tg6+a) that trivial and meaningless associations between items in a
list can improve rate of learning the list by rote: for example the pair of
nonsense syllables QEM POG was more easily learnt than POG
QEM,

3o2

Fundamentak of Skill

seemingly because the letters MP occur more freguently together than

GQ in normal English.
KNOWLEDGE OF THE RESULTS OF ACTION

The basic fact that, as Sir Frederick Bartlett once put it, "It is not
practice but practice the results of which are known that nrakes perfect'
has been shown for many tasks since the pioneer experiments of
Thorndike (lgll) on the increase of accuracy in drawing lines blindfold
when the subject is told whether each attempt was too short or too
long. Performance shows no improvement, without 'knowledge of results', but begins to do so as soon as it is introduced. If it is then
removed, perfbrmance at once deteriorates. Dees and Grindley (195r)
found the deterioration was due to subjeas tending to overshoot - a
finding which implies that they were acting as an over-sensitive or
over-responsive mechanism held in check by the feedback provided by
the knowledge of results. The same point is implied in a finding by
Annett (lgSg) whose subieas regarded a bar which they had to press
to a given extent as 'much stiffer' when knowledge of results was
removed. Dyal (1966) also fotrnd a tendency to overshooting when
knowledge of results was discontinued in a line-drawing task. It occurred, however, only with subjects who had shown an initial tendency
to undershoot before knowledge of results was given: subiects with an
initial tendency to overshoot tended to undershoot when knowledge of
results was discontinued. It seems, perhaps, that knowledge of results
produces a general tendency to overcorrect initial biases and that this in
turn is modified and held in check by the detailed knowledge given

trial by uial.
Baker and Young (196o) suggest that the improvement of performance with knowledge of results is in two stages: in the first the
approximate limits of the action are learnt, and in the second finer
adiusments are achieved. What is learnt in the first stage survives the
removal of knowledge of results, but the fine adiusunents of the second
stage are quickly lost. The same implication follows from work by
Bilodeau and Bilodeau (r958a) who snrdied the irnprovement of accuacy in moving a lever a required distance over a series of trials when
knowledge of results was glven either every uial, or one in three, or one
in four, or one in ten. They found that acclrracy improved accorcling to
the absolute number of trials given with knowledge of results. During
intervening uials without knowledge of results performance tended to
fall off, but the extent to which it did so diminished as the series of

Acquisition of Skill

3o3

trials proceeded. In short, although there was no learning dtuing trials
without knowledge of results, something was carried over from the uials
with knowledge of results to those without.
We have already seen (p. zT8) from the work of Chinn and Alluisi
(tg6+) using a vigilance task that, in terms of signal-detection theory,
knowledge of results can raise d' and change f, implyrng that it can
both improve discrimination and alter criteria of iudgment. Further
evidence of the latter is provided by Nakamtua and Kaswan $962) who
fotrnd that glving subjeas information about the speed of their responses

in a reaction-time task shortened latencies but increased errors.
The timing of knowledge of results

In the typical experiment on knowledge of results, and in many uaining
situations, the subject makes an attempt or 'uial' at the task and is then
glven information about the acctracy of his attempt before making a
further mial. There is thus a gap between making each attempt and
receiving knowledge of results dtrring which the subiect has to retain
the 'orders' given to the effector mechanism - the aim and feel of the
movement-for comparison with the results attained. There is a second
gap, benreen receiving knowledge of results of one trial and rnaking
the next, dtuing which he has to retain both the original orders and the
comparison of these with the results so as to frame orders for the next
movement. It has often been suggested that some loss of data carried
over these memory links is likely to occur if knowledge of results is at
all delayed or if there is an appreciable interval between trials. From
what has been said of short-term retention in Chapter T we should
expect lapse of time as such to have little or no effect, but that intervening activities and shifts of attention could be disruptive, and that
time might be important as giving oppornrnity for these to ocsur.
The experimental evidence on the effect of time interval before knoryledge of results is gtven comes from several experiments in which
subiects had to make a movement and then after a time varying ia
most qlses from o to 30 sec were given an indication of whether the
movement had been of the correct length or too short or too long. The
results divide into two sharply contrasting groups, one of which finds
that delaying knowledge of results increases the number oftrials required
to attain a consistently correct response (Saltzman et al., 1955, Greenspoon and Foreman, 1956, Denny et al., t96o, Dyal, $64), while the
other finds no effect of delay up to 20 or 30 sec (Bilodeau and Bilodeau,
r958b, McGuigffi, 1959, Bilodeau and Ryao, r96o)althoughsomeeffects

3o4

Fundamentals of Skill

occur with delays of an hour or more (Bilodeau and Bilodeau, r958b).
The conflict betreen these two groups seems to be resolved by results

from an experiment by Dyal et al. (1965) who fotrnd a substantial effect
of delay if the score taken was the number of responto lxlling within
a zone designated as correct, but no effect if the acnral amounts of error
were measured. All the studies in the first group used the first method
of measluement, snd all those in the second group the second method.
This result presumably means that delaying knowledge of results has
little effect on the accuracy of most of the attempts but tends to result
in rather more relatively large deviations, which is what one might
indeed expect if the subiect's attention was sometimes diverted during
the period of delay. Results by Becker et al. (lg6f) run counter to this
pattern in that they found no effea of delay although scoring 'number
correct'rather than absolute error. Their procedure was more complex,
however, including the administration of an electric shock at the time
of the knowledge of results, and this makes their experiment hardly
comparable with the others.
Clear evidence of the disruptive effect of activity intervening berreen
a trial and knowledge of its results is the much slower learning found
by Lorge and Thorndike (tgfS) and by Bilodeau (lgS6) when knowledgs was delayed until one or more further trials had taken place. This
result has been confirmed by Lavery and Suddon (1962) and by Lavery
(r964).

The princrple that information about the correctness of action should
be available quickly has been used with success in teaching Morse code:

corection letter by letter was found to lead to quicker learning than
correction of several letters at a time (Keller, t943). The same principle
appears to apply to various tlpes of equipment such as certain process
plants in industry, where the operator has to take actions which do not
have their full effects until several minutes or even an hour or so later.

In such circunstances it is difficult for him to acquire and maintain
adequate control, although astonishingly expert performances are sometimes attained. Experiments in which the subiect has been given immediate indications, by means of a computing device, of what the evenfiral
effects of his astions will be, have shown very substantial benefits to
both gaining and performance (e.g. Taylor, r95T).
The effects ofthe gap betvyeen giving knowledge of results and making
the next attempt have been less systematically sttrdied, but evidence
that learning becomes slower as the gap is lengthened has been obtained

by Macpherson et al. (rg+g) and by Bilodeau and Bilodeau (r958b).
The latter authors argue that the time betrn een the giving of knowledge

Aquisitim of Skill

3o5

of results and the next trial has more effect than a delay of the same
duration in Srving knowledge of results. Why this should be so is not
clear, but the memory load is presumably higher benreen one trial and
the next than it is bettyeen the end of one attempt and the giving of
knowledge of its results. Alternatively if the subiect ttrinks of his task
as a number of discrete trials each followed by knowledge of results,
he might well be more likely to sffi his attention benreen one trial
and the next than benreen completing an action and receiving information about its effect. Such a shift of auention might have more effect
than a relatively simple intervening aaivity, such as making another
movement, which has been shown by Blick and Bilodeau (1963) to have
a negligible effect.

The quality of data fed back

It is perhaps obvious that, other things being equal, the more precise
the knowledge gven of the results of action, the more accurate actions
will become over a series of trials. Thus the lines drawn blindfold by
McGuigan's (1959) subiests were more acctuate if they were told their
error to the nearest + in than if they were told only to the nearest $ in
or r1 in. Again Trowbridge and Cason (tg3z), using a similar task,
showed that subiects learnt more quickly and attained greater accuracy
if they were told the direction and ertent of their errors than if they
were merely told whether or not they were within * in of the correct
length.
It seems obvious also that the manner of conveying knowledge of
results should be important. On general grounds one would expect
effectiveness to be greatest when the information is clearly and simply
related to the action concerned. Any distortion or equivocation in the
information fed back to the subject will reduce its effectiveness (Morin,
1955, ShellS r96t, Hunt, ry64). On the other hand, unduly full or
comple:r information may be partly ignored (Crafts and Gilbert, 1935)
or may confirse the subiea (e.9. Katz, r96il: an example has been
noted by Singleton (see .flaining Made Easier', r9fu) who found that
a scale at the side of an industrial sewing machine indicating how fast
it was running although of some help to a trainee, caused difrculties
by taking her attention from the article she was sewing.

\[hat is not so obvious is that the information given should indicate
the disuepancy between what is required and what has been achieved
rather than merely grve a reminder of requirement or some broad
measure of achievement. The fornrer point is illusuated in an experi-

306

Fmdamentals af Skill

ment by Lincoln (tgS6) who trained subjects to wind a handwheel at
a given rate. He forurd it more effective to give his subjeas knowledge
of results by having them, berween trials, hold the rvheel while it was
driven at a rate equal to the difference between the required rate and
that at which they had been winding, than for them to hold it while it
was driven at the required rate. The point that a broad measure of
achievement is insufficient is illusuated by several studies of tracking a
target which have shown that merely indicating when the subject has
remained 'on target' for a given length of time make relatively little
difference to his perforrnance (Reynolds and Adarns , rg13rArcher et al.,
t956, Archer and Namikasr 1958, Villiams and Briggs, 1962, Bilodeau
and Rosenquist, 1964, Karlin, ry65). The small improvement of performance that does occur may perhaps be attributed to the incentive
effect of knowledge of results which we shall discuss later.

Effeas of withdrawing knowledge of results

The puqpose of the early experiments on knowledge of results was to
snrdy educational 1.sfurique, srd for this purpose the indications seem
clear. More recent interest has been engendered by the question of
whether some industrial and military training can be improved by
giving detailed knowledge of results during the training period which

will not be available afterwards on the actual iob. Experimental studies
are unanimous in finding that when knowledge of results is given during

training and then removed, performance deteriorates in subsequent
trials. The extent to which it does so differs in different cases, however,
and the problem is to find whether some methods of giving knowledge
of results during training lead to better maintained performance than
do others when knowledge of results is removed.
Research on this problem is scanty but can be taken tentatively to
indicate that performance is best maintained when the conditions are
such as to emphasise the need for the subject to observe the feel of his
acions in order to relate them to their results. Thus Annett (lg5g), using
a task in which the subiect had to learn to exert a particular pressure
on a plunger, found that those who, as they pressed, observed the
approach of a spot on a cathode-ray tube to a marked target, showed
greater deterioration when the tube was removed than did those who
had been told verbally how they had done or had seen an indication on
the tube only after their movement had been completed. I7hen the
results of the pressure could be seen as it was actually being made, there
was no need to remember the feel in order to link it to the indication of

Acquisition of Skill

3o7

results, md it is therefore understandable that the subiect retained little

knowledge of how hard he had pressed.
Provided, however, the indication of results is not so direct as to
remove the need to remember the feel of the task, performance seems to
be better maintained if the knowledge of results has been direct and
immediate. Thus Goldstein and Rimenhouse (rgS+) found that subiects
in a tracking usk who heard abuzzer sound all the time they were 'on
target' retained their skill better than those who were merely told how
they had done after a period of practice. Again Dyal et al. (lq6S)
fotrnd that performance deteriorated less after immediate than after
delayed knowledge of results. Presumably the clearer the association
bennreen action and result that has been learnt, the bemer it is retained.
Probably consistent with this general view is the finding by Macpherson
et al. (rg+g) that performance was bemer retained with somewhat longer
rather than very short intervals betrreen trials: a new trial following
immediately might well interfere with the association of knowledge of
results with the feel of a preceding action. These findings appear at first
sight to co'lflict with the evidence of Lavery and Suddon (1962) and
Lavery (tg6+) who found performance was better maintained when
knowledge had been delayed until after one or more trials had been
made than when it had been gtven after each trial. We Bry, however,
suppose that in their case the relative uselssness of the knowledge of
results had forced subiects to pay attention to other cues, and that these
still remained available after knowledge of results had been withdrawn.
Several authors have made the complementary point that a subiect
must have some cues to the results of his actions if he is to perform
accurately at all, and training procedures will be effective in so far as they
help him to observe and use such cues as are inherent in the task for
which he is being trained. They will fail in so far as they provide him
with extra cues on which he comes to rely but which are not available
when he changes from uaining to the acnral iob (Goldstein and Rittenhouse, gS4, Annett and Kry, 1957, fuinett, 1959).

AIMS AND INCENTIVES

It is clear from snrdies of animals that speed of learning is substantially
influenced by incentives. Motives in human beings are less suaightforward but can probably affea several phases of learning and rnemory.
\Ve shall briefly mention three points, two of them directly linked with
the concepts of arousal and activation mentioned in the last chapter,
one of them less so but still within the general area of motivation.

308

Furdamentals of Skill

The incentive effect of knowledge of results
Knowledge of results, as well as being a corrective, also acts as an incen-

tive. Usually the two effects are inextricably mixed, but they can in a
few cases be separated: for example Macpherson et al. (tg+g) found
that when a subject was told that knowledge of results was to be discontinued, performance fell on the very next trial despite the fact that
this ought to have benefited from knowledge of results of the previous
trial. The two effects may be further separated in terms of the type of
knowledge of results required. We have already seen that, for the corrective effect, an indication of error is needed and that an overall measure

of achievement is of litde value. For the incentive effect, however, a
simple indication of achievement appears to suffice (Gibbs and Brown,
r956).
The way in which knowledge of results exerts its incentive effect is
not yet fully clear. If the theory is valid that decline of vigilance during
a watch is due to decrease of aaivation or arousal, knowledge of results,
because it prevents this decline, must tend to increase arousal. Such a
view is entirely consistent with the parallel view that knowledge of
results affects the goal or aim the subiect sets himself - an increase of
arousal may be mediated by raising the goal. Helmstadter and Ellis
(tgSz) found that a simple indication of achievement was as effective as
sening a subject a goal or having him set himself a goal. Locke (tg61,
see also Locke and Bryan, r96il fotrnd in an adding task that setting a
goal which was rco/o higher than comparable subiects had achieved
with the insuuction to 'do your best' substantially increased the speed
of performance and made the task more interesting, but that telling
subjects how many additions they had made every ro or 15 min had no
effect. He suggests, in consequence, that knowledge of results acts
mainly by setting goals, although in his case, since subiects wrote their
answers on sheets which they filled at approximately r min intervals,
they had a fair knowledge of achievement in any case, without being
told by the experimenter. Locke refers to a number of previous sttrdies
showing similar results extend.ing back to one by Mace (1935).

Activation level and rate of learning

Bills (rgzil in a classical study found that if the subiect squeezed a
hand dynamometer while learning both the rate of learning and the
level of subsequent recall were improved. Sidowski and Eason (196o)
and Sidowski and Nuthmann (1961) failed to confirm this result, finding

Acq,risition of Skill

3o9

instead that learning was slower when a dynamometer was squeezed.
They, however, arranged for a signal light or buzzer to sound all the
time the subiect was maintaining the correct pressure and this signal

may have disuacted the subiect's attention from the main task of
learning. Other investigators have confirmed Bills' result but added
the fiuther finding that if the tension is increased beyond a certain
point, performance falls again (Stauffacher, 1937, Cotuts, 1939).
Whether this is due to overarousal: or whether the effort of squeezing
a dynamometer very hard distracts the subject's attention, is not clear.
All these snrdies used the method of learning in which subjects were
shown words or nonsense syllables in serial order and after the first
rial had to try to anticipate which item was the next on the list. Learning was thus inextricably bound up with attempts at recall. Levin Gg6l)
who separated the two processes by alternately showing and requiring
recall of, the whole list fotrnd that rate of learning rose progressively with
level of induced tension during the learning phase, but that during the
recall phase acquacy rose with increase of tension to an optimtrm and

thereafter declined. A rise to an optimum and subsequent decline in
both rate of learning and accuracy of recall has also been found with
increasing skin conductance (Berry, 196z). Levin noted that increased
tension during recall resulted not only in more correct responses but
more errors due to intnrsions from previous lists. If, as is reasonable,
these can be regarded as a kind of 'false positive' in the process of
recovering material from memory, the result is consistent with the
change of P which accompanies a rise of activation as shown in Fig. 8.9
(p. 268) - the increased activation expands both the 'signal-plus-noise'
and the 'noise-alone' curves so that more signals are detected but at the
same time more false positives are made.
Meastrres of spontaneous muscle tension taken during learning appear
to be broadly consistent with the foregoing results: Stauffacher and
Courts both fotrnd that better learners tended to be more tense, snd in an
experiment on rnaze-learning Stroud (rglr) found that rapid learners
pushed the stylus somewhat harder than slow learners. It appears, in
short, that part at least of the differences of learning between individuals
can be accounted for in terms of their level of activation or arousal.
Consistent with this view is the frnding of Shenrood (rg6S) that subieas
for whom a cube like that of Fig. 6.5A (p. t66) reversed frequently
tended to be better learners. Again several studies have found that more
anxious subieas learn better than less anoious, at least when the material
is not such as to conflict with associations induced by previous rnaterial.
Where such conflict occrus, more anxious subiects tend to be poorer

3ro

Fundarnmtals of Skill

learners (Spence et al., t956a, b, Lovaas, r96oa, b, Sarason and Palola,

196o, Lee, t96t, Gaudry and Champion, tg6z). Since such conflict
makes the task more difficult, this is perhaps another example of the
Yerkes-Dodson effect discussed in the previous chapter (p. 2To).
Suoud found in his maze learning task that pressure on the stylus
tended to be higher at points of special diffictrlty and Sherwood noted

that rate of learning tended to rise and fall with arousal level in the
sameindividuals tested on different occasions. Measures of skin resistance
taken during the learning by rote of a list of nonsense syllables have
shown that resistance differed at different points in the list and that the
resistance levels corresponded closely to the speed at which particular
syllables were learnt - lower resistance indicating higher arousal was
associated with more rapid learning (Brownr rg31). The ends of the
list were, as is usually found, learnt more rapidly than the middle, and
skin resistance was lower when the ends of the list were being presented.
This finding provides an interesting line of evidence which isnot usually
taken into account when discussing reasons for the quicker learning of
the ends of a list.
Several snrdies have shown that the beneficial effect of raising the
level of activation during learning appears much more in long-term than

in immediate recall. The former is markedly improved while the latter
is either unaffected or actually impaired (Walker and Tarte, 1963,
Kleinsmith and Kaplan r 1963, 1964, Berlyne et al., t966, Batten, t967).
The improvement of long-term recall is presumably due to activation
facilitating long-term registration and thus reducing the rate of forgetting. Why it should do so is not clear, but a moderate level of activation
would probably strengthen the signals entering the long-term memory
store and perhaps strengthen those already there: Russell (1959) has
put forward the interesting idea that random or rhythmic activity in the
brain srill tend to channel through established pathways and to confirm
them, so that if memory traces can be conceived as pathwsys, they will
be strengthened. The impairment of immediate recall is prestrmably due
to the muscle-tension, noise or other activating agent used during
learning tending to produce neural noise which takes some time to die
away snd, until it has done so, confuses recall.

The order of learning more and less difficult tasks

In the early r95os several sflrdies were published which addressed themselves to the problem of whether, when two tasks of unequal diffiorlty

have to be learnt, it is more efficient to begin with the easier or the

Acryisition of SbiA

3rr

harder. These sttrdies looked back to one by Cook (rglilwho found that

if the harder was tackled first the easier was learnt very quickty afterwards, but that if the easier was learnt first, little time was saved on the
harder afterwards. Cook's results were open to the obiection that his

subiects spent longer learning the harder task and thus, if they learnt
the easier subsequently, came to it at a later stage of practice. Most of
the strdies in the early r95os endorsed Cook's findi.g while being free
of this obieaion (Baker et al., rg5o, Szafran and Welford, r95o, Gibbs,

r95r).
Let us consider Szafran and Welford's experiment by way of exarrple.
Subjects threw small loops of chain at a target on the floor under three
conditions. In the easiest they threw directly; in a condition of intermediate difficulty they threw over a bar; and in the most .rifrcult they
threw over a screen the same height as the bar, which hid the target
from direct vierr so that it could be seen only via amirrorplacedbehind
it. Different groups of subiects performed the tasks in different orders.
The results are set out in Table 9.r: the total effor for allthree condiTABLE 9.

r Mean eruors made when throuting at a target in inches per throw

(frm Szafran and Welfurd, rgso). Each mcan in A is based on r So and in
B on 5o throws by each of z8 xtbjects

A. Mean errors for all three conditions taken together
Direct condition first

Bar condition first
Screen condition first

S.g3

5.73
5.35

B. Mean errors for the three conditions separately

Presented first Presented second Presented third

Direct
Bar
Screen

5.59
S.zg
6.99

4.So
4.95
7.t6

4.49
4.96

7.t3

tions taken together was least when the subiects began with the most
difficult condition, and greatest when they began with the easiest.
Moreover, while performance in the easiest condition was much more
accurate if it came after one of the others than if it came first, perforuurnce in the most difrcult condition was slightly zt)orse. On the basis
of their results Szafran and Welford suggested that when a more difficult
task precedes an easier, the uansfer effect is lpositive, but when an
easier precedes a more difficult, the transfer effect tends to be negative.
Subsequent work (for a summary see Holding, ry62) has shown that

3t2

Fundammtals of Skill

this statement was a little too sweeping. In particular, several e4periments in which subjeas tracked a moving target have shown that practice with a slow moving easy coruse benefits performance at a fast,
difficult course more than vice versa. A possible reason for the conflict
between ttrese uacking results and those of other experiments is that,
with a very fast-moving target, accurate tracking is impossible and is

therefore not attempted. If so, a subiect practising first on a fastmoving target will adopt less suingent standards of accuracy than will
one whose first experience is with a slower-moving target. We might
therefore reconcile these results \rith the others by saying that if two
or more tasks haoe to be learnt, it is most bmeficial to begin with the one
which elicits the gteatest care and. effort toanrds the attainment of a hgh
standard of perforrnance. Some direct support for this view is given by a
finding in Szafran and \[elford's experiment not reported in their paper,
that although subieas who began with the most difficult task were, on
average, the most accurate, they also tended to be slightly slower:
beginning with the most difficult task seemed to engender a more
precise and deliberate performance.
Results published subsequently to Holdi.g's review all conform to the
same pattern. The tasks involved have included the discrimination of
gaps in circles in the experiment by Annett (r966b) already menrioned,

identification of patterns (Coules et al., 1965), and various tracking
tasks (Goldstein and Newton, 1962, Bilodeau, 1965, Dooley and
Newton, 1965).
One further proviso needs to be mentioned regarding the relationships between difficult and easy tasks. The experiments we have mentioned have all carried practice at the more difficult task to the point
at which reasonable mastery has been attained. If the subject were not
allowed to continue to this point, he would be left with an inadequate
comprehension of the task, so that his subsequent performance at an
easier task might be confused and less satisfactory than it would have
been if he had tackled the easier task first. Apart from this proviso, the
rather surprising principle that has been outlined here seems to hold
remarkably well.

CHANGES OF PERFORMANCE DURING LONG PRACTICE

It is well known that the initial attainment of reasonable competence at
athletic, industrial, artistic and many other skills is followed by a long
period of further improvement during continued exercise of the skill.
Vhat is the precise nature of this further improvement is by no means

Ac$risition of Skill

3r3
wholly clear, but several snrdim agree that if the time taken by a task
is split up into movement times and times benreen movements, it is the
latter which decrease more with practice: for example if a subiect has to
move his hand to grasp an object and then convey the obiect to a box
where it is dropped, times taken in grasping and dropping improve
most (Wehrkamp and Smith, 1952, Rubin et al., t952rvon Trebra and
Smith, 1952, Seymour, 1959). Presumably the central decision processes
are modified with practice to a greater extent than the execution of
movements.

De long Ggsil, following suggestions made by some previous
workers, has proposed that the time taken to perform a repetitive task
falls exponentially until it approaches some'incompressible' minimum,
so that if cycle time is plotted against cycle-number on log-log paper,
time decteases approximately linearly until it approaches the incompressible minimum. He proposes that

Tn: T- + T, -,-T*
nlc

(9.r)

where T* is the nth cycle time, T - is the incompressible minimum
time - that is the 'rrle that would be taken if the task was continued
for an infinite number of cycles - and I, is the time taken by the first
cycle. The exponent & expresses the rate at which improvement takes
place with practice.

De Jong found this formulation to fit reasonably well for a number of
indusuial operations. The approach was taken up by Crossman (lgSg)
who found that this type of forurula gave a reasonably good fit to the
data from several laboratory snrdies and industrial operations. Examples
are shown in Fig. 9.3. He suggested that the improvement is due to
the operator sarting with a range of slightly different methods of doing
the task and gradually coming to select the quickest to ttre exclusion of
others. He set out a statistical formula which asstrmed a progressive
fall in the probability of usirg less efficient, and rise in the probability
of using more efficient methods, with time. It differed from De ]ong's
formula mainly in predicting that the first few cycles will be a little
quicker than they would be if a strict exponential formula held. The
data in Fig. 9.3 glve some support to this view.
Crossman pointed out a ntrmber of consquences of his model:
(i) It implies that successive performances of the same task \ilill tend
to become more uniform. Some evidence that this is so for tracking has
been provided by Relmolds (t952) and for car-driving by Lewis (rgS+)
who fotrnd that highly skilled drivers performed more consistently than
less skilled.

3r4

Fundamentals of Skill

(ii) Rate of improvement with practice will depend on the variety of
methods from which selection has to be made, the variability of the

-.\ \\ Sz
\
\,\

U
=J
=
(.

lrl

'-' r''B- -r \\-

\r-

(L

2

u)

SI

zo
o

\ \

-. _-\r

- -: o

\D--s..-.o_.o
lloo%eo6
o -o-6.,r;.}*-

C)
lrJ

u)
tJJ

.tu.o.q

=
tr
r.5

12510

20 50 t00 200 500 1000

10000

ADDITION NIUMBER
( a)
100

-z,
=

C)

0

trl

MACHINE

=
F
I

TIM E

CYCLE

- - - ---

- _ ____

UJ

J(J

C)

l0 000

100

000

I YR

2YRS 7 YRS

M

l0 M

I

NUMBER PRODUCED ( UOG SCALE )

(b)

Figure 9.3. (a) Plot by Crossman (rgSg) of data by Blackburn (lgg6)
showing improvements with practice in time taken to add digits. O"t"
for two subiects, Sr and Sz.
(b) Plot by Crossman (rgSg) showing improvements with practice in

time taken at cigar making.

time taken by each method, and the pressrue to select - that is upon the
incentive to achieve a high rate of working.
(iii) The more sub-tasks there are in the overall task, and the more

Acquisition of Skill

3r5

they interact with one another, the more opportunity there will be for
improvement, and therefore the longer improvement will continue.
(iv) Transfer of skill from one task to another will depend not so
much upon the extent to which methods posslDle for one are applied to
the other, but the extent to which methods which haoe bem selected for
the one are applied to the other. This point provides a rational explanation of the finding by Singleton GgSil that, when learning a new method
of work, there was little interference from an established skill among
thoroughly experienced subiects, although there was among recent
trainees: one can think of the former as having selected methods so
precisely adapted to the old method that it contained nothing likely to
interfere with the new.
Crossman posed the question of how such selection takes place, and
suggested that it might do so in the short-term retention we have already
noted as necessary to bridge the gap benn een the decision to act and the

receipt of knowledge of results. He argued that if the short-term
memory trace has any tendency to decay or be disrupted with time,

quicker methods would automatically be favoured because they would
provide more immediate knowledge of results.
In one respect Crossman's theory as stated is almost certainly too
rigid. It assumes that a subiect selects from a range of discrete, readymade methods, whereas all we know of complex skilled performance
suggests rather that we 'compute' a method unique to each occasion
from the data present in the situation combined with various facets of
percepnral knowledge, motor experience, airns, strateges and so on.
This point does not invalidate the princrple of Crossman's theory because these factors can be regarded as iointly defining the methods of
which he speaks, and refinement of the computations made with them

would be approximately equivalent to seleaing among various discrete
methods. If this point is granted, it makes Crossman's approach easy
to link with a number of other observations about the progress of skill

with long-continued practice :

(") On the perceptual side, practice seems to enable the skilled performer to select from among the mass of data impinging on his senseorgans so that he neglects much of what is, to an unskilled person,
strikin& and reacts strongly to data that a normal obseryer would fail
to notice. Practice may also enable him to make absolute judgments
with much greater precision than would otherwise be possible. We
have already seen in Chapter 2 that the number of categories of absolute iudgment is normally very srnall but can, at least in the case of
musical tones, be very greatly increased in those who possess 'absolute
pitch'. The same seems to be true of other sensory iudgments, for

316

Fundamentals of Skill

example the ability to make absolute iudgments of colotr seems to be
much greater than normal among those engaged in some industrial tasks.
(b) On the motor side, action becomes more precisely adapted to the
requirements of the task, so that reliance on detailed knowledge of the
results of each individual action is reduced.
G) rn some cases, as we have discussed in Chapter 3 (p. 85), long
praaice greatly reduces the time required to make a choice of response
once a signal has been identified.
(d) One of the most striking effects of repeated experience is that a
subiect recognises ways in which sequences of events hang together

and establishes routines of action. In consequence he comes to deal
with both incoming data and outgoing action in larger 'units' as we have
outlined in Chapter 6 G). r93).
All these four tendencies with practice are towards increased efficienry in dealing with incoming data and initiating action. It is thus
understandable that the skilled man seems to have, as Bartlett (tg+l)
noted, 'all the time in the world', because he has in a very real sense
less to do than an unskilled man in order to achieve the same end. He
is in consequence less likely to suffer from fatigue, ard in many industrial tasks the speed of working will come to be limited by the machine
he is using rather than by his own capacities.
At the same time these means of attaining efficienry may endanger
'flexibility' in the sense that methods should differ .sgslding to circumstances. All the four tendencies we have listed owe much of their
effect to the attainment of uniformity and often to a iudicious neglect of
minor variations between one sinration and another. At best this can
result in a routine for a job which does not exactly fit the circumstances,
but is good enough and very much quicker than one precisely tailored
to the requirements of the sinration. At worst it can lead to 'rigidity'
in the sense that action is carried through in the face of clear evidence
that it is inappropriate.
Our knowledge of the effects of long practice is less than could be
desired because laboratory experiments can seldom be carried on long
enough for the full effects of practice to be realised, and detailed sttrdies
in industry, where long-practised performances are available, are very
difficult to make. The questions of how skill develops over the years,
of what conditions favour the attainment of efficiency and the retention

of flexibility, and of how all these matters are related to events early in
training are cleilly, however, of the greatest importance to all who are
concerned with the design and conduct of training schemes, whether in
industry, in athletics, or in any other branch of skilled performance.

x
Individuals and Social Groups

The research surveyed in the foregoing chapters has been within the
traditions of experimental and physiological psychology. The snrdy of
h rman performance includes not only this but also work on individual
differencesr on abnormalities of mental funaion and on social behaviour.

These branches of the subject have in the past developed different
concepts and terminology from those of the main stream of experimental

work. It seems obvious, however, that all represent different parrs of
what is fundamentally the same body of knowledge and that all must
therefore eventrurlly link up. Iv1any of the ideas developed from the
study of skill appear to have applications in these other areas of psychology and could play some part in bringing them together. No attempt
still be made to cover the whole field systematically here: instead an
approach will be outlined to three areas for which it seems especially
appropriate and timely. They are certain facets of personality, motioatiot
and social relntionships.

PERSONALITY

When we speak of 'personahty' we mean that certain trniformities can be
detected running through an individtral's behaviourr which qpicalty
seem to have no obvious connection with capacity but appear in a
variety of circtrmstances and affea modes of dealing with other people
and the environment generally. Research on personality has usually
described syndromes of behavioru and devised measures of the extent
to which they are displayed by different individuals (e.9. Cattell r rg46).
Most of these syndromes or 'personality traits' have been arrived at
by factor-analyses of correlations between ratings made by iudges or
statements of attinrde obtained from subiects, snd the labelling of
traits with names such as 'Rigdity', 'Ascendancy', 'Emotionality', or
'Authoritarianism' has been largely innritive. With a few striking exceptions, two of which will be discussed later, little attempt has been made
317

3r8

Fundamcntals of Skill

to relate such syndromes to other aspects of behaviour and function
except to note various correlations as, for example, between scores on
personality inventories and certain occupational performances.
Principles oudined in previous chapters suggest a number of possible
uaits which, if they were proved to run through wide areas of an individtral's performance, would provide much more precise behavioural
measures of personality than those hitherto available. It would, for
example, seem worth finding out how far traits could be defined in
terms of individuals' characteristic settings of cutoff in the signal detection model (Price, ry66), or of the balance struck between speed and
accuracy, or of the attention paid to feedback from performance, or of
tendency to use existing codes to guide behaviour rather than work out
new ones. Such traits would, however, almost certainly imply that personality was bound up with capacity and ability: to some extent all
these characteristics reflect ability to cope with situations presented - for
instance, a low criterion may compensate for poor discrimination if
false positives are not serious. Some links between personality and
ability have indeed been indicated already by factor-analytic methods:
for example Chown (rg6o) identified three types of 'rigidity' of which
one, implyrng an opinionated approach to life, correlated negatively
with intelligence. The argument that the characteristic ways of acting
which we call personality must, at least in part, reflect the extent to
which capacities and skills developed in the course of experience make
those ways possible, has been put forward by Iilallace (1966).

Personfrty, physiological funaion and behaviour
There have been two notable attempts at explanation of personality in
more basic biological terms. The first of these has been the relating of
traits to endocrine functions, localised brain actiaities and drug effects.

Many of these have been associated with characteristic styles of behaviour (e.9. Mottram, 1944, Uhr and Miller, t96o, Morgan, 1965).
How a particrrlar funaion leads to a corresponcling pattern of behaviour
has not always been clear, but important progress towards specifying
the systems or mechanisms linking them has been made recently (e.g.
Eayrs, 1964, Arttrnkal and Togrol , 1964).
The second example of a personality variable being linked to more
ftrndamental human capacities is provided by the trait of extraztersionintrooersion as defined by Eysenck (1947, rg52) on the basis of snrdies
correlating a wide variety of clinical ratings and test scores. There
seem to be clear links between this trait and autonomic activity. Those

Indittiduals and Social Crroups

3t9

who score high for extraversion tend to display lower normal levels of
autonomic activity than do those with high introversion scores. Davies
et al. (rg6f) who produced this finding suggested that a second per-

sonality variable, which we may tentatively term stability-instabiliy
was indicated by change of level when stress was applied. Autonomic
activity is, as we have noted in Chapter 8, closely bound up with arousal,
and it is now very widely assumed that inuoverts are more chronically
aroused than extraverts and drat'unstable' people becorne aroused more
easily than stable.

If so, we should expect extraverts to perfonn many tasks less well
than introverts and stable extraverts less well than unstable. Ftrrneaux
(1962) found this to be strikingly so in a study of fust-year examination
results obtained by a group of engineering students. We should also
expect unstable inuoverts to do well urder easy conditions, but to be
liable to breakdown under severe stress. The evidence in this whole
area is complex and not always easy to interpret, but several recent
experimental snrdies have indicated that extraverts perform less well
than introverts under conditions in which arousal tends to be poor as,
for example, after loss of normal sleep (Wilkinson, t962, Corcoran,
1965), with tests done early in the day (Colquhotrn, 1960, Colquhoun
and Corcoran, 1960 and with vigilance tasks (Bakan, t959, Hogan,
1966). Davies and Hockey (t966) have found the viglance of extraverts

but not of introverts to decline with time and the decline among
extraverts was prevented by noise which may be regarded as arousing.
Fine (rg6f) found extraverts to have higher rates of traffic accidents
and offences, as would be expected if they were less dert. There is also
physiological evidence in the larger amounts of narcotic drugB required
to produce a given degree of sedation in more introverted individuals
(Claridge and Herrington, 196o) and their tendency to have a stronger
salivary reastion to acid in the mouth (Corcoran, 1964). Results are not,

however, entirely straightforward. For example, although the performance of extraverts tends to be poorer than that of inuoverts in the
morning, the positions are reversed in the evening. The most plausible
reason for this appears to be that arousal level is associated with body
temperature and that the diurnal rise and fall of this occurs earlier in
introverts, so that their temperanre tends to be higher than that of
extraverts during the morning and lower during the svsning (Blake,
rs6z).
The chain of evidence is not yet quite complete because it is not
clear why the insensitivity resulting from under-arousal should produce
the ebul[ent, outgoing personality associated with extraversion, whereas

3zo

Fundarncntals of Skill

high sensitivity leads to the quietness and inhibition of introversion: at
first sight the opposite is more likely. Probably the simplest explanation
is, firstlp that those whose normal level of arousal is low can tolerate
noisy environments and lively contacts with other people without being
driven into a state of over-arousal - indeed it can be argued that they
need such stimulation if they are to be optimally aroused to meet the
exigencies of life. Secondly, they will tend not to perceive the effects
of their own actions or other people's reactions to them, so that they
will appear 'thick-skinned' and unheerting of others. In servo terms,
their performance will be largely open-loop so that when they are
stimulated, the resulting behaviour will tend to be unchecked. Those of
them who react strongly to stress will be kept lively and well-adiusted
by stimulation from the envirorunent. Those who do not so react will
tend to be inert.
People whose normal level of arousal is relatively high will be sensitive and need to keep their levels of external stimulation down if they
are not to become over-aroused. They will be perceptive of their own
behaviour and of the reactions of others to it, so that in servo terms they
\ilill possess closed-loop, negative feedback characteristics. They ruill
thus tend to be quiet and controlled. Their quietness will, however,
have a tense, dynamic quality as opposed to sheer inactivity. Provided
such inuoverts do not react very strongly to stress they will be stable
and tend to make the most of their intellectual and other capacities.
If they do react strongly to stress, or if their nonnal level of arousal is
unusually high, they will appear 'highly strung', 8rd it has been suggested that extreme cases suffer from anxious or depressive neuroses
(Davies et al., 1963).
Work by Spence and Taylor (see Taylor, 1956) has suggested that
there is an association betreen anriety and general 'drive'. More recently it has come to be held that this is probably to be regarded as an
association between anxiety and some kind of aaivation or arousal
comparable to that betvreen inuoversion and arousal. It seems reasonable to suppose that anxiety, which implies the presence of mental
activity resulting from unresolved fears, would go with higher chronic
levels of both autonomic and brain activity. Evidence that this is so is
contained in frndings such as those of higher palmar sweating longer
after-effects of viewing a rotating spiral, higher blink rate and, in some
cases although not all, higher cardiac activity in more an:rious subiects
(Haywood and Spielberger , t966, Levy and L*g, 1966, Harris et al.,
1966). The clearest effects of anxiety seem to be on intellecttral activity
where they seem to follow the Yerkes-Dodson principle (see p. z7o):

Indioifimls and Social Crroups

32t

performance by highly anxious subieas tends to be superior to that of
less anxious on easy tasks but inferior on diffisrrlt ones which involve
the recodirg of data (e.9. Tecce, 1965). They seem prone to fall back
on familiar codings and routines rather than attempting to form new
ones more precisely fitted to the sittration confronting them, and so tend
to produce stereotyped responses (e.9. Weiss and Silverman, 1966), to
be reluctant to seek change (Howard and Diesenhaus, 1965) and to have
difrc,rlty in making decisions (Riedel, 1965). We saw in the previous
chapter that highly arrxious subiects may have diffiortty in learning
material which conflicts with what has previously been learnt (Spence
et al., t956a, b, Lovaas, r96oa, b, Lee, 196r). The picnrre is complicated, however, by the fact that the ease or diffiailty of any given task
depends on the subiect's ability. More intelligent subieas are thus
likely to be less affected by anxiety because the task itself will engender
less arousal for them than it would for those less able. The level of
activation produced by the sum of the effects of the task and of anxiety
will thus be optimnm \rith a more difrctrlt task in the case of abler
subieas than with those less gfted (Denny, ry66).
It must be emphasised that personality tests yreldi"rg scores for
extraversion-inuoversion, an:riety and other traits are relatively crude
insuuments, and that measures of autonomic aaivity and other indices
of activation and arousal are far from preciser So that any tieup between
the two types of measure can at present be made only in very broad
tenns. Furttrer, the overall scores of personality tests cover uuny subtly
different patterns of answers, tnd the autonomic reactions to any given
stress vary in detail from one subjea to another and from one stressful
sittration to another in the same subiect (Lacey et al., 1953, Lacey and
LaceS 1958, Kling and Schlosberg 196r). If any unitary factor lies
behind these varied patterns, it should perhaps be identified as a general
saritioity andreactioity whrch shows itself in different forms according
to the operation of several other factors as yet not fully understood.

Capacity and environment

The analysis we have offered of these personality uaits has essentidly
been in terms of ftrndamental capacities and characteristics of the
organisrrlr embodied in a dynamic system of which the individual is
part together with his environment and other people in it. To 'explain'
such uaits we have therefore two fundamental tasks: firstly to identify
the trnderlying capacities and characteristics, rnd secondly to speciff
the sfstem in which their effects are displayed.

322

Fundamentals of Skill

The operation of such systems to produce personality effects is
further illustrated by some of the personallty changes that come with
advancing age. Let us consider three examples:
(o) While most of the personality characteristics seen in old age have
probably been present throughout adult life (Reichard et dl., 196z),
they tend to become exaggerated with advancing years. In many cases
this seems to result from a change of capacity leading to lack of control.
For example, it is well known that some of those who suffer from deaG
ness develop mild paranoid symptoms, coming to believe that the people
they can see but not hear talking are criticising or plotting against them.
What appears to happen is that when sensory input is cut off the interpretative aspects of perception are left uncontrolled. The subject's
beliefs then run along the lines of his underlying interpretative tendencies, and if these are in the direction of insecurity and suspicion,
paranoid thinking results.
(b) Chansng capacity with age may lead to compensatory adjustrnents
in the method, manner or strategy of performance which tend to minimise adverse effects and optimise the use of capacities that remain
(Griew and Tucker, 1958). Perhaps the clearest example is in certain
tlpes of sensory-motor performance where speed and accuracy are
compensatory in the sense that one may be gained at the expense of the
other (Brown, see lflelford, 1958, pp. 6748rWelford et al., 1963). In
these qlses older people tend to choose accuracy, producing more
deliberate and meticrrlous performances than those of yotrnger. As a
result they waste less time correcting errors, so that their overall
achievement changes less than measures of speed alone would suggest:
indeed in some industrial and other everyday situations where accuracy
is vial, older people may, on balance, be at an advantage.
G) Such compensatory changes imply attempts to keep one facet of
performance cutstanl while accepting changes in others. An older man
may concentrate on a smaller range of activities and so become narrow
in interests and restricted in outlook. Many of those in industry seem
to maintain achievement in later middle age by stepping up the effort
they make, working more continuously and with greater concentration,
and in consequence come to show the over-arousal, impatience and
intolerance that result from chronic stress. On the other hand, some
older people appear to try to keep effort constant as well as achievement
by recruiting the activities of others to supplement their ovun, and so
become demanding and dependent.
One of the most interesting potential applications of the approach
which has been outlined here is the analysis of responsibility and its

Indivi&nls atd Social Cnrups

323

personal counterparts reliability and intqrity. ]aques (tgS6) suggested
that responsibility could be measured by the length of time for which
a man had to work without supervisory check. Ve might generalise
such a time span to include not ody supenrisory checks but any knowledge of the results of his actions. On this view, two facets of capacity
are likely to be involved within the overall seryo system. Firstly, since
a task is seldom fully dismissed from the mind until its succmsful completion has been recognised, a long span of responsibility will often
imply that a uran will not be able to concentrate on one task at a time
but will have to deal with several incomplete tasks almost simultaneously,
retaining enough data about each to be able to pick it up quickly when

need be, while avoiding confusion berreen one task and another.
Secondly he will, for the duration of the spo, have to function openloop snd, if he is to do this successfully, his perforurance will have to
be accurate enough to be consistently effective without being checked
by observation of results. Responsible performance in this sense must
therefore be prediaable within the limits of accuracy required. The
need to take action with no immediate prospect of its being checked
also involves special problems of motivation. We shall refer to these
during the discussion of motivs to which we now turn.
MOTIVES

It is obvious that, for any astion to take place, the organism must have
the necessary capacity and that there must be an occasion provided by
environmental circumstances. It is cornnronly argued, however, trat
these are not enough and that there must, in addition, be some drive
ot motizte. Thus, for example, the mere capacity to eat, coupled with
the presence of food, is not enough to ensure that eating will take place:
there must also be htrnger or some social pressure to partake.
Several theories of motivation have attempted to trace all motives
back to primary biological necessities, either for individual survival,
such as food, water or air, or for the continuation of the race, such as
sexual intercourse and care of the young. Normallp howeverr 8 person
does not eat in order to preserve life or engage in senral activity with a
view to preserving the race. Motives may occasionally be of this kind
at a highly sophisticated level, but are usually of a more sensual nature:
a person eats because of sensations of hunger and enioys food for its
taste, smell and perhaps appearance. The biological necessity of eating
is not in mind, and the sensory gratifications precede any satisfastion of
bodily need: hunger is reduced long before food is digested and made

324

Frmdamentals of Skill

available to the tissues of the body. In short, the mechanism of biological motivation is of a more immediate nature: conditions in various
tissues and neural centres put the organism into a state in which it is
restless until particular sensory stimuli or perceptual data have been
received.

Many motives such as interests in work or hobbies, play activities
and desires for social contacts seem, however, far removed from any
biological mechanisms of this kind, or at least from the traditional
biological appetites. Ways out of this difficulty which have been proposed, such as that activities which once served a direct biological purpose may become'functionally autonomous' so that they are continued
for their own sake (Allport , 1937)., or that the achievement of sub-goals
on the way to primary satisfactions may itself be satisfying (HulI, 1943),
have an ad hoc qualrty which destroys their explanatory value. The problem is further complicated by the faa that the morives behind any
action are usully mixed, and the ways of satisfying any one motive are
many, so that an activity which stafts for one reason may be continued
for quite others: for example, a man may ioin a sports club because he
enioys exercise and may continue as a member for the sake of the friendships he has formed.
Perhaps the most radical solution is that proposed by Woodworth
(tgS8) who suggested that the organism possesses an inherent and

ftrndamental tendency to 'deal with its environment' and to develop
its capacity for doing so. He indeed argued that this was the one fundamental motive and that biological appetites, such as those for food,
water and sex 'use' this tendencyr and are thus essenti"lly subsidiary.
This approach has been developed subsequently to some extent (e.g.
McCall, 1963) but not as much as it appears to deserve. It has the
important implication that motives will be linked to abilities in the sense
that the most successful dealing with the environment will be through
the skills at the subiect's command (White, 1959). Hebron (1966) has
amassed a wealth of data to show that, all through the span of human
life, capacity, motivation and learning to go hand in hand, each depending for its development upon the progress of the others.
Whatever view is correct, the listing of motives and placing of them
in orders of importance according to circumstances is an immensely
Iaborious task, so that it is worth asking whether there are ways in
which it could be by-passed - whether principles could be formulated
which would transcend the several motives hitherto regarded as basic.
In particular, it seems possible that if we knew how motives operate,
knowledge of whal they are would become of secondary importance.

Indhtid,uals ad Social Crrrups

325

Servo concepts of motivation
An obvious starting point is that whenever the organism seeks to satis$
a need, whether biological or social, primary or derived, alone or in
co-operation with others, there is an exanrple of the classical seryo
system in which aaion is initiated by signals arising from the need and

is modified by the results of the organism's own aaivity. It is generally
agreed that the initial signals grve rise to general activity or arousal,
together with some more specifically directed action. Both depend in
part on the intrerited stnrcnrre of the nervous system but both, especially
the latter, depend also on techniques and skills acquired in the course

of experience.
Many featnres of motivation are straightforwardly accounted for by
the simple servo model. To take only one example; if a subject is
deprived of food, water or air his actions become more and more
vigorous as the deprivation becomes more acute until physical exhaustion
supervenes. SimilarlS in a number of more complex human activities
such as problem-solving and memory, the level of arousal, vigour of
action and reacliness of recall rise with increased incentive to the extent
that in extreme cases achievement may be impaired by over-activation
(e.9. Toppenr\965r!Teiner and Walkerr t966, Nakaunua and Krudis,
t96T).

Four elaborations to the simple servo model are, however, necessary.

(") Several variables rnay combine to determine the effective signal
strength. In pafriailar, the effect of any discrepancy bennreen present
state and optimum seems to be inversely related to the effort, difficulty

or other unpleasantness involved in correcting it; or to put the point
in more uaditional terms, the incentive effect of a reward is partty offset
by the cost of achieving it. The effect is, however, complicated in some
cases by the faa that effort seerns to enhance the value of the reward
eventually achieved (for a review see Lewis, 1965). Readiness to undertake an action may nevertheless be conceived as depending in part on
some kind of ratior or diffrutce, between rerult and effmt or other cost.
Such a relationship, besides providing an obvious explanation for the
conrmon reluctance to trndertake laborious or difficult tasks, suggests a
reason for certain gpes of delinquent behaviollrr such as oandaliwt,
since the ratio of effect to effort is usually much greater with destnrctive
than with constnrctive activities.
(b) A distinaion needs to be drawn benreen cases in which each
action taken contributes to the reduction of the conditions which
initiated activity, and cases in which a whole chain of actions has to

326

Fundanmtals of Skill

be completed before any reduction is achieved. To the first, the simple
servo model app[es well in the sense that motivation tends to diminish

as activity proceeds: the reduction in the rate of eating towards the
end of a large meal has its counterpart at the laboratory level in the
lowering of arousal as learning proceeds (Freedman et al., 1966).
In the second type of case, however, the speed and vigour of performance tend gradually to increase up to the point at which the goal is
attained and action ceases. How this positive feedback effect arises is
not at present clear, and any or all of several possible reasons for it
may be tnre. It could, perhaps, be an additional effect of the inverse
relationship benneen incentive and effort: at the beginning of a long
chain of actions the effort to be expended before the final result is
achieved will be large, but it will progressively climinish as the task
proceeds.

The classical explanation is based on the finding from many experiments that the incentive effect of a reward diminishes if it is delayed
after the completion of the achievement for which it is given (Hull,
1943). On this view the gap, in terms of time or intervening events,
before obtaining the reward is relatively great at the beginning of the
task and the incentive effect should therefore be low. As the task proceeds the gap shortens and the incentive effect should correspondingly

rise.

Two other very different qpes of explanation are, perhaps, possible.
One lies in the incentive effect of knowledge of results of action we have
already discussed in Chapters 8 and 9 (pp. 276 and 3o8): knowledge

that the goal was being approached should have a stimulating effect
which would progressively enhance performance so long as the conditions originally giving rise to it remained-'n being. The other explanation
lies in the finding that successful achievement of one task lowers, atrd
failure raises, the level of tension with which a subsequent task is
approached (Leshner, 196r): in these terms all action short of the goal
would be a kind of temporary failure which would tend to raise the
level of tension in the subiect.

All these explanations, except perhaps the last, can account for the
well-known difficulty of taking action before the need for it has become
pressing and, in so far as this is a feanrre of responsible behaviour'
emphasise that the corresponding facets of personality - reliability and
integrity - depend on the responsiveness of the human servo system
trnaided by regenerative loops.
(r) It is implicit in the view of extraversion-inuoversion which we
have outlined earlier that the human servo reacts not only to external

Indfuidwk and Social Ctroups

327

stimuli but also to its wl, state of activity. Essentially the same point
has been made, as we have already noted in Chapter 8, by Hebb (rgSS)
and by Berlyne (lg60) who regard the organism as trying to mainain
an optimum state of arousal by seeking stimulation if trnder-aroused
and quiet if over-aroused, although whether the optimum sought
should be defined in terms of arousal-level or throughput of information
or some other variable is a question meriting further thought. The
result in any case seerns to be that, when overloaded, the subiect spontaneously sheds paft of the load by neglecting certain aspects of his
task (e.9. Davis, 1948) or by retiring from it altogether. Strhen underloaded, however, he craves for stimulation, as in the conditions of
€xtreme monotony produced by the now famous experiments on perceptual deprivation (for a review see Zubek, r964rsee also Petrie et al.,
1960, Jones et al., 196r, Zuckerman and Haber, 1965, Smith and Myers,
1966). In everyday life, people seem clearly to prefer a moderate level
of activity to either a very high level or complete idleness, and tend to
fill their time with social contacts, gamesrprtizzles and suchlike pursuits
if there is nothing more presslng to do. Differences between individtrals
in the levels ofactivity and natlue ofptusuits required to avoid boredom
suggests that the optimum level rises with intellecnral capacity (Wyatt
and Fraser, 1929)., although this may not always be apparent because
much of the aaivity of more able people is in the form of ttrinking
unaccompanied by overt behaviour.
(d) In any real-life performance there are not one but mtmy sery)o
loops Qerating simultarcously. We have already in Chapter 6 (p. r93),
mentioned the skilled turner who, moving the tool of his lathe over the
face of a casting, is ordering and co-ordinating a series of actions which
iointly accomplish the task of machining the face concerned: this in
nrn is only one of several involved in the larger task of machining the
whole casting, strd the casting may be only one of several required for
a single iob of construction. There Brer in short, a whole hierarchy of
tasks of different magnimdes and time scalm all being performed simultaneously, the larger embracing the smaller and contro[ing the tempo,
accuracy, order and nranner in which they are carried out. It seems fair
to argue that each larger task provides the immediate motiae for its
component smaller tasks.
The hierarchical principle does not stop at the individual unit of
production. If we were to ask the turner why he was engaged on his iob
of construction, he might reply that it was part of his work at the factory, that this in turn was a means of earning a high wage, that his
earnings would enable him to buy a car, and that this would entrance

328

Fundammtals of Skill

the opportunities enioyed by his family. Each stage in this series can
again be regarded as a task and as motivating those which lie below. ft
will be noted that the higher order tasks are of a social nature while dre
lower are more individual, but they shade into each other without any
discontinuity of princlple. In essentially the same way as a construction
iob co-ordinates skilled action, the higher, social tasks co-ordinate
wider aspects of behaviour, determining priorities such as that money
should be saved for a car instead of being spent on entertainment'
deciding the choice of a iob offering high wages and overtime rather
than comfort or sectrrity, and so on.
Such a method of conceiving motivation may at first sight appear
excessively complex, yet it is really very much simpler than many
methods currently used. Instead of proceeding from a hlryothetical list:
of basic needs downwards to their manifestations in detailed behaviollr,
we are free to give oru main attention to the more immediate obiectives"
of action and to the tracing of these in the service of larger units of performance. We still do not, of course, know a priori what it is that uldmately gives direction to a subiest's dealing with his environment, but
for many practical issues, we do not need to do so: it is enough to
identiff motives in immediate individual, family and social aims and to
examine the precise ways in which they control behaviollr. When a more
ftrndamental identification is required, the tracing of reasons for action
forward to larger tasks provides a promising method of gening at itSuch a procedure would probably reveal that the simple reflexes which
must be regarded as 'ends in themselves' have their analogues in more.
highly organised behaviour: tracing forward might often carry us a very
long way, but would eventually arrive at activities which seem to have
no real aim beyond themselves. When such a point is reached we have
a plausible candidate for recognition as a basic human motive.

In these tenns the effects of any incentive are likely to be complex
since they will occur at several different levels, but they will nevertheless

be identifiable. Let us look, by way of example, at the possible nature'
and causes of job satisfact'ion. We should expect that an important factor
in the satisfaction glven by a iob would be the extent to which its feedback loops were closed and revealed substantial results of actions taken:
in plain language, one facet of iob satisfaaion is the extent to which a
man's actions have a recognisable effect on his work or work situation.
There will thus be an immediate satisfaction in operating large and
powerful machines, and a more subtle, longer-term satisfaction in being
able to influence the organisation within which one's work is done - it.
is, indeed, tempting to spectrlate that strikes and other indusuial actions

Indiztidruls and Social

Cnoups

32g

which bting powerful feedback effects arise when the feedback inherent
in the iob is deficient.
Satisfaction will, however, depend not only on the characteristics of
the iob itself, but also on the extent to which it ministers to broader
personal aims - it is well known how a compelling long-term purpose
can bring profotrnd satisfaction from an otherurise tedious or distasteful
routine; converseln the fulfilment of such an aim can greatly affect
the satisfaaion given by everyday activities. Probably the most frequent
example is the loss of a wouran's interest in work when she marries or
starts a family. Other examples occtrr when a nun or woman reaches a
coveted position to which their lives have long been oriented: the completion of such a 'lifetime' task can leave a person devoid of any strong
motives until he discovers a fresh maior obiective. Perhaps it is not too
fanciful to suggest that wider satisfaction in Ufe lies in the pursuit of
tasks which are sufficiently broad in scope to grve coherence to the lesser
aims and achievements of daily living, while having a reasonable chance
of fulfilment and replacement each decade or so during the adult years.

SOCIAL RELATIONSHIPS AND BEHAVIOUR

Most of the sttrdies surveyed in previous chapters have been of a subiect
confronting a piece of apparahrs (taking 'apparattrs' in a wide sense)
which presented a task, and the total performance of subiect and apparahrs together has depended on the characteristics of both. Such interaction benneen nran and machine is seldom discrrssed explicidy in relation to laboratory experiments although it has figrred prominendy in
some recent treatments of indusuial work (e.g. Singleton et al., tg0l).
It seems fair to argue that there is a close analogy between such rnsnmachine systems and social groups in which individtrals are in communication with one another, snd that knowledge we already possess
of the first could help in understandirg the second. Let us take a suaightforward, perhaps obvious, example by comparing the relations of, on
the one hand, an operator with a semi-automatic process plant and, on
the other, of a foreman with a skilled tradesuran in a production shop.
Just as a procss plant and its operator form a closed-loop servosystem in which each action by one influences the other and the

behaviotr of the two in combination develops with time, so the same
is tnre of any social situation in which there is two-way communication
betrreen the individuals concerned. All the well-known problems of
information-transmission are present in such'conversations', whether
between uran and machine or rran and rnan. The limited capacity of

33o

Fundamentals of Skill

both human and machine courmunication channels will set maximum
rates at which messages can be accurately received and generated, the
maximum depending on factors such as the range of possible messages
that may be conveyed and the extent to which each is clearly discrimin-

able from others and free from irrelevancies and random disnrrbance.
Ease of communication will be profoundly affected by the receiver's
understanding of the 'language' in which the information is coded and
by the extent to which knowledge and familiarity enable decisions and
insuuctions to be given in terms of broad sequences of events and
routines of action, rather than about individual detailed items. Diffioilties of short-terrn retention may arise when data are not presented all at
once but spread over a period of time, and all the effects of continued
performance such as learning, fatigue, loss of vigilance and boredom
observed in machine operation have their close analogues in hnman
interaction.

This approach has been implicit in a number of attempts to simulate
the interaction between two or more individuals in terms of mathematical models (e.9. Restle and Davis, t962, Richards, t96z) or by
programming computer's (e.9. Siegel and Wolf, t962, Gullahorn and
Gullahorn, 1963, Loehlin, 1965) working in each case only wift characteristics of individuals and of tasks and without any additional
characteristics attributable to groups as such. Probably the most direct
example of the present approach has been in the work of fugyle and his
colleagues on soc'ial skills. They have attempted not only to identify
the principles common to social and sensory-motor performances, but
also to sttrdy in detail the behaviour of subiects in social situations to
see how skill is shown in the strategies used during conversation (Argyle,

t967, Kendon, 1967).
The same methods of analysis seem to be potentidly powerful tools
for the snrdy of larger scale social organisations such as in factories or
offi,ces (Welford, r96ob, t96zb, t966, Stager, ry66). For example, the
snrdy of comrruntications in an organisation would involve asking a range
of detailed questions about each individual in the chain. I[hat is the
variety of decisions as well as the number he has to make? How many
sources of data does he have to co-ordinate? How directly do the data he
receives indicate the appropriate action to take? Are all the data for a
given decision present at one time, or do some have to be carried in
memory until other data arrive? Are the data clear and precise, or may

they be unreliable or vacillating? Even if the average rate at which
decisions have to be made is well within a man's capacity, are there

periods of overload? How far can he differentiate the results ofhis actions

hdividuak and Social Cooa6

33r

from those of actions by ottrers ? How long does he have to wait for
feedback to arrive? Does he receive sweral feedbacls at different

times? If so, do the indications of those which come quickly differ from
those which are slower - for example, does action have to be taken
which is immediately unpopular in order to bring long-term benefits?
Such comnnrnication can, to some extent, be measured in ternrs of
information in a manner similar to that of Eq. 3.6 (p. 6j). To calculate
the inforrnation transmitted by a partictrlar person in the chain we can
make a table \rith, ssy, a coltunn for each type of incoming message

and a row for each qpe of aaion taken. We calculate Eps log

r:

I

the different columns, then Z p*log,?n for the different row, and finally
X p",

fog* individual cell. The information transmitted is then cal-

culated in exactly the same way as the information gained in Eq. 3.6
by adding the first two sums and subuacting the third. To assess the
load on the person concerned we need to consider the average and
distribution of information transmitted per message in relation to the
rate and intervals at which the messages arrive. More elaborate analyses
can be made in the manner outlined by writers on multi-dimensional

information uansmission (e.9. Attneave, 1959, Garner, t96z). An
interesting application of this type of analysis to reports of faults and
actions taken to deal with them has been outlined by Leuba (tg67).

Some further applications

To illustrate the possibilities of this approach further we shall apply it
briefly to four other areas either of well recognised concern to indusuy
or of uaditional interest to social psychology. It cannot be claimed that
any radically new conclusions are reached in any of these cases, but the
treatment suggests ways of looking at some present problems which
open up new methods of tackling them.

(") Flexibility of an organisat'ion. If a working group is to be adaptable
to changing conditions, or is even to maintain a high standard in stable
conditions, it must not only have clear insuuctions, but feedback tf
inforrwtion about the quality of performance attained. This can be
achieved to a limited extent by observation, on the part of a foreman
or manager, of the end product of the group he controls, but much
fuller information will obviously be obtained when there is rapid and

332

Fundammtals of Skill

easy two-way commtrnication berween manager and managed which
gives detailed insight into difficulties encountered and ideas about how

they might be overcome. In stable conditions, once the methods for a
iob have been established, it is possible to run on satisfaaorily for an
appreciable time without such feedback, and there may indeed be a
temptation to avoid it as an unwelcome addition to the information load
imposed by the iob. When, however, conditions are rapidly changing,
or in the early stages of developing an operation, such feedback will
gtve a flexibility and rapidity of adjustment not otherwise possible.
(D) Behaoiour of crowds. One of the uaditionally difficult problems of
social psychology is to account for the uniformity of the behaviour of

crowds. Accounts in terms such as 'social facilitation' are not truly

explanatory, since they are essentially ad hoc, and thus little more than
descriptive. The present terms, however, provide the simple hypothesis
that, when the members of a crowd are all aaing alike, the sight and
sotrnd of others provides each individual with a kind of augmented
feedback of his own behaviolr. This would be especially so if the
crowd were, say, shouting slogans in unison. It has its analogue on a
smaller scale in the feeling of 'unusual power' reported by crews of
rowing eights when their boats are going well: when all are pulling
precisely together each man is said to get the feeling that the whole
result is due to his own effort. The seeming ratio of effect to effort is
thus gready enhanced, and it is reasonable to suppose that the satisfaction gained from the aaivrty is correspondingly increased.
How far other aspects of crowd behaviour could be accounted for
in the same terms is not at present clear, but seems to be worth
serious study. For example, how far is the effect of size of crowd to be
explained in terms of signal-to-noise ratio, in the sense that, as the crowd
becomes larger, augmented feedback increases in suenBtr, and individual differences of behaviour tending to blur the unison are smoothed
out ?

(r) Social norms. The broad princrples of coding outlined in Chapter 6
seem especially applicable to social norms and customs. These essentially

represent the codi$ing of insights and building of routines which
avoid complex ad hoc decisions by providing ready-made solutions to
recurrent problemsr and enable individuals to predict the behaviotrr of
others. The learning of such norms and customs is an important process
in the structuring of individual perception and behaviour. Modifications
to norms and customs take place in the course of time, presumably as a

Indhtiduals and Social Crroups

333
response to changrng circumstances or fresh insights into methods of

dealing urith the problems of living. There is, however, usually a substantial time-lag before adjustments become effective, so that they are
often out of phase with the current needs of society and oscillation may
occur with swings of over-adjustment and readiustment taking place
over a period of years.
We have already noted that the building-up of routines in individuals
as the result of experience makes for greater efficienry in stable coDditions but greater rigidity in the face of change. The gradual accumulation of norms and customs in a society seems likely to have similar
effects, making a mature society more stable, since fewer decisions of
policy are open to question, but less capable of absorbing new ideas
quickly than societies in which rules of conduct are less developed and
entrenched.

(d,) Leadqship and size of group. Harcotut (rg59) has suggested that the
number of people one man can lead will depend on the amount of
information they iointly generate: if they produce too much, the leader
will be overloaded and will be liable to take hasty, ill-considered decisions. The amount of information individual members produce will
depend on the nature of the iob, the conditions under which it is done
and their personal characteristics. If they are performing a routine iob
under stable conditions and are of even temperament, one man will be
able to lead a large number. If, however, the work is not of a routine
nature so *rat it requires constant detailed attention, or if various members are doing different iobs, or if working conditions are unpredictable
or there are unstable personalities among the team, the maximum size
of group that one man can lead effectively will be reduced.
Consideration of the capacities of both leader and group members
together suggsts possible reasons for relationships which have been
claimed to exist benreen morale and size of working group. ft has been
urged that a small group is more 'democratic', while a large group
tends to generate an 'authoritarian' leadership which is less acceptable.
Evidence about which type of organisation leads to better results and
greater iob-satisfaction is, however, conflicting. A possible reconciliation
is provided by the results likely to follow if a group generates more
information than a single leader can deal with. In the terms we have
been using, a 'democratic' system means that there is a possibitity of
effeaive feedback from group members to leader, whereas an 'authoritarian' system means that members of the group are unable to influence
the leader's decisions. As we have seen in Chapter 8, s common reaction

334

Furdammtals of Skill

to overload in laboratory tasks is that the subiea sheds part of the load
by ignoring some of the signals he should observe and omining some
of the actions he should take. Usually these are the less frequent ones,
so that the subiect concentrates on the main feanrres of his task and
ignores side-issues. In this way he often nranages to put up a reasonable,
although not, of course, wholly adeqtrate performance. One way in
which an overloaded leader could readily 'shed load' would be to rely
on routines established in his group to keep it ftrnctioning while ignoring feedback from members. This would in effect mean that leadership
would become 'authoritarian'. It would doubtless have little adverse
effect for a time, so at first it would seem to be working well, but an
important means of recogmsing the need for modification or change
would have been lost, and the group's performance would thus tend to
become progtessively less effective.

On the other hand, the alternative procedures available when a group
becomes too large for one person to lead are not without difficulties of
their own. There seem to be two main possibilities: firstly members of
the group can be required to refer particular problems to one of a range
of specialists added to the group as deputy leaders. The disadvantage of
this system is that individual members may have to make difficult decisions about which specialist to go to, and may find themselves going a
tedious round from one specialist to another until the right one to deal

with their particular problem is found. They may well feel that a
reasonably benevolent authoritarianism is preferable.

Alternatively, the group can be split into smaller sub-groups, each
with a leader who can call on specialists and is responsible in turn to a
leader higher in a hierarchy. In this way the members are still able to
refer all their problems to one person, and although the difficulty of
gening the right specialist service may merely be uansferred to him, the
morale and effectiveness of the group as a whole is likely to be higher
than if such decisions have to be made by individual members. The
superiority of the second system is likely to increase with the frequency
of problems requiring decisions for individual members. Thus, with a
routine operation under stable conditions, there may be little to choose
benn een the effectiveness of the two methods of organisation, but if
technical and management functions are highly complex the second
could be expected to yield better results.

Indivifuals and Social Ctroups

33s

Towards unification of the htunan sciences
Over and above partictrlar applications, perhaps the main importance of
the approach which has been outlined here is that it links the snrdy of
human performance, and of the physiological mechanisms underlyrng
it, to shrdies of social behaviour. Psychology occupies a position intermediate between the social sciences on the one hand and the older
human biological disciplines on the other. In his sturdies of individual
human behaviour the psychologist has been forced from time to time
to reco$use that certain social factors may exert important effects. On

the other hand, he knows that he is snrdying a biological organism
whose behaviour is based on a nervous system reaaing via sense organs
and muscles with the environment, so that he must sooner or later
look to physiology for his explanations. Both physiologist and psychologist recognise that these explanations cannot at present be supplied,
and indeed that *re goal is not fully attainable because the detailed
breakdown of behaviour into physiological terms would be impossibly
complex.

The need to maintain contact nevertheless remains. The division
benreen the two disciplines is in a sense arbimary, in teuns of the size of
unit studied - betrnreen, for psychology, the whole organism and, for
phpiology, individual cells and stnrctures. fn many ways this division
is both convenient and necessary, nevertheless there are occasions
upon which consideration by the psychologist of the detailed mechanism
of the human brain and body can tie together many faas at first sight
disconnected or even discordarrt. At the same time, the physiologist
considering the action of large masses of nerve cells has often to resort
to the snrdy of behaviour at the level normally within the realm of
psychology. There is thus a two-way traffic between the disciplines in
which psychology seeks theory and explanation downwards, and physiology seeks the testing of hypotheses upwards in the scale of funaional
units.
It is reasonable to suggest that psychology can and should play the
same role in relation to social studies as physiolory does in relation to
psychology, providing the means of concepnralising the deailedmechanisms of the behaviour of groups and organised social units. For psyche
Iogy to play this role, however, the principle must be recognised that
social units are composed of individtrals, and that it is their interaction
\dth their environment and each other that produces social phenomena.
Such recognition implies that accounts of social phenomena need to be
broken down from steady states or slow changes into processes in which,

$6

Fundammtals of Skill

ideallp chains of detailed individual actions can be described. It is
often impossible to make this kind of breakdown, iust as it is impossible
to analyse individual behaviour into detailed physiological processes,
but the attempt needs to be made, and even if it fails for a time, the
ultimate aim must be acknowledged. The psychologist must in ttun
remember that, iust as psychological findings have sometimes pointed
to matters requiring physiological research, so social sttrdies are likely
to direa his attention to problems in his own field that might otherwise
pass unnoticed.
The opportunities for co-operation appear to be very substantial. A
sustained attempt to exploit them by the disciplines concerned would
be very much to their munral advantsge, lending to social studies an
often needed precision and to psycholory a desirable perspective.

References

ABBEyT D. s . (tg64) 'Conuol-display-subject interaction and performance

on a complex perceptual-motor task', Ergutontics, 7, l5r-t64.
ADAMS, f . A. (1954)'Multiple versus single problem uaining in human
problem solving', j. exp. Psychol., 48, t5-t8.
ADAMS, f. A. (rgSS)'A source of decrement in psychomotor performance'rJ. exp. Psychol., 49t 39v394.
ADAMs, f . A. (tg6z) 'Test of the hypothesis of psychological refractory
period' , j. exp. Psychol., 64, z8o-287.
ADAMS, I. A., snd BouLTER tL. n. (1962) 'An evaluation of the activationist hypothesis of human vigilance', j. exp. Psychol., 64, 495-504.
ADAMS, f . A., and cHAMBERS, R. w. (1962) 'Response to simultaneous
stimulation of two sense modalities', j. etc,. Psychol., 63, t98-2o6.
ADAMs, f . A., and DIIKsTRA, sANNE (t966) nshort-terrr memory for
motor responses', j. exp. Psychol., 7l, 3r4-3r8.
tMonitoring
ADAMS, I. A., STENsoNT H. H., and HUMES, I. M. (lg6t)

of complex visual displays: 2. Effects of visual load and response

complexity on human vigilance', Human Factors, 3t 2r3-22r.
ADTsESHTAH, w. (r95r) Unpublished Ph.D. Thesis, Llniversity of
Cambridge.
ADRTAN, E. D. (tg+il The Phy$cal Backgrotmd of Perceptian. Oxford
University Press.
AIKEN, L. R., and LIcHTENsTEIN, M. (1964)'Reaction times to regularly recurring visual stimuli', Percep. mot. Skills, r8r 7t3-72o.
ALLroRTI G. w. (r937) Personality: a Psychological Interpretatiot New
York: Henry Holt & C,o.
ALLUrsr, E. A., and HALLe T. I. (1963)'Declines in auditory vigilance
dtrring periods of high multiple-task aaivity', Percep. mot. S&e7ls, r5t
739-74a..

AMMoNS, R. n. (1947) 'Acquisition of motor skill: r. Quantitative analysis and theoretical formulatiorr', Psychol. REtt. 54,263-28r.
ANDERs oN, NANcy s. (196o)'Poststimulus cuing in immediate memory',
j. exp. Psychol., 60,, zr.Gzzt,
ANDERsoN, NANcy s., and LEoNARDT f . A. (tgS8) 'The recognition,
naming and reconstruction of visnal figrrres as a firnction of contour
rednndancy', j. exP. Psychol., 56, z6z-27o.
ANDREAsSTT I. L. (1965) 'Effects of induced muscle tension and auditory
337

338

References

stimulation on tachistoscopic perception', Percep. ?not. Skills, zo, 82984t.
ANDREAssT, I . L, (t966) 'Skin-conductance and reaction-time in a cootinuous auditory monitoring task', Amer. J, Psychol., 79t 47v474.
ANNETT, I. (1959) 'Learning a pressure under conditions of immediate
and delayed knowledge of results', Quart. j. exp, Psychol., rrr 3-r5.
ANNETT, f. (1966a) 'A note on Davis's refutation of the expectancy
hypothesis', Quart. J. exp. Psychol., r.8, r79-r8o.
ANNETT, I. (r966b) 'Training for perceptual skills', Ergonomicsr gr 459468.

ANNETT, f ., GoLBy, c. w., and KAy, H. (1958)'The measurement of
elements in an assembly task - the information output of the human
motor system', Quart. j. exp. Psychol., r.o, r-rr.
ANNETTT J., and KAy, H. (tgSil'I(nowledge of results and "skilled
performance"', Occupational Psychol., 3r, 69-79.

ANTRoBUS, I. S.e and sINGER, f . L.(1964)'Visual signal detection as a
function of sequential variability of simultaneous speech, J. exp.
Psychol., 681 6o3-6ro.
ARcHER, E. f ., KENT, G. w.r and MorE, F. A. (1956).Effectof longterm practice and time-on-target information feedback on a complex
tracking task', J. exp. Psychol. St, ro3-r12.
A R c H E R, E . J ., and N A M r K A s, c . A . (r 958)'Pursuit rotor performance as a
function of delay of information feedback', l. exp. Psychol.r16r321-327.
ARcYLE, M. (tg6il The Psychology of Interpersonal Behaaiour. Harmondsworth : Penguin Books.
ARTUNKALT s., and TocRoL, n. (1964) 'Psychological snrdies in hyperthyroidisrrl', in Brain-Thyroid Relationships (ed. Margaret P. Cameron

and Maere O'Connor). Ciba Foundation Study Group No. r8.
London: f. & A. Churchill, pp. g2-roz.
AsH, r. E. (r9r4)'Fatigue and its effects upon control', Arch. Psychol.,
4, No. 3r.
ATKrNsoN, R. c. (1963)'A variable sensitivity theory of signal detection', Psychol. Ra,)., 7ot 9t-ro6.
ATTNEAvE, F. (tgSo) 'Dimensions of similarity', Amer. 1. Psychol., 63,

5r6-556.

ATTNEAvET F. (tgS+) 'Some informational aspects of visual perception',
Psychol. Reo., 6r, t83-193.
ATTNEAvE, F. (tgSil 'Transfer of experience with a class-schema to

identification-learning of patterns and shapes', J. exp. Psychol., S4t

8r-88.

ATTNEAvEe F. (lgSg) Applications of Information Theory to Psychology.

New York: Henry Holt & C,o.
AUDLEv, R. f ., and wALLrsr c. p . (t964) 'Response instructions and the
speed of relative iudgments. r. Some experiments on brightness discrimination', Brit, J. Psychol., SSt 5y73.

Refermces

339

AvERBAcH, 8., and spERLINc, c. (196r) 'Short tenn storag3 of infor-

mation in vision', in Information Theory (ed. C. Cherry). London:

Butterworth.
BADDELEY, A. D. (tg6+a) 'Language-habits, S-R compatibility and
verbal learning', Amer. J. Psychol., 77t 463-468.
BADDELEY, A. D . (tg64b) 'Semantic and acoustic similarity in short-term
memory't Naturer zo+ 1116-r tt7.
BADDELEY, A. D. (r966a) 'The influence of acoustic and semantic similarity on long+erm memory for word sequences', Quart. j. exp.
Psychol., 18, 3o2-3o9.
BADDELEY, A. D. (r966b) 'Short-terrn memory for word sequences as a
ftrnction of acoustic, semantic and formal similarity', Quart. J. exp.
Psychol., r8, 362-365.
BADDELEY, A. D.e coNRAD, R., and HULLr A. J. (1965)'Predictability

and immediate memory for consonant sequences', Quart. j.

exp.

Psychol., r7t rTS-r77.
BAHRTCKI H. P., BENNETT, w. F.e and FITTS, p. M. (tgSSa) tAccuracy
of positioning responses as a function of spring loading in a control',
J. exp. Psychol., 49, 432-444.
BAHRICK, TI. P., FITTS, P. M., and SCHNEIDER, R. (Tgssb).Repro.

duction of simple movements as a function of factors influencing
proprioceptive feedback', J. exp. Psychol., 49t M5-454.

BAHRTGKT H. P.r NOBLE, M., and Frrrs, p. M. (lgS+) tExtra-task per-

formance as a measure of learning a primary task'r}. exp. Psychol., 48,
298-3o2.
BAHRIcKI I{. p., and sHELLv, cARoLyN (rgS8) 'Time sharing as an
index of automatization', 1. exp. Psychol., 56, 288-293.
BAKANT P. (tgSz) 'Preliminary tests of vigilance for verbal material',
U.S.A.F. Hwttan Resources Research Cmffe, Research Note Sz-7,

Lakeland Air Base.
BAKAN, P. (lgSg) 'Extroversion-introversion and improvement in an
auditory vigilance task', Brit. J. Psychol., Sot 325-332.
BAKER, c. H. (tgSg) 'Attention to visual displays during a vigilance
task:2. Maintaining the level of vigilance', Brit. j. Psychol., 5or 3036.

BAKERe c. H. (196o)'Factors affecting radar operator efficienCy', J. Inst.
Naaig.., r3r 148-163.

BAKER, c. H. (tg6t) 'Maintainirg the level of vigilance by means of
knowledge of results about a secondary vigilance task', Erganontics, 4,

3rr-3r6.

BAKERT c. H., and youNc, p. (t96o)'Feedback during training and
retention of motor skills', Canad.
Psychol., r'4; 257-264,
BAKER, KATHERINE E., wyLIE, RUTH c., and GAGNEe R. M. (rgSo)
'Transfer of training to a motor skill as a function of vrariation in rate
of response', j. exp. Psychol., 4o,, 72r-732.

j,

34o

References

BAKER, R. A., and wARE, J.R . (t966)'The relationship between vigilance
and monotonous work', Ergonomics, 9, rog-rr4.
BARMACKT J. E. (tggg) 'The length of the work period and the work
curve', I. exp. Psychol., 2St ro9-r15.

BARRY, E. (rg6+)'Effects of prolonged deliberation on psychophysical
iudgment', Amer. J. Psychol., 77, z7o-275.
BARTLETTT F. c. (tglz) Runembering. Cambridge lJniv. Press.

BARTLETT, F. c. (rg+3)'Fatigue following highly skilled work', Proc.

Roy. Soc. Il., r3rt 247-257.
BARTLETT, F. c . (t947) 'The measurement of human skill', Brit. med. J.,
i, 8gS and 877. Reprinted (rg+s) Occupatiutal Psychol., 22t 3r-38

and 83-9r.
BARTLETTI F. c. (tgSo) 'Programme for experiments on thinking',
Quart J. exp. Psychol., 2, r41-t52.

BARTLETTI F. c. (tgS8) Thinkingt cwt Expertmental and Social Study.
London: Allen & LInwin.
BARTLETT, F. c.e and MAcKwoRTH, N. H. (rgSo) Planned Seeing.
London: Ff.M.S.O. Air Publication 31398.
BARTLEY, s. H., and cHUTE, Elorsn Gg+l) Fatigue and Impairment in
Man. New York and London: McGraw-Hill.
BARTZ, w. H., SATZ, p.r FENNELL, EILEEN, and LALLy, f . R. (t967)
'Meaningfulness and laterality in dichotic listening'. j. etcp. Psychol.,
73r 2a4-2to.
BASSETT, M. F. and wARNE, c. f . (r9r9) 'On the lapse of verbal meaning
with repetition', Amer. j. Psychol., 3,ot 4t5-4r8.
BATTEN, D. E. (1967) 'Recall of paired associates as a function of arousal
and recall interval', Percept. mot. Skills, 24, ro55-ro58.
BECKER, p.'\il'., MUSSTNA, cARoLyN M., and PERSONS, R.'w. (1963)
'fntertrial interval, delay of knowledge of results and motor performance', Percep. tnot. Skills, t% 559-563.

BEcBrE, c. H. (tgSg) 'Acctrracy of aiming in linear hand-movements',
Quart. J, exp. Psychol., rrr 65-75.

BETLIN, H., and HoRN, RHEne (1962) 'Transition probability effects in
anagram problem solving', 1. exp. Psychol., 61, 5r4-5r8.
BErsHoN, R. I. (r.967) 'Problems of task description in process control',
Ergonotttics, ror t77-t86.

BErsHoN, R. f ., and cRAwLEy, f . E. (1965) 'The control of a continuous
baking oven, an analysis of a complex skill', Ergonomics, 8t 375.
BEKESY, c. voN (tg6il Sutsory Inhibition. Princeton University
Press.
BELBTN, EUNTcE (lgSo) 'The influence of interpolated recdl upon recog-

nition', Qtnrt.|t'. etcp. Psychol., 2, t63-t69.
BELBTN, EUNTcE (tgS8) 'Methods of training older workers', Ergonomics,
r.r 2o7-22r.

Refermces

34r

BELBIN, EUNIcE (tg6+) Training the Adult Worker. D.S.I.R. Problems
of Progress in Industry No. 15. London: EI.M.S.O.
BELBTN, EUNICE, BELBTN, R. M.e and HrLL, F. (t957) 'A comparison
between the results of three different methods of operator training',
Ergonomics, rr 39-50.
BELBTN, E., and DowNS, s. M. (t964)'Activity learning and the older
worker', Ergonomics, % 429-437.
BELBTN, E.r DowNS, s., and MooREl B. (tg6+) "'Lfnlearning" and its

relationship to age' , Ergonornics, % 4t9-427.

BELLe c. R., pRovrNS, K. A., and HroRNs, R.'w. (tg6+) tVisual and

auditory vigilance during exposure to hot and humid conditions',
Ergonomics, % 279-288.

BENsoN, A. f ., and cEDyE, J. L. (tg6z) 'Some supraspinal factors
influencing generalised muscle activity', in Symposiwn on Skeletal
Muscle Spaun (ed. P. C. Turnbull). Leicester: Franklyn, $7ard and
W'heeler.
BERGER, c.e and MAHNEKE, A. (rgS+)'Fatigue in two simple visual

tasks', Amer. J. Psychol., 67, Sog-Srz.
BERGSTRoM, f . A. (tgo7)'Effects of changes in time variables in memo-

rizing, together with some discussion of the technique of memory

experimentation', Arner. 1. Psychol., l.t, zo6-238.
BERGUM, B. o.1 and LEHR, D. f. Q96z)'Vigilance performance as a
function of interpolated rest', 1. appl. Psychol., 46 425-427.
BERcuM, B. o., and LEHR, D. J.(r963a)'Effects of authoritarianism on
vigilance performance'rJ. appl. Psychol., 47t 75-77.
BERGUM, B. o.e and LEHR, D. J. (rg6fb) 'End-spurt in vigilance',
J. exp. Psychol., 661 383-385.
BERcuM, B. o., and LEHR, D.I. (1964)'Monetary incentives and vigilance', 1. exp. Psychol., 67, t97-t98.
BERLYNET D. B. (rg6o) Confltct, Arousal and Cuiosity. New York:

McGraw-Hill.
BERLYNE, D. E., BORSA, DONNA M.1 HAMACHER' JANE H., and

KoENrc, rsoLDE D. v. (tg66) 'Paired-associate leaming and the
timing of arousal', j. exp. Psychol., 74 r4.
BERLYNET D. E.e CRAW, MARGARET A., SALAPATEK, p. fl.e and
LEwrs, IUDr rH r,. (1963)'Novelty, complexity, incongnrity, extrinsic
motivation and the GSR'rI. etcp. Prychol.r 66, 56u567.
BERRYe R. N. (1962)'Skin conductance levels and verbal recall', j. exp.
Psychol., 61, z7S-277.
BERTELSoN, P. (tg6t) 'Sequential redr:ndancy and speed in a serial
two-choice responding task', Quart.l. exp. Psychol., t3,t gc-roz.
BERTELSoN, P. (rg6E) 'S-R relationships and reaction times to new
versus repeated sigBals in a serial task', 1. exp. Psychol., 65, 478-484.

BERTELsoN, P. (tg6S) 'Serial choice reaction-time as a function of
response versus sigRal-and-response repetition', Nature, 2016, 2t7-2t8.

342

References

BERTELSoN, P. (t966) 'Central intermittency twenty years later',

Qttart. j. exp. Psychol., r8, t53-r63.
BERTELSoN, n. (r967a) 'The refractory period of choice reactions with

regular and irregular interstimuli intenrals', In Attention utd Per-

formance (ed. A. F. Sanders). Amsterdam: North-Flolland Publishing
Co.
BERTELSoNT p . (t967b) 'The time course of preparation', Qtmrt.l. exp.
Psychol., T9t 272-279.
BERTELSoN, p., and BARZEELE, f . (rg6S) 'Interaction of time-uocertainty and relative signal frequency in determining choice reaction

,#;;I;ii'; :3*"";,1';#:X;' RENKTN, A. (rg6s) .vitesse ribre

et vitesse impos6e dans une tache simulant le tri mdcanique de la

correspondanc e', Ergonunics, 8, 3-22.
BERTELSoNI p., and JoFFE, RAcHEL (lg6l) 'Blockings in prolonged
serial responding', Ergonomics, 6, ro9-r 16.
BERTELSoN, p., and RENKIN, e. (1966)'Reaction times to new versus
repeated sigpals in a serial task as a function of response-signal time
intenal', Acta Psychol., 2St t3z-t36.
BEVAN, w., AVANT, L. L.r tnd LANKFoRDe H. c. (1967)'Irrfluence of
interpolated periods of activity and inactivity upon the vigilance dectement', J. appl. Psychol., S\ 352-356.

BEVANT'w.1 HARDEsry, D. L.1 and AVANT, L. L. (tg6S) tResponse
latency with constant and variable interval schedules', Percep. mot.
S&//s, zor 969g7z.
BILLS, A. G. (tgzil'The influence of muscular tension on the efficiency

of mental work', Atner.l. Psychol,, 38, zz7-25t.
BILLs, A. G. (lglr) 'Blocking: a new principle in mental fatigue', Amer.

J. Psychol., 4lt z3o-245.

BILLS, A. c., and BRowNe c. (tgzg) 'The quantitative set', .7. exp.
Psychol., r'2, 3or-323.

BrLLs, A. G., and sHAprN, M. f. (rg:6)'Mental fatigue under automatically controlled rates of work'rI. gen. Psychol., rSt 335-347.
BILoDEAU, E. A. (1966) Acquisition of Shill. New York: Academic
Press.

BILoDEAU, E. A., and BTLoDEAU, INA McD. (r958a) 'Variable frequency
of knowledge of results and the learning of a simple skill', ]. exp.
Psychol., SSt 37gi83.

BILoDEAUI E. A.e and BTLoDEAU, rNA McD. (r958b) 'Variation of
temporal intervals among critical events in five studies of knowledge
of results',I. exp. Psychol., SSt 6o34tz.
BILoDEAU, E. A., foNES, M. B., md LEvy, c. M. (tg6+) tl-ong-teun
memory as a funaion of retention time and repeated recalling', j. exp.
Psychol., 671 3o3-3o9.
BTLoDEAU, E. A., and RvAN, F. f . (196o)'Atest for interaction of delay

kfermces

343

of knowledge of results and two t)?es of interpolated activity'r 1. exp.
Psychol., 59, 4r4-4r9.
BILODEAU, rNA McD. (1956) 'Accuracy of a simple positioning response
with variation in the number of trials by which knowledge of results is
delayed', Amer. 1. Psychol., 69t 434-437.
BTLODEAU, INA McD. (1965) 'Transfer of training across target sizes',
j. etcp. Psychol., 7ot r35-r4o.
BILODEAU, INA MCD., and ROSENQUTST, H. S. (1964) 'supplementary
feedback in rotary-pursuit uackinr'r I. etcp. Psychol., Gg, s3-57.
BINDER, A., wol.IN, B. R., WEICHELT ROSEMARIT, and TEREBTNSKI,
s. f . (1966) 'Llncertainty and stage of decision', Amer. I. Psychol.r 79,
89-96.
BTNETT A., and HENRT, v. (1894) 'La m6moire des mots', Annie Psychol.,

r, t-23.

BINFORD, f . R., and LoEBe M. (t966)'Changes within and over repeated

sessions in criterion and effective sensitivity in an auditory vigilance
task', j. exp. Psychol., 72t 339-345.
BIRRENT I. E.e and BorwrNrcK, I. (1955)'Speed of response as a function of perceprual difficulry and age', J. Gerontol., tot 433-436.
BLACKBURN, f . M. (1936)'Acquisition of skills: an analysis of learning
curvest, Indust. Hhh. Res. Bd. Rep., No. 73. London: H.M.S.O.

BLAKE, M. f . F. (tg6il 'Relationship between Circadian Rhythm of
body temperature and introversion-e:ftraversion', Naturet 2t1t 8g6897.
BLANKENSHTP, A. s. (1938) 'Memory span: a review of the literature'

Psychol. Bull.r 35r t-z1.
BLrcK, K. A., and BTLODEAU, E. A. (1963)'Interpolatedactivityandthe
learning of a simple skill', j. exp. Psychol., 651 5r5-5r9.
BLYTH, K. w'. (lg6g) 'Ipsilateral confusion in 2-choice and 4-choice
responses with the hands and feet', Naturet r99r t3tz.
BLYTHT K. w. (1964)'Errors in a further four-choice reaction task with

the hands and feet', Naturer 2ot, 64t442.
BODERT D. P. (1935) 'The influence of concomitant

activity and fatigue
upon certain fonns of reciprocal hand movernent and its fundamental
components', Cotttp. Psychol. Monogr.t rr, No. 4.
BooKr w. r. (r9o8) The Psychology of SkiA. LIniv. Montana Studies in
Psychol., Vol. r. Republished rgz1. New York: Crregg.
BooNSr f . P., and BERTELSoN, p. (r96t) 'L'influence de l'inceninrde
temporelle sur le temps de rdaction de choix', Armde Psychol., 6r,
36r-376.
BoRcER, R. (1963) 'The refractory period and serial choice-reactions',

Quart. j. exy. Psychol., tSt t-tz.
BORNEMANN, E. (tg+z)'[.Intersuchungen uber der grad der geistigen
beanspnrchung', Arbeitsphysiologu, r.2, t4z-r9r.
BoswELL, J. f ., and BILODEAU, E. A. (t964) 'Short-term retention of a

344

Refermces

simple motor task as a function of interpolated activity', Percep. mot.
Skills, r.8, zz7-23o.
BorwrNIcK, J., and BRINLEy, J. F. (r96za)'Aspects of RT set during
brief intervals in relation to age and sex', J. Gerontol., r7t 295-3or.

BorwrNIcK, f ., and BRINLEv, f. F. (rg6zb) 'An analysis of set in
relation to reaction time', J. exp. Psychol., 63, 568-574.
BorwrNrcK, J., BRTNLEy, f . F., and RoBBrN, f . s. (1958)'The interaction effects of perceptual difficulty and stimulus exposure time on
age differences in speed and accuracy of response', Gerontologia, z,

I-IO.
BorwrNrcK, ;., and sHocK, N. w. (tg1z) 'Age differences in performance decrement with continuous work', .7. Gerontol., Tt 4t-46.
BorwrNrcK, J., and THoMpsoN, L. '\il'. (rg66) 'Premotor and motor
components of reaction time', J. exp. Psychol., 7rt 9-r5.
BoULTER, L. R., and ADAMS, f. A. (rg6g)'Vigilance dectement, the
expectancy hypothesis and intersignal interval', Canad. Jt'. Psychol., r7t
2OT-2O9.

BRADLEy, J. v. (rgSg) 'Direction-of-knob-turn stereotypes', j. appl.
Psychol., 4!t 2r-24.
BRATNARD, R. \il.r rRBY, T. S.e Frrrsr'P. M.randALLUrsr, E. e. (1962)

'Some variables influencing the rate of gain of information', I. exp.

Psychol., 6!, ro5-r ro.

BRAy, c. 'w. (rgz8) 'Transfer of learninr', J. exp. Psychol,, t\ M3467.

BREBNER, f ., and coRDoN, r. (tg6z) 'Ensemble size and selective
response times with a constant signal rate' , Qtnrt. j. etcp. Psychol., r$
t t3-r t6.

BREBNER, f., and coRDoN, r. Gg6+)'The influence of signal probability and the number of non-signal categories on selective response
tirnes', Quart. j. exp, Psychol., l.6, 56-6o.
BRIcHcIN, M. (t966) 'Teoreticke a metodologicke problemy rryzkumu
prubehu volnich pohybu', Acta Unia. Carolinae, Philos. List. Monogr.
No. r 2. Prague : LIniversita Karlova.
BRTCKER, P. D. (r955a) 'Information measurement and reaction time: a
review', in Information Theory in Psychology (ed. Henry Quastler).
Glencoe, Illinois: The Free Press, pp. 350-359.
BRTcKER, P. D. (tgSSb) 'The identification of redundant stimulus patterns', J. exp. Psychol., 49t 73-8t.
BRTGGS; G. E., Frrrs, p. M.1 and BAHRTCK, fI. p. (lgSZ) tEffects of
force and amplitude ctres on learning and performance in a complex
tracking task', J. exp. Psychol., S4s z6z-268.
BRrccse G. E., and wATERs, L. K. (tgS8) 'Training and transfer as a
ftrnction of component interactioD', J. exp. Psychol., 56, 4g2-Soo.
BRoADBENTe D. n. (1952) 'Speaking and listening simultaneously', j.exp.
Psychol., 43t 267-273.

Referutces

345

BRoADBENT D. E. (tgSz) 'Failures of attention in selective listening',
j. exp. Psychol., 44, 428-433.
BRoADBBNT, D. r. (1954) 'The role of auditory localisation in attention
and memory span', j. exp. Psychol., 47t t9t-t96.
BRoADBENT, D. E. (rgS6) 'Listening between and during practised
auditory distractions', Brit. j. Psychol., 47, 5t4o,
BRoADBENT, D. E. (tgSZa) 'Imrrrediate memory and simultaneous
stimuli', Quart J. exp. Psychol., 9; r-rr.
BRoADBENTT D. E. (rgSZb) 'A mechanical model for human attention
and immediate memory', Psychol. Reu., 64, zo5-215.

BRoADBENT, D. E. (lgSS) Perception and Cuwnunication. London:
Pergamon Press.
BRoADBENT, D. E., and cREcoRy, MARGARET (tg6t) 'On the recall of

stimuli presented alternately to two sense-organs', Quart. I.

exp.

Psychol., t3s ro3-ro9.

BRoADBENT, D. E.r and cREGoRy, MARGARET (t.g6z) 'Donders' Band C-reactions and S-R compatibility', 1. exp. Psychol.r 63, 575-578.
BRoADBENT, D. E.r and cREcoRyr MARcARET (tg6ta) uDivision of
attention and the decision theory of signd detegtiorr', Proc. Roy. Soc.
.B., r58r 22z-23r.
BRoADBENT, D. E., and cREcoRy, MARGARET (r.g6lb) 'Vigilance considered as a statistical decision', Brit. 1. Prychol., 54t 309-323.
BRoADBENT, D. E.e and cREGoRyr MARcARET (r.g6+) 'Stimulus set
and response set: the alternation of attentiof , Quart. j. exp. Psychol,t
16, 3o9-3r7.

BRoADBBNTT D. 8., and cREGoRyr MARcARET (rg6S) 'On the inter-

action of S-R compatibility with other variables affecting reaction
tirne', Brit. J. Psychol., 56, 6t47.

BRoADBENT, D. 8., md cREGoRyr MARGARBT (tg6il'Psychological
refractory period and the length of time required to make a decision',

Proc, Rojt. Soc. Il., t68, r8r-193.
BRoADBENTI D. E., and HERoNT A. (tg6z)'Effects of a subsidiary task

on performance involving immediate memory by yotrnger and older
men', Brtt. J. Psychol., Slt r89-r98.
BRoADHURsT, p. L. (1957) 'Emotionality and the Yerkes-Dodson I-aw',
1. ex7. Psychol., S$ 345-352.

BRowNl c. H. (tglil'The relation of magninrde of galvanic skin responses and resistance levels to the rate of learninr', 1. eNp. P$tchol,,

zo, z6z-278.

BRowN, r. D. (196o) 'Many messages from few sources', Ergononticsr 3,
r 59-r 68.

BRowNe r. D. (tg6z) 'Measuring the "spare mental capacity" of car
drivers by a subsidiary auditory task', Ergonamics, 5, 247-250.
BRowN, r. D. (rg6Sa) 'A comparison of two subsidiary tasks used to
measnre fatigue in car drivers', Ergonamics, 8, 467-473.

346

Refermces

BRowN, r. D. (r965b) 'Effect of a car radio on driving in traffic', Ergonomicsr 4 475-479.

BRowNr r. D.e and pouLToN, E. c. (rg6r) 'Meastrring the spare
"mental capacity" of car drivers by a subsidiary task', Ergonomics, $
35-40.

BRowN, f . (tgS+) 'The nature of set-to-learn and of inua-material
interference in immediate memory', Qunrt. J. exp. Psychol., 6, r4rr48.

BRowNr I. (1958) 'Some tests of the decay theory of immediate memory',
Quart. j. exp. Psychol., !o, tz-zt'.

BRowNr J. (rgSg) 'fnformation, redtrndancy and decay of the memory
trace', in Mechanisatton of Thought Proce.sse.s. National Physical Laboratory Symposium, No. ro. London: Ff.M.S.O.
BRowN, RUTH A. (1957) 'Age and paced work', Occupational Psychol.,

3rl I r-2O.

BRowN, R. L., GALLowAyr w. D., snd sAN GIULIANo, R. A. (rg6S)
'Effects of time-sharing and body positional demands on cutaneous
information processing', Percep. moL Skills, zo, rozr-r 026.
BRUcE, D. f . (rgS8) 'The effect of listeners anticipation on the intelli-

gibility of heard speechn, Lang. and Speecht lt 7997.
BRvAN, w. L., snd HARTERT N. (1897)'studies in the physiology and
psychology of the telegraphic language', Psychol. Rq)., 4, zZ-53.
BRvAN, w. L., snd HARTER, N. (t8gg) 'Studies on the telegraphic
language. The acquisition of a hierarchy of habits', Psychol. Reo., 6,
345-375.

BUcKe L. (rg6f) 'Errors in the perception of railway signals', Ergutomics,

6, r8r-tgz,

BUcK, L. (1966) 'Reaction time as a measure of perceptual vigilance',
Psychol. Bull.., 65, zgr-3o4.
BURKE, D., and crBBS, c. B. (t:g6S)'A comparison of free-moving and
pressnre levers in a positional control system', Ergonorticsr S, z3-29.
BURSTLL, A. E. (tgS8) 'The restriction of peripheral vision during
exposure to hot and humid conditiorrs', Quart. J. exp. Psychol., ror

tt3-r29.

cALLANTTNET M. F.e and wARREN, f. M. (rgSS)'Learning sets in
human concept formation', Psychol. Rep., \ 363-367.

cAMERoN, c. c. (tg6+) Unpublished B.A. Thesis. IJniversity of
Adelaide.

CARDozo, B. L., and LEopoLD, F. F. (tg6l)'Human code transmission: letters and digits compared on the basis of immediate memory
enor rates', Ergonomics, 6, t33-r4r.

CARMTCHAEL, L., HoGAN, rr. p.e and WALTERe A. A. (tglz) 'An
experimental study of the effect of language on the reproduction of
visnally perceived form'r1. ercp. Psychol., r5t 73-86.
cARrENTER, A. (tg+6)'The effect of room temperaftEe on performance

kfrmces

347

of resistance box test', Med. Res. Cotilu Appl. Prychol, Res., Unit
Rep. No. 50.

cARrENTER, e. (r95r) 'A case of absolute pitch', Quart.l. exp. Psychol.,
3t 92-93.
cARTERETTET E. c., and coLE, M. (rg6f) 'Repetition and confirmation
of messages received by ear and €ye', Quart. j. exp. Pgrchol., r5r
r55-172.
cASTANBDE, A.e and LIrSITT, L. p. (rgSg) 'Relation of stress and
differential position habits to performance in motor learning', j. exp.
Psychol., 57, 2S-3o.

cATTELL, f . M. (r.8gf) 'On errors of observation', Amer. J. Psychol.,
5,285-293.

cATTELLT R. B. (tg+6) Description and Measurement of Personality.
London: G. G. Harrap & Co.
cHAMBERSI E. c. (rg6r) 'Industrial fatigue', Occupational Psychol., 3St

4-57.

cHANEy, F. 8.1 and TEEL, K. s. (tg6il'Improving inspector performance through training and visual aids', l. appl. Psychol.r1tt 3r r-3r5.
cHAsE, R. A.e cuLLEN, J. K.1 oppENsHAw, f. w.1 and suLLrvAN,
s. A. (1965) 'Studies on sensory feedback: III. The effects of display
gain on tracking performance', Quart. 1. exp. Psychol., r7t r93-2o8.
CHASET R. A., HARVEY, S.r STANDFAST, SUSANT RAPIN, ISABELLE,
and surroN, s. (r96ra)'Snrdies on sensory feedback: I. Effect of
delayed auditory feedback on speech and keytapping', Quart J. exp.

Psychol., rlt t4t-r1z.
CHASE, R. A.1 RAPINT ISABELLE, GILDEN, L.2 SUTTONT S.e and
culLFoyLE, c. (tg6rb) 'Studies on sensory feedback: II. Sensory
feedback influences on keytapping motor tasks', Quartj. exp. Psychol.,
r3r r 53-t67.

cHEATHAM, p. c. (tgSz) 'Visual perceptual latency as a function of
stimulus brightness and contour shape', I. exp. Prychol., 4lt 369-38o.
cHBRRy, G. (1953) 'Some experiments on the recognition of speech with
one and with two ears', J. Acoust. Soc. Amer.r 2St 97Sg7g.
cHrNN, R. Mcc., and ALLUIST, E. A. Gg6q)'Effect of three kinds of
knowledge-of-results information on three measures of vigilance performanc€', Percep. mot. Skills, 18, 9or +tz.
cHocHoLLE, R. (tg+o) 'Variation des temps de reaction auditifs en
fonction de l'intensit6 a diverses fr6quences', Annie Psyclnl., 4v4t,

65-t24.

cHocHoLLE, R . (t943) 'Relation de la Latence d'une sensation auditive
diff6rentielle avec l'amplitude d'une variation brusque de frdquence:
. . . d'intensit6', Bull. Aiol., r3,7t 6+14++1 75t-752.
cHowN, sHEILA M. (rg6o) 'A factor analysis of the Sfesley Rigidity
Inventcry: its relationship to age and nonverbal intelligence', 1. aht.
Psychol., 6r, 49r-494.

348

Refermces

cHowN, SHEILA, BELBIN, EUNIcE, snd DowNS, syLvIA (tg6l)
'Programmed instruction as a method of teaching paired associates to
older learners', j. Gerontol., 22, 2r2-2t9.
cHRrsrrE, L. S.e and LUcEe R. o. (1956)'Decision structure and time
relations in simple choice behaviour', Bull. ntath. Biophy.s., r8, 89-llr.
CLARTDGE, c. s., and HERRrNcroN, R. N. (196o) 'sedation threshold,
personality and the theory of neurosis', J, ment. Scf., ro6, i568-r583.
CLARKI n. J. (1965)'Recognition memory for random shapes as a function of complexity, association value and delay', j. etcp. Psychol., 69,
5go-595.

CLAY, HTLARY M. (r95+) 'Changes of performance with age on similar
tasks of varying complexity', Brit. J. Psychol., 4St 7-r3.
CLAY, HTLARY M . (t957) 'The relationship between time, accuracy and
age on similar tasks of varying complexity', Gerontologia, \ 4T-49.
CLAY, HTLARY M. (tg6+) Research in Relation to Operator Training.
Monograph available from the Librarian, Science Research Council,
London, S7.C.r.
coHENe B. u. (1963)'An investigation of recoding in free recall', l. exp.

Psychol., 65, 368376.
coHEN, L. H., and LTNDLEy, s. B . (t936) 'The relationship of muscle
tonus changes to vibratory sensibility' , Psychol. Monogr.r 47, No . zrz,
pp. 83-93.
coLQUHouN, w. P. (tg6o) 'Temperament, inspection efficienry, and
time of day', Ergonomics, 3t 377-378.
coLQUHouN, lry. P. (t96t) 'The effect of "unwanted" signals on per-

forrnance in a vigilance task', Ergonomics, 4, 4r-Sr.
coLQUHouN, 'w'. P. (t962) 'Effects of hyoscine and meclozine on
vigilance and short-term memory', Brit. j, indust. Med.t t9t 287-296.
coLQUHouN, w. P. (t966) 'The effect of "unwanted" signals on per-

formance in a vigilance task: a reply to ferison'r&rgonomicsrgt1r7-4r9.
coLQUHouNr'w'. P. (tg6il'Sonar target detection as a decision pro-

cess', j. a?pl. Psychol.r 5rr r87-t9o.
coLQUHouN, 'w. p.r and BADDELEvI A. D. (tg6+) 'Role of pretest
expectancy in vigilance decrement', J. exp. Psychol., 68, 156-16o.

coLQUHouN, w. p., snd BADDBLEv, A. o. (1967) 'Influence of signal
probability during pretraining on vigilance decrement' ,1. ex?. Psychol.,
7* r53-r55.
coLQUHouN, w. p., and coRcoRAN, D. w. f . Gg64)'The effects of
time of day and social isolation on the relationship between temperament and performance', Brtt. J. soc. clin. Psychol., 3, zz6-23t.

coNRADr R. (r95r)'Speed and load stress in sensori-motor skill', Brit.
J. indu.str. Med., 8, r-7.
coNRAD, R. (tgS+a) 'Missed signals in a sensorimotor skill', j. exp.
Psychol,, 48, ra.

coNRAD, R. (r954b)'Speed stress', in Sympositmt ut Human Factors in

kf*mces

34g
Equipmant Design (ed. $tr. F. Floyd and A. T. Welford). London:
ff . K. Lewis & Co. for The Ergonomics Research Society, pp. 95- roz.

coNRAD, R. (lgS6) 'Performance of telephone operators relative to
traffic level', Nature, r78t r48o-r48r.
coNRAD, R. (t957)'Decay theory of immediate memory', Nature, 11g,

83r-832.
coNRAD, R. (1958) 'Accuracy of recall using keyset and telephone dial,
and the effect of a prefix digit', J. appl. Psychol., 42, z}5-zgg.
coNRAD, R. (r96oa) 'serial order intnrsions in immediate memory',

Brit. J. Psychol., 5r, 45-48.

coNRAD, R. (r96ob) 'Letter sorting machines - paced, "lagged" or rrrrpaced?', Erganomics, 3t r49.r17,

coNRAD, R. (r96oc) 'Very brief delay of immediate recall', Quart. J.
exp. Psychol., r.2, 45-47.

coNRADr R . (t96za) 'Practice, familiarity and reading rate for words and
nonsense syllables', Quart. j. exp. Psychol., r,4t 7t-76.
coNRAD, R. (t96zb) 'An association bennreen memory errors and errors
due to acoustic masking of speech', Nature, rgSr r3r4-r3r5.

coNRAD, R. (1963) 'Acoustic confi,rsions and memory span for words',
Naturet r91, Tozg-ro3o.
coNRADr R . (t96&) 'Acoustic conftrsions in immediate memory', Brit.
J. Psychol., SSt 75-84.
coNRAD, R. (tg6S) 'Order error in immediate recall of sequences',

J. zterb. Learn. uerb. Beh., 4 t6t-t6g.

coNRAD, R., FREEMANe p. R., and HULL, AuDREyr I. (1965)'Acoustic
factors versus language factors in short-term memory', Psychon. Scf.,
3r 57-58.

coNRAD, R.1 and HTLLE, BARBARA A. (tgSil 'The decay theory of
immediate memory and paced recall', M.R.C. Applied Psychology
Research Unit Report No. 3o7.
coNRADr R., and HULL, A. J. Gg6+)'Information, acoustic confusion
and memory span', Brit. j. Psychol., S$ 4zT43z.
coNRAD, R.1 and HULL, AUDREy, f. (1966) 'The role of the inter-

polated task in short-teun retentior', Quart. j. exp. psychol., 18, 266269.

cooKr T.

(tglil 'Amotrnt of material and difficulty of problem

solving'r 1. exp. Psychol., zo, 288-296.
coRBALLrs, M. c. (1966) 'Memory span as a function of variable presentation speeds and stimulus dtrrations', j. exp. Psychol.r Trt 46t-465.

coRBALLrs, M. c. (tg6il 'Serial order in recognition and recall',

j. exp. Psychol., 74t 99-ro5.

coRcoRANr D. 'w. f . (lg6+) 'The relation between introversion and
salivation', Amer. J. Psychol., 7% 298-3oo.

coRcoRANr D. 'w. f . (lg6S) 'Personality and the inverted-Lf relarion',
Brit. J. Psychol., 56, 267-273.

35o

Refercnces

coRcoRAN, D. w. f . (rg66) 'Prediction of responses to multidimensional from responses to unidimensional stimuli',}. exp. Psychol,, 7rt
47-54.

cosrA, r. D., VAUGHAN, H. G., and GTLDEN, L. (tg6S) tComparison
of electromyographic and microswitch measures of auditory reaction
time', Percep. tnot. Skills, 2o,, 77r-,772.
coul,ESr I., AvERy, D. L., and MESKTL, A. (1965)'Information trsnsmission in a pattern discrimination task as a function of initial task

diffiorlty', Percep. mot. Sfu//s, 2r, g2TA3g.
couRTS, F. A. (rggg) 'Relations between experimentally induced muscular tension and memorization', j. exp. Psychol., 2St 235-256.

couRTS, F. A. (tgp) 'Relations between musculiar tension and pGr-

formance', Psychol. Bull.t 39t 347-367.
cRAFTS, L. \fi/., snd cILBERTe R. w. (f935)'The effect of knowledge of
results on maze learning and retention', J. educ. Psychol.r 26, t77-r87.
cRAIK, F. r. M. (t967) Personal commrurication.
cRArKl K. f . 'w. (rg+f) The Nature of Explanation. Cambridge IJniv.
Press.

cRArK, K. f . vy . (t947)'Theory of the human operator in control systems.
I. The operator as an engineering system', Brit.J. Psychol.r 3& 56-6r.
cRArK, K. f . w. (r9+8) 'Theory of the human operator in control systems.
II. Man as an element in a control system', Brit. J. Psychol., 38,

r42-r48.
cRANNELL, c.'\tr/'., snd rARRISHT J. M. (tg17)'A comparison of immediate memory span for digits, letters and words' , i . Psychol. r 4+3rg-327 .
cRAwFoRD, A. (r96t) uFatigue and driving', Ergonamics, 4, t43-r54.
cRAwFoRD, IUNE, HUNT, E., and rEAK, G. (tg66) 'Inverse forgetting in
short-term memory', 1. exp. Psychol., 72t 415-422.
cREAMERT L. R. (rg6g) 'Event unceftaintyr psychological refractory
period, and human data processing', J. exp. Psychol., 66, t87-t94.
cRossMANl E. R. F. tv'. (rgSf)'Entropy and choice time: The effect of
frequency unbalance on choice-response', Quart. J. exp. Psychol., 5,

4r-5r.

cRossMANe E. R. F. w'. (1955) 'The measurement of discriminability',
Quart. j. exp. Psychol., 7t t76-t95.

cRossMAN, E. R. F.'rry'. (lgS6) 'The information capacity of the human
operator in symbolic and non-symbolic control processes', in Infor"ntation Theory and the Human Operator. Min. of Supply Publication
\fR/D2 / 56.
cRossMAN, E. R. F.'w. (tgST) 'The speed and accuracy of simple hand
movements', in The Nature and Acquisitton of Indusnial Skills, by
E. R. F.
Crossman and !7. D. Seymour. Report to M.R.C. and
D.S.I.R. ]oint Cornmittee on Individual Efficiency in Industry.
cRossMAN, E. R. F. w'. (1959) 'A theory of the acquisition of speed-skill',
Ergonortics, 2, r 53-r 66.

Refermces

3jr

cRossMAN, E. R. F. w'. (r96oa) 'The inforrration-capacity of the human
motor-system in pursuit tracking', Quart. J. exp. Psychol., t2t r-16.
cRossMAN, E. R. F.'w. (r96ob) Automation and Shill. D.S.f.R. Problems
of Progress in Industry No. 9. London: EI.M.S.O.
cRossMANe E. R. F.'lJV. (196r)'Information and serial order in human
immediate memory', rn Informatio?t Theory (ed. C. Cherry). London:
Butterworth & Co., pp. r47-r59.
cRossMANr E. R. F.'w., and cooDEREr P.I. (1963) Feedback contol of
hand-movement and Fitts' Laut. Communication to the Experimental
Psychology Society.
cRossMANe E. R. F. \tv., and SZAFRAN, I. (1956)'Changes with age in
the speed of inforrrration intake and discrimination', Experientia Suppl.,

q tz8-r35.

cRowDER, R. c. (tg6il 'Reciprocity of retention and interpolated-task
scores in short-term memory', Percep, mot. Skills, 24, go3-9o9.
DALEr H. c. A. (1964) 'Retroactive interference in short-term memory',
Naturer 2olt r4o8.
DALE, H. c. A. (1967) 'Familiarity and free recall', Quart,J. exp. Psychol.,

19, ro3-ro8.
DALEr H. c. A., and cREcoRy, MARGARET (rq66)'Evidence of semantic
coding in short-term memory', Psychut. Scl'., St 75-76.
DALLETT, K. M. (tg6+)'Effects of a redundant prefix on immediate
recall', I, exp. Psychol., 67, z9Gzg8.
DARcuse H. D. (tgS+) 'The range and strength of ioint movement', in
Hwnan Factors in Equipmant Design (ed. V. F. Floyd and A. T.
S7elford). London: fI. K. Lewis & Co. Ltd.
DARDANO, f . F. (t962)'Relationships of intermittent noise, intersignal
interval, and skin conductance to vigilance behaviour', J. atplPsychol.,

46, ro6-r14.
DARRow, c. \f,r'. (tgfS) 'Emotion as relative functional deconication:
the role of conflict', Psychol. Ret)., p, 56G578.
DAvrEs, D. R., and HocKEy, G. R. !. (1966)'The effects of noise and
doubling the signal frequency on individtral differences in visual
vigilance performance', Brit. 1. Psychol., S% 38t-389.
DAvrEs, M. H., CLARIDGEe G. s., and wAwMAN, R. f . (1963) 'sedation
threshold, autonomic lability and the excitation-inhibition theory of
personality: III. The blood pressure response to an adrenaline antagonist as a measure of autonomic lability', Brit. j. Psych,tat., rogr 5i8567.

DAvrs, D. R. (tg+8) Pilot Error. Air Ministry Publication A.P.3r39A.
London: H.M.S.O.
DAvrs, D. R. (rg+g) 'The disorder of skill responsible for accidents',
Quart. 1. exp. Psychol., r, t3Gt4z.

DAvrsr D. R. (1958) 'Human errors and transport accidents', Ergonomics,
2t 24-33.

352

Refermces

DAvrs, D. R. (tq66) 'Railway signals passed at danger: the drivers,
circumstances and psychological process es', Ergutamics, 9t 2tr-222.
DAvrs, R. (1956) 'The limits of the "psychological refractory period"',
Qtnrt J. exp. Psychol., 8, 24-38.
DAvrs, R . (t957) 'The human operator as a single ctrannel information
system', Quart. j. ercp. Psychol.r g, rr9-T2g.
DAvrs, R. (1959) 'The role of "attention" in the psychological refractory
period', Quart. J. exp. Psychol,, u, zrr-22o.
DAVIS, R. (1965) 'Expectancy and intermittency', Qtnrt, j. exp. Psychol,,

r7t 75-78.
DAvIs, R.5 MORAYT N.r and TRETsMAN, ANNE (r96ra)'Imitative
responses and the rate of gain of information', Quart. j. etcp. Psychol.,

r3r 78-89.
DAvrs, R., SUTHERLANDe N. s., and IUDD, B. R. (r96rb) tfnformation
content in recognition and recall', 1. exp. Psychol,r 6r, 422-429.
DAvrs, R. c. (tg+o) 'Set and muscular tension', Indiana (J. Publ. Sa.
Ser. No. ro.

DAvrs, s. vv. (t955)'Auditory and visual flicker-fusion as meastrres of
fatigue' , Arner. J. Psychol., 68, 654-657.
DEANET G. E. (196r) 'Human heart rate responses dtrring experimentally

induced anxiety', J. exp. Psychol., 6t, 48g-4g3.
DEEs, vALERTE, and cRrNDLEy, G. c. (r95r) 'The effect of knowledge

of results on learning and performance: IV. The direction of the
error in very simple skills', Quart. 1. exp. Psychol., 3, 36-42.

DEESE, f . (rgSS) 'Some problems in the theory of vigilan,ce', Psychol.

Rq).r 62, 359-368.
DEESE, f ., and KAUFMAN, R. e. (1957) 'serial effects in recall of unorganised and sequentially organised verbal material', l. exp. Pytclnl.,
54, r8o-r87.

v. H.e Bfid ranAs, c. c. (tg6o) 'Supplementary report:
The Yerkes-Dodson law and shift in task difficulty', 1. eNp. Psychol.,

DENENBERG,

59t 429-43r.
DENNY, f . P. (1966) 'Effects of anxiety and intelligence on concept
formation', j. exp. Psychol.r T2t 5964o2.
DENNY, M. R., ALLARDe M.e HALL, E., and ROKEACHT M. (rg6o)
'Supplementary report: Delay of knowledge of results, knowledge of
task and intertrial interval', j. exp. Psychol.r 6o,1 3z7,
DENNY-BROWN, D. (tg+g) ufnterpretation of the electromyogram',
Arch. Neurol. Psych'iat., 6t, gg-r28.
DEUTScH, J. A., snd DEUTscH, D. (1963) 'Attention: Some theoretical
considerations', Psychol. Reu., 7o, 8o9o.
DrMoND, s. f . (1966) 'Facilitation of performance through the use of the

timing system', j. ezcp. Psychol., 7tt r8r-r83.
DOHERTY, M. E. (1966) 'Response bias and Murdock's D Scale', Psychol.
Bull.r 66,289-29e..

Refrences

353

DoNDERS, F. c. (t868) 'Die Schnelligkeit psychischer Processe' Arch.

Anat. Phys., 65748t,
DooLEy, R. p., and NEwroN, f . M. (1965) 'Transfer of uaining between
quickened and unquickened displays', Percep. mot. Skills, 2r.) r.t-t1.

Douct,Asl c. c., and KocH, A. c. E. (r95r)'Carbohydrate metabolism

and muscular exercise', j. Physiol., tr4t 2o8-22r.
DRAzrN, D. H. (196r) 'Effects of foreperiod, foreperiod variabiliqy, snd
probability of stimulus occurrence on simple reaction time', j. etcp.
Psychol.r 62, 43-So.
DUDLEvI N. A. (rgS8) 'Output pattern in repetitive tasks', Inst. Prod.
Engrs. j., 37, 3o3-3r3.
DUFFy, E. (tgSil 'The psychological significance of the concept of
ttarousal" or "activation" ', Psychol. Rer,)., 64, 265-275.
DUNcAN, c. p. (1958) 'Transfer after maining with single versus multiple
tasks', J. ex?. Psychol., SS,-63-72.

DvAL, f . A. (rg6+)'Effects of delay of knowledge of results in a linedrawing task', Percep. mot. Skills, r9t 433-434.

DvAL, f . A. (1966) 'Effects of delay of knowledge of results and subiect
response bias on extinction of a simple motor skill', j. exp. Psychol.,

7\ 559-563.

DyAL, f . A.5'wILSoN, w. f ., and BERRy5 K. r. (1965)'Acquisition and
extinction of a simple motor skill as a function of delay of knowledge
of results', Quart.l. exp. Psychol., t7t 158-162.
EAGLE, M. N. (tg6il'The effect of learning strategies upon free recall',
Amq. J. Psychol., ht er-p1.
EAsoNr R. c., BEARDSHALL, ANN, and IAFFEE, s. (t965) 'Perfonnance
and physiological indicants of activation in a vigilance situation',
Percep. mot. Skills, 2o,, 3-r3.

EASoN, R. c., and BRANKS, f . (tg6f) 'Effect of level of activation on the
quality and efficiency of performance of verbal and motor tasks',
Percep. mot. Skills, r.6, 525-543.
EAsoN, R. c., and wHrrEe c. T. (tg6t)'Muscular tension, effort and
tracking difficulty: studies of parameters which effect tension level and
performance efficiency', Percep. mot. Skills, r2r 33r-372.
EAsrERBye R. s. (tg6il 'Percepnral organization in static displays for
man/machine systems', Ergonotnics, ror rg1-2o5.
EAvRS, f . r. (1964) 'Effects of thyroid hormones on brain diftrentiation',
in Brain-Thyraid Relationships (ed. Margaret P. Cameron and Maere

O'Coruror). Ciba Foundation Study Group No. r8. London: f. &. A.

Chtrrchill, pp. 6o-7r.
EccLEs, f . c. (tgSE) The Natrophy$ological Basrs of Minds. Odord
Llniversity kess.

EDwARDS, r. (1963) 'The integration of spaced signals', Erganonics, 6,

r43-t52.

EDWARDST Er. T., MARGARIA, R., and DrLL, D. B. (tgf+)

tMetabolic

354

References

rate, blood sugar and the utilization of carbohydrate', Amer. j. Physiol.,
ro8, 2o3-2o9.
EDWARDS, w. (196r) 'Behavioural decision theory', Ann, Reu. Psychol.,

r2t 473-498.

EGAN, J. p.e CARTERETTEI E. c.1 and THWINGe E. f. (lgS+)'Some
factors affecting multi-channel listening', J. Acoust Soc. Amer., 26,
774-782.
EGAN, f . p., SCHULMAN, A. r., and GREENBERG, G.z. (t959)'Operat-

ing characteristics determined by binary decisions and by ratings',
J. Acoust. Soc. Amer.t 1tt 768-773.

ELrrHoRN, A. (196r) 'Central intermittency: some further observations',
Quart. J. ex?. Psychol., t3,t 240-247.

ELrrHoRN, A., and BARNETT, T. f . (tg67)'Apparent individual differ-

ences in channel capacity', in Attqttion and Performance (ed. A. F.
Sanders). Amsterdam: North-Holland Publishing Co.
ELTTHORNe A., and LAwRENcE, cATHERTNT (1955) 'Central inhibition

- some refractory observations', Quart. j, exp. Psychol.r T, ttGltzT.
ELLror, P. B. (t964)'Tables of d", in Segzal Detection and Recognition
by Hutnan Obserwers (ed. f . A. Swets). New York: Viley.
ELLrorr, R. (1965)'Reaction time and heart rate as functions of magnitude of incentive and probability of success', J. Pers. soc. Psychol., 2,
6o44o9.

EL-TEMAMY, M. A. A. (tg66) Unpublished Ph.D. Thesis. Llniversity
of Birmingham.
EMMERICHI D. S.' GOLDENBAUM, D. M., HAYDEN, D. L.e HOFFMAN,

LINDA s.1 and TREFTS, JEANNE L. (rg6S)'Meaningfulness as a

variable in dichotic hearimg', j. ezcp. Psychol., 69, 433-436.
ERTKSEN, c. 'w., and HAKE, H. '\ry'. (1955) 'Multidimensional stimulus
differences and accuracy of discrimination', J. exp. Psychol.r 5or
r 53-r6o.
ERTKSEN, c. \M., and JoHNSON, H. f. Gg6+)'Storage and decay
characteristics of nonattended auditory stimuli', J. exp. Psychol., 68,
z8-36.
ERTKSENT c. vr., and LApprN, f . s. (tg67)'Selective attention and very
short-term recognition memory for nonsense forms', J. exp. Psychol.,
73t 358-364.
ERIKSEN, c. \i[r.: and srEFFye R. A. (tg6+) 'Short-terrn memory and

retroactive interference in visual perception', J. exp. Psychol., 68,
423-434.

ESTES, w. K., and Polrroe F. DA (tg6i 'Independent variation of
information storage and retrieval processes in paired-associate learning', j. ex?. Psychol., 75, t8-26.

EvANs, S. n. Gg6il'Redundanry as a variable in pattern perception',
Psychol. Bull., 67, ro4-r13.
EYSENCK, H. f . (t947) Dimmsions of Personality. London: Kegan Paul.

References

355

EysENcK, H. J. (lgSz) The Scimtific Study of Persuulity. London:
Routledge & Kegan Paul.
FINE, B. f . (rg6f) 'Introversion-extraversion and motor vehicle driver
behaviorlr', Percep. mot. Shills, rG 91-roo.
FISHERe R. A., and vATES, F. (rgf8) Statistical Tables for Biological
Agrtcultural and Medical Research. London: Oliver & Boyd.
FITTS, p. M. (rgS+) 'The information capacity of the human motor
system in conuolling the amplitude of movement', j. exp. Pqtchol., 47t
38r-39r.
FITTs, p. M. (t966) 'Cognitive aspects of information processing: III.
Set for speed versus accuracy'r 1. exp. Psychol., 7tt 84W57.

FrrrS, p. M. e and' BIEDERMAN, r. (rg6S) 'S-R compatibility and
information reductior', J. exp. Psychol.r 69, 4o8-4t2.

FITTs, p. M.1 and DEININGER, R. L, (tgS+) 'S-R compatibility: correspondence among paired elements within stimulus and response
codes', j. exP. Psychol., 48, 483-492.

FITTs, p. M.; and rETERSoN, f . R. (tg6+) 'Information capacity of

discrete motor responses', j. exP. Psychol., 67, ro3-ttz.
FITTST p. M.e pETERsoN, f , R., and woI,pEl G. (rg5f) 'Cognitive
aspects of information processing: II. Adiustments to stimulus
redundancy', j. exp. Psychol., 65, p3-432.
FITTS, p. M., and RADFoRD, BARBARA r. (1966) 'Information capacity
of discrete motor responses under different cognitive sets', j. exp.
Psychol., 7rt 475-482.
Frrrs, p. M.r and sEEGER, c. M. (rgSf)'S-R compatibility: spatial
characteristics of stimulus and response codes', j. ucp. Psychol., 46,

t99-2rc.

FITTS, p. M., and swrrzER, c. (1962)'Cognitive aspects of information
processing: I. The familiarity of S-R sets and subsets'r1. exp. Psychol.,
63, 3zr-329.
FITTS, P. M., WEINSTEIN, M., RAPPAPORT, M., ANDERSON, NANCY,
and LEoNARD, J. A. (tgS6) 'Stimulus correlates of visual pattern
recognition: a probability approach', J. exp. Psychol., Stt r'-rr.
FLETsHMANe E. A., and rARKER, f . F. (t962) 'Factors in the retention
and relearning of perceptual-motor skill'r j. exp. Psychol., 64, zt5-226.
FLETsHMAN, E. A., and RrcH, s. (1963)'Role of kinesthetic and spatialvisual, abilities in perceptual-motor learnin g', I. exp. Psychol., 66, 6-r r.
FLoyDr w. F., and wELFoRD, A. T. (eds.) (tgSf) Symposium on Fatigue.
London: ff. K. Lewis & Co. for the Ergonomics Research Society.
FoRREST, D. w. (rgS8)'Influence of length of task on rate of work and
level of muscular tension' , Occupational Psychol., 32, zS3-257.
FoRRrN, 8., and MoRrN, R. E. (t966) 'Effect of contextual associations
upon selective reaction time in a numeral-naming task', j. exP. Psychol.,

7r, 4vq6.

FoRRIN, B,, and MoRIN, R. E. (tg6il'Effects of context on reaction

356

Refrences

time to optimally coded signals', in Attention and Performance (ed.

A. F. Sanders). Amsterdam: North-Holland Publishing Co.
FosrER, HARRIET (t962)'The operation of set in a visual search task',
J. exp. Psychol., 63, 74-83.

FowLER, B . (t964) Unpublished B.A. Thesis. Llniversity of Adelaide.
Foxe R. Ir., GOLDSMITH, R., HAMPToN, r. F. G., and wrLKrNsoN,
R. T. (1963) 'The effects of raised body temperattrre on the performance

of mental tasks'r 1. Physiol., t67t 22-23.
FRAISSEe P. (tg+z)'Etudes sur la memoire immediate: II. La rdproduction des formes rhythmiques', Annde Psychol., 43, ro3-r43.
FRArssEs P. GgqD'Etudes sur la memoire imm€diate: III. L'influence
de la vitesse de presentation et de la place des elements. La nature du
present psychologique', Annde Psychol., 4* 29-42.

FRA r s s E, P . (r9 5T)'La P6riode r6fractoire psychologique', Annde Psychol.,

S% 315-328.
FRAISSE1 P., and BLANCHETEAU,

M. (tg6z) 'The influence of the

number of alternatives on the perceptual recognition threshold',

Quart. J. exp, Psychol., t4, S2-SS.
TRASER, D. c. (tgS:) 'The relation of an environmental variable to
performance in a prolonged visual task', Quart. J. exp. Psychol., St

3l-32.

,FRASER, D. c. (lgS8) 'Decay of immediate memory with age', Nature,

l.8z, r 163.
FREEDMAN, N. L.1 HAFER, BRENNA M.e and DANIELT R. s. (1966)
'EEG arousal decrement during paired-associate learning', j. comp.
physiol. Psychol., 6t, r5-r9.
FREEMAN, c. L. (tg:l) 'Mental activity and the muscular processes',
Psychol. Ra)., 38, 428-4p;g.
FREEMAN, c. L. (1933) 'The facilitative and inhibitory effects of muscular tension upon performance', Amer. J. Psychol., 4St 17-52.
FREEMAN, c. L. (lgf8) 'The optional muscular tensions for various
performances', Amer. 1. Psychol., 5t, t46-r5o.
FREEMAN, c. L. (t:,g+o) 'The relationship between performance level
and bodily activity level', j. exp. Psychol., 26, 6oz-6o8.
FRIEDMAN5 M.;., and REyNoLDS, f . H. (tg67)'Retroactive inhibition
as a function of response-class similarity'13. exp. Psychol.r74135r-355.
FURNEAUX, w. D. (t962)'The psychologist and the university', lJnia.
Quart., r7t 33-47.
GARNER, 'w. R. (t962) Uncertainty and structure as Psychological Concepts. New York: John ITiley & Sons.
GARNER, w. R.l and HAKEr H.'\ry'. (tgSr)'The amount of information in
absolute iudgements', Psychol. Rea., 58, 446-459.
cARvEy, w. D., and KNOWLES, w. B. (rgS+)'Response time patterns
associated with various display-control relationships', j. exp. Psychol.,
47t 315-322.

Refrmces

357

cARvEY, w. D., and MITNIcK tL. r.. (1955) 'Effect of additional spatial
references on display-control efficiency', j. exp. Psychol., 5or z7G
z8z.
GARvEY, w. D., urnd TAyLoR, F. v. (1959)'Interactions among operator
variables, system dlmamics and task-induced stress', j. appl. Psychol.,

43t 79-85.
cATES, A. r. (1917)'Recitation as a factor in memorizing', Arch. Psychol.
Af. Y., 6, No. 40.

GAUDRY, E.e and cHAMpIoN, R. A. (t962) 'The identification of
learning-curve parameters with intelligence and motivation', Austal.
J. Psychol., t4r 87-94.
vAN DE GEER, J. p., and LEvELT, w. f . M. (1963)'Detection of visual

patterns disturbed by noise: an exploratory study', Quart. J. exp.
Psychol., tSt r:,92-2o4.
crBBs, c. B. (tgSr)'Transfer of training and skill assumptions in tracking tasks', Quart. j. ex?. Psychol., 3t 99-r ro.

crBBS, c. B. (lgS+) 'The continuous regulation of skilled response by
kinaesthetic feed back', Brtt. J. Psychol., 4* 24-39.
GIBBS, c. B.e and BRowN, r. D. (rgS6)'fncreased production from
information incentives in an uninteresting repetitive task', Manager,
24t 374-379.

cIBsoN, J. f . (r95o) The Perception of the Visrul lVuld. Boston, Mass.:
Houghton Miffi,in.
GrBSoN, J. I., and RAFFEL, cERTRUDE, (1936)'A technique for investigating reuoactive and other inhibitory effects in immediate recall',
1. gen. Psychol,, t1t to7-tt6.
cLICKMAN, s. E. (196r) 'Perseverative neural processes and consolidation of the memory trace', Psychol. BulI., 58, zt}-233.
GLUCKSBERG, s., KARSH, R., and MoNTy, R. A. (tg6il 'sequential
memory: keeping track performance as a function of irrformation
exposure time and interstimulus noise', Percep. mot. Skitls, z+ 65r6s6.
GOLDMAN-EISLERT FRTEDA (lgS6) 'The determinants of the rate of
speech output and their mutual relations', J. Psychosomatic Res., 2,

r37-r43.

GOLDMAN-ErSLER, FRTEDA (tgS8) 'Speech production and the predictability of words in context', Quart. J. exp. psychol., ro, g6-to6.
GOLDMAN-EISLER, FRIEDA (tg6t) 'Hesitation and inforrrration in
speech', iD Information Theory (ed. C. Cherry). London: Buttenrorth,

pp. t6z-r74.
GoLDMAN, JosEpHrNE (r9rz) Fatigue and Efici*tcy: A Study in
Industy. New York: Suryey Associated fnc.
GoLDSTEINe D. A., and NEwroN, f . M. (t962)'Transfer of uaining as
a function of task difficrrlty in a complex control sinratioD', J. exp.
Psychol., 63, 37o-3T5.

358

Refrmces

coLDsrEIN, M.1 and RITTENHoUsE: c. H. (rgS+) 'Knowledge of results
in the acquisition and transfer of a gunnery skill', j. etcp, Psychol., 48,
t87-t96.
corrsDANKER, R.1 BRoADBENTe L., ond vAN sANTr c. (1963) 'Reaction time to single and to first signals', J. exp. Psychol.r 66, 163t67.

cour.D, f ., and sMITH, K. Ir. (rg6g) 'Angular displacement of visual
feedback in motion and learning', Percep. mot. Skills, t7t 6gg7lo.
cour.D, I. D . (tg6S)'Differential visual feedback of component motions',
1. exp. Psychol., 69, 263-268.
cour,Dr I. D., and scHAFFER, AMy (rg6S)'Partial visual feedback of
component motions as a function of difficulty of motor control', j. exp.
Psychol., 7ot 564-569.
cRANDJEAN, 8., and rERRET, E. (196r) 'Effects of pupil aperture and
of the tirne of expostue on the fatigue induced variations of the flicker
fusion frequency', Ergonwtics, 4, 17-23.
cRAy, f. A., and wEDDERBURN, A. A. r. (tg6o)'Grouping suategies
with simultaneous stimuli', Quart. j. exp. Psychol., \2t r8o-r84.
cREEN: D. M., and BTRDsALL, T. G. (rg6+)'The effects of vocabulary
size on articulation score', in Signal Detection and Recogvition by Hwnan
Obsemters (ed. f. A. Swets). New York: ITiley.
tsignal
GREEN, D. M., BIRDSALL, T. G., Snd TANNER, W. P. (tgSil
detection as a function of signal intensity and duration', I. Acoust.

Soc. Amer., 29t 523-53r.
cREENe D. M., and swETs, f. A. (t966) Signal Detection Theory and

Psychophysics. New York: Wiley.
cREENSpooN, f ., and FoREMAN, sALLy (lgS6) 'Effect of delay of
knowledge of results on learning a motor task', j. exp. Psychol., 5rr

zz6-228.

cREcoRyl R. r. (1956) 'An experimental treatment of vision as an information source and noisy channel', in Information Theory (ed. C.
Cherry). London: Butterworth & Co.
cRrEWr s. (r958a)'Age changes and information loss in performance of
a pursuit tracking task involving internrpted preview', j. exP. Psychol.,

SS, 486-489.
cRrEWr s. (r958b) 'Information gain in tasks involving different stimulusresponse relationships', fVafure, r.8z, 18t9.

cRrE\M, s. (lgS8c) 'LJncertainty as a determinant of performance in
relation to age; Gerontologia, z, 284-289.
cRrEw, s. (rgSg) 'Set to respond and the effect of intermpting signals
upon tracking performance'r 1. exp. Psychol., 57t 333-337.
cRrEwe S. (rg6+) 'Age, information transmission and the positional
relationship between signals and responses in the performance of a
choice task', Ergonomics, 7, 267-277.

Refnmces

359

cRIEW, s., and TUcKERT w. A. (rgS8)'The identification of job activities associated with age differences in the engineering industry',
J. appl. Psychol., 42,278-282.
cRrMr p. F.e and wHrrEr s. rr. (1965)'Effects of stimulus change upon
the GSR and reaction time', J. exp. Psychol., 6gt, z7Gz8t.
cRUBER, H. E.e and DTNNERSTETNT A. f . (1965)'Theroleof knowledge
in distance-perception', Amer. J. Psychol., 78, 575-58r.
cULLAHoRN, f . T., and cULLAHoRN, f . E. (1963)'A computer model

of elementary social behavior', in Com?uters cnd Thought (ed.
E. Feigenbaurr and ]. Feldman). New York: McGraw-Hill.

HABERT R. N. (1965)'Effect of prior knowledge of the stimulus on word-

recognition processes', I. exp. Psychol., G9, z8z-286.
HABER, R. N.r and HERsHENsoN, M. (rg6S) 'Effects of repeated brief
exposures on the gfowth of a percept', j. ex|. Psychol.r 691 4b46.
HATDER, M.e and DrxoN, N. F. (196r) 'Influences of training and
fatigue on the continuous recording of a visual differential threshold',

Brit. J. Psychol., 52, 227-237.
HALEr D.I. (1967)'Sequential effects in a two-choice serial reaction task',
Quart. J. exp. Psychol., t9r r33-r4r.
HALLIDAv, A. M.r KERR, M.e and ELrrHoRNe A. (tg6o)'Grouping of
stimuli and apparent exceptions to the psychological refractory period',
Quart. j. exp. Psychol., t2, 72-89.
HAMMERToN, M. (tg6z)'An investigation into the optimal gain of a
velocity control system', Ergonotttics, 5, 53y543.
HAMMERToN, M.1 and TIcKNER, A. H. (1960 'An investigation into the
comparative suitability of forearm, hand and thumb conuols in acquisi-

tion tasks', Ergonotnics, 9, rz1-t3o.
HARcouRTr R. A. F. (1959) Personal communication.
HARcuM, E. R., and FRTEDMANT s. M. (1963) 'Reproduction of binary
visual patterns having different element-presentation sequences',

j. exp. Psychol., 66, 3oo-3o7.

HARDEsry, D., TRUMBo, D., and BEvAN, w. (t963) 'Influence of
knowledge of results on performance in a monitoring task', Percep.
mat. Shills, r6, 629434.

HARRrs, c. s.e THACKRAY; R. r., and SCHoENBERGER, R. w. (1966)
'Blink rate as a function of induced muscrrlar tension and marrifest
anxiety', Percep. mot. Skills, zz, 155-16o.
HARRIsoN, c. f . (t:.967) 'Some additive results in short term memory',
in Attmtion and Performance (ed. A. F. Sanders). Amsterdam: NorthHolland Publishing Co.
HARTLTNE, n. K. (rgf8) 'The response of single optic nerve fibres of
the vertebrate eye to iUumination of the retina', Amer. J. Physiol.,
tzr., 4oo-4r5.
HAUTye G. T., TRITEST D. K., and BERKLEy, 'w'. f . (tq6S) Biomedical
sun)ey of ATC faciltttes: z. Expeience arid age. FederalAviationAgenCy,

360

References

Office of Aviation Medicine, Civil Aeromedical Res. Inst.r Oldahoma
City. Rep. No. AM 65-6.
HAywooD, H. c.: and SpIELBERGER, c. D. (1966) tPalmar sweating

as a function of individual differences in manifest anxiety', J. Pers.
Soc. Psychol., 3r ro3-r'o5.
HEBB, D. o. (tg+g) The Organization of Behaztior. New York: John
Wiley & Sons. London: Chapman & Hall.
HEBB, D. o. (lqSS) 'Drives and the C.N.S. (Conceptual Nervous
System)', Psychol. Reo, 62, 243-254.
HEBRON, MTRTAM r. (1966) Motiztated Learn:ing: a Deaelopmental Study
from Birth to the Seniu,m. London: Methuen & Co. Ltd.
HELMSTADTER, G. c., and ELLrs, D. s. (tgSz)'Rate of manipulative
learning as a function of goal-setting techniqu€s'r 7. exp. Psychol., 4lt

t25-r29.

HELSoN5 H. (tg+il'Adaptation-Level as frame of reference for prediction of psychophysical data', Amer. J, Psychol., 6o, t-29.
HELsoNe H., &nd srEGER, I. A. (t962)'On the inhibitory effects of a

second stimulus following the primary stimulus to r@ct', ,. exp.

Psychol., 64, zor-2o5.

HEMrNcwAy, A. (1953) 'The Physiological background of fatigue', ir
Symposiwn on Fatigu.e (ed. S7. F. Floyd and A. T. S7elford). London:
H. K. Lewis & Co., for the Ergonomics Research Society.
HENDERSoN, A.1 GoLDMaN-ErsLERr FRTEDA, and SKARBEK, A.
(rg6S) 'The common value of pausing time in spontan@us speech',
Quart, 1. exp, Psychol., r7t 343-345.
HENDERsoN, f . G. (fgSg) 'The estimation of the transfer function of a
human operator by a correlation method of analysis', Ergonontics, 2,
274-286.
HENMoN, v. A. c. (rgo6) 'The time of perception as a measure of
differences in sensation', Arch. Philos. Psychol. Sci. Meth., No. 8.
HENRv, F. M. (igSz) 'Independence of reaction and movement times
and equivalence of sensory motivators of faster response', Res. Qttart,,
23t 43-53.
HENRv, F. M. (rg6t) 'Reaction time-movement time correlations',
Percep. moL Skills, r.z, 6346.
HENRv, F. M., and RocERS, D. E. (196o) 'Increased response latenry for
complicated movements and a "memory drum" theory of neuromotor
reaction', Res. Quart., ll, M8-458.
HENsHAw, E. M., and HoLMAN, p, (1933)'Manual dexterity-effects of
traininB II, distribution of practice in manual dexterity', Indust. Hhh.
Res. Bd. ReD. No. 67. London: ff.M.S.O.
HERoN, e. (1962) 'Immediate memory in dialling performance with and
without simple rehearsal', Quart. J. etcp. Psychol., r.4t 94-r'o3.
HrcK, \trr. s. (t948) 'The discontinuous functioning of the human operator in pur'suit tasks', Quart. j. exp. Psychol., rr 36-5r.

Refumces

36r

HrcKr w. E. (tg+g)'Reaction time for the amendment of a response',
Quart. j. exp. Psychol., t, t75-t79.
HrcK: w. E. (r952a) 'On the rate of gain of information', Quart. 1. exp.
Psychol., 4, tt-26.

HICK, w. E. (tgSzb) '!fhy the human operator?', Trans. Soc. Instrtnn.
Technol., 4, 67-77.

HrcKr'w'. E.r and wELFoRDe A. T. (lgS6)'C,omment on the paper
"Cenual intribition: some refractory observations" by Alick Elithorn
and Catherine Lawrence', Quart. j. exp. Psyclwl., 8, 3g-4r.
HILGENDoRF, LTNDEN (t966) 'Information input and response time',
Ergonortics, g, 3r-37.

HrLLrx, w. A. (t96o) 'Visual pattern identification as a function of fill
and distortion', 1. exp. Psychol., S9t tgz-rg7.
HrLLrxr'fil'. A., and MARX, M. fr. (t96o)'Response strengthening by
information and effect in human learning', j. exp. Psychol., 6o,, gTro2.
HIRSH, I. f . (tgSo) 'The relation between localization and inte[igibility',

J. Acoust. .Soc. Amfr., 22, tgGzoo.
HocHBERG, f ., and BRooKS, vrRcrNra (196o) 'The psychophysics of
form: reversible-perspective drawings of spatial obiects', Amer. I.
Psychol.r T3t 337-354.
HocHBERG, f . E.r and McaLrsrER, E. (rgSS) 'Relative size versus
familiar size in the perception of represented depth', Amer. j. Psychol.,
68,294-296.
HocHBERG, f ., and McALISTERe E. (rgSg) 'A quantitative approach to
figural "goodness" ', j. exp. Psychol., 46, 36t-364.
HoDGE, M. H., and poLLACK, r. (1962)'Confusion matrix analysis of
single and multidimensional auditory displays', 7. exp. Psycltol,, 63,
t29-142.
HocAN, f . A. (196r) 'Copying redundant messages'r}. etcp. Psyclnl.r 62,
r53-r57.

HocAN, M. J. (t966) 'Influence of motivation on reactive inhibition in
extraversion-introversion', Percep. mot. Skills, 22t t87-t92.
HoHLE, R. rI. (tg6S)'Detection of a visual signal with low backgrotrnd
noise: an experimental comparison of two theories', I. exp. Psychol.,
7ot 459-463HoLDTNG, D. H. (1962) 'Transfer berween difficult and easy tasks', Brit.

J. Psychol., Slt 397-4o7.

HoLDrNce D. H., and MAcRAE, A. \trr. (rg6+)'Guidance, restriction and
knowledge of results', Ergonomics, 7t 289-295.
HoLDTNG, D. H., and MAcRAE, A. vrr. (1966)'Rate and force of guidance
in perceptual-motor tasks with reversed or random spatial correspondence', Ergonwtics, 9, z8yz96.

HowARDr K. r., and DIESENHAUs, H. r (rg6S) nPersonality[correlates
of change-seeking behaviot', Percep. mot. S&il/s, zt, 655464.

362

References

HowARD, T. c. (1963) 'The relation between psychological and mathematical probability', Amer. j. Psychol., 76 335.
HowARTHe c. r.e and ELLrs, K. (1961) 'The relative intelligibility
threshold for one's own name compared with other names', Quart. j.
exp. Psychol., rlt 236-239.

HowARTH, c. r., and TREIsMAN, M. (196r) 'Lowering of an auditory
threshold by a near threshold warning signal', Quart. J. exp. Psychol.,

t3, r2-18.

HowE, M. f . A. (lg6S) 'Intra-list differences in short-term memory',
Quart. J. exp. Psychol., r7t 338-342.

HowE, M. J. A. (1966)'A note on order of recall in short-term memory',

Bit. J. Psychol., S% 435-436.

HowELL, w. c., and KRETDLER, D. L. (1963) 'Information processing
under contradictory instructional sets', .7. etcp. Psychol., 65, 39-46.
HowLANDe D., and NoBLE, M. E. (1953) 'The effect of physical coDstants of a control on tracking performance', 1. exp. Psychol.r 461 35336o.

HUBELI D. fI. and wIESELe T. N. (tg6z) 'Receptive fields, binocular
interaction and functional architecture in the cat's visual cortex',
1. Physiol., 16o, ro6-154.
HUBEL, D. H. and wrESEL, T. N. (1968) 'Receptive fields and functional
architecrure of the monkey striate cortex'r1. PhystoL, r95r 2rS-243.
HucHEs, r. M. (lg6f) Srudies in midimensional Discimination Ph.D.
Dissertation at Cambridge University Library.
HULBERT, s. F. (t957) 'Drivers' GSRs in traffic', Percep. mot. Skillsr T,
305-3r5.
HULTCKA, rRENE M.1 srERNs, rr.r and GRossMAN, f . Gg6il 'Agegroup comparisons of paired-associate learning as a function of paced
and sef-paced association and response times', ]. Gerontol., 22, 274280.

HULL, c. L. (r9 43) Principles of Behauior. New York: Appleton-Cenrury.
HUNT, D. P. (tg6+) 'Effects of nonlinear and discrete transformations
of feedback information on human tracking performance', j. exp.
Psychol.r 67, 48G494.
HuRwrrzl L.;., and ALLrsoNe R. S. (lg6S)'Factors influencing perforrnance in psychological testing of the aged', rn Behaaior, Aying and
the Nerz)ous Syst*n (ed. A. T. STelford and J. E. Birren). Springfield,
Illinois: Charles C. Thomas.
HYMANT R. (tgSf) 'stimulus information as a determinant of reaction
tirne', j. exp. Psychol., 45, 188-196.
rNGLrs, f . (tg6S) 'Immediate memory, age and brain function', in
Belruvior, Aging and the Nerl)ous System (ed. A. T. Velford and J. E.
Birren). Springfield, Illinois: Charles C. Thomas.
rNGLrsr I., and ANKUs, MARv N. (1965) 'Effects of age on short-term
storage and serial rote learning', Brit. j. Psychol., 56, r83-r95.

Refwmces

363

rNGLIs, f., and cAIRD,'w. K. (rg6f)'Age differences in successive
responses to simultaneous stimulatioD', Canad. J. Psychol., t% g8r05.

rRroN, A. L. (t966) 'A brief history of research on the acquisition of
skill', in Acquisition of Skill,(ed. Edward A. Bilodeau). London and
New York: Academic Press.
IACoBS, J. (r882) 'Experiments in prehension', Mind, r.2t 75-79.
JACoBSoN, E. (tgzg) Progressioe Relaxation. University of Chicago
Press.

IAeuES, E. (tgS6) Measurement of Responsibility. London: Tavistock
Publications.
IARRARD, L. E. (tg6o) 'The role of visual cues in the performance of
ergographic work', 1. exp. Psychol.r 6o,, 57-63.
JEEvEs, M. A. (tgSZ) Unpublished Ph.D. Thesis. LJniversity of Cambridge.
JEEvES, M. A. (t96t)'Changes in performance at a serial-reaction task
under conditions of advance and delay of information', Ergonomics, 4,
329-338.
JENKTNS, w. L.e and coNNoR, MrNNA s. (1949) 'Some designfactors
in making settings on a linear scale', j, a?pl. Psychol., 3* 395-409.
JENKTNsT'w. L., and oLsoN, M. w. (tg1z) 'The use of levers in making
settings on a linear scale', J. appl. Psychol.r 36,269-27t.
IENKTNS, w. o. (tg+il'The discrimination and reproduction of motor
adiustments with various types of aircraft controls', Amer. J. Psychol.,
6o, 397-406.
IERoME, E. A. (tg6z) In Aging aromd the Vorld: Proc. Sth Cong.
Internat. Ass. Gerontol., San Francisco, 196o, Vol. r (ed. C. Tibbitts
and W'. Donahue). New York: Columbia Univ. Press.r pp. 8o8-823.
JoHN, r. D. (t964) 'The role of extraneous stimuli in responsiveness to
signals: refractoriness or facilitation?', Austral. J. Psychol., r.6, 87a6.

JoHN, r. D. (1966)'Intensity of non-key stimuli in the Donders'c-t]rye
reactiort', Austral. J. Psychol., 18, r48-r53.
JoHN, r. D. (r96la)'A statistical decision theory of simple reaction time',
Austral.|t'. Psychol., t9t 27-34.
JoHN, r. D. (tg6lb)'sequential effects of stimulus characteristics in a
serial reaction time task', Austral. 1. Psychol., r9t 35-40.
JonNsroNr'w. A.e HowELL, \u'. c.1 and GoLDSTEIN, r. L. (1g66)
'fluman vigilance as a function of signal frequency and stimulus
density', j. exp. Psychol.r T2t 736-743.

IoHNSToN, w. A., HowELL, \ry'. c., and zLlKowsKr, M. M. (tg6l)
'Regulation of attention to complex displays', j. etcp. Psychol., 73t
48r-482.
'WILKINsoN, H. JEAN, and BRADEN, Iue (196r) 'InfonnaJoNEs, A.e
tion deprivation as a motivational variable'r 1. exp. Psychol., 62, tz6t37.

16+

Refermces

IoNG, I. R. DE (lgSZ) 'The effects of increasing skill on cycle time and its
consequences for time standards', Ergonomics, t, 5l-6o.
KALTL, R. E.e and FREEDMAN, s. f. (tg66a)'fntermanual transfer of
compensation for displaced vision, Percep. mot. Skills, 22, r.23-126.
KALTL, R. 8.1 and FREEDMAN, S. J.(t966b)'Persistence of ocular rotation following compensation for displaced vision', Percep. mot. Skills,
22, r35-r39.
KALSBEEK, f . 'w. rr. (tg6+) 'On the measurement of deterioration in
performance caused by distraction stress', Ergonomics, 7, t87-t95.
KALSBEEK, f . w. n. (lg6S)'Mesure obiective de la surcharge mentale:
nouvelles applications de la m6thode des doubles tAches', Traaail

Hwnain, r-2t tz2-r32.

KALSBEEK, f . 'w. rr.e and ETTEMA, f . H. (tg6+) 'Physiological and

psychological evaluation of distrastion stress', Proc. znd Internat.
Cong. Ergonomics, Dortmund. Suppl. to Ergonortics, pp. 443-447.
KALSBEEK, f . w. Ir., and syKEs, R. N. (1g6il'Objective measurement
of mental load', in Attmtion and Performance (ed. A. F. Sanders).
Amsterdam: North-Holland Publishing Co.
KArLAN, rRA T., and cARvELLAST T. (rg6S) 'Scanning for multiple
targets', Percep. mot. Skills, 2t, 23y243.

KArLAN, r. T.e CARVELLAS, T., and METLAyT'w. (t966)'Visual
search and immediate memory', J. exp. Psychol., 7tr 488-493.
KAPPAUF, '\[/'. E.e and powE, 'w. E. (tgSg) 'Performance dectement

at an audio-visual checking task', j. exp. Psychol., 57, 49-56.
KARLTN, L. (1959)'Reaction time as a function of foreperiod duration
and variabiliry', j. exp. Psychol., 58, r85-r9r.
KARLTN, L. (tg6S) 'Effects of delay and mode of presentation of extra
cues on pursuit-rotor performanc€', J. exp. Psychol., 7o, 438-440.
KARLTN, L. (t966) 'Development of readiness to respond during short
foreperiods', J. exp. Psychol., 72t 5o5-5o9.
KAssuM, D. A. (tg6il Unpublished report. Psychological Laboratory,
IJniversity of Cambridge.
KAswAN, f ., and youNc, s. (rg6Sa) 'Effect of luminance, exposure
duration, and task complexity on reaction time', j. exp. Psychol., 69,
393-40,0.

KAswAN, f., and youNG, s. (t965b) 'Effect of stimulus variables on
choice reaction times and thresholds', J. exp. Psychol.r 6g1 5rr-5r4.
KATaI D. (lgfS) The World of Colour. London: Kegan Paul, Trench

Trubner & Co. Ltd.
I<ATZ, D. (tgqg)'Gestalt laws of mental work', Brit.1. Psychol., 39r t75r 83.

KATZ, r-. (t964)'Effects of differential monetary gain and loss on sequential two-choice behaviour', j. exp. Psychol.r 6S) 245-249.
KATa' M. s. (tg6il'Feedback and accuracy of target positioning in a
homogeneous visual field', Amer. J. Psychol.r 8o, 4o5-4to.

References

365

KAUFMAN, H., and LEvy, R. M. (1966) 'A fufther test of Hick's law
with unequally likely alternatives', Percep. mot. Skills, 22, 967-970.
KAY, H. (lgSt) 'Learning of a serial task by different age groups',
Quart. J. exp. Psychol., 3, 166-183.

KAY, H. (1953) Experimental studies of adult leaming. Lfnpublished Ph.D.
thesis - Cambridge Llniversity Library.
KAY, u. (1954) 'The effects of position in a display upon problem solving',
Quart. J. exp. Psychol., 6, 155-169.
KAYI H. (tgSS)'Some experiments on adult learning', in Old Age in the
Modern World: Rep. 3rd Cong. fnternat. Ass. Gerontol., London
tg14. Edinburgh: Livingstone, pp. 259-267.

KAY, fr., and poul-ToNe E. c. (tgSr) 'Anticipation in memorbhg',
Brit. J. Psychol., 42r 34-4r.

KAy, rl.e and wErss, A. D. (tg6r)'Relationship berween simple and
serial reaction times', JVarure, rgrr 7gv7gt.
KELLER, F. s. (tg+)'Studies in international Morse code: f. A new
method of teaching code receptiorr', j. appl. Psychol., 27t 4o7-4r5.

KENDoN, A. (tg6il'Some functions of gaze-direction in social interaction', Acta Psychol., 26, 2243.
KERRI M., MrNGAy, RosEMARy, and ELrrHoRN, e. (1963) 'Cerebral
dominance in reaction time responses', Brit. I. Psychol., S4t 325-336.
KERRe M.r MINGAY, ROSEMARv, and ELTTHORN, e. (1965) 'Patterns
of reaction time responses', Brit. 1. Psychol., SQ 53-59.

KrMorsuKr, K. (tg6il 'On the measurement of the speed of finger
motion',1. Scr. Lab., 43,244-25t.

KrNG, PEARL H. M. (tg+8) 'Task perception and interpersonal relations
in industrial training: II', Human Relations, !, 373-412.
KIRCHNER, w. K. (tgS8) 'Age differences in short-terrn retention of
rapidly changing infornration', J. exp. Psychol., SS, 3SziS8.
KrRKl R. E., and HEcHT, ELTzABETH (rg6f) 'Maintenance of vigilance
by programmed noise', Percep. mot. Skills, 161 553-56o.
KLEINsMTTH, L. f ., and KArLAN, s. (1963). 'Paired-associate learning
as a function of arousal and interpolated interval', j. exp. Psychol.r 65,

r9o-r93.
KLEINsMTTHT L. J., and KArLAN, s. (1964) 'Interaction of arousal and
recall interval in nonsense syllable paired-associate learnin9', 1. exp.
Psychol., 67, tz4-t26.

KLEMMER, E. T. (tgS6) 'Time uncertainty in simple reaction time',

j. exp. Psychol., Sr't t79-t84.
KLEMMER, E. T. (t957)'Simple reaction time as a function of time uDcertainty', j. exp. Psychol.r 54, rg1-zoo.
KLEMMER, E. T. (1961) 'The perception of all patterns produced by a
seven-line matrix' , J. exp. Psychol., 6t, 274, 282.
KLEMMER, E. r. (1964) 'Does recoding from binary to octal improve the
perception of binary patterns?', j. exp. Psychol., 67, tg-2r.

366

References

KLEMMER, E. T., and FRICK, F. c. (1953) 'Assimilation of information
from dot and matrix patterns', J. exp. Psychol., 45, r5-r9.
KLING, J. w., and scHLosBERG, H. (196r)'The uniqueness of patterns
of skin-conductance', Amer. j. Psychol., 74t 74-79.
KNAPP, BARBARA (tg6g) Skill in Sport: the Attainment of Profi,ciency.
London: Routledge & Kegan Paul.

KNIGHTe A. A. (rg6il Unpublished Ph.D. Thesis. Llniversity of Birmingham.

KNTGHTe A. A., and DAGNALL, p. R. (tg6il'Precision in movements',
Ergonomics, tq 3zr-33o.
KNoWLES, w. B.l GARvEy, w. D.e and NEwLrN, E. p. (1953)'The effect
of speed and load on display-control relationships', 7. exp. Psychol.,

46, 65-75.

KopLrN, J. H., Fox, R., and DozrERl F. (1966)'A failure to replicate
the inhibitory effects of a second stimulus following the primary
stimulus to react', J. exP, Psychol., 72, 9t4-9t6.
KoRN, J. H., and LINDLEv, R. u. (1963)'Imrnediate memory for consonants as a function of frequency of occurrence and frequency of
appearance', J. exp. Psychol., 66, I49-r54.
KoRNBLUM, S., and KosrERr'w. c. (1967) 'The effect of signal intensity
and training on simple reaction time', in Attmtion and Performance
(ed. A. F. Sanders). Amsterdam: North-Holland Publishing Co.

KosrER, w. G., snd BEKKER, f. A. M. (tg6il'Some experiments on
refractoriness', in Attmtion and Performance (ed. A. F. Sanders).
Amsterdam : North-Holland Publishing Co.

KRUEcER, w. c. F. (tglil'The influence of amount limits and time
limits upon the rate of work', j. appl. Psychol., 2t, rr3-rr8.
LABERGE, D., and TwEEDy, I. R. (tg6+) 'Presentation probability and
choice time', 1. exp. Psychol., 68, 477-48r.
LAcEy, f. r.e BATEMAN, DoRorHy E., and vAN LEHN, RUru (1953)
'Autonomic response specificity : an experimental study', Psychosornatic

Med., 15, No. r.
LACEv, f. r., and LAcEy, BEATRIcE c. (rqS8)'The relationship of
resting autonomic activity to motor impulsivity', in The Brain and
Human Behaaior. Res. Pub. Assn. Nerv. Dis. Baltimore: Williams &
Wilkins.

LAcHMAN, R., and TUTTLE, ABrcArL, v. (rg6S) 'Approximations to
English (AE) and short-term memory: construction or storage?',

J. exP. Psychol., 7or 386-393.

LAMB, f ., and KAUFMAN, H. (tg6S) 'Information transmission with
nnequally likely alternatives', Percep. mot. Skills, 2r, 2SS-259.

LAMBERT, w. E., and lAKoBovrrs, L. A. (196o)'Verbal satiation
and changes in the intensity of meaning', j. exp. Psychol., 6o, 376383.

LAMING, D. R . y, (t962) 'A statistical test of a prediction from informa-

kfermces

$l

tion theory in a card-soning situation', Qunrt. J. exp. Psychol., t4,
38-48.
LAMTNG, D. R. J. (t966) 'A new interpretation of the relation between
choice-reaction tirne and the number of equiprobable alternatives',
Bit. J. math. stattst. Psychol., r9, t3g-r4g.
LASzLo, JUDTTH r. (1g66) 'The performance of a simple motor task
with kinaesthetic sense loss', Quart. j. exp. Psychol., r8, r-8.
LAUERe A. R., and suHRe v.'w. (rgS8) Road adaptation of a laboratory
technique for studying driaing fficiency with and aithout a refreshment
pantse. Chicago, Illinois: National Safety Council.

LAvERY, f . f . (lg6+) 'The effect of one-trial delay in knowledge of
results on the acquisition and retention of a tossing skill', Amer. J.
Psychol.r TTr 437-U3.

LAVERY, J. J., and suDDoN, FLoRENcE H. (t962) 'Retention of
simple motor skills as a function of the number of trials by which KR is
delayed', Percep. mot. S&i//s, r* 23t-237.
LAwRENcEe D. fr.e 8r1d LABERGE, D. L. (tgS6) 'Relationship between
recognition accuracy and order of reporting stimulus dimensions',

j. exp. Psychol., Stt r2-r8.

LAwsoN, EVERDTNA A. (196t) 'A note on the influence of different
orders of approximation to the English language upon eye-voi,ce span',
Qtmrt. J. exp. Psychol., 11 53-55.
LAwsoN, EVERDTNA a. (1966) 'Decisions concerning the reiected ctrannel', Quart. j. exp. Psychol., 18, z6o-265.
LEEI L. cHARLorrE (lg6t)'The effects of auiety level and shock on a
paired-associate verbal task', J. exp. Psychol,, 6t, 2t3-2t1.
LEONARD, J. A. (tgSf) 'Advance information in sensori-motor skills',
Quart. J. ex?. Psychol., S, t4r-r49.
LEoNARD, f . A. (1958) 'Partial advance information in a choice reaction
task', Brtt. j. Psychol., 49, 89a6.

LEoNARD, f . A. (lgSg) 'Tactual choice reactions: I', Quart. J. exp.
Psychol., ltr 7G83.
LEoNARD, f . A., NEWMAN, R. c., and cARpENTERe A. (1966)'On the
handling of heavy bias in a self-paced task', Quart. J. exp. Psychol., r8,

r3o-r4r.

LESHNER, s. L. (196r) 'Effects of aspiration and achievement on muscular

tensions'rI. etcp. Psychol., 6t, t33-t37.
LEUBA, H. n. (1967) Measuing the size of an ergonomics probletn. Paper
to 3rd. fnternat. Cong. Ergonomics, Birmingham.
LEvrN, r. P . (t967) 'Induced muscle tension and response strift in pairedassociate learning', j. exp. Psychol., 71, 4zz-426.
LEVYI P., and LANG, P. f. (1966)'Activation, control, and the spiral
after movement', J. Pers. soc. Psychol.r l, ro5-rr2.
LEwrs, M. (lg6S) 'Psychological effect of effort', Psychol. BuIl., 64,
r 83-r9o.

368

Refermces

LEwI s, R. E. F . (tg14) 'Consistenry and car-driving skill', Brtt.J. industr.
Med,, r3r r3r-r4r.
LICKLIDER, f . c. R. (tg+8) 'The influence of interaural phase relations
upon the masking of speech by white noise', J. Acoust. Soc. Amer.' 2o1
r 5o-r 59.

LTDDELL, H. S. (tgU) 'Conditioned reflex method and experimental
neurosis', in Persanality and Behaaior Disorders, Vol. II (ed. t. Mc V.
Hunt). New York: Ronald Press.
LINcoLN, R. s. (t956)'Learning and retaining a rate of movement with
the aid of kinesthetic and verbal cues', -7. exp. Psychol., 5rr rgg2o.4.

LTNDLEY, R. r{. (lg63) 'Association value, familiarity and pronounciability ratings as predictors of serial verbal learning', J. exp. Psychol.
6sr 34T-35r.
LTNDSLEY, D. B. (r95r) 'Emotion', in Handbookof Experimental Psychology (ed.S. S. Stevens). New York: John tUfiley & Sons. London:
Chapman & Hall, pp. 473-516.
LIppoLDl o. c. J., REDFEARN, f.'w'. T.e and vuco, J. (tg6o)'The
electromyography of fatigue', Ergonomics, 3, r:2r.-r3r.
LrPsrrr, L. p.e and ENGEN, T. (196r) 'Effects of presentation of paired
and single-stimulus on discrimination of length', Amer. J. Psychol.,

74t 274-277 LLOYD, D. P. c. (1942)'Stimulation of peripheral nerve terminations by
active muscle' , J, Neurophysiol., 5r 153-165.

LLoyD, K. E., REID, L. s.e and FEALLoCK, J. B. (t96o)'Short-term
retention as a function of the average number of items presented,

j. exp. Psychol., 6o,, 2ot-2o7.

LocKEe E. A. (tg6il 'Motivational effects of knowledge of results:
knowledge or goal setting?', j. appl. Psychol., Sr, 324-329.

LocKE, E. A.e and BRvAN, JUDITH, F. (tg6i'Performance goals as
determinants of level of performance and boredom', J. appl. Psychol.,

5rr I2O-I3O.

LoEBe M., and BTNFoRD, J. R. (t964) 'Vigilance for auditory intensity
changes as a function of preliminary feedback and confidence level',
Human Factors, 6, 445-458.
LoEB, M., and scHMrDT, E. A. (rg6f) 'A comparison of the effects of
different kinds of information in maintaining efficiency on an auditory
monitoring task', Ergonontics, 6, 75-8r.
LoEHLIN, I. c. (tg6S) ' "Interpersonal" experiments with a computer
model of personality', j. Pers. soc. Psychol., 2, 58o-584.

LoNG, E. R.e HENNEMAN, R. H.1 and GARvEyr'w. D. (t96o) tAn
experimental analysis of set: The role of sense-moddity', Amer. J.

Psychol., 7lt 563-567.
LoRGE, r. (r93o) 'The irffuence of regularly interpolated time intervals
upon subsequent leaffiirig', Teach. Coll. Contr. Educ. No. 438.

Refermces

16g

LoRGE, r., and THoRNDIKE, E. L. (rgfS) 'The influence of delay in
the after-effect of a connection', j. ex?. Psychol., r.8, 186-194.
LovAAs, o. r. (rg6oa) 'The relationship of induced muscle tension,
tension level and manifest anxiety in learning', J. exp. Psychol., S9t

r45-r52.
LovAAS, o. I. (tg6ob) 'Supplementary report: the relationship of induced muscle tension to manifest anxiety in learning'rJ. exp. Psychol.,
S9r 205-2e,6.

LovELESS, N. E. (tg6z) 'Direction-of-motion stereotypes: a review',
Ergonomics, 5, 357i83.

LUNDERvoLD, ARNE (tgS8)'Electromyographic investigations during

typewrithg', Ergonomics, t, zz6-233.

McALLrsrER, DoRorHy E. (1953) 'The effects of various kinds of relevant verbal preuaining on subsequent motor performance', j. exp.
Psychol., 46, 329-336.

MccALLl R. f . (1963) 'fnvested self-expression: a principle of hurnan
motivation, Psychol. Rev., 7o, 289-303.

MccoRMAcK, p. D. (tg6z) 'A two-factor theory of vigilance', Brit. J.
Psychol., 53, 357363.
MccoRMAcK, p. D., BTNDING, F. R. s.1 and MCELHERAN, w. G.
(rg6f) 'Effects on reaction time of partial knowledge of results of performanc€', Percep. mot. Skills, t% 279-28r.

MccoRMAcK, p. D.e and McELHERAN, w. G. (rg6f) 'Follow-up of
effects on reaction time with partial knowledge of results', Percep, tt@t.
Skills, r7t 565-566.
Mccoy, 'w'. K. (tg6f) 'Test of the accuracy of Crossman's ConfusionFnnctior', Percep. mot. Skills, r7t 3o3-32o.
MAcE, c. A. (tgfS) Incenthtes: some experimmtal studies. Industrial
Health Research Board (Great Britain) Report No. 72.
McFARLAND, R. A.1 and MosELEy, A. L. (rgS+) Hwnan Factors in

Highuay Transport Safery. Boston: Harvard School of Public Health.

MccEocH, J. A.e and rRroN, A. L. (tgSz) The Psychology of Human
Learning. New York: Longmalts.
MccHIEe A., CHArMAN, f., and LAwsoN, f. s. (tg6S)'Changes in
immediate memory with Bge', Brit. 1. Psychol., 56, 6y75.
MccurcAN, F. f . (tgSg) 'The effect of precision, delay, and schedule
of knowledge of results on performance', 1. exp. Psychol., 58, 7984.

McculGAN, F. f ., and MACcAsLIN, EUcENE, F. (tgSS) 'I[hole and
part methods in learning a percepnral motor skill', Arner. J. Psychol.,
58, 65846r.
MAcKAy, D. M. (1965) 'Visual noise as a tool of research', J. gen.Psychol.,
72t r8t-t97.

MACKwoRTH, IANE F. (1959) 'Paced memorizing in a continuous task',

j. ex7, Psychol., 58, zo6-ztr.

37o

Refrences

MACKwoRTH, f . F. (tg6+a) 'Interference and decay in very shortterm memory'r 1. uerb. Learn. vterb. Beh., 31 3oo-3o8.

MACKy6RTH, f . F. (rg6+b) 'The effect of true and false knowledge of
results on the detectability of signals in a vigilance task', Canad. J.
Psychol., r8, lo6-l t7.

MACKrygRTH, f . F. (1954c) 'Performance decrement in vigilance,

threshold and high-speed perceptual motor tasks', Canad. j. Psychol.,
18, zo9-223.

MAcKwoRTHT J. F. (1965) 'Decision interval and signal detectability in
a vigilance task', Canad. 1. Psychol., r9r rrr-tr7.
MAcKwoRTH, J. F.e and MACKwoRTH N. n. (tgS6)'Theoverlapping

of signals for decisiorls', Amer. i. Psychol-, 69, z6-47-

MAcKwoRTH, f. F.r and TAYLoRT M. M. (rg6f) 'The d'measure of

siggral detectability in vigilance-like situations', Canad. J. Psychol., t7,

302-325.

MAcKwoRTH, N. H. (rgSo) Researches ut the Measuretnutt of Human
Performance. Medical Research Corurcil Special Report Series No. 268.
London: H.M.S.O.
MACKWORTH, N. fI.e KAPLANe I. T.e and METLAYT'W. (tg6+)'Eye
movements during vigilance', Percep. mot. Skills, t\ 397-402.
MACKWORTH, N. H.e and MAcKwoRTH, JANE F. (tgSB)'Visual
search for successive decisiorlS', Brit, J. Psychol., 49, 2Io-22I.

MACKySRTH, N. H.e and MAcKwoRTH, f . F. (tgSg) 'Remembering
advance cues during searching', Brit. J, Psychol., Sot 207-222.
McNrcoL, D. (rg6il Unpublished Ph.D. Thesis. LIniversity of Cambridge.
MACpHERSoNI S. f ., DEES, vALERIE, and GRINDLEY, G. C. (tg+g)
.The effect of knowledge of results on learning and performance: III.

The influence of the time interval betrreen trials', Quart. J. exp.

Psychol., t, t67-t74.
MACRAE, A.'\ry'., snd HOLDING, D. u. (1965)'Guided practice in direct
and reversed serial tracking', Ergonomics, 8, 487-492.
MASRAE, A.'w.e and HoLDING, D. H. (1966) 'Transfer of training
after gurdance or practice', Quart. J. exp. Psychol., I.8, 327-333.

McREyNoLDS, p.r and AcKER, MARY (tgSg) 'Serial learning under
conditions of rapid presentation of stimuli', Amer. J. Psychol., 72t
589-592.
MgRUER, D. T., and KRENDELe E. S. (1959)'The human operator aS a
servo system element. Parts I and II', J. Franklin Inst.r 2671 38r-4o3

and 5rr-536.

MALMS, R. B. (lgSg) 'Activation: a neuropsychological dimension',
Psychol. Reo., 66, 367-386.
MARGRATN, susAN A. (tg6il 'Short-term memory as a function of
input modalitY', Quart. i. exy. Psychol,, r9t ro9-r14.

References

3Tr

MARILL, T. (1957) 'The psychological refractory phase', Brit.J. Psychol.,
48, 93-97 MARKS, M. R., and IAcK, ol,l,tr (1952) 'Verbal context and memory
span for meaningful material', Amer. J. Psychol., 65, 298-3oo.

MARTIN, p. R.5 and FERNBERGER, s. 'w. (tgzg) 'Improvement in

memory span', Amer. 1. Psychol., 4\ 9rn4.
MARTZ, R. L. (t966) 'Signal presentation rate, auditory threshold and
group vigilance', Percep. mot, Skills, 23, 463-469.
MARTZ, R. L. (tg6il'Auditory vigilance as affected by signal rate and
intersigRal interval variability', Percep. mot. Skills, 24, tg1-2o3.
MELToN, A. w., and IRwINT I. Mce. (r94o)'The influence of degree of
interpolated learning on retroactive inhibition and the overr transfer
of specific responses', Arner. J. Psychol., 53,, r73-2o3.
MERKELT f . (t88S) 'Die zeitlichen verhilurisse der sillensthiitig-keit',
Philos. Stud., 2, 73-t27.
MERToN, P. A. (1956) 'Problems of muscular fatigue', Brit. med. Bull.,

r2r 219-221.

MICHAELS, R. M. (196o) oTension responses of drivers generated on
trrban streets', Bull. Higw. Res. Bd. lYash., No. 27tr 29-44,
MIcHoN, f . A. (tg64a)'Temporal strnction of letter groups and span of
perception', Qttart. j. exp. Psychol., r6, z3z-24o.
MrcHoN, f . A. (tg64b)'A note on the measurement of percepnnl motor
load', Ergonomics, 7, 46t-463.
MICHoN, J. A. (1966) 'Tapping regularity as a measure of perceptual

motor load', Ergonomics, g, 4ot-4r2.
MrcHoN, f . A., and vAN DooRNE, H. (1967)'A semi-portable apparatus
for the measurement of perceptual motor load', Ergonomics, to, 67-72.
MrcnorrE, A. (t946) In Perception de la Causalitd. Louvain: Instinrt
Sup€rieur de Philosophie.

MrcHorTE, A. (1963) The Perception of Causality. London: Methuep.
MIcKo, H. c. (t966) 'Vigilance -arousal us. reinforcement', Quart. J.
exp. Psychol., 18, 39-46.

MILLER, G. A. (tgS6)'The magical number seven, plus or minus two:
some limits on our capacity for processing information', Psychol. Ra).,
63, 8t-97.

MILLER, G. A., GALANTER, E., and PRTBRAM, K. H. (196o) Plans and
the Structure of Behauior. New York: Henry Holt & Co.
MTLLER, G. A.1HEIsE, G. A.e and LICHTEN,'w'. (rqSr)'The intelligrbility of speech as a function of the contqxt of the test materials',

j. exp. Psychol., 4r, 3zg-335.

MILLER, G. A., snd SELFRTDGE, JENNTFER A. (r95o) 'Verbal context
and the recall of meaningful material', Amer.1. Psychol., 63, t76-r85.
MTLLER, r . (t957) 'Perception of nonsense passages in relation to irmount
of inforrnation and speech-to-noise ratio'rJ. exp. Psychol.r SSt 388-393.
MrRA, E. (rg4) Psychiatry in Var. London: Chapman & Hall.

372

Refrences

MITcHELLe M. f . H., and vINcE, MARGARET A. (rgSr)'The direction
of movement of machine controls', Quart. J. exp. Psychol., 3t 24-35.

MONTPELLTERe G. or (r93) Les Altdrations Morphologiques des Motntements Rapides. Louvain: Institut Sup6rieur de Philosophie.
MoNTy, R. A., KARSH, R., and TAUBe rr. a. (r967a)'P?cing of rehearsal

in sequential short-term memory', J. exp. Psychol., 74r 3oo-3o2.

MoNTyr R. A., KARSH, R., and TAUBl fI. A. (t967b) 'Keeping track of
sequential events: irrelevant information and paced rehearsal', Percep.

mot. Skills, 24t 99-ro3.

MoNTy, R. A., TAUBl fI. A., and LAUGHERYe K. R. (tg6S) tKeeping
track of sequential events: effects of rate, categories, and trial length',

j. exp. Psychol., 69, 224-229.

MoRAYT N. (lgSg)'Attention in dichotic listening: affective cues and the
influence of instructions', Quart. j. exp. Psychol., l,l, 56-6o.

MoRAY, N. (196o) 'Broadbent's filter theory: Postulate H and the

problem of switching time', Quart. J. exp, Psychol., tz, 2r4-22o.
MoRAYe N., and TAYLoR, A. M. (rgS8) 'The effect of redundancy in
shadowing one of two dichotic messages', Lang. and Speech., 3r 7-ro.
MoRGAN, c. T. (lg6S) Physiological Psychology. New York: McGraw-

HilI.
MORGAN, C. T., COOK, J. S., CHAPANTS, A., and LUND, M.'\il',. (eds.)
(tg6a) Human Engineering Guide to Equipment design. New York:

McGraw-Hill Book Co., Inc.
MoRIKIyol y., snd rIDA, H. (tg6il 'Choice behaviour and reaction

times', j. Sci. Lab., 41, 46r-467.
MoRrN, R. E. (1955) 'Factors influencing rate and extent of learning in
the presence of misinfonnative feedback'rj. exp. Psychol.r 4gt 343-35r.
MoRrNr R. E.e DERosAe D. v.e and srul.TZ, v. (tg6il 'Recognition

memory and reaction time', in Attention and Performance (ed. A. F.
Sanders). Amsterdam: North-Holland Publishing Co.
MoRrNr R. E., FoRRrN, B., and ARCHER, w. (196r)'Information processing behaviour: The role of irrelevant stimulus information', J. exp
Psychol., 6t, 89-96.
MORTN, R. E.r KONTCK, A., TROXELL, NOLA, and MCPHERSON,
sANDRA (tg6S) 'Information and reaction time for "naming" responses'r.7. exp. Psychol., 7ot 3og-3r4.
MoRRTsETT, L.e and HovLAND, c. r. (lqSg) 'A comparison of three
varieties of training in human problem solving', J. exp. Psychol.r 58,
52-55.

MoRTENSoN, F. J., and LoESS, n. Gg6+)'Effect of very brief interpolated activity on short-term retention', Percep, mot. Skills, t\ 7978o3.

MorrRAMe v. H. Ggq+) The Physical Basis of Personality. Efarmondsworth: Penguin Books.
MowBRAYI G. H. (1952) 'Simultaneous vision and audition: the detection

Refermces

373

of elements missing from overlearned sequences', j. exp, Psychol., 44,
292-3OO.

MowBRAy, G. H. (tgSf) 'Simultaneous vision and audition: the comprehension of prose passages with varying levels of diffic'tilty'r1. exp.
Psychol., 46, 365-372.
MOWBRAY, G. n. GgSD 'The perception of short phrases presented

simultaneously for visual and auditory receptioD', Quart. j. exp.

Psychol., 61 8@2,
MOWBRAY, c. H. (t96o) 'Choice reaction times for skilled responses',
Quart. J. exp. Psychol., l,2, tg3-2o2.
MOWBRAY, c. II. (tg6+) 'Subiective expectancy and choice reaction
times', Quart. j. exp. Psychol., l.6, ztGzz3.
MowBRAy, c. fI.e and RHoADES, M. v. (tgSg)'On the reduction of
choice reaction times with practice', Quart. j. exp. Psychol., l,l,, tilz3.
MUNsoN, w. A., and KARLTNT J. E. (tg14) 'The measurement of human
channel transmission characteristics', j. Acoust Soc. Amcr., 26, 542553.

MURDocKe B. n. (r96oa) 'The Distinctiveness of stimuli', Psychol. Rez).,

67, rG3r.

MURDocKe B. n. (r96ob)'The immediate retention of unrelated words',
1. exp. Psychol.r 60,, zzz-234.
MURDOCK, B. B. (1961) 'The retention of individtral items', j. exp.
Psychol., 62, 6t8425.
MURDocK, B. r. (1962)'The serial position effect of free recall', j. exp.
Psychol., 64, 482-488.
MURDOCKI B. B. (1963a) 'Short-term retention of single paired associates', J. exp. Psychol.r 65, 433-443.
MURDOCK; B. B. (tg6lb) 'Interpolated recall in short-term memory',
J. exp. Psychol., 66, SzS-532.
MURDOCK, B. n. (r965a) 'A test of the "limited capacity" hlpothesis',
j. exy. Psychol., 69, 237-240.

of a subsidiary task on short-terrn
memory', Brit. 1. Psychol., SQ 4t3-4r9.
MURDOCK, B. B. (r965c) 'Signal-detection theory and short-terrn
memory', 1. exp. Psychol., 7ot 443-M7.
MURDOCK, B. B . (t966a) 'The siterion problem in short-term memory',

MURDOCKT B. B. (tg6Sb)'Effects

j. exP. Psychol., 72, 317-324.
MURDOCKI B. B. (1966b) 'Visual and auditory stores in short-term
memory', Quart. j. exp. Psychol., r8, 2o6-2tt.
MURDOCKI B. B. (tg6il 'Auditory and visual stores in short terrn
memory', ir Attqttion and Performance (ed. A. F. Sanders). Amsterdam:
North-Holland Publishing Co.
MURDOCKI B. B., and voM SAAL, w. (t967)'Transpositions in shortterm memory', J. exp. Psychol., 74t r37-r43.
MURRAY, D. f . (tg6S) 'Vocalization-at-presentation and immediate re-

374

Refuences

call, with varying presentation-rates', Quart. etcp, Psychol., r7t 4756.

MURRAy, D. J. (1966) 'Vocalization-at-presentation and immediate recall,
with varying recall methods', Quart. j. exP. Psychol., r8, 9-r8.
MURRELLT K. F. H. (tgSl)'Data on human performance for engineering
designers : Controls', Engineering, r84, 3o8-3ro.
MURRE LL t K. F. H. (tg6S) Ergonomics. London: Chapman & Hall.
MURRELL, K. F. H.1 Brrd rNTwrsLE, D. G. (t96o)'Age differences in
movement pattern', Nature, r85, 948.
NAKAMURA, c. y., and KASwAN, f . w. (t962) 'Effect of stimulus condition and reaction time information on spatial stimulus generalization',
J. exp. Psychol., 64, 67-76.
NAKAMuRA, c. y., and KRUDIs, B. R. (1967)'Evaluation of a response
speed measure of incentive value of reward',i. exp. Psychol., 74, 4449.

NAyLoR, f .c., and BRIccsl G. r. (1963) 'Effects of task complexity and
task organization on the relative efficiency of part and whole training
methods', j. ex7. Psychol., 65, 217-224.

NE r s sERl rJ. (rq6f)'Decision-time without reaction-time : experiments in

visual scanning', Amer. 1. Psychol., 76, 376185.
NErssER, u., and BELLER, H. r. (1965)'Searching through word lists',
Brit. J. Psychol., 56 349-358.
NEUMANN, EVe (196o) 'Frequency and usefulness of verbal and norlverbal methods in the learning and transfer of a paired-associate serial
motor task', J. exy. Psychol., 60,, ro3-rro.
NEyMAN, E. B. (1g66) 'Speed of reading when the span of letters is
restricted', Amer. J. Psychol.r Tgt 272-278,
NTSKERSoN, R. s. (rg6Sa) 'Response times for "same"-"different"
iudgrnents', Percep. rnot. Skills, zo, 15-18.
NrcKERsoN, R. s. (r965b) 'Response tirne to the second of two successive
signals as a function of absolute and relative duration of intersignal

interval', Percep. mot. Skills, 2tt 3-ro.
NTSKERSoN, R. s. (1966) 'Response times with a memory-dependent
decision task', 1. exP, Psychol., 72, 76t-769,
NrcKERsoNl R. S. (t967a) 'Categoization time with categories defined
by disiunctions and coniunctions of stimulus attributes', j. ex/. Psychol., 7* 2rr-2t9.
NTcKERSoNe R. s. (tg6lb) 'Expectsncy, waiting time and the psychological refractory period', in Attmtion and Performance (ed. A. F.
Sanders). Amsterdam: North-Holland Publishing Co.
NTSKERSoN, R. s. (l96lc)'Psychological refractory phase and the functional significance of signals', J. exp. Psychol., 73t 3o3-3r2NISKERS9N, R. s., and FEEHRER, c. E . (t964)'Stimulus categorization
and response time', Percep. mot. Skills' 18, 785-793.

Referenas

3TS

NISHIOKA, A., AKIBA; N., snd vAMAHIRA, I. (196o) 'Experiments on
the prolonged vigilance task (I)', Institute for Science of Labour, Tokyo,

Rep. No. 17, pp. 23-3e..
NoBLE, M.e FITTS, p. M., and WARREN, c. E. (lgSS)'The frequency

response of skilled subiects in a pursuit tracking task', j. exp. Psychol.,

49t 249-256.
NOBLE, M.e TRUMBOe D., and FowLER, F. (t967)'Further eviderrce of
secondary task interference in tracking', j. exp. Psychol.r T* t46'-lt4g.
NORMAN, D. A. (1966) 'Acquisition and retention in short-term memoryr,
J. exp. Psychol., 72t 369-38r.
NORMAN, D. A. (tg6l)'Temporal confusions and limited capacity pro-

cessors', in Att*ttion and Performance (ed. A. F. Sanders). Amsterdam:
North-Holland Publishing Co.
NORMANI D. A., and WTCKELGREN, w. A. (lg6S) 'Short-term recognition memory for single digits and pairs of digits',I. exp. Prychol., 7ot

479-489.

NORTH, I. D., and LoMNrcKr, z. A. (196r) 'Further experiments on
human operators in compensatory tracking tasks', Ergonomics,
339-353.

4,

NoRTH, f. D.e LOMNICKI, z. A.e and ZAREMBA, s. K. (lgSS) .The
design and interpretation of human control experiments', Ergonomics,

\ 314-327.

OLDFIELD, R. c. (tgS+)'Memory mechanisms and the theory of
schemata', Bit. j. Psyclnl., 45, t4-23.
OLDFTELD, R. c. (1955) 'Apparent flucnrations of a sensory threshold,,
Quart. J. exp. Psychol., 7t ror-r15.
OLDFTELDe R. c. (rg66) 'Things, words and the braiD', euart.

J. exp.

Psychol., r8, 34o-351
OLDFTELD, R. C.e and wINGFTELDI A. (rg6S) 'Response latencies in
naming objects', Quart. j. ex?. Psychol., r7t 273-28r.
OOSTLANDER, A. M., and DE swART, n. (1966) 'search-discrimination

time and the applicability of information theory', j. exp. psychol.,
72t
423-P8.

osBoRNE, E. r. (r9r9) The outpttt of wontm workers in relation to hours
of work in shell-making. Industrial Fatigue ResearchBoardReportNo. 2.
II.M.S.O., London.
PATLLARD, I. (196o) 'The patterning of skilled movements', rn Handbook
of Physiology - Neurophysiology.III (ed. J. Field, ff. !7'. Magoun and
V. E. Hall). Vashington: American Physiological Society, pp. 1679r7o8.

PAILLARD, I. (1968) Lecture at the Cambridge University Psychological
Laboratory.
pARDUccr, A. (1965) 'Category iudgement: a range-frequency modelr,
Psychol. Rev., 72t 4o7-4r8.

376

Refrmces

PARKS, T. E. (lg6S) 'Post-retinal visual storage', Amer. J. Psychol.r 78,

r45-t47.

PETERS, R.lil'. (tg!+a) Competing messa,ges: the effect of interfering messages
upon the recePtion of primary rnessages. Lf.S.N. School of Aviation

Medicine. Proiect NMooro64or, Rep. No. 27.
nETERSI R. \ry'. (tgS+b) Message reception as a function of the time of
occtfiTence of ercffaneous rnessages. tf.S.N. School of Aviation Medicine.

Project NM ooro64or, Rep. No. 33.

rETERS, w., and wENBoRNE, A. A. (1936)'The time pattem of voltrntary movements', Brit. J. Psychol., 26, 388-396 and 27, 6v73.

pETERsoN, L. R. (r966a) 'Short-term verbal memory and learning',
Psychol. Reu., 73t r93-2o7.
pETERsoN , L. n. (r966b) 'Reminiscence in short-term retentioD', J. exp.

Psychol., 7r, t r5-r 18.
rETERSoNT L. R., and KRoENER, susaN (1964) 'Dichotic stimulation
and retentiott', j. exp, Psychol., 68, rz1-r3o.
rETRIEI A., coLLINS, w'.e and soLoMoN, p. (196o) 'The tolerance
for pain and for sensory deprivatiorr', Amer. J. Psychol.r 23, 8ryo.
pEw, R. w. (1966) 'Acquisition of hierarchical conuol over the temporal
organization of a skill', j. exp. Psychol., 7rt 764-,77r.
prERCE, f . R. e and KARLTN, f . E. (tgSil 'Reading rates and the
irrformation rate of a human channel', Bell System Tech. j., 16, 497-

5r6.

p r EnoN, u. (tgzo) 'Nouvetles recherches sur l'analyse du temps de latence

sensorielle et sur la loi qui relie ce temps i l'intensit6 de l'excitation',
Annde Psychol., 22, S8-t42.
pIERoN: H.(1936)'Recherches sur la latence de perception des accroissements de luminosit6', Annde Psychol., 3Tr r-16.
rTERREL, RosEMARy, and MURRAv, cATHERTNE s. (rg6g) 'Some
relationships between comparative judgment, confidence and decisiontime in weight-lifting', Amer. J. Psychol., 76t z8-38.
rIERSoN, w. R. (196r) 'Body size and speed', Res. Quart.r l2t t97-2oo.
pILLSBURv, w. B., snd syl,vESTER, ANNE (rg+o) 'Retroactive and
proactive inhibition in immediate memory', j. exp. Psychol., 27t 532545.

prNNEo, L. R. (tg6r) 'The effects of induced muscle tension during
tracking on level of activation and on performance'r 1. exp. Psychol.,
62, 523-53 r.

pINNEo, L. R. (1966)'On noise in the nervous system', Psychol. Rev.,
73t 242-247.

prrzrG. F.e and DowNrNG, r,. (1967)'Optimal behaviour in a decisionmaking task as a function of instnrctions and payoffs'rl. exp. Psychol.,

7* 549-555.

poLLAcK, r. (1952) 'The information of elementary auditory displays',

J, Acoust. Soc. Amer., z4r 745-149.

fufrmes

3ll

poLLAcK, t. (r953a) 'The information of elementary auditory displays:

II', J. Acoust. Soc. Arner., 2* 765-769.

PoLLAcK, l. (r953b) 'Assimilation of sequentially encoded information',
Amer. 1. Psychol., 66, 4zt-435.

POLLAcKT r. (r963a)'Message-uncertainty and message-reception : IIf.
Effect of restriction of verbal context', J. verb. Learn. uerb. Beh., r,
392-395.
PoLLAcK, t. (r96fb) 'Speed of classification of words into superordinate
categories'rJ. uetb. Learn. zterb. Beh.r 21 159-165.
POLLAcK, r. (t96tc) 'Verbal reaction times to bridy presented words',
Percep. mat. Skills, r7t r37-r38.
PoLLAcKT r.r and DECKER, L. R. (r.gS8) 'Confidence ratings, message
reception and the receiver operating characteristics' , j. Acoust Soc.
Amer., 3o, 286-292.
PoLLAcK, r., and FrcKS, L. (t954) 'Information of elementary multidimensional auditory displays. II', J. Acousr. Soc. Amer.,26, t55-r58.

PoLLACKI r., snd IoHNsoN, L. (rg6EA) 'Monitoring of sequential
binary patterns', Pucep. mot. Skills, 16, 9rr{r3.
PoLLAcKT I.r and JoHNSoN, r. (r963b) 'Continuing memory span for
digits', Percep. mot. Skills, r.7t 73t-734.

PoLLAcKT r.r and IoHNsoN, L. B. (1965) 'Memory-span \rith efficient
coding procedures', Amcr. j. Psychol., Z\ 6og4t4.
POLLACKT r., IoHNsoNr L. B.r and KNAFF, p. R. (tgSg)'Running
memory span', j. exp. Psychol., ST, t37-r46.
PosNER, M. r (rg6g) 'Irnmediate memory in sequential tasks', Psychol.

Bull.r 6o, 333-349.

M. r . (t964) 'Rate of presentation and order of recall in immediate memory', B'rit. j. Psychol., 55r 3o3-306.
PosNER, M. r. (1965)'Memory and thotrght in human intellectual perP osNER,

formance', Brit. J. Psychol., 56, tg7-zt1.
PosNER, M. r. (tf7a) 'Short term memory systems in human infiorma-

tion processing', in Attentian and Pufomtance (ed. A. F. Sanders).

Amsterdam : North-Elolland Publishing Cro.
PosNERl M. r. (r967b) 'Characteristics of visual and kinesthetic memory
codes', j. exP. Psychol., 7St ro3-ro7.
posNER, M. r., and KoNrcKe A. r. (1966)'On the role of interference in
short-terrn retentioo', j, exp. Psychol., 72t z2t-23t.
posNER, M. r., and MrrcHELL, R. F. (1967)'Chronometric analysis of
classification', Psychol. Reo., 74t 392-409.
posNER, M. r., and RossMAN, ELLEN (rg6S) uEffect of size and location
of informational transforms upon short-term retentioo', j. exp.Psychol,,

7ot 496-505.

posrMAN, L., and BRUNERT I. s. (1949) 'Multiplicity of set as a determinant of perceptual behaviour', j. exp. Prychol.r 39r 36y377.
posrMAN , L., and cocclN, IUDTTH (t960'S7hole versus part learning

375

References

of serial lists as a ftrnction of meaningfulness and intralist similarity',

1. exp. Psychol., 68, r4o-r5o.
posrMAN, L., and coccrN, IUDTTH (t966) 'Whole versus part learning of paired-associate lists', 1. exP. Psychol., 7rt 867-877 -

pOSTMAN, L., KEPPEL, G., and STARK, KAREN (1965) 'Unlearning as
a function of the relationship between successive response classes',
J. exP. Psychol., 69, rIt-tr8.
POSTMAN, L., TURNAGE, T. 'W'.e and SILVERSTEIN, A. (tg6+)'The
running memory span for words', Quart. i. exp. Psychol., r61 8r-89.

pou1,ToNr E. c. (rgSo) 'Perceptual anticipation and reaction time',
Quart. j, exp. Psychol., 2, g9-1r2.
rOULToN, E. c. (1953) 'Two-channel listening', J. exP. Psychol,, 461 9r96.

pouLToN, E. c. (tgS}'Eye-hand span in simple serial tasks',J. acp.
Psychol,, 47t 4o3-4to.

pouLToN, E. c. (rgS8a) 'Copying behind during dictation', Quart. J.
exp. Psychol,, l.o, 48-55.

pouLToN, E. c. (rgS8b) 'Measuring the order of difficulty of visualmotor tasks', Ergonomics, t, 234-239.
pou1,ToNr E. c. (196o) 'The optimal perceptual load in a paced auditory
inspection task', Brit. J. Psychol., 5t, t.27-139.
poulToNr E. c. (1963) 'sequential short-term memory: some tracking
experiments', Ergonuttics, 6, t'17-t32.

poulToNr E. c. (1964)'Poswiew and preview in tracking with complex
and simple inputs', Ergonamics, 7, 257-266.
pou1,roN, E. c. (t966)'Tracking behavior', in Acquisition of Skil/ (ed.
E. A. Bilodeau). New York and London: Academic Press, pp. 36r4ro.

pRrcE, R. Ir. (rg66) 'signal-detection methods in personality and perception', Psychol. Bull., 66, 5542pRovrNSr K. A. (tg1l)'Sensory fastors in the voluntary application of
pressure', Quart. i. ex7. Psychol,r g, z8-4t.
pRovrNs, K. a. (1958) 'The effect of peripheral nerve block on the appreciation and exectrtion of finger movements', i. Physiol., r43t 55-67.
pUGHe L. A., OLDROYDe C. R., RAY, T. S.e and CLARK, M. L. (1g66)
'Muscular effort and elecuodermal responses', j. ex/. Psychol., 7rt
z4t-248.
RAAB, D. H. (tg6z)'Effect of stimulus-duration on auditory reactiontime', Amer. J. PsYchol.r 75, 298-3or.
RAABe D. H., and cRossBERGl M. (lg6S) 'Reaction time to changes in the
intensity of white noise', i. exy. Psychol., 69, 6ofit1
RABBrrr, p. M. A. (1959) 'Effects of independent variations in stimulus
and response probability', Nature, r.83, !2r2.
RABBrrr, p. M. A. (t962) 'Short-term retention of more than one aspect
of a series of stimuli', Nafire, r95r To2.

Referurces

379

RABBITT, p. M. A. (tg6+) 'Ignoring irrelevant information', Brit. J.

Psychol., SSt 4o3-4r4.
RABBITT, P. M. e. (1965) 'Response-facilitation on repetition of a limb
movement', Brit. J. Psychol., SQ 3o3-3o4.

RABBITT, P. M. A. (r966a)'Errors and error correction in choiceresponse tasks', J. exp. Psychol., 7r.t 264-272.

RABBrrrr P. M. a. (r966b)'Times fortransitions benreen hand and foot

responses in a self-paced task', Quart. j. exp. Psychol., r8, 334-339.
RABBrrr, P. M. e. (r966c)'Error correction time without external er:ror

signals', Nature, 2r.2, 438.
RABBITT, P. M. A. (tg6la) 'Learning to ignore irrelevant information',
Amer. J. Psychol.r So, r-r3.

RABBITT, P. M. a. (r967b)'Time to detect errors as afunction of factors
affecting choice response time', in Attention and Performance (ed.
A. F. Sanders). Amsterdam: North-Holland Publishing Co.
RABBITT, P. M. A.e and PHILLrps, sHrRLEy (tg6il 'Error-detection

and correct latencies as a function of S-R compatibility', Quart.
1. ex?. Psychol., r9t 37-42.

RAPOPORT, A. (1959) 'A sturdy of disiunctive reaction times', Behaoioural

Scr., 4t z9*3I:5.
RAY, w. s. (1965) 'Mild stress and problem-solvinE', Amer. 1. Psychol.,
78, zz7-234.
REICHARD, suzANNE, LIvsoNr FLORINE, and PETERSEN, p. G.
(tg6z) Ageing and Persorulity. New York: fohn $7iley & Sons.
RErD, c. (t928) 'The mechanism of voluntary musctrlar fatigue', Quart.

j. exp. Physiol., r9t t7-42.
RErD, L. s.e HENNEMAN, R. H.e and LONG, E. n. (196o) tAn experimental analysis of set: The effect of categorical restriction', Amer. J.

Psychol.r T3t 568-572.
RESTLE, F., snd DAvrs, J. H. (tg6z)'success and speed of problem
solving by individuals and groups', Psychol. Ran., 69, 52c..536.
REsrLE, F.r and EMMERTcH, D. (tg6O) 'Memory in concept attainment:
effects of giving several problems concrurently', j. exp. Psychol., 7tt
794-799.
REY, P., and REY, f.-P. (1963)'Les effets compar6s de deux dclairages
fluorescents sur une tache visuelle et des tests de "fatigtr€" ', Ergonomics, 6, 393-4ot.

REYNOLDS, B. (tgSz) 'The effect of learning on the predictability of
psychomotor performance', J. exp. Psychol., 44, r89-r98.
REYNOLDS, B.e and ADAMS, f . A. (rgS:) 'Motor performance as a
function of click reinforcement', j. exp. Psychol., 4St 3tl-32o.
REYNOLDS, D. (tg6+)'Effects of double stimulation: temporary inhibition of response', Psychol. Bull., 62, 333-347.
REYNOLDS, D. (1966) 'Time and event uncertainty in trnisensory retction time', 1. exp. Psychol., Trt 286-293.

38o

Refermces

RrcHARDs, f . M. (tg6z) 'The c"tre additivity principle in a restricted
social interaction sinration', j. exp. Psychol., 61, 4Sz-457.
RIcHARDSoNT pATRrcrA, and voss, J. F. (rg6o) 'Replication report:
Verbal context and the recall of meaningfui material'r 1. etcp. Psychol.,

fut 4r7-4r8.

RIEDEL: vr. w. (1965)'Anxiety level and the "doubtful" iudgement in a
psychophysical experiment', j. abt. Psychol., 7ot 462-464,
RoBINsoN, c., and LoEss, H . (t967) 'Short-term retention of individual
paired associates as a function of conceptual category'r1. etcp. Psychol.,
75t r33-r35.
RoBrNSoNr f . S.r BRowN, L. T.r and HAYES, w. fr. Gg6+)'Test of
effects of past experience on perception', Percep. mot. Skills, rE,
953-956.
RoDwAN, A. s., and HAKE, tI. w. (1964)'The discriminant-function as a
model for perception', Anter. J. Psychol., 77t 38o-392.

RosENeursT, H. s. (lg6S) 'Visual response component of rotary

pnrsuit tracking', Percep. mot. Skills, 2\ 555-56o.
Ross, HELEN E., and cREGoRy, R. L. (t964): 'Ir the !7eber fraction a
function of physical or perceived input?', Quart. J. exp. Psychol., r.6,
TT6-I,22.

RUBIN, G., voN TREBRA, pATRrcrA, and sMrrH, K. u. (tg1z)'Dim€rlsional analysis of motion: 3. Complexity of movement pattern', j. appl.
Psychol., ?6, z7z-276.
RUBINSTETN, L . (t964)'Intersensory and inrasensory effects in simple
reaction time', Percep. mot. Skills, r8r t1g-r72.
RUCH: T. c. (r95r) 'Motor systems', rn Handbookof ExperhnentalPsychology (ed. S. S. Stevens). New York: John! Sfiley & Sons. London:
Chapman & Hall, pp. r54-2o8.
RUssELLe \il'. n. (1959) Brain Memory Learn'ing. Oxford University Press.
RvAN, J. Gg6il Unpublished Ph.D. Thesis. Lfniversity of Cambridge.
sALDANHA, E. L. (rg15) 'An investigation into the effects of prolonged
and exacting visual work', M.R.C. Applied Psychology Research Unit
Report No. 243.
sALDANHA, E. L . (tg17) 'Alternating an exacting visual task with either
rest or similar work', M.R.C. Applied Psychology Research Unit Report
No. 289.
SALTZMAN, r. J., KANFER, F. Ir.e and GREENSPooN, f. (lqSS)
'Delay of reward and human motor learning', Psychol. ReP.r 5r r39-t52.
sAMpsoN, u. (1964) 'Immediate memory and simultaneous visual stimulation', Quart. j. exp. Psychol., l6, r-ro.
sAMpsoN, H., and spoNG, n. (r96ra) 'Binocular fixation and immediate
memory', Brit. J. Psychol., S2t 239-248.
sAMpsoN, H., and spoNG, n. (r96rb) 'Handedness, eye-dominance and
immediate memory', Quart. I. exP. Psychol., t3t r73-r8o.

Refermces

38r

SANDERS, A. F. (r96ta) 'Rehearsal and recall in immediate memory',
Ergonomics, 4, 29-34.
SANDERS, A. F. (t96tb) 'Aandachtsverschuiving en lawaaihinder', Ned.

Tijd. Psychol., r.6, 46v474.

SANDERSe A. F. (tg6+) 'Selective strategies in the assimilation of suc-

cessively presented signals', Quart. J. exp. Psychol,, 16, 36g-372.

SANDERS, A. F. (1967) 'Some aspects of reaction processes', in Attention

and Performance (ed. A. F. Sanders). Amsterdam: North-Holland

Publishing Co.

SARASoN, r. c., and PALoLA, E. G. (196o)'The relationship of test and

general anxietY, difficulty of task and experimental instnrctions to performance', 1. exp. Psychol., S9t r85-r9r.
SAUFLEY, '\ry'. I{.e and BILODEAU, rNA McD. (.t963)'Protective selfpacing during learninE'rJ. exp. Psychol., 661 596-6oo.
scHAUB, G. R.1and LTNDLEY, R. H. (tg64)'Effects of subiect-generated
recoding cues on short-term memory'rl. exp. psychol.r 6g, t7r-r75.

scHIFF, w. (196r)'The effect of subliminal stimuli on guessingaccuracy', Amer.1. Psychol., 74t 54-60.

scHourEN, f. F.e and BEKKER, f. A. M. (tg6il'Reaction time and

accuracY', ir Attentian and Performance (ed. A. F. Sanders). Amst€rdarn: North-Holland Publishing Co.
scHourEN, f. F.e KALSBEEK, f.'w. H., and LEopoLD, F. F. (t962)
'On the evaluation of perceptual and mental load', Ergonomics, 5, 25r26c..

scHwAB, R. s. (1953) 'Motivation in measurements of fatigue', iD Symposium on Fatigue (ed. Str. F. Floyd and A. T. lTelford). London:
rr. K. Lewis & Co., for the Ergonomics Research Society.
SEARLE, L. v., snd TAYLoR, F. v. (1948) 'studies of tracking behavior.
f . Rate and time characteristics of simple corrective movements',7.
exp. Psychol., 18, 615-63r.
sErBEL, R. (tg6f) 'Discrimination reaction time for a rro23-alternative
task', 1. exp. Psychol., 66, zt1-226.
SEIBEL, R., CHRISTg R. E.9 and TEICHNER, w. TI. (rg6s) .Short.term
memory under work-load stress', J. exp. Psychol., 7ot t54-t62.
SETDENSTETN, S.e cHERNIKoFF, R.; and TAyLoR, F. v. (rg6o) The
relationship of a retinal-gain index to systeff, performance, Lf.S. Naval
Res. Lab., Rep. No. 5548.
SELYET rI. (lgSo) The physiology and Pathology of Exposure to Srress.

Montreal: Acta.

SENGSTAKE, c. B. (t965) 'Perception of deviations in repetitive patterns',
j. exp. Psychol. 7or 2rvzt7.
SEYFFARTHT rr. (lg+o) 'The behaviour of motor units in voluntary con-

traction', S&r. Norke Vidensk akad. I Mat-Nat Kl., No. 4.
sEYMouR, 'w'. D. GgSry) Industrial Training for Mamml Operations.
London: Pitrnan.

382

References

sEyMouR, w. D. (r954b) 'Experiments on the acquisition of indusuial
skills', Occupational Psychol.,28,77-89. (tgSS) Part z. Ibid.,29t 8z-98.
(rgS6) Part 3. Ibid., 3or g4-ro4. (1959) Part 4, assembly tasks. Ibid.,
33r 18-35.

sHALLrcEl T., and vIcKERsl D. (tg6+) 'Theories and experiments on
discrimination times', Ergonomics, 7t 37-49.
5HANNoNT c. E., and wEAvER, w. (1949) The Mathematical Theory of
Cwnmunication lJrbana, IK: University of Illinois Press.
sHELLy, M. \ry'. (lg6l) 'Learning with reduced feedback information',

j. exy. Psychol., 62, 209-222.
5HEpARD, R. \M.5 and sHEENAN, MAUREEN l\[. (1965)'Immediate recall
of numbers containing a familiar prefix or postfix', Percep. mot. Skills,
zr,263-273.

sHERRTNGToNI c. (19o6) The Integrathte Action of the Nentous Systern,.
Cambridge University Press fi'eset Edition rg47).
sHERwooD, J. J. (1965) 'A relation between arousal and performance',
Amer. J. Psychol., 78, 46t-465.
sIDowsKr, I. B., snd EASoN, R. G. (196o)'Drive, verbal performance
and muscle action potential', i. exp. Psychol., 6o,, 365-37o.
sIDowsKr, f . B., and NUTHMANN, c. (196r) 'Induced muscrrlar tension,
incentive and blinkrate in a verbal learning task', j. ex/. Psychol.r 6r,
295-299.
srE6EL, A. r., and woLF: I. I. (r962)'A model for digital simulation of
two-operator man-machine systems', Ergonomics, 5t 557-572.

sIMoN, H. A., and KorovsKy, K. (rg6f) 'Fluman acquisition of coficepts for sequential patterns', Psychol- Reo., 7ot 534-546.
srMoN, J. R. (196o) 'Changes with age in the speed of perforrnance on
a dial setting task', Ergonornics, 3t T69-174.
srMoN, J. R.e and sIMoN, BETTx p. (1959)'Duration of movements in
a dial setting task as a function of the precision of manipulation',

1. appl. Psychol., 43t 389-394.
srMoNsoN, E. (lg6S) 'Performance as a ftrnction of age and card.io-

vascular diseases', in Behaaior, Agetng and the Nem.torzs {ystem (ed. A. T.
Sfelford and J. E. Birren). Springfield, Illinois: Charles C. Thomas,

pp. 40r-434.
srMoNSoN, E., and BRo zEK, I. (1952) 'Flicker Fusion frequency. Backgronnd and applications', Physiol. Ren., 32t 349-378sINcLEToN, 'w. T. (rgS:) 'Deterioration of perforrnance on a shortF.
term perceptual-motor task', in Symposium on Fatigue (ed.
Ergothe
for
&
Co.
Lewis
H.
K.
London:
Floyd and A. T. \[elford).
nomics Research SocietY.
sTNGLETSN, w. T. (1954) 'The change of movement timing with aBe',
Brtt. J. Psychol., 45, t66-!72.
SINGLET9N, w. T. (rgSS) 'Age and perf,ormance timing on simple

kferances

383

skills', in Old Age in the Modem lVorld: Rep. 3rd Cong. Internar.
Ass. Gerontol., London, 1954. Edinburgh: Livingstone, pp. zzt231.
STNGLETON, w. T. (tgSil 'An experimental investigation of sewingmachine skill', Brit. J. Psychol., 48, r27-r32.
STNGLETON,}v. T. (1959) 'The training of shoe-machinists', Ergonomict,

z, t48-r1z.

SINGLETON, w. T., EASTERBYT R. s., and WHITFTELD, D. J. Gg6l)
'The human operator in complex systems', Ergottotttics, ror gg-2gz
(also published in book forrn).
srPowtcz, R. R., WARE, I. R.e and BAKER, R. a. (1962) tThe effects of
reward and knowledge of results on the performance of a simpte
vigilance task'r I. exp. Psychol,, 641 58-6r.

SLACK, c. 'w'. (tgSf) 'Learning in simple one-dimensional tracking',

Amer. J. Psychol., 66, 33-U.
SLAMECKA, N. f . (1963) 'Choice reaction-tirne asafunaionof meaningful
similariry', Amer. J. Prychol., 76, 274-28o.

SLAMECKA, N. J. Gg6il'Recall and recognition in list-discrimination
tasks as a function of the number of alternatives', 1, exp. Psychol., 74,

t87-r92.

SLATER-HAMMEL, A. T. (t958) 'Psychologicalrefractorypedod in simple
paired responses', Res. Quart. Amer. Ass. Hlth. Phys. Educ. Rec., zg,

468-48r.

sMrrH, E. E. (t961)'Effects of familiarity on stimulus recognition and

categorization', j. exp. Psychol., 74t 324-332.
SMITHT K. Lr. (t966) 'Cybernetic theory and analysis of learning', in

Acquisition of Skill (ed. E. A. Bilodeau). New York and London:
Academic Pressr pp. 425-482.
SMITH, K. u., and. SMITH, w. M, (t962) Percegtion and Motion Philadelphia: Saunders.
sMrrH, L. A. (1965)'Individual differences in ma:rimal speed of muscular
contraction and reaction time', Percep. mot, Skills, 2r., r.gr2z,
SMITH, M.e and wrlsoN, EDNA A. (rgsr) 'A model of the auditory
threshold and its application to the problem of the multiple observer,,
Psychol.?Monogr.t 67, No. 9.
sMrrH, MARILYN c. (tg6la)'Theories of the psychological refractory
period', Psycltel. Bull., 67, 2o2-2t3,
sMrrH, MARILYN c. (t967b)'Reaction time to a second stimulus as a
function of intensity of the first stimulus', Quart. I. exp. Psychol., r9t

t25-r32.

sMrrH, MARILYN c. (r967c) 'The psychological refractory period as a

function of perforrrrance of a first response', Qttart. j. exp. Psychol., y,9,
350..352.

sMrrH, s., and MYERS, T. r. (1966)'stimulation seeking during sensory
deprivatiorr', Percep. mot. Skills, 23, rr5r-r tG3.

l8+

References

sMITH, s., MvERS, T. r., and MURpHyr D. B. (t967)'Vigilance during
sensory deprivation', Percep. mot. Skills, 24t 97r+76.
sMITH, w. M.e MccRARy, J. \il'.5 and sMrrH, K. rr. (t96o) tDelayed

visual feedback and behavior', Sciencet tl2t ror3-ror4.
sMrrH, w. M., SMrrH, K. u., STANLEy, R., and HARLEy, \il. (rg56)
'Analysis of performance in televised visual fields: preliminary report',
Percep. mot. Skills, 6, r95-r98.

soLoMoNS, L. M. (r9oo) 'A new explanation of S7eber's law', Psychol.
Rev.r 7r z34-24o.

spENCE; K. \il., FARBER, r. E.1 and McFANN, H. H. (r956a) tThe
relation of anxiety (drive) level to performance in competitional and
non-competitional paired-associates learning'r j. exp. Psychol., 52, 296305.

spENcEr K. 'w.e TAyLoRr y., and KETcHEL, RHoDe (r956b) 'Anxiety
(drive) level and degree of competition in paired-associates learning',

1. exp. Psychol., S2r 306-310.

spENcER, J. (tg6r) 'Estimating averages', Ergonomi,cs, 41 317-328.
spENcER, J. (1963) 'A further study of estimating averages', Ergonomics,

6, 255-265.
snENCER, J. (rg6S) 'Experiments on engineering drawing comprehension', Ergonontrcs, 8, 93-r ro.

spERLrNGe G. (196o) 'The information available in brief visual presentations', Psychol. Monogr.t 74t No. fr.
spERLrNG, G. (rg6f)'A model for some kinds of visual memory tasks',
Hwnan Factors, St t9-3r.

spERLrNG, G. (tg6il'successive approximations to a model for short

term memory', in Attmtion arid Perforwtance (ed. A. F. Sanders).

Amsterdam: North-Holland Publishing Co.
SrTELBERGERe c. D., BERNSTETN, rRA H., and RATLTFF, R. c. (1966)

'fnformation and incentive value of the reinforcing stimulus in verbal
conditioning', j. exp. Psychol., 7r., z6-3t.

srTELBERGERT c. D., and DENNv, f . p. (tg6g) 'Visual recogRition
thresholds as a function of verbal ability and word frequency', l. exp.

Psychol.r 65, 5974o2.

sprrz, H. H. (t967)'Information transmission in an absolute iudgment
task with feedback, using normal and retarded subiects', J. con p.
physiol. Psychol., 64, 8S+2.
STABLER, f . R., and DvAL, f . A. (1963)'Discriminative reaction-time as
a ioint ftrnction of manifest auiety and intelligence', Amer. J. Psychol,,

76, 484-487.
srAGER, p. (1g66) 'Note on use of information concepts in the assessment of group structure', Percep. tnot. Skills, 2lt 239-242.
sTANTLAND, A. c. (1966) Patterns of Redundancy. Cambridge University
Press.

sTAUFFAcHER, J. c. (tglil

'The effect of induced muscular tension

Refrences

385

upon various phases iof the learning process', j. ex?. Psychol,, 2t, zG
46.

srEFFYr R. A.r and ERTKSEN, c. 'w. (lg6S) nshort-term, perceptualrecognition memory for tachistoscopically presented nonsense forrns',

j. exp. Psychol., 7ot 277-283.

srErNMANr ALBERTA n. (r94 l) 'Reaction time to changecompared with
other psychophysical methods', Arch. Psychol., JV. Y., 4tt No. zg2.
STENNETT, R. c . (t957) 'The relationship of perfonnance level to level
of aroussl', j, exp. Psychol,, 54, 544t.
sroNEe G. c., and cALLAwAy, r. (1964) 'Effects of stimulus probability
on reaction time in a nunber-naming task', Quart. j. exp. Psychol., l:6,
47-55.
sroNE, M. (t96o) 'Models for reaction time' , Prychontetrtkat 2St z5t-26o.
srRouD, f. B. (rggt) 'The role of muscrrlar tensions in stylus maze
learning', J. exp. Psychol,, r,4, 6offi3t.
suclr G. f ., DAvrDoFF, M. D., and suRwILLo, w. w. (196o) tReaction
time as a function of stimulus information and arge'rl. exp. Psychol.r 6o,,

2p-2M.
suRwrl.l.o, \tv'. w.r and eurLTER, R. E. (rg6+)'Vigilance, age and
response-time', Amer. J. Psychol., 77s 6t44zo.
swETSr I. A. (1959) 'Indices of signal detectability obtained with various
psychophysical procedures', J. Acotait. Soc. Amfi.r 3rr 5rr-5r3.
swErs, J. A. (ed.) (tg6+) Sigwl Detection and Recognition by Hwnan
Obsenters. New York: ITiley.
swETs, f . A., and sEwALL, susAN;T. (1963) 'Invarianceof signal detectability over stages of practice and levels of motination'r1. exp. Psychol.,

66, tzvr26.
SWETS, f. A.e SHIPLEY, ELIZABETH F.e MCKEY, MOLLY f., and
GREEN, D. M. (1959) 'Multiple obsenrations of signals in noise',
J. Acoust. Soc. Amer., 3t, Sr4-52t.
swETSr A., TANNER, w. p., and BIRDSALLT T. a. (196r) tDecision

I.

processes in perception', Psychol. Rev.r 68, 3or-34o.

sYMoNDs, srR cHARLES (rg66) 'Disorders of memory', Brain, 89,
625444.
sYMoNs, f . R., and MAcKAy, c. K. (t962)'An investigation of changes
in auditory sensitivity during the performance of a mental task',
Quart. 1. exp. Psychol., t4t to4-r'o8.
SZAFRAN, I. (1963) 'Age differences in choice reaction time and cardio-

vasqrlar status among pilots', Nature, 2oo, go4no6.
SZAFRAN, f. (1966) 'Age, cardiac output and choice reastion time',
Nature, 2Ws 836.

SZAFRAN, f., and wELFoRDT A. T. (lgSo)'On the relation benreen
transfer and difficulty of initial task', Quart. 1. exp. Psychol., 21 88-94.
TAKAKUWA, r. (undated) Thefmction of caficmtration ntaintettance (TAF)
- an eoaluation of fatigue. Hokkaido University School of Medicine.

T

386

References

TANNER, vI. p., snd swETs, f . A. (r9S4)'A decision-making theory of
visual detection', Psychol. Ra)., 6l, 4ot-4o9.
TAUB, rI. A.e MoNTy, R. A., and LAUGHERY, K. R. (tg6il tKeeping

track of sequential events: effects of stimulus on-tirne and interstimulus off-time', Percep. rnot. Shills, 24t tS9-t66.
TAUB, H. A., and MyERs r!.L. (196r) 'Differential monetary gains in a
two-choice situation'r 1. exp. Psychol., 6t, tST-t62.
TAYLoRT D. H. (1964) 'Drivers' galvanic skin response and the risk of
accident', Ergonomics, % 439-45r.
TAYLoRT D. H. (1966) 'Latency components in nro-choice responding',

l. exp. Psychol.r T2t 48t-487.
TAYLoR, F. v. (1957) 'Simplifying the controller's task through display
quickenin g', Ocanpational Psychol., 3r, rze-t21.
TAyLoR, F. v., and BTRMTNGHAMT H. p. (tg+8) 'Studies of tracking
behavior. If. The acceleration pattern of quick manual cor:restive
responses', 1, exp. Psychol., 38, 783-,lg1.
TAyLoRr F. v., and cARvEy, w. D. (1959)'Thelimitations of a "Proctrustean" approach to the optimization of rlrn-rlachine systems', Ergonotnics, 2s r87-t94.

TAYLoR, IANET A. (1956) 'Drive theory and manifest an dety', Psychol.
Bull.r 53r 3o3-32o.
TAYLoRT M. M. (1965) 'Detectability measures in vigilance: comment
on a paper by S7iener, Poock and Steele', Percep. not. Skills, 2ot

t2r7-t22r.

TAyLoRr M. M.rLrNDsAy, p. H., and FoRBES, s. M. (tg6il tQuantification of shared capacity processing in auditory and visual disctimina-

tion', in Attention and Performance (ed. A. F. Sanders). Arnsterdam:
North-Holland Publishing Co.
TEccE, f . f . (1965) 'Relationship of anxiety (drive) and response competition in problem solving', j. abtn. Psychol., Tot 465-467.
TBccEr f. f., and TEsrA, D. H. (rg6S)'Effects of heightened drive
(shock)r on 2-t 4- and 8-choice card-sorting', Percep. tnot. Skills, zo,

7t51r6.

TELFoRD, c. w. (r93r) 'The refractory phase of voluntary and associative
responses', I. exp. Psychol., t4t r-36.
THrEssEN, D. D., and RoDGERsT D. e. (196r)'Population density and
endocrine ftrnction', Psychol. Bull.., 58, 4.4.r-45r.

IHrEssEN, D. D., zoLMAN, f . F., and RoDGERse D. A. (tg6z)'Relation between adrenal weight, brain cholinesterase activity and hole-inwall behavior of mice under different living conditions', j. comp.
physiol. Psychol., SS, 186-19o.

THoMAs, E. A. c. (tg6il 'Reaction-time studies: the anticipation and
interaction of responses', Brit. J. math. statist. Psychol., 2o,, t-zg.
THORNDTKE. E. L. (tgar) Hwnan Learning. New York: AppletonCenttrry.

fuferences

387

THouLEssr R. H. (r93r) 'Phenomenal regression to the "real" obiect. II',
Brit. J. Psychol., 22, r-3o.

THURSTONET L. L. (t9z7a) 'A law of comparative iudgment', Psychol.

Rev.r 34t 273-286.

THURSTONE, L. L . (t9z7b) 'Psychophysical Analysis', Amer.J. Psychol.,

38, 368-389.

THYLEN, I. o. (1966) 'The effect of initial pointer position relative to the
control on directional relationships in the presence of two ssnficring
stereotypes', Ergonomics, % 469-474.
TOPPEN, f . T. (1965) 'Effect of size and frequency of money reinforcement on human operant (work) behaviour', Percep. mot. Sh7/s, 2ot
259-269.
Training Made Easier: A Reviett of Four Recettt Studies. D.S.f.R. Problems of Progress in Industry No. 6. I-ondon: ff.M.S.O.
TREBRA, rATRTCTA VON, and SMrrH, K. rr. (tgsz)'The dimensional
analysis of motion: 4. Transfer effects and direction of movement',
j. ap?I. Psychol., 36, 348-353.
TRETSMAN' ANNE M. (196o) 'Contexttral cues in selective listening',
Quart. 1. exp. Psychol., l,2, z4z-248.
TRBTSMAN, ANNE M. (tg6+a) 'Verbal cues, language and meaning in
selective attention', Amer. 1. Psychol., 77t zofiztg.
TRETSMAN, ANNE M. (t96ab) 'The effect of irrelevant material on the
efficiency of selective listening', Amer. j. Psychol.r TTt 533-546.
TRETSMAN, ANNE M. (1965) 'The effects of redtrndancy and familiarity
on translating and repeating back a foreign and a native language',
Bit. J. Psychol., 96, 369379.
TRETSMAN, ANNE M. (1g66) 'Our limited attention', Adv. Sctr'., 22t
6oo-61 r.

TRETSMAN, ANNE, and cEFFEN, crNA (1967) 'selective attention: Perception or response?' Quart, j. exp. Prychol., l.9t t-t7.

TRETSMAN, M. (196+) 'Noise and lfeber's law: The discrimination of
brightness and other dimensions', Psychol. Rqt.t 7\ 3r4-33o.
TRETSMAN, M.: and WATTS, T. R. (t966) 'Relation benneen signal
detectability theory and the uaditional procedures for meastrring
sensory thresholds: Estimating d' from results given by the method of
constant stimuli', Psychol. Bull., 66, 438-454.
TRESSELT, M. 8., and MAYZNER, M. s. (1965) 'Anagrarn solution times:
a function of individual differences in stored diagram frequencies',
j. exp. Prychol., 7ot 6o6-61o.
TRESSELTe M. E., and voLKMANN, f . Gg+z)'The production of uni-

form opinion by non-social stimulation', 1. abn. soc. Psychol., !7t 234243.

TROwBRTDGE, M. H., and cAsoN, rr. (1932) 'An experimental study of

Thorndike's theory of learning', j. gen. Psycltol., 7t 245-258.

388

References

TRUMBo, D.e NoBLEe M., and swrNKr I. (1967)'Secondary task interference in the performance of tracking tasks', j. etcp. Psychol., 73t
232-240.

TsAo, f. c. (tg+8)'Studies in spaced and massed learning: r. Time

period and amount of practice', Quart. I. exg. Psychol,, tr 29-36.
TULvTNG, E., and ARBUcKLE, T. y. (rg66)'Input and output interference in short-term associative memory', j. exp. Psychol., 72t t4Sr50.

TULvrNc, E., and coLD, cEcrLLE (tg6l) 'Stimulus information and
cortextual information as determinants of tachistoscopic recognition
of words', j. exp. Psychol, 66, 3tg-327.
TUNE, c. s. (tg6+) 'Sequential errors in a time-sharing task', Brit. J.
Psychol., SSt 4t1-4r9.
TUNTURT, A. R. (tg++) 'Audio frequency localization in the acoustic
cortex of the dog', Amer. j. Physiol., t4\ 397-403.
TURvEy, M. T. (1967) 'Repetition and the percepttral information store',
1. exp. Psychol., 74, 289-293.
TUsrINe A. (tg+il 'The nature of the operator's response in manual
control and its implications for controller design'rJ. Instn. elect. Engrs.,
94 [IA), rgc.lzoz.

uHR, L., and MTLLER, f . e. (eds.) (rg6o) Drugs and Behansior. New

York: John STiley & Sons Inc.

ULEHL A, z. I. (1966) 'Optimality of perceptual decision criteria', j. ex?,

Psychol., 7rt 564-569.

ULICH, r. (1967) 'Some experiments on the function of mental training
in the acquisition of motor skills', Ergonwnics, ro, 4rr-4r9.
uNDBRwooD, B. f . (196r)'Ten years of massed practice on disuibuted
practice', Psychol. Rez)., 68, zzgrz47.
vERNoN, H. M. (r9r9) The influmce of hours of work and of omtilation ut
output in tinplarc manufacture. Industrial Fatigue Research Board
Report No. r. Ff.M.S.O., London.
vERNoN, H. M . (rgzoa) Fatigue and fficimcy in the iron and steel industry.
Indusuial Fatigue Research Board Report No. 5. H.M.S.O., London.
vERNoNr H. M. (tgzob) The speed of adaptation of output to ahered
hours of tlork. Industrial Fatigue Research Board Report No. 6.
H.M.S.O., London.
vERNoN, H. M. (tgz+) On the extent and fiects of oariety in repetitiae
work. Indusuial Fatigue Research Board Report No. 26. H.M.S.O.,
London.
vERNoN, H. M., and BEDFoRD, T. (lgz+) Two studies on rest pauses in
industry. Industrial Fatigue Research Board Report No. 25. Ff.M.S.O.,
London.
vERvrLLE, E.r and cAMERoN, N. (1946) 'Age and sex differences in the
perception of incomplete pictures by adults', j. genet. Psychol., 6E,
149-t57.

kftmces

l8g

vIcKERs, D . (t967) Unpublished Ph.D. Thesis. Llniversiry of Cambridge.
vINCE, MARGARET A. (r948a) 'The intermittency of controlmovements
and the psychological refractory period', Brit.J. Psychol.r S\ t49-rsl.
vrNCE, MARGARBT A. (rg48b) 'Corrective movements in a pursuittask',
Quart. j. ex?. Psychol., rt 85-ro3.
vrNcE, MARGARET e. (1949) 'Rapid response sequences and the psychological refractory period', Brit. j. P$,chol., 4o,, 23-4c..
vrNcE, MARGARBT A. (r95o) Sorze exceptions to the psychologicalrefractory pertod in mskilled manual responses. Medical Research Cotrncil
Applied Psychology Research Unit Report No. tz4/5o.
vrNCEr MARcARET A.1 and wELFoRDT A. T. (tg6il'Time taken to
change the speed of a response', Naturet 2r3t S3z-533.

vrrELBSr M . (t932) In&sfial Psychology, New York: !7. !7. Norton &
Co.

WALD, A. (rg+il Sequ*ttial Attalysis. New York: Iohn \triley & Sons.

London: Chapman & Hall.

wALDr A. (r95o) Statistical Decision Functions. New York: fohn STiley &

Sons. London: Chapman & flall.
WALKER, E.L.2 and TARTET R. o. (1963)'Memory storage as a function
of arousal and time with homogeneous and heterogeneous lists',
J. verb. Leam, verb, Beh.r 2t lr3-rr9.
WALLACB, IEAN G. (lgS6) 'Some sttrdies of perception in relation to
8g€', Brit. j. Psychol,, 47t 283-292.
WALLACET I. (1966) 'An abilities conception of personality: some implications for personality measurement', Arner. Psychologist, zr., t3zr38.
wALLIs, c. P.e and AUDLEY, R. f . G964) 'Response instnrctions and the
speed of relative iudgments: If. Pitch discrimination', Brtt.J. Psychol.,
55, t2r-t32.
wALLrs, D., and SAMUEL, I. A. (tg6r)'Some experimental sttrdies of
radar operating', Ergonortics, 4r 155-168.
wARDr N. (tgSo)'Speed as a function of distance: An analysis of roadcycling performances', Brit. j. Psychol., 4o,, zt2-2t6.
WARRINGTON, ELTZABETH K., KINSBOURNBT M.r and IAMEST M.

(tg66) 'LIncertainty and transitional probability in the span of

apprehension', Brtt, J. Psychol., S7t 7-t6.
WAUGHT N. c. (rg6o)'Serial position and the memory span', Amer.J.
Psychol.r Tlt 68-79,
WAUGH, NANCY c. (tg6il nPresentation time and free recall', j. exp.
Psychol., 73t 3yM.

wAUGH, N. c., and NoRMANT D. A. (tg6S) 'Primary memory', Psychol.
Reu.r 72t 8*ro4,

wEBsrER, R. G., and HASLERUDT G. M. (tg6+) tfnfluence on extreme
peripheral vision of attention to a vistral or auditory task', j. exp.
Psychol.r 68,269-272.

39o

References

WEENE, P. L. (lg6S) 'Effects of random and orderly presentations of
visual sequences', Prcep. mot. Skills, zor 897-903.
wEHRKAMrI R., snd SMITHT K. fr. (rgSz) 'Dimensional analysis of
motion i 2. Travel-distance effects', f. appl. Psychol., ?6, 2c.r-2o6.
wErNERr B., and wALKER, E. L. (t966) 'Motivational factors in shortterm retention', J. exp. Psychol., 7tt r9o-I93.
wErss, A. D. (tg6S) 'The locus of reaction time change with set, motivation and aBe', J. Geruttol., zo, 6ffi4.
wErss, B. (lgS+) 'The role of proprioceptive feedback in positioning
responses', 1. exp. Psychol,, 47r zt7-224.
wElss, R. L., and STLVERMAN, I. (1966)'Arudety and response strreotypy: an experimental critique', Percep. mot. S&il/s, 22t 91-to4.
wELFoRD, A. T. (rgSr) Skill and Age: An expeimqttal Approach.

O:dord tlniversity Press for the Nuffield Foundation.
wELFoRDe A. T. (1952) 'The "psychological refractory period" and the
timing of high-speed performance - a review and a theory', Brit. J.
Psychol,, 43t 2-rg.

wELFoRDI A. r. (lgS8) Ageing and Human Skill. Oxford University
Press for the Nuffield Foundation.

wELFoRD, A. T. (1959) 'Evidence of a single-channel decision mechanism
limiting performance in a serial reaction task', Qtnrt. 1. exp. Psychol.,

rr, r93-2to.

wELFoRD, A. T. (r96oa) Ergonomics of Automation. D.S.f.R. Problems of
Progress in Industry - 8. London: [I.M.S.O.
wELFoRDI A. r. (r96ob) 'Administrative work and human relations from
the standpoint of an experimental psychologist' (abstract), Ergonomics,
3t 179.
wELFoRD, A. T. (r96oc) 'The measurement of sensory-motor performance: survey and reappraisal of nrelve years' progress', Ergonomics,
3r r8g-23o.
wELFoRD, A. T . (t96za) 'Arousal, channel-capacity and decision',
Naturet r94t 365-366.
wELFoRD, A. T. (r96zb) 'Experimental psychology and the sttrdy of
social behaviour', in Society: Problerns and Methods of Stu.dy (ed. A. T.
Welford, M. Argyle, D. V. Glass and f. N. Morris). London: Routledge
& Kegan Paul, pp. 153-168.
wELFoRDe A. r. (r96zc) 'Changes in the speed of performance with age
and their indusuial significance', Erganontics, * r39-r45.
wBLFoRDI A. T. (r96zd) 'Changes of perforrnance time with age: a
correction and methodological note', Ergonomics, 5, 58r-582.
wELFoRD, A. T. (rg6S) 'Performance, biological mechanisms and age:
a theoretical sketch', in Behaztior, Ageing and the Neraous Systetn (ed.
A. T. Welford and f . E. Birren). Springfeld, Illinois: Charles C.
Thomasr pp. 3-2o,
wELFoRDe A. r. (1966)'Individual capacity and social demands: a new

Referances

39r

look at social psycholory'rin OperatiotwlResearch and the Social Sciences
(ed. f . R. Lawrence). London: Tavistock Publications, pp. 53r -542.
WELFoRD, A. T.e BRowN, RUTH A.r and GABB, f. B. (tgSo)'Two
experiments on fatigue as affecting skilled perforrnance in civilian airc:rew', Brit. 1. Psychol., 4ot t9S-ztt.
wELFoRDe A. T.r NoRRrs, A. fr., and sHocKr N. w.(r963) tMovement,
time and age: a preliminary report' (abstrae.), Ergonomics, 61 3ro.
wEsr, L. J. Gg6il'Vision and kinesthesis in the acquisition of typevrriting skill', j. appl. Psychol., 5r, 16r-166.
wEsrHoFF, I. M. (tg6+)'In search of a measure of percepttral work',
Phil'ips Tech. Reo.,25, 5ffi4.
wESToN, H. c. (rgSg) Vistul fatigttc-tpith special refermce to lighting,

in Symposium on Fatigue (ed. !f. F. Floyd and A. T. \[elford).

London: H. K. Lewis & Co. for the Ergonomics Research Society.
wnrrEr R. w. (1959) 'Motivation reconsidered: the concept of competence', Psychol. Reu., 66, 297-333.

wHrrrENBURG, I. A.l Ross, s., and ANDREWS, T. G. (1956)'Sustained

perceptual efficiency as measured by the Mackrnrorth "clock" test',

Percep. mot. Skills, 6, ro9-rt6.

wTcKELGREN, w. A. (tg6+) 'Size of rehearsal group and short-teun
memory', j. exp. Psychol., 68, 4t3-4t9.
wTcKELGRENI 'w. e. (r965a) 'Acoustic similarity and intnrsion errors in
short-term memory', j. exp. Psychol., 7ot ro2-ro8.
wTcKELGREN, w. A. (rg6Sb) 'Short-terrn memory for phonemically

similar lists', Arner. J. Psychol., 78t 567-574.
wTcKELGREN, vr. e. (r956a) 'Short-term recognition memory for single
letters and phonemic similarity of retroactive interference', Quart. J.
exp. Psychol., r8, SS4z.
wrcKELcRENe 'w. a. (r966b) 'Numerical relations, similarity and short-

term recognition memory for pairs of digits', Bit. J. Pgtchol., 57t
263-274.

wrcKELcREN, w. A. (r966c) 'Phonemic similarity and interference in

short-terrn memory for single letters', j. exp. Psychol., 7r, 39G4o4.
wTcKELGREN, w. A. (r966d) 'Consolidation and retroactive interference
in short-term recognition memory for pitch', j. exy. Pqtchol., 72;
250-|259.

wTcKELGREN, \ry'. e. (r967a)'F-xponential decay and independence from

irrelevant associations in short-term memory for serial ordef , J. exp.
Psychol., 73t t65-t7t.

wTcKELGREN, w. A. (rg6Zb) 'Rehearsal grouping and hierarchical orgailzation of serial position cues in short-tenn memory', Quart. Jf . exp
Psychol., r9t g7-roz.

WIENER, E. L. (lg6S) 'Tables of the ftrnction I - (t - p)Iv' , Percep.
mot. Skills, zr, 887-89r.

392

References

WIENER, E. L., PoocKr G. K., and srEELE, M. (t964) 'Effect of time-

sharing on monitoring performance: simple mental arithmetic as a
loading task', Percep. mot. Skills, r9t 435-Mo.
WIENERT N. (1948) Cybemetics. New York: Technological Press of M.I.T.
and lohn \[iley & Sons.
wILKrNsoN, R. T. (196r)'Comparison of paced, mpaced, irregular and
continuous display in watchkeepiog', Ergonontics, 4, 259-267.

wrLKrNSoN, R. T. (tg6z)'Muscle tension during mental work under
sleep deprivation', j. exp. Psychol., 64, 565-57r.

wrLKrNsoN, R. T. (t964) 'Artificial "sigRals" as an aid to an inspection
task', Ergonontics, 7, 63-72.
wILKrNsoN, R. T. (rg6S) 'Sleep deprivation', in The Phytiology of
Hwnan Suntiztal (ed. O. G. Edholm and A. L. Bacharach). London &
New York: Academic Press, Ch. t4.
WILKTNSoN, R. T. (tg6il 'Evoked response and reaction tirne', in
Attention and Performance (ed. A. F. Sanders). Amsterdam: NorthHolland Publishing Co.

WILLIAMS, A. c., and BRrGGs, G. E. (t962) 'On-target versus offtarget information and the acquisition of tracking skill'r j . etcp. Psychol.,
64, Sr9-SzS.

wrLLrAMs, fr. L.1 KEARNEY, o. F.e and LUBTN, A. (lg6S) 'signal
uncertainty and sleep loss', J. exp. Psychol.r 69, 4ot-4o7.
wrLLrAMS, IUDTTH a. (t966) 'sequential effects in disfunctive reaction
time: implications for decision models', j. exp. Psychol.rt, Tr, 665672.

WINNICK, wrLMA A.e LURIA, f ., and zvKoR, 'w. J. Gg6il .Two

srgnal detection approaches to tachistoscopic recognition', Percep. ffiot.
Skills, z4t 795-8o3.
wooDHEAD, MURTEL M. (1959) 'Effect of a brief loud noise on decisionmaking', j. Acoust Soc. Amer., 3r, t32g-t33r.

woODHEAD, MURIEL M. (rg64a)'Searching a visual display in inter-

mittent noise', j. Somd Vib., r, t57-t6t.
wooDHEAD, MURIEL M. (tg6+b) 'The effect of bursts of noise on an
arithmetic task', Amer. J. Psychol., 7% 627-633.
wooDHEAD, MURTEL M. (1966) 'Varying the number of alternatives in
short-term recall', Brit. J. Psychol., S% 45-52.
wooDRow., Ir. (tgr+) 'The measurement of attention', Psychol. Monogr.,
17, No. 76,

wooDwoRTH, R. s. (t8gg) 'The accuracy of voluntary movement',
Psychol. Rerr). Monogr. Suppl,r 3, No. 3.

wooDwoRTH, R. s. (rgS8) Dynamics of Behaoior. New York: Henry
Holt & Co. London: Methuen.
wooDwoRTH, R. S., and scHLosBERGl rr. (rgS+) Experimental Psychology. New York: Henry Holt & Co. London: Methuen.

WRIGHT, J. M. voN (tgSla) An Experimmtal Study of Hwtan Serial

References

393

Learning. Soc. Sci. Fennica Commentationes Humanorum Littera-

rum, 23, No. r.
wRrcHT, f. M. voN (tgSZb) 'A note on the role of "guidance" in
learning', Brit. J. Psychol., 48, r33-t37.
wyATTr s. (tgzil Rest-pauses in industry. Indusuial Fatigue Research
Board Report No. 42. H.M.S.O., London.

wyATTr S., and FRASER, f. A. (tgz) Studies in repetitive work with
special refermce to rest$auses. Industrial Fatigue Research Board
Report No. 32. H.M.S.O., London.
wyATTe s., and FRAsERT f . A. (1928) The conparctthte effects of aariety
and miformity in work. Indusuial Fatigue Research Board Report
No. 52. ff.M.S.O., London.
wyATTr s., FRASER, f. A., and srocK, F. G. L. (tgzg) The effects of
tnonotony in stork. Indusuial Fatigue Research Board Report No. 56.
ff.M.S.O., London.
wyATT, S.r and LANcDoN, I. N. (t.glz) Inspectim Inocesses in industry.
M.R.C. Industrial Health Research Board Report No. 63. H.M.S.O.,
London.
wyATTr s., and LANGDoN, f . N . (t937) Fatwue and boredotn in repetitiae
ztork. M.R.C. Industrial Health Research Board Report No. 77.
ff.M.S.O., London.
yATEs, A. f. (rg6Sa) 'Delayed auditory feedback and shadowing',
Qtrcrt. 1. exp. Psychol., l.7t tz1-t3t.
yATEs, A. f. (rg6Sb) 'Effects of delayed auditory feedback on morse
transmission by skilled operators', j, exy. Psychol'., 69, 467-475.
vENSENT R. (1965) 'A factor influencing motor overflow', Percep. mot.
Skills,2or 967j68.
vERKES, R. M., and DoDSoN, f . D. (r9o8) 'The relation of suength of
stimulus to rapidity of habit formatioD', J. contp. Neurol. Psychol., r.8,
459-482.
vNTEMAT D. B., Bnd MUEsER, c. E. (rg6o)'Remembering the present
states of a number of variabl€s', j. etcp. Psychol,, 6o,, 18-22,
xNTEMA, D. 8., and MUEsERT c. E. (t962) 'Keeping track of variables
that have few or many states', j. exp. Psychol., 631 39r-395.
vNTEMA, D. 8., and TRAsK, FRANcES p. (rg6E) 'Recall as a search
process', l. verb. Leatrt. uerb. Beh., zt 65-74.
zANcwILLro. L . (t937) 'An investigationof the relationship betrneen the
processes of reproducing and recognizing simple figrrres, with special
reference to Koftka's trace theory', Brit.J. Psychol.r 2Tt z5vz76.
zANGwrLL, o. L. (rgfg) 'Some relations bennreen reproducing and
recognizing prose material', Brit. 1. Psychol., 29t 3To-382.
zANcwrLL, o. r,. (r95O'A note on immediate memory', Quart. j. ezcp.
Psychol., 8, r4c-r43.

ZEAMAN, D.e and DENEGRET f . Gg6il Yariability of irrelevant discriminative stimuli', 1. exp. Prychol., 7lt 574-580.

394

R{uences

zrMNY, c. H. (tg6S) 'Effect of flicker-periodicity and arousal during a

rotarjr-pursuit task', Atner. j. Psychol., 78, 75-82.
zuBEK, f . P. (1964) 'Effects of prolonged sensory and perceptual deprivation', Bnt. med. Bull., zo, (r), 38-42.
ZUcKERMAN, M.e and HABER, M. M. (1965)'Need for stimulation as a
sotrrce of stress response to percepnral isolation'r 7. abfl. Psychol., 70

37r-377.

Appendix

TABLE A. r Crossman's Confusion Fmction
Fa t+a
>a
ti
rr

{

U2

EE

('l

'E

e^
#Eg

!0

o
I

g

-Eru

a

G.

o0

r<

.Io

EE
'tro

G'

o0

o

'99

'or5

68'97

'98
'97
.96
'95

'o2g

343r

22'75

'05g

r6.98

!nu

ul

3oH

o

'49

t.o29
r'o5g

'97o

r.o89

'gr8

.48

'47

.46

I3.5I
tt-20

'844
.82r

r.252

t.286

'7gg
'777

'39

r.322
r'358

'756
'736

'37

l'434

8'r3

'90

7'35
6's8

'40-

.89

.

t68

5'e5
5'42
4'98
4'60
4'27

'r84

'87
.86

.2o1

.8S

'234

.84
'83
.82

'252

.8r
'80
'79
'78

'3o,4

'77

'76
'75
'74

.2r8
'269
.286

3'rr

'34g^

2,94

396

'4t5

'435
'455

'73
'72

'474

.7O

.sr (

.69
.69
.67
-66
.65

'535

.7r

'495

'556

'578
'599

'4t
.38

.36

'35

'34
'33

.32

3'49
3'29

'322

'359
'377

'43

3'98
3'72

.3r
.30
.29
.28
.27
.26

2'78
2'65

2'53
2'4r

.25

.24

2,30
2.20

.23
.22
.21

2.TT

2.02

r'94
t'87

.20
.19

r'73

.17

r.8o

.r8

.6zt

r.6f

t'67

.r6
.r5

'644

r'55
r'50

.14

.6o

'713
'737

r'40.

.II
.ro

'59

'76r

.09

'57
.56
'55

.8r r
'837
'863

r'3r
l'27

.64
.63
.62

'6r
.58

'54

'667

'69o

'786

'889

'53

'gr6

'5r

'97r

.52

.50

'943

r.ooo

.r3

t'45

'12

r'36

.08

.97
.06

l'23

I.19
r.16

.05

r'13
r'o9

.04

r'o3

.or

I.OO

o.o

.03

r.o6

'02

3e6

'945

r.zr8

.42

9'52

Or0

o

r. r84

-r29
'136
'152

.gg

i

00

'44

-92

'9r

I

U

v7

'8gt

'08g

'ro5

I

o

1.12g

'45

'94
'93

cf

b0

o

6a

o

'o44

0{

6l

00

o

6a

b0

o

a{

U A,a,

€86

I

(,

a

HOg

'o74

Er
ql

a{

r.r52

r'396

t'474

.868

'7t6

'697

'678

I.5I5

.66o

r'556
r'599
t'644
t.69o
t'737
t.786
r'837

'643
'6zs

r.889

t'943

2.OOO

2'o59
2.T2o

.608

'592
'576
'560
'544
'529

'5t4

'5@
'486

2.r84

'472
'458

2.252
2.322

'444

2'396
2'47+
2'556
2'644
2'737
z'831
2'943

'4t7

3'o5g
3'r 84
3'322

'327

3'474
3'644

.288

3'837
4'O5g

4'322
4'64/.
5'o5g

'43o
'4o.4

'39r
378

'365

'353
'34o

'3t4
'30r

'274

.26r
'246
'231

'2t5
'r98

5'6U

'177

@

o

6'6M

'15r

A?pndix

3e7

TABLE A.z Nortnal dwiates wrd ordirutes for calatlating d' and B whm
the aariwtces of the distributiotts beiry compared are equal

p
.or
'02
.03

Normal Ordinate
deviate

2.326
2'054

'o27
'o48
.o68

0
.26

'27

.o86

.28
.29

t'645

'ro3

.30

.06
.97

r'555

.I I9

.3r

.08

r'405

.04
.05

.09

.ro

.rr
'12

.r3

.14

.r5
.r6

.17

.r8

.19
.20
.21
.22
.23

.24
.25

r.88r
t'751

'r34
'r49

r'476

.162

r'34r

.32

Normal Ordinate
deviate
'643

'6r3

'583
'553

'337

'524

'348

'496
'468

'353
'358

'4r2

'44o

'362
'371
'374
'378

't76

'35

'385

r.227

.

r88
.2@

.36

'358

t.tz6

.212
'223

.38

r.o8o
r.o36
'994
'954
'9r 5
'878

'842
.8o6
'772

'39

'272
.28o

'44
'45

.I,26

.288

.46
'47
'48

.roo

'3o,4

'7c,6

'332
'305
'279

.40

.3I I

'3r8

'61+

'37

'233
'243
'253
'263

'296

'739

'342

'34

'33

r.282
r'175

'325

'33 r

'3&T

'38 r

'384

'253

'386

.41
.42

.zz8

'389

'43

't76

'49

.50

-2o.2

'15r

'c/75

'050
'o25
.ooo

'39r
'393
'394

396
'397
'398
'398

'3W
'399

To obtain d'add the normal derriates ootresponding to the proportions (p) of

misses and false positives, i.e.

d' : ND for pNOsiv * ND for pfSa'
To obtain F divide the ordinate cornesponding to the proportion (p) of misses

by the ordinate oornesponrling to the proportion (p) of false positives; i.€.
o _ Ordinate for PNOsrv
P-

o

Appmdix

3e8

TABLE A.3 Logrn for numbers from r to roo
n

log, n

n

I

o.ooo

z6

r.ooo
r'585

logrn

n

logrn

4'7o,o

5r
52
53

5

2-322

27
z8
29
3o

6
7

2'585

z.8sT

3r

4'954

8

5.ooo

3.OOO

32
33

9
IO

3'r70

34

3.322

35

3'459
3'585

2
3

4

II
t2
r3

r4
r5
r6
r7
r8
r9
20

2t
22
23

24
25

2'OOO

4'755
+'8rcz

+'8s8

4'W7

54

n

log, n

s'6Tz

76

6.248

5'7ao
5'728
5'755

77
78
79

8o

6'267
6.285

6'3o4
6'3zz

55

5'78r
s'8o7

8r

634o

5'o4

56
57

s'8gl

8z

6'gs8

S8

s'8s8

83

5'o87
5'129

59

5'883

84

5'9o,7

85

36
37

5'r70

6r

5'g3 r

86
87

6'p6

3'7o,o^

E8

63

39

88
89

3'9o7

40

65

5'977
6.ooo
6.ozz

6'+sg

3'8c.7

5'248
5'285
5'322

5'954

9o

6'476
6'4gz

5'358
5'392
5'426
5'459
5'492

66

6'ou

9r

6.5o8

4'Ooo

4r

4'o87

42
43

4'r70

4'248
4'322

4

4'392
4'459

46
47
48
49
5o

4'524
4'585

4'644

45

5'2o,9

5'523
5'555
5'585

6o
6z
64

67

6.o66

68

6.o87

69
7o

5. ro9

7r

6.r5o

72
73

5'6r5

74

S.6M

75

6'rz9

6.r7o
6.r9o

6'zw

6.229

92
93

94
95

g6
97
98

99

IOO

6'lts

6'392
6'4og

6'ul

6'sz4

6'srg
6'sss

6's7o
6's8s
6.600

6.615
6'629
6'6++

Apmdix

399

IABLE A .4 p lrsr$) and p toer} * r)

? a rrs,$) p ror,G + ,) p ,,"r,G) prcr,G+ ,)
'or

.02
.o3

.04

.066

.I I3
'152

'55

'249

.56

'292

'275
'300

'57

'468
'462

'S8

'332

'324
'346

'+s6

'367

'59
'60
.6r

'442
'435

'4c6
'423

.64

.06
.07

'244
'z69

.ro
.II
'12

.r3

.14

.r5

.16
.17

.r8

'19

.20

'21

.22
.23

.24
.25

.26
.27
-zB

.29
.3O

.3r
.32

'33

'3r3

'350
'367
'383
'397

'r53

'lBt

'53

'54

.62
.63

'4rr

'44r

.65

'423
'435

'457

'473

'455

'488
'503

'M5

'464
'473
'48 r
'488

'494

'5r7
'53r

'5M
'556

'569

.7O

'36o

'896

.71
.72

'35r
'34r
'33r
'32r

'go5
'gog

'73

'387

'892

'90r

'500
'505

'58 t

'74
'75
.76

'5 ro

'592
'6o3

'30r

'77

'921

'29o
.280

'925

'5 r8

'624

'514

'6t4

'78
'79

'246
'235

'gu

'64s

'8r
.83

.88

'39
'4(J^

'529

'723

'89
'90

'527

'73r

'526
'524
'521

'5r8
'5r5

'738
'745

'752

,929

'82

'524
'526
'528
'529

'7r5

'gt7

'932
'936

'63s

'699

'z69

'gr3

'zs8

'521

'6s4

.3I I

.80

'707

'49

'869

.888

'530
'530

.50

'844
'849
'8s+
'8sg
'864

'378
'369

'53r

.48

'833
'839

.69
.69

'37

'47

.828

'874

'69o

.46

'42o

'4r2

.8zz

'828
'883

'53 r

'45

'+7

.8 16

'396

'664
'673
.682

.4r
.{2
'43
'44

'4y',9

'7gg
'8o5
.8r r

.66
.67

'530

.38

'491
'+8s
'48o
'474

'4o,4

'35
.36

'34

'495

.188
-22o

.t86

.216

.09

.5r
.52

.05

.08

'o6T

'rI,4

'223

.84

.2ll

.85

't99

.86
.87

't8T
'175
.162

.92

'r50
'r37
,r24
.TII

'93

'og7

'9r

'94
'95
.96

'5r2

'759
'766
'773

'508
'504
'500

'78o
'786
'793

'97
.98

'99

I 'oo

'o84

'qlo

'o57

'o43
'o29

'or4
o.o

'94q-

'947

'g5 r

'954

'957
'96o

'964

'967
'97o

'973
'977
'98o
'983
'986
'g8g

'992
'9p,4

'gn

r.ooo

4oo

Apputdix

rABLE A.5 Probability of response (R) rz paced tash with differtn7 sigflal
rates and Gmowts of storage
,c

\[ithout

Ifith storage
over I stage

Vith storage

\fith storage

'9o,9

'995
'982

'9998
'gg8

'9999
'9995

,7r4

'994

'9992

'934

'986

'997

'992
'982
'966

storage

'I

over 2 stages

'2
'3
'4
'5

'833
'769
'667

'9o.4

'973

.6

'6zs

'87o

'953
'928

'7

.588

'9

'556
'526

.8

'96t

'895
.8oo

't6s

I.O

'500

'73r

I.I

'4',76

'698

1.2

'455

r.3

'435

.666
'636

1.4

'4r7

r.5
r.6
r.7
r.8
r.9

'400

'6o7
'580

'385

'555

'897

'862
'824
'784
'745
'707
'67o

over 3 stages

'g4l

'9A9

'87o
.826

'78r
'737

'63s

'694
,654

.6ci2

'53r

'6t7

'37o-

'572

'357
'345

'509
'488

'5M

2.O

'333

'468

'5 r8

'683
'553

2.1

,323

'459-

'472

2.2
2.3
2,4
2.5

'3r3

2.6
2.7
2.8
2.9

3'o

'432

'494
'451
'432

'524
'499
'475
'454
'434

'303

'4r7

'294
.286

'4o.2

'4t5

'387

'400

'278
'270

'36r

'399
'384

369

'37o.

'338

'3M

'328

'3330

'345
'3333

'374

'350

'263
'256
'250

'356

'4t6

'385
'357

Use of this table.

r. TofrndRwhentime peritemrrsignal frequency aqd ) are known:-the ploqgct
of r and signal frequency gives i against which R may be read direcdy. For
example, if t : 'Siec andsignal frequency : 4 per s99r-t : ), and R with
storage over one stage (l : ,) is .4681 i.e. iesponies will be made to '468 of
the signals presented, ind the remairurlg_'s3z 1vill be miss-ed.
z. To fiid r dhen signal frequency, I and?-are known: find R *-q^e appropriate I column; ilivide th-e * corresponding to this by the signal frequency
to obtain ,.
To find signal frequency when r, ) and R are known: find R irl thg appro;
r3. priate
tr cofumn; divide-the * corresponding to this by r to obtain the signal
frequency.

be infeqred providgd values of R are obtainable
4. \[hen )t is not known it may
over a fairly wide range of i. At high anil very low_ vdues of t, R-is Practically indepehdent of )r- provided atJeas_t one signal can be stored.- If signd

frequen&, t and R can be calcrrlated for one or both extremes, the I may
be ihoseir- which gives the best fit for intennediate values.
For values not given in the table interpolation _is accurate enough for most
purposes. V[en - 2 or.more, R fofvalues of fr above 3'o can be calsulated
approximately as^R -

t/*.

Appetdix

40r

THE DERMTION OF TABLE A.5

The model on which Table A.4 is based assumes (i) a single channel
which, whenever a response is made, is ocorpied for a fixed length of
"ne / during which it cannot deal with other signals, snd (ii) that in
certain circumstances signals arriving dtrring the 'nte the channel is
busy can be held in some form of short-term storage until the channel
is free.

The case when there is no storage

Let us consider first the case when no signals can be stored. This is
shown diagrammatically in Figrre A(r). The first signal (Sr) initiates
the first response which occupies the single channel for a time interval tr.
.S, occurs dtuing this interval and is missed. At the end of /r, the subiect
has to wait dtrring an intervd w, until a further sigual, S, arrives. He
then initiates a second response, and so on. Every response \pill be
followed by a waiting time and thus the subiect's whole time can be
regarded as divided into two categories t and w and the total
"ne
over the task $rill thus be
(a + zlJ * Ur+ urz).. . + (t"+ wn)

and since q - tz : t., the proportion of time spent making responses
g) may be expressed:

T6:0): t+tfr
+

(Ar)

where i is the number of signals that can be stored and sf is the mean sr,.

All signals occturing during the riu e spent making responses will be
missed.

The proportion of signals responded to (R) may be calculated thus:

&r:o) -

rtSo'
x

(A2)

when f, is the average ntrmber of signals fafling within a period ,.

The clrse when one signal can be stored
An example of the case when one signal, but not more than one, can be
stored is shown in Figure A(z). S, initiates the first response and
ocorpies the channel for a time f,, during which .Ss occurs. Because of
the storage, & is raponded to in a time /r, which follows immediately
after the end of /r rrithout any waiting time. S, and Sn occur drrring t,

Figtre A

sl

S2

I

I

t

Sa

I

I

I

+

wt

tl
S2

S1

I

S5

wz

tz

I

\
tl

I

wt

t3

t2

I

l

3a

:'
3b

t4

S-5

s3 s4

s2

I'

t3

Ss

I

I

lI

Sg So

I

2

S3

\

t3

t1

t2

s2

s3 s4 s5

ta

t5

t4

t5

It

I

\

t3

t2

tl

ss

sr sz% so

lrr

I

\

4a

t3

t4

t3

ta

t5

Ss Sz Sa

Sg

Sro

t2

tl
Sr Sz Sg

S+ Ss

tl

ttl

4b

t2

t1
Sr SeSgS+ Ss

4c
( neLF scALE)

SIGNALS IN
STORE

o

tl t2 t3 % ts
3

3

2

4o.2

2

t6

3

t7 t8 ts

trc

1 l'

0

2 2

Ap?endix

4oj

and of these one (say Sr) is missed, and one (say S) is stored to be
responded to during ts, again without any waiting time intervening. No
ftrrther signal occurs dtrring /, so that a waiting time intervenes benreen
the end of /r and the occturence of Su which initiates t o. It is clear that,
with storage, less waiting times will occtr and the proportion of signals
responded to will be higher.
To calculate the proportion of responses when there is storage we
need to calculate fust the average length of chain of times t which
occur without an intervsning waiting time. When one signal can be
stored, /2 will follow /, without a waiting time whenever one or more
signals occur during ty Let us call the probability of this p(x,2 t)
where q is the ntunber of signals occurring during /r. The probability
of /r following t2 again without a wai 'ng time will be the probability
of one or more signals occurring both in /, and t2, i.e.

?(q2 \ x22 r)
The probability of further responsm following without a waiting time
may be calculated in a similar manner. The total length of chain will
thus be:

r * ?(q2 t) * p@r ) r, )cz2 r) . . . p(xr2 t . . . xn 2 t)
We will for futtrre convenience denote this tlpe of summation by the
symbol ,p; followed by the basic term, in this case (x 2 r). We will
also for future convenience write a for (x 2 r) so that we may denote
the whole suulmation by ,p(o).
We can think of the subject's time as being taken up with chains of
responses of an average length r x mean V@) each followed by a waiting
time eD. Putting A for mean ,l@) the proportion of time during which
the single channel is busy may, by analogy with Eq. Ar, be written
T1t-r)

: fi+a
='4

--

(Ar)

and by analogy with Eq. PLz the proportion of signals responded to
may be written

Rtr:r): ry
,c

(A4)

The case when two signals can be stored
When two signals can be stored the average length of the chains of
responses which follow one another without an intervening waiting
time will be increased. This is because responses can be made to both

4o4

Apperdix

of nn o siguals which occur during an interval / which is followed by an
interval t during which no signals occur. An example is shown in
Figrrre AGa). .S, initiates /1, and & is stored until /2e 8s in the case with
one stage of storage. During t* two further signals occur: Sg is
responded to during /r, dnring which no further signals occur. .S4 is
still in store, however, and is responded to during /a, which follows
immediately upon /r. The whole cycle may be repeated without a
waiting time intervening if ftrther signals occur during /n.
In Figure A$a) the blank interval /, is shown as following immediately upon the interval tz during which two signals occur. The
example in Figrre A(fb) shows that this need not necessarily be the
case and that the second signal during /, will not be missed if one or
more intervals with one sigual intervene. Thus in Figrrre A$b), tshas
one slgnal which can be carried over the blank interval tnuntil /u.

This means that in the srunmation for determining the average length

of chain two possibilities have to be considered: firstly, as beforet ai
secondly, and additionalln the probability that two or more siguals will
occur and that a blank interval will follow immediately or after one or
more intervals in which one signal occurs. In the same notation as
before we may write this possibility
p(x2 z).V(x t).p(x - o)
Let us call this expression b. Then the whole g summation will be

v@+b)
and if we call the mean for this expression B, we can by analogy with

Eq. A3 write

'Jq zfr
T1t:z): tB
+

(AS)

R1,r:z\:ry

(A6)

--

and
tc

The case when three signals can be stored

Vhen three siguals can be stored, the storage capacity may be filled in
one of nro ways: either when an interval / occurs during which there
are three signals or if rwo intervals occur with two signals and no blank
interval benreen. fn either case stored responses will fill two subsequent
blank intervals. Figure A(+) illustrates these possibilities. In Figrrre
A(+a) three signals occur during rr. Two of these are responded to dtrring
two subsequent blank intervals r, and rr, the third during rn. In Figure

A?Mdix

4os

A(+b) two signals occru in each of r, and r, being worked off during the
subsequent blank intervals /r, /n and /u. We should note that any number
of intervals with one signal could be inserted between any of the intervals in Figrre A(+a) and A(+b) without overloading the storage, and
after the first blank interval the store \pill not be overloaded if an
interval with two signals occurs. A possible more complex sequence of
this kind is shown in Figrre A(c) together with the numbers of signals
in store at the end of each period /.
In the same notation as before we may thus write these two possibilities

p(x > i.V(x - r) .p(x - o) .V(x
and
ptc(

- r)

x ylp@ - z).y\* - r) .p(x - o) .V(x - t)lp(* - o)

- t) .P(x - o) .V(x - l)
- z).V(x - t).p(x 2 z).rdx
x VIP@ - z).V(x - t).p(x - o) .yt(x : t)lp@ - o)

Summing and simplifying we obtain

lp(x> ) * p(x - z).p(x2 z).V(x - r)l,p'(x - r).p'(x - o)

x ylp@ - z).rp'(x - t).?(x - o)I

Let us call this expression c. The whole g summation thus becomes

v@+b*c)
and if we call the mean for this expression C, then by analogJ with
Eq. A5 we may write

T1t:B):tc+fi
='9 =

(Az)

and

R1,r:s):

b7
,c

(As)

Expressions for storage of more than three signals are cumbersome

but similar in principle to the foregoing.
APPLICATION OF THE MODEL TO THE RANDOM CASE
We shall now consider the model as applied to the case where the
signals arrive at strictly random intervals.

No storage
Dnring each interval (, + tfr) one signal will be responded to and fr
gr

4oG

Appendix

will be missed. The duration of rJ is therefore the average interval
during which one signal ocsanrs i.e.

a):'=
x

(A9)

For the proportion of time spent responding, Eq. Ar may be rewritten
I1r:01

t+t/fr
fr

f+r

(Aro)

and for the proportion of signals responded to Eq. Az becomes
R1r:o;

- -rf+r

(Arr)

Vith storage
The calculation of fi enables Eqs. A4, A6 and A8 to be rewritten
respectively
R1,l:1) :

,4,
frA+r

(Arz)

R1,1,,=2)

: frB*r
=-1

(Arr)

R1r:B)

: *C*r
=^C-

(Ar+)

One stage of storage: calculation of A
Since in the random case p(xr) : ?(xz) . . . : p(x*),

A - r * p(x 2 t) * pr(x 2 r). . . + p"(x 2 r)
or approximately (exactJy rf n

- o):
A-I I

-a

(Ars)

Substinrting accordirg to the expression for a Poisson distribution, i.e.

*ifr

P - *l'*
we obtain

A-#:sd

(Ar6)

Appendbc

4o.7

Two stages of storage: calctrlation of B
By analogy with Eq. Ar5 we may write

R-

r

r-(a*b)

(Arz)

Substinrting as before we obtain after reduction

B-e@-*)

(Ar8)

Three stages of storage: calculation of C
By analogy with Eq. Arj we may write

c-#

(Arg)

Substinrting and reducing as before we obtain

C

- eK* - *), - +frrl

(Azo)

Table A3 gves R1,l:oy, R1,l:r;1 R1,l:21 and R1;:ay for various values of

f. By means of this uble it is possible, knowing any three of the quan-

tities, signal frequency, t, R and 1", to calculate the fotrth. Strialy
speaking, it only applies to the case where signals occur at random
intervals and the task is of infinite length, but when storage is present
moderate departures from randomness seem to make very little difference, and the inaccuracy due to asstrming infinite length is small
with tasks of moderate duration. A number of trials were made using
a random signal generator with an electronic recording machine* acting
as 'subject'. The recorder could be set to respond \rith a constant t
and with o, r or 2 stages of storage. Recorded results for a range of
signal frequencies from below * - .5 to about * - 3.o showed agreement with Table A5 to within z or 3o/o. The recorded figures tended
to be a little too high, probably owing to the signal generator producing
a series which was not quite random, having itself a short resolving
time which would cause signals at the very shortest intervals to be
run together and thus treated by the recorder as one.

The effect of variability in r

If t is not fixed but has appreciable variance the foregoing model will
require modification. Firstly the mean probabilities of different frequencies of r will be affeaed since f will vary linearly with ,. Consider

* N. T. ITelford (tgSz) 'An elecuonic digrtal recorcling machine - the

SETAR' , 1. sci. Instruln.e 29t r-4.

4o8

Appmdix

the probability of obtaining $ : o in the random case. Since when

x-or/l(x-o):s-0

mean fr - mean (- log I,(x: o))

(Azt)

so that the arithmetic mean of f wifl corresp,ond to the geometric mean
of p(x - o). Since p(x - o) is less than r, mean p(x o) will{be
greater than if r had no variance. As the probability of a chain ending
increases with p(x: o), the mean length of chain will decrease.
Secondly the product of r and length ol'chain will tend to be increased. Let us consider the qrse when there is one stage of storage.
In place of tAin,Eq. A3 we have to use mean (t'.rp@)) where r'is the
mean r for any partictrlar chain V(a). Ve qurnot usually calculate this
since we do not know the individual ,'s and V(a)s, We can however
estimate rough limits since

(lez)
- iA + ctr .6v(ol .rt. v(o)
where i is the overall mean t, ot is the' standard deviation of t'
(: ot/\/@, cv--t is the standard deviation of y(a) and rt,.v@) is the
mean (t' .rlr(o))

product-moment correlation between ,' ancl V@). Mean (t' .rp(o) will

thus vary benreen tA when either 6{t oakD or rr.v@) - o, and
iA * ay.Oy@ywhen ot, 8Dd 6v@)are greater than o and f{.v@l: *r.
Since rt'.,t@) is likely to be positive, mean (t' .y,(a)) is likely to be greater

tfuntA.

I

if

The overall result of these effects on anrd thus on R can be seen
we express r and mean ,l@) as proportions ,cf , and A n Eq. A3 - say
t t/t nd mean rp(a) A/q. Eq. A3 then becomes

-

,v :t):W t/t.A/n

t(1

_

,A

(Azf)

tA + rxri

The overall effea will therefore diminish as tA nses relatively to tfr. In
other words, it will become relatively small:r as f and .1, increase.

Author Index

Abben D. S., r9o, 337
Acker, Mary, 2o3r 37o
Adams, f. A., rr4, rt7, 136, 2oz,
276, z8o, zBt, 289, 293, 3er6,
337r 3M, 379

Adiseshiah, !7., 237, 332

Adrian, E. D., 40, 337

Aiken, L. R., @,337
Akiba, N., 374

Bakan, P.r 274t 3t9t 339
Baker, C. H., l8t, 275, 276, 3o2,
339

Baker, Katherine 8., 3rr, 33g
Baker, R. A., 34or 383
Barmack, f. 8., z48t 34o
Barnett, T. J., loSr rr3r 354

Barry, E., 48, 34o

Allard, M., 3Sz
Allison, R. S., 98, 362
Allport, G. !7., 3241 337
Alluisi, E. A., 278, 28rr 3o3, 337,
3Mr 347
Ammons, R. B.r 26c,1 337
Anderson, Nancy S., tZ7, 2o8,
337, 355
Andreassi, J. L,2671 33Tr 338
Andrews, T. G., 39r
Ankus, Mary N., 2o7r 362
Annett, f., rr2, r43, r52r 299, 3o2t
306, 3O7, 3r2r 338

Antrobusr f. 5.r 274, 338
Arbuckle, T. Y., 2o8, 382
Archer, E. f., 306, 338
Archer, !7., 372
Argyle, M., 33o, 338

Banlett, F. C., r8, t6l, t6z, tT3,

2r4, 237, 249, 257, 27rt 294,
3O2t 3161 34o

Bartle5 S. ff., 24rr z49r 34o

Bartz, \f. ff., 23o, 34o
Barzeele, f., 7ot 3p
Bassett, M. F., 258, 34o
Bateman, Doro*ry E., 366
Batten, D. E., 3ro, 34o

Beardshall, Ann, 353
Becker, P. !f., 3o4r 34o
Bedford, T., z8zr 388
Begbie, G. H, t3T, 34o

Beilin, H., r75 t 34o
BeishoD, R. 1., t92, 196, 34o

Bekesg G. von t 79, 34o
Bekker, J. A. M., 49, 57, 66, 79,
ro8, ro9, rr4, 3661 38r
Belbin, Eunice, 2oo, 288, 2891 29r,
292, 295, 3OO, 3Or, 34Ot 34r,

Artunl€I, S., 3r8, 338

Ash, I. E., 245, 2461 338
Atkinson, R. C., 32r 338

348

Attneave, F., 44, 163, t64, 165,
r74, r75t r77, r78r 33r, 333

Audley, R. f., 58, 338, 389

Avant, L. L., 342

Averbach, E., 2291 339
Avery, D. L., 35o
Baddeley, A. D., t97t z2o, 224,
225, 22g, 276, 3Orr 33g, 348
Bahrick, fI. P., r33, r39, t4or 339t

3U

4q

Belbin, R. M., 34r
Bell, C. R., z9tr 34l
Beller, Ef. K., 9or 374
Bennett, \V'. F., 339
Benson, A. l.r z47r 34r

Berger, C., 2491 279, 34t
Bergstrorn, f. A., 2o3, 34r
Bergum, B. O, 275, 276, 341

Berkley, V. f., 359
Berllme, D. E.rz64t277t 3ro, 3zl,
34r
Bernardelli, Betty M., 22

Indcr

4ro
BernsteiD, Ira ff., 384

Berry, K. K., 353
Berry, R. N., 3o9, 341
Bertelson, P., 70, 80, 8r, I t7, l,2o,
t23, r25, t28, 255, 257t 285t
34rr 342, 343
Bessel, F. 'W'., r r

Bevan, \tr., Tor 2751 342t 359
Biederman, I., 9r, 355

Bills, A. G., 248, 255, z6t, 263,
2851 3o8, 342

Bilodeau, E. A.,2oo,286,3o2r 3o3r
3o4t 3o5, 342, 343
Bilodeau, fna McD., 2181 2961 3o2,
3o3, 3O4, 306, 3r2t 342: 343t
38r
Binder, A., 95r 343

Binding, F. R. S., 369
Binet, 4, 2161 343

Binford, f. R., 278, 3431 368
Birdsall, T. G., 3r, 45r 358., 385
Birmingham, EI. P., r4r 386
Birren, f. E., 5t, 54, 57, 343
Blackbum, J. M., 3t4r 343

Blake, M. f. F., 3t9r 343
Blancheteau, M., 95, ro3, 356
Blankenship, A. B., r98, 343
Blick, K. A., 3o5r 343
Blyth, K. \[., 8o, 343

Boder, D. P., 2631 343

Book, 17. F., rt, tgz,343
Boons, J. P., 7or 342t 343

Borger, R., ro8, r 13, r25t 343
Bornemann, 8., r 3z, 343
Borsa, Donna M., 34r

Boswell, J. f., zoot 343
Boturinick, J., 5r, 54, 57, 59, 69,
79' 259, 343' 344
Boulter, L. R., z76r z8tr 337t 3M
Braden, Ina, 363

Bradley, f. !7., r89, 34
Brainerd, R. !7., 82r 86, 3M
Branks, J., 2631 3S3
Bray, C. W., r84, 3M
Brebner, f., 89, 344
Brichcin, M., t37r 3M
Bricker, P. D., 67, t77,344
Briggs, G. E., T4o, 29Tt 306, 344,
3741 392

Brinley, J. F., 69,3M

Broadbent, D. E., 34, 47, 70, 88,

98, ror, rr5, T.2ot 122, r28,
r33, 2O2, 2c,6, 227, 23O, 231,
238, 256, 274, 275, 276, 277t
278, 344, 3451 358

Broadbent, L., 358
Broadhurst, P. L.r 27tt 345
Brooks, Virginia, 166, 36r

Brown, C., 2481 342
Brown, C. H., 3 rot 345
Brown, I. D., 67182, t33r 2761 3o8,
345' 346, 357
Brown, f., r98, 2oor 2o6t 2@' 2to,
346

Brown, L. T., 38o

Brown, Ruth A., r3o, r35, r85,
258, z6o, 296, 322, 3461 39o

Brown, R. L., r 33, 346

Brozek, f., 25o, 25r, 382

Bruce, D. f., ro3, 346
Bruner, f. S., 96, 377
Bryan, Iudith F., 3o8, 368
Bryan, !7. L., rr, tg2,346
Buck, L.r 2741 276, 346
Burke, D., r39, 346
Burns, f. T., 116

Bnrsill, A. E.r 2S3,346

Caird, \U7. K., 2.,6' 2o7r 362
Callantine, M. F., 289,346
Callaway, E., 82, 385
Cameron, C. G., 9r, 961 346
Cameron, N., 388
Cardozo, B. L., r98 t 224t 346
Carmichael, L., ro3, T74, 346
Carpenter, A., 24 4e,,3461 347r 161
Carterette, E. C., 48, 347, 354
Carvellas, T., 9or 364
Cason, H., 3o5, 387
Castaneda, A,r zTrt 347

Cattell, J. M. t 27t 347
Cattell, R. 8., 3 T7, 347

Chambers, E. G., z8,zr 347
Chambers, R. I(/., Tr7r 136, $7

Champion, R. A., 3ro, 357
Chaney, F. B., 288, 347
Chapanis, A., 372
Chapman, !., 369

Chase, R. A., rro, r38, 347
Cheatharn, P. G., rr3, 347

4tr

Index
Chernikoff, R., 38r
Cherry, C., 99, rr5, 347
Chinn, R. McC.,278., 3o3, 347
Chocholle, R., 58r 342

Chown, Sheila M., 2951 3r8, 342,
348

Christ, R. 8., 38r
Christie, L. S., 76, 94, 348
Chute, Eloise, 24rr 249t 34o
Claridge, G. S., 3r9, 348, 35r
Clark, H. f., 2281 348
Clark, M. L., 378
Clay, Hilary M., 22, z31t 29ot 348
Cohen, B. H.r 226., 348
Cohen, L, H., 268, 348
Cole, M., 48r 347

Collins, !7., 376
Colqutroun, !7. P., 274, 276, zT8,
zBt, 3r9r 348
Connor, Minna 8., r44,363
Conrad, R., 85, ro4r r3or r3r, r35,

tgg, 2o2, 2o3, 2o4, 2c,6, 22O,
224, 225, 22g, 339, 348, 349
Cook, J. S., 372
Cook, T. S7., 3rr, 349
Corballis, M. C.r 2a4, 2161 349
Corcoran, D. 'W'. 1., 44, 3r9, 348,
3491 35o

Costa, I. D., 6z1 35o
Coules, f., 3r2t 35o
Courts, F. A, 263, 3a9r 35o
Crafts, L. Itr., 3o5, 35o

Craik, F. I. M., 216,35o

Craik, K. J. I7., 13, 15, 24, ro5,

rt4, r,26, rg4, 2521 35O
Crannell, C. !7., 2241 35o
C,raw, Margaret A., 341
Crawford, A., z6t, 284,35o
Crawford, Junq 2ro,35o
Crawley, J. 8, t92, 1961 34o

Creamer, L. R., tz5, 35o
Crossman, E. R. F. !7., l.4, rZ,29,
30, 48, 50, 5rr 53r 55, 57, 63,
64, 65, 68, 7r, 75, 82, 9c,, 93t
r3o, r4o, r&t r43, r44, 146,

r47, r48, r49, r5O, r52, r92t
196, 2t8, 22o, 223, 225, 226,
29rr 3r3, 3t4r 35Or 35rr 396

Crowder, R. G.r 2e,6., 35r
Cullen, f. K., 342

Dagnall, P. R., r49, 366
Dale, ff. C. A., r99, 2t9, 2zo, 229,
3or, 35r
Dallett, K. M., T99r 35r
Daniel, R. S., 356
Dargus, ff. D., r37r 35r
Dardano, f. F., z8o, 35r
Darrow, C. S7., 2691 35r

Davidoff, M. D., 385

Davies, D. R., 3r9, 35r
Davies, M. fI., 3r9, 32or 35r

Davis, D. R.e 252, 254, 255, 276,
z9r, 327,35r, 3Sz

Davis, J. EI., 33or 379
Davis, R., 85, 86, 9o, ro8, ro9, trz,

3or, 352

Davis, R. C., 79t 352
Davis, S. !tr., 25t, 352

Deane, G. 8., 2To, 352

Decker, L. R., 34, 377
Dees, Valerie, 3c2t 352, 37o
Deese, I.r zr3t 225:275, 276, 352
Deining€rr R. L., r86, 355

Denegre, J., 961 393
Denenberg, V. ff., 27tr 3Sz
Denny, J. P., 96, 32t, 352, 384

Denny, M. R., 3o3, 352

Denny-Browne D., 246, 352
Derosa, D. Y., 372

Deutsch, D., To2,352
Deutsch, f. A., ro2, 352
Diesenhaus, FI. I., 32rr 36t
Diikstra, Sanne, zoz, 337
DiU, D. B., 353

Dimond, S. J., r33, 136, 3Sz

Dinnerstein, A. f., r7r, 359
Dixon, N. F., z49r 3S9
DodsoD, f. D., z7or 393
Doherty, M. E., 30, 3Sz
Donders, F. C., 60, 88, 353
Dooley, R. P., 3 rzt 353
Doorne, ff. van, t34r 37r
Douglas, C. G.r 2431 353
Downing, L., 361 376
Downs, Sylvia M., 2951 34\ 348
Dozier, F., 366
Drazin, D. H., 70, t23t r24r 3S3

Drew, G. C., 252

Dudley, N. A., 283,353

Duffy, 8.,264, 353

Index

4t2

Duncan, C. P.,2891 353
Dyal, f . A., 27rr 3a2r 3o3, 3o4,3o7,
353, l8+
Eagle, M. N., 3or, 353
Eade, A. 8., 14
Eason, R. G., 262, 263, z8o, 3o8,

353' 382

Easterby, R. S., r 6r, 353, 383
Eayrs, J. T., 3r8, 353
Eccles, f. C., 272t 353

Edwards, 8., t6zr 353
Edwards, Ff. T., 243, 353
Edwards, !7'., 36, 353
Egan, f. P., 47r 99, 354

Elithom, A., ro8, rr3, tr4, rz2)
354, 359t 365
Elliot, P. B., 461 354

Elliott, R.r z7r, 354

Ellis, D. S., 2761 3o8, 36o
Ellis, K., ro3, 36r
El-Temamy, M. A. 4, r44,354
Emmerich, D. S., 23o, 234, 354,
379

Engen, T., 40, 368

Entwisle, D. G., r43, 374
Eriksen, C. \[., 30, 44, zoz) 229,

23r, 3541 384

Estes, !7. K., 3or, 354
Ettema, f. H., r35, 364
Evans, S. Ef., r75, 354
Eysenck, H. f., 3r8, 354, 355

Farber, I. E., 384

Feallock, f. B., 368

Fechner, G. T.r z8r 29
Feehrer, C. E., 89, 324
Fennell, Eileen, 34o
Fernberger, S. !tr., 37o
Ficks, L., 43r 377
Fine, B. f ., 3r9r 355
Fisher, R. A., 33, r48, 355
Fitts, P. M., 661 72, 74r 82r 84, go,

g5, r43t r44, r45, 146, r4g,
r5o, r5rr r53, t54, 16o, 176,
r83, 186, 339, 3M,355, 374

Fleishman, E. A., 286, 2891 355

Floyd, !7. F., 24rr 3SS

Forbes, S. M., 386
Foreman, Sally, 3o3, 358

Forrest, D. I7., 248, 27or 355
Forrin, B., 88, 93, 355, 372
Foster, Harriet, 90, 356

Fowler, B., 7or 356
Fowler, F., 374
Fox, R,, 366
Fox, R. fI., z8t, 356

Fraisse, P., gS, roz., ro8, r ro, 2o3,

2r3r 356

Fraser, D. C., 2oz,2o3, 275, 356
Fraser, f. A., z\z, 283, 327, 392t
393

Freedman, N. L., 27o, 3261 3j6
Freedman, S. f ., r 84, 363, 364
Freeman, G. L., 262, 264, 266, 26Z,

27o, 356

Freeman, P. R., 349

Frick, F. C., 43r 365

Friedman, M. I., 293, 356
Friedman, S. M., t8o, 359
Furneaux, I7. D., 3r9r 356
Gabb, f. 8., 258, z6or 296, 3go
Gagne, R. M., 339

Galanter, E., 37r
Galloway, \U7. D.r 346
Garner, S7. R., 66, r75, 237, 33rr
356

Garvey, I7. D., r33, t8r, r83, 186,
r9r, 356, 357, 3661 368, 386
Gates, A. I., zoo,294, 357
Craudry, 8., 3ro, 357
Gedye, f . L., 247, 34r
Geer, J. P. van der, r7S, 3SZ

Geffen, Gina, ror, 387

Gibbs, C. 8., r39, r44, 276, 3o8,

3rrr 3461 357

Gibson, f. f., t69, \7ot r72, 2rZ,
357

Gilbert, R. T[., 3o5, 35o

Gilden, L., 347r 35o
Glickman, S. 8., r97r 3S7
Glucksberg, S., 234 t 357
Goggin, Judith, 29zt 3or, 3TT
Crolby, C. Str., 338
Crold, Ceciller ro3r 388
Goldenbarun, D. M., 354

Goldman, ]osephine, 357
Goldman-Eisler, Friedat t27r 357,
36o

Indcx
Croldsmith, R., 356
Croldstein, D. A., 3t2, 357

Goldstein, I. L., 363
Goldstein, M., 3o7t 357
Croodeve, P. f., r4o t t42t r43, t44t
r5o, 35r
Gordon, I., 89, 344

413

Harvey, S., 347

[Iaslerud, G. M., ror, 389

HautS G. T., 2721 3Sg
Hayden, D. L.r 354
Hayes, \tr. ff., 38o

Hayrood, H. C.r 3zor 359
Hebb, D. O., r97, 264, 265, 272,

Gottsdanker, R., r 19, 358
Crould, J., r84, 358
Crould, J. D., r38, 358
Crrandiean, E., 25o, z5r, 358

Heise, G. A., 37t

57, 358
Greenbergr G. Z.r 3S4
Greenspoon, f., 3o3, 358, 38o
Gregory, Margare1 34, 47, 7c,,98,

Henderson, A., t27, 360
Henderson, f. G., r4r 360

Gray, J. A., 23or 358
Green, D. M., 3r, 45, 47, 49t 55,

ror, tzo, t22, 2tg, 229, 23O,

23r, 275, 278, 3451 35r
Gregory, R. L., 29r 358., 38o

Griew, S., 14, 671 8zr 2rz, 3221 358
Grim, P. F., 2641 35g

Grindley, G. C.r 3o2, 352, 37o
Grossb€rg, M., 59, 378
Grossmsrr, f., 362

Gruber, Ff. E., r7tr 3S9

Guilfoyle, G., 347
Gullahorn, f. E., 33o, 3Sg

Gullahorn, J. T., 33o, 3Sg

Haber, M. M. t 327, 3g3
Haber, R. N., 48r 359
Hafer, Brenna M., 356
Hake, FI. V., 3e,, 44r 66, 354, 356,
38o

Hale, D. J., 8o, 359

Hall, 8.r 352
HalI, T. J., z8tr 337
Halliday, A. M., ro8, r 13, 3Sg
Hamacher, Iane H., 34r

Hammerton, M., t37, t4r 3Sg
Hampton, f. F. G., 356
Harcourt, R. A. F., 333, 3Sg
Harcum, E. R., r8o, 359
Hardesty, D. L.,276, 342, 3Sg

Harleg V., 383

Harris, C. S., 32or 3Sg
HarrisoD, G. 1., 224, 359
Harter, N.e rt, rg2, 346
Hartline, fI. K., 79, 3Sg

277, Z8O, 287r 327t 360

Hebron, Miriam 8.r 324., 360
Hecht, Elizabeth, z8o, 365

Helrnholtz, f{. von 28
Helmstadter, G. C., 2761 3o8, 36o
Helson, ff., 43t tt9, 36o
Hemingway, Aq 2431 360
HenrroD, V. A. C., 39r 47r 48, 50,
5r
Henneulan, R. ff., 368, 3Zg

Henri, Y., 2r6t 343
Henry, F. M., t37r 36o
Henshaw, E. M., 2941 360
Fleron, A., r33 , 2or, 238, 3451 360
Herrington, R. N., 3 t9, 348

Hershenson, M., 48r 359
Hick, S7. E., 6tr 63, 64, 65, 661 7t,
72, 73, 76, 78, 82, 84r 86, 9o,

ro8, rrr, rr3, 116, 16o, t6z,
36e,,36r

Hilgendorf, Linden, 67, 87, 36r
HiU, F., 34r

Hille, Barbart A., 2o4r 34g
Hillix, \[. A., t7S, z95r 36r

Hiorns, R. \[., 34r
Hirsh, I. J., 99r 36r
Hochberg, f., 165, t66, r7r, 36r
Hockey, G. R. J., 3r9r 35r
Hodge, M. ff., 4r, 43r 36t
Hoffinan, Linda S., 354
Hogan, ff. P.r 346
Hogan, J. A., 213,36t

Hogan, M. f., 3 tg, 36r
Hohle, R. Ef., 35, 36r

Holding, D. fI., z99r 3rrr 3611 3To
Holman, P., 2941 36o
Eforn, Rheba, t75, 34o
Hovland, C. 1., z89r 37z
Howard, K. I.r 32r, 36r
Howard, T. C., 36, 36t

Index

414

Howarth, C. I., lo3, 36r
ffowe, M. J. A.r 2o4r 2o8r 362
Howell, Itr. C., 66, 362,363
Howland, D., r39,362
Hubel, D. EI., t6z, 362
Hughes, I. M., 50, 55, 362
Hulbert, S. F., 27o, 362
Hulicka, Irene M., 29o, 362

Hull, A. f., 2o4r 22or 2241 3391 349

Hull, C. L., 324, 326, 362
Humes, J. M., 337
fftrnt, D. P., 3o5, 362
Flunt, 8., 35o
Hurwitz, L, f ., 98, 362
H5mran, R., 65, 671 73r 82r 861 87,
362

Iida, II., 751 372
Inglis, 1., zo6t 2o7t 292, 362
Irby, T. S., 3M
Irion, A. L., z6or 2861 2921 362t 369
Irwin, I. McQ.r 3oo, 37r

Kaplan, Ira T., 90, 3641 37o
Kaplan, S., 3ror 365
Kappauf, !tr. E., 2761 364
Karas, G. G., 27rt 352

Karlin, f. E., 3T3r 3T6
Karlin, L., 3t, 69, 79, 84, 85, tz3,
306, 364
Karsh, R., 357, 37rt 372

Kassum, D. A.r 2r4t 2t9r 364
Kaswan, f. V., 58, 3o3, 3641 374
Katz, D., r73, 2481 364
Katz, L., 36, 364
Katz, M. S., 3o5, 364

Kaufman, FI., 87, 3641 366
I(aufman, R. A.r 2r3, 225, 352
Kay, fI., r24, r84, r87, zo8, 2o9,
2rr, 2t2) 23O, 232, 234, 259,
296, 297, 298, 3o7t 338, 364,
365

Kearney, O. F., 392

Jaffee, S., 353
Jakobovits, L. A., 2581 366

Keller, F. S., 3o4, 365
Kendor, A., 33o, 365
Kent, G. \f., 338
Keppel, G., 377
Kerr, M., ro8, r 13, tr4, 359, 365
Ketchel, Rhoda,384
Kimotsuki, K., r37, 365
King, Pearl ff. M., 288, 365

Jarrard, L. 8,r 245, 363

Kirk, R. E., z8or 365

Jack, Ollie, 225, 37o

Jacobs, J., r98, 224, 232, l6E
]acobson, E., 266, 363

James, M., l8g
Jaques, 8., 323, 363

]eeves, M. A., 8t,237, 363

]enkins,

'!?'.

L, T44,363

Jenkins, !7. O., r39 t 14% 363
Jerome, E. 4.,236, 363
Joffe, R., 256, 257t z8S, 342

fohn, I. D., 4Tt rr9,363
Johnson, H. 1., zoz, 354
JohnsoD, L. B., 82, t8o, 2tz, 224,
377

Johnston, W. 4.r 277, 363
Jones, 4., 3271 363
]ones, M. B., 342
Jong, f . R. de, 3 B, 363

Iudd, B. R., 3Sz

Kalil, R. E., r84, 3631 364
Kalsbeek, J. Str, H., r33, r35, 136,
364,38r
Kanfer, F. H., 38o

Kinsbourne, M., 389
Kirchner, !0'. K, zr2t 292, 365

I(leinsmith, L. f., 3ro, 365
Klemmer, E. T., 43, 69, 70, t23t
t8o, 2291 365
Klirg, J. !7., 32r, 365
Knatr, P. R., 377
Knapp, Barbara, 2gr, 293, 295, 366
IGright, A. A., 75r BS, t49, 366
Knowles, \V. B., r8r, 186, 356 r 366
Koch, A. C. E.r 2431 353
Koenig, Isolde D. V., 34\
Konick, A. F., r99r 372, 377

Koplin, f. H., rr9r 366
I(orn, J. H., 2t9r 366
Kornblum, S., ro9, 366
Koster, \Utr. G., Io8, Iogr II4, 366
Kotovsky, K., 2381 382

Kreidler, D. L, 66, 362

Krendel, E. S., 14, 37o
Kroener, Susan, roo, 376

Indcx

415

Krudis, B. R., 325, 374

Lippold, O. C. f ., 243, 2461 368

La Berg€, D. L.r 92r 23t, 366, g6l

Livson, Florine t 379
Lloyd, D. P. C.r 244, 368
Lloyd, K. E., 2241 368

Kruegar, S7. C. F., 2481 366
LaceS Beatrice C., 32rt 366
Lacey, f. I., 32r, 366
Lachman, R., 2251 366

Lally, f. R., 34o

Lamb, 1., 87, 366

Lambert, !f. 8., 258, 366
LaminB, D. R. J., 36, 50, 366
L*9, P. f., 32or 367
Langdor, J. N., 283,393
Lankford, Ff. G, 342
Lappin, J. S., 2z9r 3S4

I-aszlo, Judith I., r 39, 367

Lauer, A. R., 284, 16l
Laughery, K. R., 37zr 385
Lavery, J. J., 3o4r 3o7r 36T

Lawrence, Catherine, ro8, rr4,
r22, 354

Lawrence, D. H., 23Tr 36T

Lawson, Everdina A., ror ,2251 367

Lawson, J. S., 369
Lee, L. Charlotte, 3ror 32\ 367
Lehn, Ruth van, 366
Lehr, D. f., 2751 2761 34r
Leonard, f. A.r74r 8o, 83, tr2rr77t
r8r, 29o, 337,355, 16l
Leopold, F. F., r98, 224t 3461 38r
Leshner, S. S., 326, 362
Leuba, fI. R., 33r, 367
Levelt, Itr. f. M., t71r 3ST

Levin, I. P., 3o9r 36T

L.1ry, C. M., 342

L.olf, P.r 3zor 367

Levy, R. M., 87r 364
Lewis, Judith L.r 34r

Lewis, M., 3251 367
Lewis, R. E. F., 3r3, 362
Lichten, S7., 37r
Lichtenstein, M., 69, 337
Licklider, f. C. R., 99r 367
Liddell, ff. S.r z7z, 368

Lincoln, R. S., 3o6, 368
Lindley, R. ff., 2oo, 2t9r 3or, 366,
368, 38r

Lindley, S. 8.,2681 348
LindsaS P. ff., 386

Lindsley, D. 8., 2691 368

Lipsitt, L. P., 40, 27rt 3411 368

Locke, E. A., 3o8, 368
Loeb, M., 275, zT8, 3431 368
I-oehlin, f. C., 33o, 368
Loess, H., r99r 3orr 372, 379

Lomnicki, Z. A., t39r 3Ts
Long, E. R., 95, 368, 379

Lorge, L, 294, 3o4r 368
Lovaas, O. I., 3ro, 32rr 368

Loveless, N. E., r 89, 369

Lubin, A.r 39z

Luce, R. D., 76, 94r 348
Lnnd, M. \f. t 3Tz
Lrrndervold, Arne, 246, E6g
Luria, I., 392

McAlister, E., 165, t66, t7r, 36t
McAllister, D., 289, 369
McCall, R. 1., 324, 369

MacCaslin, Eugene t 29tt 369
McCormack, P. D., 2741 275, 369

McCog !tr. K., 54r 369

McC;ary, f. I7., 383

Mace, C. A., 3o8, 369
McElheran, !7. G., 2751 369

McFann, H. H., 384

McFarland, R. A., 2841 369
McGeoch, f. A., z86t 292t 369

McGhie, A., 2281 369
McGuiBsnr F. f., z9r, 3o3, 3o5,
369

Mackay, C. K.,267, 385
MacKay, D. M., 561 369

McKey, Molly I., 385
Mackwofth, fane F., r 28, zo8r zr2,
22gr 234, 278, 27g, 36g, 37o

Mackworth, N. H., t9, r28, t6z,
2r2, 234, 273, 274t 275, Z8t,
34or 37o

McNicol, D., 2zo, 2zz, 222, 3To

McPherson, Sandra, 372
Macpherson, S. f., 3o4, 3o7t 3o8,
37q^

Macrae, A. V., 2991 36r, 37o
McRelmolds, P ., 2o3, 37o
McRuer, D. T., 14, 3To

Index

4I.6
Mahneke, A., 249, 279, 34r

Malmo, R. 8,264,37o
Margaria, R., 353
Margrain, Susan, zor, 37o
Marill, T., ro8, rr2, rr4, 37o

Marks, M. R., 2251 37o
Martin, P. R., 2or, 37o

Martz, R. L., z74r 37r
Marx, M. H., 2951 36r
Mayzner, M. S., r75, 387
Melton, A. I7., 3oor 37r
Merkel, f., 6tr 73r 8z, 37t
Merton, P. A.r 243t 37r
Meskil, A., 35o

Metlay, V., 3641 37o
Michaels, R. M., 27o
Michon, f. A., r 34, zz9, 3Tr
Michotte, 4, 24, 17 4, 37r
Micko, H. C., z8tr 37T
Miller, G. A., 2r, 4e,, 43, 95, 96,
ro2, r79, 2t7, 225, 238, 37r
Miller, f., 95, 37t
Miller, f. G., r79, 2r7r 3r8, 388

Mingag Rosemary, 365
Mira, 8.,269,37r

Mitchell, M. f. fI., r89, r9r, 37r
Mitchell, R. F., 94, 377
Mitnick, L. L., r83, 357
Montpellier, G. de, r52, r95, 37r
Monty, R. A., 234, 357, 37r, 372,
385

Moore, 8., 34r
Moray, N., 66, roo, 229, 352, 372
Morgan, C. T., r37r 3r8, 372

Morikiyo, Y., 75r 3Tz
Morin, R. E., 82, 88, 9r, 93, 2tt,
3o5, 3551 372

Morrisett, L., 289, 372
Mortenson, F. f., r99r 37z

Moseley, A. L.,284, 369

Mote, F. A., 338
Mottram, V. ff., 3r8, 372
Mowbrag G. ff., 83, 84r 89, t32,
372t 373

Mueser, G. 8.r 233, 393
Munson, \f. A., 3 r, 373
Murdock, B. 8., 29, 3e,, r33, zo3,

2cl6, zo8, 2to, zrr, 2t3, 226,
228,3orr 373
Murphn D. B., 383

Mtrrray, Catherine S., 48, 376
Mnrray, D. 1,, 2oo, 229, 373

Mnrrell, K. F. H., r37, t4o, r43t
r89, 2551 3731 374

Mussina, Carolyn M., 34o
Myers, J. L., 36, 385
Myers, T. I., 3z7t 383
Nakamurs, C. Y., 3o3 t 325, 374
Namikas, E. f., 306, 338
Naylor, f. C., 29T, 374
Neisser, ff., 9ot 374
Neumann, Eva, 288, 374

Newlin, E. P., 366
Newman, E. 8., r8o,374
Newman, R. C.r 36T

NewtoD, f. M., 3r2r 353t 357
Nickerson, R. S., 81, 89, 9or 93,

ro8, t23r 374

Nishioka, N., z8o, 374

Noble, M. E., r33, t39, t43, 339,
362, 374, 387

Norman, D. A., r 13, r99t 2o4t zo8,
2ro, 216, 375t 389

Norris, A. H., 39o
North, f. D., T39, t44r 37S
Nuthmann, C., 3o8, 382

Oldfield, R. C.r 27r 8z, t79, 375
Oldroyd, C. R., 378

Olson, M. \U(/., r44, 363
Oostlander, A. M., 9e,, 375
Oppenshaw, f. \[., 347
Osborne, E. E.r z82, 375

Paillard, f., r37t 16o, 3TS
Palola, E. G., 3ro, 38r
Parducci, A., 43, 375
Parker, f. F., 286, 3SS
Parks, T. E., z3r, 3Ts

Parrish, f. M.tz24t 35o

Peak, G., 35o
Perret, 8,, z5o, 25rt 358
Persons, R. S7., 34o
Peters, R. W., roo, 375
Peters, S7., rp, t44r 3T6
Petersen, P. G, 379

Petersor, f . R., r44 t r49, r5rr 355

Peterson, L. R., roo, r98, zro, 316

Petrie, 4,, 3271 376

Indcx
Pew, R. S7., r95, 376
Phillips, Shirley, 116, 3lg
Pierce, f. R., 84, 85, 376
Pierone H.e 58, 376
Pierrel, RosemarJr, 48, 376
Pierson, \[. R., r37, tM, 376

Pillsbury, \f. B.r 2el6, 376
Pinneo, L. R., 32, 263, 376
Pitz, G. F., 36, 376
Polito, F. da, 3orr 354

Pollack, I., 34, 4% 4\ 42, 43t 82,

92, ro3, l8o, zot, 2o3, 2cl6,
2t2r 224, 36r, 376, 377

Poock, G. K., 3gt

Posner, M. I., 94, r98, r99, zo2,
237,288, 377
Postman, L, 96, 2r9, 292, 293,

3or, 377

Poulton, E. C., 15, 16, 99tu 6, r33,

r38, zo8, 2o9, zr2, 2t3, zZ4,

346, 3651 378
Powe, Str. E., 2761 364

Pribram, K. H.r 37l
Price, R. Ef., 3r8, 378
Provins, K. A., r39, 14\ 34\ 378
Pugh, L. A., 27or 378

Quilter, R. E.r 274, 385
Raab, D. H., 58, 59, 378

Rabbitt, P. M. A., 8t, gz, gS, r:6,
\r7,2Ogr 378, 379

Radford, Barbara K., t49t 355
Raffel, Gertru de, zt7, 357
Rapin, Isabelle, 347
Rapoport, A.r 76, 379
Rappaport, M., 3Ss

Ratliff, R. G., 384
Ray, T. S., 378

Ra1l, S7. S., 27r, 379
Redfearn, J. !7. T., 368
Reichard, Suzanne, 3zz, 37g
Reid, C.r 244t 379
Reid, L. S., 368, 3Tg

Renkin, A., 8or 3p
Restle, F., 234., 33or 3Tg
Rey, J. P., 25Tr 3Tg

Rey, P.r 25Tt 379
Reynolds, B., 306, 3r3, 3Tg
Reynolds, D., r22, r24r 37g

4t7

Relmolds, J. H.r 293, 356
Rhoades, M. V., 83, 373
Rich, S., 289, 355
Richards, f. M., 33o, 379
Richardson, Patricia, 2zS, 3Tg

Riedel, Str. Str. t 3zrt 379
Rittenhouse, C. E., 3o7, 3ST
Robbin, f. S., 344
Robinson, G., 3or, 379

RobinsoD, f. S., 95, 38o
Rodgers, D. A., r3T, 386
Rodwan, A. S., 44, 38o
Rogers, D. E., 36o
Rokeach, M., 352
Rosenquist, H. S., 293, 306, 343,
380

Ross, Helen 8., 29r 38o
Ross, S., 39r
Rossman, Ellen t r99, 3TT

Rubin, G., 3r3, 38o
Rubinstein, L., ro8, 38o
Ruch, T. C., t37r 38o

Russell, I7. R., 3ro, 38o

Ryan, F. J., 3o3, 342
Ryan, 1., 222., 38o

Saal, !7'. vom, 226, 373

Salapatek, P. H., 34r

Saldantra, E. L,249., 38o

Saltzman, f. J., 3o3, 38o
Sampson, H., 23or 38o
Samuel, f, A.r 274, 389
Sanders, A. F., r t4, tzo, t34, 293,
38o

San Giuliano, R. A.,346
Sant, C. van, 358
Sarason, I. G., 3ror 38r

Satz, P., 34o

Saufle5 !tr. ff., 2481 38r
Schaffer, Amy, r38, 358
Schaub, G. R., 2oo, 38r

Schiff, ro3, 38r

Schlosberg, F{.e r34t r74r 286, 32r,
365, 392

Schmidt, E. A.r 275., 368

Schneider, R., 339
Schoenberger, R. \[., 3Sg

SchouteDr f. F., 49, 57r 661 79, r33t

r52, r53, r54r 38r
Schulman, A. l.r 3S4

Indcx

418

Schwab, R. S., 245, 38r
Searle, L. V., t39, r43r 38r

Seeger, C. M., r83, 3Ss

Seibel, R., 80, 2161 38r
Seidenstein, S., r38, 38r
Selfridge, Jennifer, 225, 3Tr
Selye, H,r 272., 38r
Sengstake, C. B., rl5r 38r
Sewall, Susan T., 36, 385

Seyffarth, H., 2461 38r

Seymour, T7. D., 288 ,29rt 3r3r 38r
Shallice, T., 50, 54r 38r

Shanno[, C. E., r 3, 382
Shapin, M. I., z6t, 342

Sheenan, Maureen M., zt6,382
Shelly, Carollm, r33t 339
Shelly, M. !tr., 3o5, 382
Shepard, R. I7., 216, 382
Sherringtonr C., 248, 382
Sherwood, f. f., 3o9, 3ror 382
Shipley, Elizabeth F., 385
Shock, N. Itr., 2591 3M, 39o
Sidowski, J. 8., 3o8, 382
Siegel, A. f ., 33o, 382
Silvermor, J., 32r, 39o

Silverstein, A, 377
Simon, Betty P., r 52, T95, 382
Simon, H. A., 2381 382
Simon, J. R., rS2, r95, 382

Simonson, E., 24rr z5or 25t, 382
$ingere f . L., z74r 338
Singleton, !tr. T., r37, 21tt zS3,
27r, 284, 2grr 3O5, 3r5, 329,
3821 383

Sipowicz, R. R.,276, 383
Skarbek, A., 36o
Slack, C. !7., 2861 383

Slamecka, N. f., 48, 3or, 383
Slater-Hammel, A. T., ro8, rr4,
383

Smith, E. E., 93, 383
Smith, K. LJ., r38, r12t r84, r95t
3r3, 358, 38o, 383, 3871 389

Smith, L. A., r37r 383
Smith, M., 3r, 383
Smith, Maril5m C., ro8, r ro, T2o,

r22, l8s
Smith, S., 28o, 327, 383
Smith, \P'. M., rro, r38, r84, 383
Solomoll, P., 376

Solomons, L. M.e z7r 384
Spence, K. W., 3ro, 32% 32\ 384
Spencer, f., t6z, r85, 384
Sperling, G.5 229, z3rt 232, 339,
384

Spielberg€rr C. D., 96, 2T6, 32o,
359, l8+

Spitz, ff. FI., 4e,, 384
Spong, P.r z3o., 38o
Stabler, f. R., 27rr 384
Stager, P., 33o, 384

Standfast, Sus aort, 347
Staniland, A. C., rZ5, 384
Stanley, R., 383

Stark, Karen, 377

Stauffacher, J. C., 3o9r 384
Steele, M., 39r

Steffy, R. A., 23r, 354, 18+
Steger, J. A., r 19, 360
Steinman, Alberta, 55, 384
Stennett, R. G.,264., 385
Stenso[, ff. H., 337

Sterns, H., 362
Stock, F. G. L,r 393
Stone, G. C., 82, 385
Stone, M., Sor TTr 385
Stroud, f. B., 3o9, 3ror 385

Stultz, Y., 372
Suci, G. J., 67, 82r 385

Suddon, Florence H., 3o4, 3oT,
367

Suhr, V. !7., 284, 16l
Sullivan, S. A, 347
Snrwillo, W'. S7., 2741 385
Sutherland, N. S., 3Sz
Sutton, S., 347
Swart, H. de, 9c, 375

Swets, f. A., 2ot 3rt 34t 36t 47, 49,

55, 57r 358, 385

Swink, I., 387
Switzer, G., 82, 84r 9or 355

Sykes, R. N., r35, 1361 364
Sylvester, Anne, 2e,6, 376
Symonds, C., r97r 385
Symons, f. R.,267, 385

Szafran, f., 82, t84, 3trr 35o, 385
Takakuwa, E., 2571 385
Tanner, S7. P.r 2or 3r, 358, 385

Tarte, R. D., 3ror 389

Indcx
Taub, fI. A., 36, z34r 3T\ 3221 385

Taylor, A. M., 661 372
Taylor, D, fI., 89, z7or 386

Taylor, F. V., r33, r37t r39, r43,

rM, rgr, 3o4, 3571 38r, 386
Taylor, J., 384
Taylor, Janet A., 3zo, 386
Taylor, M. M., 49t 66, 72t 218,
37or 386
Tecce, f. J., 27r, 32rt 386
TeeI K. S., 288, 347
Teichner, !7. ff., 38r

419

Uhr, L., 3r8, 388

Ulehla, Z. 1.r 36, 388
Ulich, 8.r 295., 388
LJnderwood, B. 1., 293., 388
Vaughan, ff. G., 35o

Vernon, ff. M,, z&z, 283, 284,
388

Verville, E., 96, 388
Vickers, D., 39r 5or 54, SS, S7t T2t
169, r7or 38r, 388

Telford, C. !7., 106, 386

Vince, Margaret A., 106, rro, rr5,
116, rr7, rr8, r2Z, tz&, t4r,
r42, r89, r9r, 2r3, 256, 37r,

Thiessen, D. D., 2721 386

Viteles, M., 24rr 389
Volkmann, f., 43r 387

Terebinski, S. f., 343
Testa, D. fI., 27rr 386
Thackrsy, R. I., 359

Thomas, E. A. C., r24r 386
Thompson, L. \[., 79t 3M
Thorndike, E. L., 3o2, 3o4, 368,
386

Thouless, R. H., r72r 386
Thnrstone, L. L., 27,386

Thwing, E. I.r 3S4
Thylen, J. O., r9r, l8Z
Tickner, A. H., r 37, 359
Togrol, B., 3r8, 338
Topper, J. T.r 325., 386
Trask, Frances P., 23or 3g3
Trebra, Patricia von, 3r3r 38o, l8l
Trefts, Jeanne L.r 354

Treisman, Anne M., 66, 83, 93,t 99t
roo, ro3, rr5, 2261 3Orr 352,
EBI

Treisman, M., 38, 362, l8Z
Tresselt, M. E., 43, r75, l8Z

Trites, D. K., 359
Trowbridge, M. ff., 3o5, 387
Troxell, Nola, 372
Trumbo, D., r33, 359t 3741 387
Tsao, f. C., 293, 387
Tucker, !7. A., 322., 358
Tulving, E., ro3, 2o8, 3821 388
Tnne, G. S., r33r 388

Tuntnri, A. R., 40, 388
Turnage, T. W,., 377

Turvey, M. T., 2or, 388
Tustin, A., 14, 388
Tuttle, Abigail Y., 225, 366
Tweedy, f. R., 92, 366

388, 389

Voss, f. F., 379

Vuco, J., 368

Itrald, A., 13, g8g
STalker, E. L., 3ro t 325t 389
\trallac€, f., 2o3r 3r8, 389
Wallace, Jean G., 48, 96, 389
\[allis, C. P., 58, 338, 389
Vallis, D.r 2T4, 389
!7dter, A. A.r 346
lfard, N., 2451 389
!7are, J. R., 274, 34% 383
\[arne, C. I., 258, 34o

\[arren, C. 8., 374
Sfarren, f. M., 289, 3.46

l[ar:rington, Elizabeth K.e 224,
226, g8g

\[aters, L. K., 29r, 3M
Watts, T. R., 38, 387
ITaugh, Nancy C., 2o3, 2o4r 2o8,
2161 38g

Sfawman, R. I., 35r

ITeaver, I[., t3, 382
!7ebster, R. G., ror, 389

\[edderbtrrn, A, A. I.r 23o., 358

S7eene, P. L., r8o, 389
lTehrkamp, R., r52, T95r 3r3r 389

STeichel, Rosemarie, 343

\[einer, 8., 325, g8g

Weinstein, M., 355
Veiss, A. D., r24, r32r 365, g8g
!7eiss, B., r39, 39o

\[eiss, R. L., 32rr 39o

Indcx

42g^

\[elford, A. T., 14, t6, 17, rg,22,
23, 32t 52r 8r, tO7, rrO, rI3,
rr7, rr8, T27, t28, 146, r5rr
r52t r53, r55, 184, r85, r87,
r92, r95, t96, zrt, 235, 24t,

lfinnick, STilma A., 34r 3gz

\trolf, J. f., 33o, 382
lfolin, B. R., 343

lfolpe, G., 355

Woodhead, Mtrriel t r34, 224t 392

$felford, N. T., r95, 4o7

\floodrowr H., 69, 392
\[oodworth, R. S., r rr t34t t4ot
r4r, r42, r44, r57, r74t 286,

\Mertheimer, M., T64, 168

$7right, J. M. vorl, 294, 297, 298,

258, z6o, 265, 2961 3rr, 322t
33o, 355, 36r, 385, 38g, 3go

Tfenborne, A. A., t42, t44r 376
!7est, L. f., 2891 39t
S7esthoff, J. M.e r52, r53, 39r
Weston, H. C.r 2491 39r
White, C. T., z6z1 3S3
\[hite, R. !tr., 3241 39r

\[hite, S. FI., 2641 359

Vhitfield, D. f., 383

S7hittenb*g, f. A.e zT4, zBt,
39r

lTickelgrenr \tr. A., T99, 2or.) 2ro,
22O,22rr 226,3751 39r
!7iener, E. L.r zBr, 39r

324, 392

392

Wyatt, S., 282, 283, 284, 3z7t 392t
393

!7ylie, Ruth C., 339
Yamatrira, J., 374
Yates, A. J., ror, rro, 393
Yates, F., 33, r48, 3Ss
Yensen, R., 2461 393
Yerkes, R. M., 27or 393
Yntems, D. 8., 23o, 2331 393

S7iener, N., 13, 39r

Yourg, P., 3o2r 339,
Young, S., 58r 364

Ifilkinson, H. Jean, 363
Wilkinson, R. T., to2, 27o, 274t

Zaikowski, M. M., 363
Zangwll, O, L., 225, 296, 393

!7iesel, T. N., t6zr 362

275, Z8tr 39rr 392
Willianrs, A. C., 3o6, 392
Williams, fI. L, 275r z&t, 392

lTilliams, Iudith A., 8r, 392
Wilson, Edna A., 383

lfilson, S7. f., 353
\fingfield, A., 82, 375

Zarcmba, S. K.r 375
Zeaman, D., 961 393

Zinvry, G. H., 27o, 393
Zolmarr., f. F., 386

Zubek, f, P., 3z7r 393

Zuckerman, M. t 327t 393

Ztrkor, S7. J., 392

Subjea Index

a-reaction, 6o
absolute judgment, 4o-3
absolute pitch, 4ot 3r5

boredorrlr r9r 24Tr 33o
brain injury, zS8, z6t

abstraction, t6z

accidents, 282, 284, 3rg
acoustic confusion, to4, t97r z2o
activation, 19, 247, z6z-73, z8o-r,

3o8, 3r9,3221 3251 327
adaptationt 43t 247, z8o

after-effects, 2641 2g3

neural, 2rt 97
visual, 229
sge, 16, 23, 98, 196, zrz, 2354,
25o,258, 272r zgz
air-traffic controllers, 272
Ames room, r7o
anagramsr r75

anoxia, 259
anticipation, z2S
anxietyt 247t 255, 27rr 3Og, 32c.

32r

aphasia, 239

apraxia, 239
athletic enduran@,243
attentiorrr rgr rg8, 2r7r 275
resEiction of, 253
authoritarianismr 3 U t 333
'automatic' performance, r 12, tg3
automation, r8o
autonomic activity t 2ot z6z, z72t
3r8
averages, estimation of, t6z
b-reaction, 6l
F, 3z-8, ror, r23, t35, 2ro, 25r,
268, 27or 3o3, 3c9, 397

ballistic action, 15, 22t r4o, t57,
16o, t96r 29o
benzedrine, 28r

c-reastion, 88-go, 98, toz
Cambridge Cockpit t 252, z8r
capacity,2o-t, 247r 318, 324
limited channel, 16, 265, 3zg

spare, r34
capstan-lathe operators, 29r
car driving t r34t z6t, z7o
car radio t r34
'cerebral dictionary', ro3, 226, 293,

30r

choice, 18, 6r-87, t76

distinction from identification,

8t+q

'churrks't rT9, 2rT

classification, zl

compatibility, see display-control
relationships
compensation, 23t rg6, z8a
speed and accuracy, 66,266., 3r8,
322

complexity, 97, t6o
computer analogy- r3r r 8, 24, ro5,
z3Z, z38r 3r5
concept formation, t6z, 2341 48
conceptual frameworks, 2r-4, r9r,

237,289
confidence levels, 3o, 34, 37

confusion function, 29, 53ar 223,
396

consoles, r8o
constancyr perceptual, t7z
@nstant method 38
constants, extraction of, 18, rg4
continuity of change, 16Z

of line 165

continuous performance, tzfig
controls, effects of, t8.yz

blocking, 255-7, z6r, z7T
body temperature, z8r, 3r9

optimtun sensitivity of, t44

42r

InCex

422

conveyor lines, n8
cramps, writer's and telegraphist's,
246

critical flicker frequen*,249, z6t
crowds, behaviour of, 332
cutoff in signal-detection theory,

324

cybernetic approach to skill, $-r9

d', 3z-8, 48, 53, 66, ror, zro, 25o,
2681 3o3, 397

deafrress, 322

decay theory in short-term memorY, 202-4
decision theory, 13
delegation, fi4
delinqu€ocy, 3zs
dichotic [stening, zo6, 2271 23o
digit-symbol substitution, r79, 256
discovery method in uaining, 295
discrimination, 2e, 27-59, 93, r33,

r83, 2671 3rz

errors b, 54

limits of, 55

multi-dimensional, 43-T

multipler 39-47

Vickers' model, 56

display-control relationships, r 8,
24, 82-7, 9e,, r33, t8onz, 289,
299

distance perception, fi8-72

distractability, 256
diurnal rhythm, 283, 3t9

'dominant details', t74
drugs, 25o, 31 8
dual-task studies, t324, zo6, 274
economy in perception, 24, t6z-79

in performance, t8l
EEG, rozr 263

electro-conrnrlsive therapy, r97
emotion,2c,, z6z
endocrine functions, 31 8
engineering drawings, r85
ergograph, 24rr 2M

errors, correction of, r 16
discrimination, 54
movement, I'47
persistent, 296-3oo
reaction times, 65, 7r

serial learning, 30
expectanc!, 24, 7r, 77, 79, 276
extraversion-introversion, 3t8-zo,
326

eye-strain, 249
facilitatory effect, 59,2261 263, 265
failure, effect of, 326
false positives, 33: 46, 268, 3o9
false reactions, 67

familiarity, effects of, 84, 85, 87,
93, 95, rO3, r7r, 173-5, Zt8,

238, z6o1 3orr 329
fatigue > 19, 23t r94, t96, 273, z8t,
316,33o

blocking in, 255-7, z6r

industrial, z8z-5
mental , 247-6zt 293

explanatory models, z6o-z
neuro-muscular, 24vT
pain as factor in, 2U
sensory changes in, 249-5r

short-term memory, 2584., z6t
slowing of performance in, z5t255, z6r
specificity of, 244, 248, 283
static, 243, 244
fear, 269
feedback, 15, ro9, rrr, 116, r38,
3o2r 3r8, 33r
delayed, ror, r ro

time required for, r4o-5

'feel', r39r 3o3
filter, perceptual, 99-ro3, rr5
'final block', 2t5

flexibility, 289, 316, 33r

foreperiod, 69, 79r 89, r2z
games, II3, 298

Gestalt theory, 24, 164, 165
glare, 249
'goal gradient', 326
gradients in perception, 169, t7r
grouping, r r 3-r 5, rr7 , r 35, t64,
176, r79, 2Or, 2O3t 224

'guidance' in training t 299
habituatiort, 247, 263, 27o, z8t
heart rate, r35, 266, 269, z8o, 32o
heat stress, z8z

Index

Hick's Law, 621 84, ro2, r3z
'higher' units of perfonnance, see
larger units

hyoscine hydrobromide, z8r
hlryothesis formatiot, 22

identification, distinction from
choice, 8Z++
images r 237

incentives, 245, 264, 269, z7r,
306-7, 3r2, 3r4, 325, 328
inertia in controls, r39
information theory, 13, 6t-82, r45,
t6r, 175, 2r7r 263, 33r

inhibitiott 79t 247, 265
'initial impulse' in movement, r4o
inspection work, 288

intelligence, r84, r9rr 2331 283,
3r8, 327
interests, r 74, 324
interference in memory, r98-2oo,

423

letter sorting, tz8
levels, mental, 266
likelihood ratiot 3zt 3T
loading, direct measurement, t29-

r32

dual-task, 13z-6
effects of co-ordinationr r35
mental, n8-36
see also overloading and underloading
lorry drivers, 284
machine minding, rz}
machine tools, r8o
man-machi:ne systems, r rr !2, 14,
329

mathematical models, 25

memory, effect of astive response,
2o,o

long term, t97,2t6r 2281 47
of movement, 2co, 2oz

2o4-r4
intermittenry of response, 15, ro5,

recoverJr of data t 2ot tz8, zt6,

r34, 257
introspection, 4T
introversion, see extraversion
inuusions in recall, zt6
invariance of objects, t69, r7t, t75
invisible mending, 288

short term, 23, 93, r29r r33, t94,
287, 29rr 3r5, 33o
basic facts, r98, 2o2
capacity of store, zt3-27

iob satisfaction, 3zB
ioy-sticksr r39

keeping uack of variables, 233-5
knowledge of result$ 2TSt z8or z85,

3o2-7r 3r5, 323

effect of withdrawing, 3cr6-T
incentive effect, 3o8, 326
qnality of, 3o5-6
timing of, 303-5
larger trnits of performance, r7-r8,
23, rT8, t924, 2r3r 257, 292,

3r6, 327Ar 33o
lateral inhibitior\ 79
leadership, 285, 333-4
leaming, stages in, 287-3oz
long term storage, 294-3oo

retrieval, 3oc.z

short term link, 2gr-4
undersgsrting task, zgg_9r

3OO-2

decay theory, 2oz-4
fatigue effects, 258-9., z6r

interference, r98-zoo, zo4-r4
limited spsnr r98
locating store, 2zT-32
rate of presentation, 2o4

running,2or, zrtr 224
set size, effects of, 223-5
stress effects, 266

mental arithmetic, t3z
micro-behaviotrral approach, z5
monitori.g, 15, 85, rrr, rr8, t27,

r38, r8o, 29o
monotoDYr r9r 247t 274,283
morale, 285, 333
motivation, 19, zTS4, 3o7t 323-4

in fatiguet 245

servo concepts, 325-9)
vigilancer zTs4
motor mechanisms, 17

movement, correction for errors,
r47

course against time, r4t-3

424

Ind,ex

movement-contd.

pilots, air line, 272

formulae, r45, 146, r47
memory of, zoor zoz
minimrun, r47
relation to reaction time, r37
sensory control of, r38-4o
tempo and context, r95
muscle tension, effect of, r4o, 262,

power assisted controls, r39

duration and accuracy, r4tt Poisson type distribution, 38
population stereotypes, r 89
r45{o
Post Office sorters, 289, 29r, z9S,
effect of target, T43., r5rr r56
force and time, tM
300

264, 266, 27or 3o9
musicians, rr3, rr4

nenral, after-effects, zrt 9T
distance, 80, ro3

noise, zc-r, 28, 55, r97, zto,
zz6, z6t, 2651 3ro
spread of effest, 80, ro3
nenrological models, 25, 77-82
nenrosis, experiment al, z7 z
noise, 34, 49, 95, 99, 27o, z8o, 3rg

neural, 2c-r, 28, 55, r9T, 2ro,
226, z6t, 2651 3ro
visual, 2ot 3r, 54

novelty, effect of,264

order, retention of, 2t9r z2z
norders'
by central mechanisms, 15,
rr4, 116, r43, r85, 29or 3o3
orientation, maintenan@ of, 237
overloading, tz8, t32r 24c-13, 32Tt
33o, 333
paced tasks, tz8, r35

effect of latitude, tz9
Parkinson's Law, r3o
perception, analysis in, 93

as running hypothesis, to4

central factors in, 16r
economy in, t6z-79
focusin E n, 79

mechanisms of, t7

quantising of, r2o

selection in, 91-ro4.r 115-16, t6z

size of set in, 95-8
social conventions, effects of, t74
personality t 2ot 3rT-23

perspective, t6g

practice, r8, 74, 83, 85, 9rt Ir2,

r24, I,26, r83, r85, rg4, 2Or,
2r3, 27rr 288, 2gg, 3I,2-16
mental, 295
spaced t 292-4
prediction, 18
pressure controls, r39
problem solvingt 22t z3S*7, 27rt
325
process plant, 3o4

programming of action, 48-9
proglessive part method t 29r
quantised perception, r2o

radar, 273
random-walk models t 5t, 72
Raven's Matrices test, 238
reaction time, t:ir 25r zto,257,274,
z8or 3o3

conceptual models, 7o-82

different senses, 60
discrimination, 29, 47-59
errors, effects of, 65, 7t

familiarity, effects of, 84

information theory, 6t-82
neurological models, 77-82
sequencing effects, 65
serial classification models, 7r4,
97

set size, effects of, 6v94
simultaneous scanning models,

76-7

Receiver-Operator Characteristic
(ROC), 34, 2ro
recoding, 179-89, r99, 2t7, 2t9,
2281 238, z861 32t

recruitment, musctrl ar, 246
redundancy in information theory,

t63, r75, t78
refractory period, 106, r2o
regularity of figrrre, 165-7
rehearsal, 2oc-lz, 2o4, 2t2, 2tT,
232t 2931 294

Idcx
relaxation,266

reliability, 3231 326

reminiscence effects, 2ro

repetitive work, rrz
responsibility t 3z2t 326
rest pauses t 275, zZz, 284
reticnlar forrnation, 263
retrograde amnesiat r97
rigrdity, 316, 3r7t 333
risk taking, zo
routine, impairment of, 2SB

running memory spanr zor, 2tt,
224

satiatio\ 247
scannin9r go
tschemata', t73, 238, 286
selectiotr, 12
selective response, see c-reaction
self-regulating systems, see servomechanisms
sensory deprivation, 28o, 327
sensory-motor system diagram,

t9

sequences of response, t4, 26, 65,

II5

Sequential Probability Ratio Test,
Sot 72

serial learning, errors in, 30
scfvo-Etechanismsr r3r t4, 25, ro5,
r38, r43, r5o, 285, 32a, 3z3t

325+

425

sinus arrythmia,266

skill, rt, tz

definition of, 12, 18, 23r 316

mental c 2r-4

pioneering sttrdiesr r r

skin conductance and resistance,

z6z, 264, 266, 269, 3o9, 3zo
sleep, z7o, z8o, 3rg
slips of the tongue t 2!
social behaviowt 32T34
conventions, effects on percep-

tion, r74
facilitatiou, 332

norlns, 332-3
skills, 33o
tasks, 3zB

'spare' capacity, r34
spatial transposition, 184-9
speech, speed of, tz7
steering wheels t r39t t44
stereoscopic viewing t r7r
strategy of performance , 17, 22, 25,
73t 75r 84, rr3r r \g, rg5r 2o8,
2r4, 22r, 245, 2871 3r5, 322,
330

stress, 2c,, tgt, z6z-,73, z8z, 322
symbo[c relationships, see trars-

lation processes, symbolic
synmetry, t6S-Zr zz4

teaching machines, 29o
temperature, effects of, zS4

temporal uncertainty, 60, 62, 677or 89, r2z-5, t49

set, 24
set size, 6oa4., 95-8, zz3-s
shoe machinists, 29rr 3o5
srgnal detectiott, rgrzrt 3t-47, 48,
rol, t23, I35, 2rO, 25O, 268,
27791 3o3, 3I8

thinking t 24t 4T-8
thresholds, sensory, zo, 2T-38, 267,

simplicity, z4

traces, 79r 272., 3@

274

Thurstone's letter-series corrlsignal-to-noise ratio, 2ot 66, 95,
pletion problem t 239
25O, Z6t
time iudgment, t34

simultaneous translation, r28
single-ctrannel hlryothesis, 105-36,
263
alternative theories, 12o-6
'gate' mechanisrDr tr4, rr9

grouping, rr3-r5r rtTt t3S
modification of responser rt6-t9

serial reaction time, tofiz6
tracking 105-6, rr7-r8

tracking, r3-r9, 105-6, 116, rr7-

rr8, 126, r33, 136, r38, tM,
r8r, rgrr zrz, z6o, 299t 3o6,

3t2t 3r3
training, tz, t36.r rgrr 299t 3o4

discovery method t z9S
progressive part method, 29r
transfer effect, 287, 3tel2., 3r5

transfer ftrnaions, 14

htdcx

426

translation processes, 18, 24r 55,
83, 87, 90, ro9, r 16, r33, t43,
160, t79-89, 2oo, 216, 232,
286, 29or 294

symbolic, 68, r84, 186-9
tremor, I38

typewriting, r ro, r 13, rr4, 289,
298

underloading, 273, 327
vandalisrnr 3zs

velocity controls, l9r, zS5
vigilance, r9-zr, 273-8r, 3t9r 33o
theories of, z7 5-8t

motivation, 275-6
signal detection, 277-9
viscous friction: r39
visual acuity, z4g
visual illusions, r7o
visual noise, 2ot 3r, 54
twarm up', z6o, z8z

watch keepin* 19
!7'eber's Law, 27-3% 5r, 53, r39,
r59
STechsler Adult Intelligence Scale,
233

work study, r r
writing, reversed, r85

activation, 28o-r
blockin1r zTT

STurtzburg School, 237

expectancyr 2T6

Yerkes-Dodson Law, 27or 3ro, 32o

METHUEN,S

MANUALS OF MODERN PSYCHOLOGY

Edited by C. A. Mace 1946-68

H. Jt'. Butcher 1968-

Fundamentals of Skill