This article was downloaded by: [McMaster University]

On: 14 October 2014, At: 10:58

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

The Journal of General Psychology

Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/vgen20

Effects of Knowledge of
Performance: A Survey and
Tentative Theoretical Formulation
a
R. B. Ammons
a
Department of Psychology , University of Louisville , USA
Published online: 06 Jul 2010.

To cite this article: R. B. Ammons (1956) Effects of Knowledge of Performance: A Survey and
Tentative Theoretical Formulation, The Journal of General Psychology, 54:2, 279-299, DOI:
10.1080/00221309.1956.9920284

To link to this article: http://dx.doi.org/10.1080/00221309.1956.9920284

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information
(the “Content”) contained in the publications on our platform. However, Taylor
& Francis, our agents, and our licensors make no representations or warranties
whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and
views of the authors, and are not the views of or endorsed by Taylor & Francis. The
accuracy of the Content should not be relied upon and should be independently
verified with primary sources of information. Taylor and Francis shall not be liable
for any losses, actions, claims, proceedings, demands, costs, expenses, damages,
and other liabilities whatsoever or howsoever caused arising directly or indirectly in
connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.
 Terms & Conditions of access and use can be found at http://www.tandfonline.com/
page/terms-and-conditions
Downloaded by [McMaster University] at 10:58 14 October 2014
 The Jonmal of General P s y d o b g y , . l 9 5 6 , S 4 j 279-299.

EFFECTS OF KNOWLEDGE OF P E R F O R M A N C E : A SURVEY

. AND TENTATIVE T H E O R E T I C A L F O R M U L A T I O N .

Department of Psychology, University of Louisville

R. B. AMMONS~

A. INTRODUCTION
When the psychologist experiments with knowledge of performance, he
ordinarily sets up a situation in which he compares the learning of a task
or performance changes by a group which receives relatively more informa-
tion about how it is carrying out the task with the learning or performance
Downloaded by [McMaster University] at 10:58 14 October 2014

of another group which receives less information. An example of this

arrangement is found where the B-29 Pedestal Sight Manipulation Test
(37, pp. 558-607) was used in training experiments with gunners. When
the gunners were not told whether or not they were making “hits,” they
shot much more poorly than when hits were indicated by a device which
made the picture of the target plane look red when they were “on target”
and ranging properly. In general, research on the problem of knowledge
of performance involves determining the effects of giving or withholding vari-
ous kinds of information about performance during and for varying amounts
of time after that performance. Although the general problem seems quite
simple and clear-cut, further discussion in this report will reveal some of its
complexity.
Studies have been made with a wide range of materials under a con-
siderable variety of conditions. Judd (30) started the work, asking his
subject to make direction estimations behind a screen. Such apparatus as a
two-hand coordination test (23), a lever-positioning device (39), a ball-
tossing arrangement (34), a ranging device (52), an ergograph (7), and a
simulated tracking device (44)have been used. Performers have been asked

*Received in the Editorial Office on July 20, 1954.

1This research was supported i n part by the United States Air Force under Con-
tract No. 33(616)-95 by the Aero Medical Laboratory, Wright Air Development
Center, Air Research and Development Command, Wright-Patterson Air Force Base.
Permission is granted for reproduction, translation, publication, use and disposal in
whole and in part by or for the United States Government. A more detailed dis-
cussion of some of the material included in the present report can be found elw-
where ( 5 ) .
SThe writer wishes to express his particular appreciation to Mrs. C. H. Ammons
and to Mr. ROSSL. Morgan of the Aero Medical Laboratory for much valuable criti-
cism and assistance in the preparation of this paper.
279
 280 J O U R N A L OF G E N E R A L PSYCHOLOGY

to learn mazes (20), aim guns ( lo), estimate lengths (25, 4 9 ) , learn Morse
code ( 3 1 ) , draw lines of a given length (49, 50, 51), write legibly (18),
turn a knob (35), and learn a complex industrial operation (33). Appli-
cations have even been made in devices which permit immediate knowledge
of the correctness of an answer in a school test (41).
T h e present paper will review and summarize factual material now avail-
able. Tasks which are avoided should probably also be specified to delimit
the paper. I t has been judged either impossible or undesirable to try to do
any of the following: ( a ) Discriminate definitively between the effects
of motivation and those of learning. T h e nature of the interaction of these
two variables is not even clearly known in the case of rats, where many in-
vestigations have been made. There is no work that the author could find
in the field of effects of knowledge of performance on humans which throws
Downloaded by [McMaster University] at 10:58 14 October 2014

any significant light on this problem. ( b ) Include incentive studies with

humans. As a rule these studies do involve knowledge of performance in
a more o r less identifiable form. Unfortunately, there have almost never
been control groups suitable for the purpose of the present paper-that is, all
groups received about the same knowledge of performance. (c) Include
material dealing with level of aspiration. T h i s material is complex and often
seems to be contradictory. Inclusion would mean that a theory of personality
would have to be included, at least implicitly. Since none of the knowledge
of performance studies reviewed made an explicit attempt to study indi-
vidual personality as related to individual performance, there seemed to
be no point in bringing in personality material on a speculative level. ( d )
Include material on guidance. Although studies of guidance involve an
element of knowledge of performance, they also bring in questions of per-
sonality interactions and lack controls to make them conclusive in the study
of knowledge of performance.
As indicated, the next section of this paper deals with the empirical gen-
eralizations made possible by experimentation and concomitant observation.
B. EMPIRICAL
GENERALIZATIONS
Each generalization is based upon one or more reasonably adequate studies.
T h e discussion will indicate what apparatus or equipment was used and
something about the experimental conditions. No rigid rule has been fol-
lowed in setting up the order of the generalizations. However, it will be
found that roughly they move from general to specific, important to less im-
portant, and better substantiated to less well substantiated. Some overlap
will be found. Overlap appears where it has seemed to the writer that two
or more aspects of a situation should be emphasized separately.
 R. B. AMMONS 28 1

1. Generalization I
T h e performer usually has hypotheses about what he is to do and how
he is to do it, and these interact with knowledge o f performance. Trow-
bridge and Cason (51) noted, in an experiment calling for drawing lines
of given lengths, that their subjects used numerous different methods, many
involving verbalization. Elwell and Grindley (23) found the same sort of
thing with subjects performing a two-hand coordination test. In a study
calling for subjects to push down a lever and hold it there for a given time,
Macpherson, Dees, and Grindley found that “the subjects reacted in an
extraordinary variety of ways, each dependent upon the hypotheses which
they had formed about the workings of the apparatus. ... None of these
ideas agreed with the simple instructions given (or with the truth about the
apparatus)’’ (35, p. 71). Bilodeau, writing about a gunnery training study
Downloaded by [McMaster University] at 10:58 14 October 2014

with the Pedestal Sight Manipulation Test, says: “There is no doubt but
that extraneous noises in the experimental situation serve to increase error
variance by leading to diverse and individual hypotheses about gunnery
behavior” ( 14, p. 4).
One can be reasonably sure that performers will have hypotheses, and
that these will interact in some way with knowledge of performance. Since
the hypotheses are often incorrect, the interaction will ordinarily lead to
below-optimum performance. Hypotheses can be about relevance of cues,
appropriateness of responses, or both. T h e performer can be led by his
hypothesis to attend to cues only randomly associated with “correct” cues,
and thus will make whatever responses he makes at the wrong times. O n
the other hand, his hypothesis may lead him to make inappropriate responses,
even though he is responding at the right times. Finally, if the hypothesis
is somehow inconsistent with the actual situation, but strongly held and there-
fore resistant to change, he will perform poorly, due to emotionality accom-
panying the feelings of frustration arising when he is “unsuccessful.”

2. Generalization 2
For all practical purposes, there is always some knowledge of his perform-
ance available to the human performer. Thorndike (49,50) and Mac-
pherson, Dees, and Grindley. (35) asked subjects to draw a line of a given
length, but gave them no information about how long a line they actually
drew each time. Even with repeated practice the lines did not show signs of
approaching the lengths called for by the experimenter. T h a t something
was happening is indicated by the results obtained by Seashore and Bavelas
(46) when they reanalyzed some of Thorndike’s data. They found that
 282 J O U R N A L OF GENERAL PSYCHOLOGY

several of the performers showed an increasing self-consistency around some

“incorrect” line length. These people were apparently setting up their own
standards even though they had no way of knowing whether their responses
were wrong, o r by how much, o r in what direction. T h e subject’s standard
failed to correspond with the experimenter’s standard ; therefore, from the
experimenter’s point of view, repeated practice revealed no “improvement”
in performance.
Several studies seem to throw light on the nature of the information the
performer is using. Arps (7, S) found that performers set up subjective
sources of information which they used when he objectively (physically)
stopped telling them how they were doing in the ergograph (weight lifting
by finger) situation. H i s people reported imagining pen markings and the
physical characteristics of individual lifts, and that this kept up their per-
Downloaded by [McMaster University] at 10:58 14 October 2014

formance. Rook and Norvelle ( I S ) studied school children performing a

variety of simple tasks, like writing “a’s’’ legibly as rapidly as possible. W h e n
the children were switched for experimental purposes to conditions where
they were not given their scores, many of the children responded by develop-
ing methods of keeping track of how well they were doing. Ross (43) gave
quizzes to college students, and then told some their exact scores, while he
told others nothing about how they had done. W h e n he found no differ-
ence in performance between the two groups, he made a further analysis of
the situation, and found that the no-knowledge-of-performance group were
able to guess their scores quite well. T h e y were therefore receiving a sort
of subjective knowledge of performance.
I n a military situation, several investigators, among them Morin and
Gagne (40) and Bilodeau (14), have used a red filter to indicate to per-
formers that they were “on target” with the B-29 Pedestal Sight Manipula-
tion test. T h a t is, the target plane appeared red when they were on target
with respect to tracking and ranging. T h i s knowledge of performance regu-
larly led to better ranging scores, but the tracking scores were high and
were not affected. O n e could speculate that the performers were receiving
all the knowledge they needed about tracking just from looking in the gun-
sight. T h e red filter added no information. I n a similar situation, Sackett
(44) found that giving scores in simulated tracking did not improve per-
formance. Here again, it is quite possible that performers received immediate
knowledge of how they were doing from the task itself.
In any case, we can be sure that in any practical situation the performer
knows he has done something, and he has an idea of how he did it. I n
experiments, he is often “perverse” enough to work out cues to use which
 R. B. A M M O N S 283

the person testing him never thought to use-spots on the ceiling or work
surface for aiming, unmasked auditory cues, etc. W e can be reasonably
certain that he will set up his own hypotheses and goals if we do not give
him any definite ones. In this situation, however, he may end up respond-
ing only randomly to the relevant cues and thus learning nothing about the
“proper” task because he is evaluating his performance by criteria unrelated
to the desired performance.

3. Generalization 3
Knowledge of Performance affects rate of learning and level reached by
learning. Almost universally, where knowledge of their performance is
given to one group and knowledge is effectively withheld or reduced in the
case of another group, the former group learns more rapidly, and reaches a
Downloaded by [McMaster University] at 10:58 14 October 2014

higher level of proficiency. This effect has been demonstrated for such
widely different tasks as range finding and college quizzes. T h e advantage
given the knowledge-of-performance group is to some extent a “directional”
one, as suggested by hlacpherson, Dees, and Grindley (35). T h e performer
can correct his performance on the basis of information about direction
and amount of error. T h e knowledge makes possible a better orientation
to the task. It should be kept clearly in mind that in many cases the modifi-
cation of amount of knowledge of performance given by the experimenter
changes the task to be learned. Thus, differences in rate of learning and
level reached in learning are due, at least in part, to actual differences be-
tween tasks.
I n the following summary of experimental material, studies are grouped
where possible according to apparatus or subject matter.
a. Pressey (41) reports that students who repeated quizzes with an im-
mediate self-scoring arrangement showed much greater learning than did
those to whom the test was merely given again without any knowledge of
results. H e worked with difficult English vocabulary items, Russian vocabu-
lary, and facts in the field of psychology. Angel1 (6) and Morgan and
Morgan (38) also found greater gain where correctness of quiz answers or
quiz scores was known to students immediately.
b. Although the results are not completely consistent, a study by Book
and Norvelle (18) seems to show that school children do better in writing
legible a’s, crossing out letters, performing mental multiplication, and in
substituting symbols when they are given knowledge of results than when
they are not.
c. Lindahl (33) found over-all improvement in adequacy of performance
 284 J O U R N A L OF GENERAL PSYCHOLOGY

of a complex industrial skill requiring arm and leg coordination when trainees
were shown graphs or recordings of their movements and it was pointed out
to them how these differed from those of better operators. His procedure
probably would have been more effective if he could have worked out an
arrangement for giving knowledge of performance sooner after the actual
performance.
d . I n studies of knob turning (15) and of lever positioning (13), Bilo-
deau showed that there was learning with knowledge of performance even
when the experimenter systematically distorted the information. His sub-
jects improved toward the “artificial” or distorted goals set up by making
the measuring scale systematically nonlinear, and apparently did not notice
the distortion or at least did not comment on it.
e. Thorndike (49, 50) found a much greater improvement in drawing
Downloaded by [McMaster University] at 10:58 14 October 2014

lines of a given length when the subjects were given information about how
they were doing. T h e greatest gain was between the first and the second
trials and this seemed to be due to an orientational effect of the information
given after the first trial. Trowbridge and Cason (51) also studied line
drawing and found the knowledge-of-performance group to be much superior
to the no-knowledge group.
f. Johanson (29) reports that subjects given knowledge of performance
about their manual reaction time with a telegraph key significantly improved
it beyond a supposed minimum. There was greater improvement when
electric shock was used to “inform” them than when they were just told
their reaction times. Other incentives may have more effect in some situa-
tions than does knowledge of performance.
g. Hamilton (25) found that estimation of length improved with knowl-
edge of performance, but did not improve when subjects were not given in-
formation about their performance. Thorndike (49) also observed more
improvement in estimation of line length when subjects had some idea of
how they were performing.
h. Using a two-hand coordination test, Elwell and Grindley (23) found
improvement only while some knowledge of performance was being given.
i. Lorge and Thorndike (34) had subjects throw a ball over the shoulder
a t a target. When they were told after each throw where the ball had gone,
they improved; when they were not told, they failed to improve.
j . Eaton’s (22) subjects tried with a rotary marker to make a circle
which was complete but whose ends did not overlap. There was as much
improvement in 100 trials where the amount of error was indicated as in
1,000 trials where there was no direct knowledge of performance.
 R. B. AMMONS 285

k. Gibson and Smith (24) found that subjects apparently can learn to
judge distance if given information as to the accuracy of their judgments, at
least when photographs are used.
1. In an investigation (52) at Tufts College with a range and height
finder, subjects reduced their constant error and variability if the instructor
sounded a buzzer whenever they made an error of more than a predetermined
amount. This improvement did not appear when no such signal was given.
m . W i t h trainees learning to point a gun, Biel, et al. (lo), discovered
that improvement took place almost only during the times when a buzzer
was sounded to indicate that the trainee was off target.
n. In a series of studies (14;28; 40;47; 53, pp. 414-415) it has been
shown that trainees acquire ability to score higher on the Pedestal Sight
Manipulation Test when they are given knowledge of performance in the
Downloaded by [McMaster University] at 10:58 14 October 2014

form of a red signal over the target plane whenever they are “on target.”
I n each of the studies cited, the acquisition was more rapid. Houston (28)
observed better performance when he used the red filter on 100 per cent
of the trials than when he used it 25 per cent or 50 per cent of the time.
Morin and Gagne (40) found that using a tone to indicate time on target
was as effective as using the red filter.

4. Generalization 4
Knowledge of Performance affects motivation. T h e most common effect
of knowledge of performance is to increase motivation. Much evidence on
this point has been collected informally or can be inferred.
a. With a two-hand coordination test, Elwell and Grindley (23) noted
that subjects were bored, more often late to experimental sessions, and less
careful when they were deprived of knowledge of performance, while they
were more alert and reported enjoying themselves when they were given
knowledge of performance.
b. Helmstadter and Ellis (26) tried various kinds of goal-setting pro-
cedures with a block-turning task, and concluded from the results that simple
knowledge of performance led to as much motivation as any of the more
complicated procedures.
c. Pressey (41) found that students who had used a “punchboard,” a
device for the immediate self-scoring of their quiz results, preferred this
way of taking multiple-choice tests, and did not like giving it up as an in-
structional device.
d . I n an early and widely cited study of ergograph performance, Arps
(7, 8) noted that subjects seemed much more highly motivated when they
 286 JOURNAL OF GENERAL PSYCHOLOGY

were given the details about their “lifts.” W r i g h t ( 5 6 ) also used the ergo-
graph and found that subjects had a “better attitude” when they could see
the record of their performance being made on the smoked drum than when
they were just told to work as long and as hard as they could.
e. Both Bilodeau ( 1 4 ) and Houston ( 2 8 ) observed that trainees on
the Pedestal Sight Manipulation T e s t found the red filter condition more
interesting than when they were getting no supplementary information about
their tracking and ranging. A possible explanation of the many and varied
findings with this test (14; 28; 40; 4 7 ; 53, pp. 414-415) is that when
they are receiving on-target knowledge of performance from the red filter
arrangement, the trainees are motivated to t r y the ranging part of the task
more often and so make a higher score. Without the knowledge of per-
formance, they perhaps pay less attention to the ranging since it is a less
Downloaded by [McMaster University] at 10:58 14 October 2014

obvious dimension of the task.

f . Book and Norvelle (18) found experimentation with the motiva-
tion growing out of knowledge of performance somewhat difficult with school
children. During the early stages of experimentation there were marked
differences in motivation among the groups, but eventually all the groups
developed ways of keeping track of their scores, even when these were not
given to them. W h e n this occurred, all of the groups became interested
and motivated, some because of scores released by the experimenter and
others because of scores which they, themselves, computed.
g. Using several different tasks calling for simple skills, Macpherson,
Dees, and Grindley (35) noticed that giving knowledge of performance led
to a more favorable general attitude toward the experiment. T h e y also
found that the incentive effects seemed to become relatively more important
later in practice.
I n general, the more a subject knows about what and how he is doing,
the more highly motivated he is likely to be. However, motivation does
not always increase with increasing knowledge. There is probably little
effect when he is already performing a t a high level of proficiency, and
increased knowledge may actually lead to a decrease in motivation when
he is doing poorly.
It is very important to keep in mind what the subject is motivated to do
when knowledge of performance increases his motivation. Often, he is moti-
vated to score higher, not necessarily to learn the task faster or better. H e
may then resort to taking advantage of weaknesses in the apparatus, learning
habits which are of no value or actually lead to poorer performances when
he later attempts to learn a similar task.
 R. B. AMMONS 287
5 . Generalization 5
T h e more specific the knowledge of performance, the more rapid the im-
provement and the higher the level of performance. Other things being
equal, the more exactly a subject knows how he has performed, the more
likely he is to be able to make appropriate corrections. T h i s effect, then, is
primarily a directive one. A study by Trowbridge and Cason (51) showed
how specificity works. T h e subjects were learning to draw a line of a given
length. Individuals given the amount and direction of their errors did much
better than those who were just told whether or not they were within one-
eighth inch of being correct in either direction. Waters ( 5 5 ) asked his
subjects to estimate the duration of a 1Zsec. time interval. H e found that
improvement was “roughly proportional to the degree of information given”
about the correctness of the estimates. T h e results from two studies by
Downloaded by [McMaster University] at 10:58 14 October 2014

Bilodeau ( 1 1, 12) appear to contradict this generalization. Using a lever-

positioning task, in each case he found no difference during a test period
between groups which had in effect received information of different preci-
sion during training. However, relatively few trials were given, and this
was a very simple skill, which factors may well account for the discrepancy.
It is very important to keep in mind that there is an optimum specificity
of knowledge, and that additional knowledge will not improve performance
or may even lead to its deterioration. Bilodeau and Morin (16, 17) re-
moved the tracking pipper from the sight of the Pedestal Sight Manipula-
tion Test and found that trainees made better ranging scores than usual
with it present. A possible explanation is that with the pipper present, too
much information is being given the trainee, at least at the start of practice,
and this is confusing to him. Crafts and Gilbert (20) found in a maze-
learning experiment that the scores they were using were probably too com-
plex. T h e so-called knowledge-of-performance group was not superior in
performance. T h e subjects apparently were confused and concentrated on
one of the scores, neglecting the other two.
T h e point of optimum specificity of knowledge of performance is related
to some extent to stage of learning. At the start of learning a new task,
the subject can use little information. As learning proceeds, he is able to
use more and more. Learning is, in fact, dependent upon the subject’s acqui-
sition of techniques for utilizing a greater range of information.
6. Generalization 6
T h e longer the delay in giving knowledge of performance, the less effect
the given information has. A number of studies seem to have some bearing
 288 J O U R N A L O F GENERAL PSYCHOLOGY

on this generalization. Saltzman (45) reported slower learning of a verbal

maze when knowledge of performance was withheld 6 sec. T h e fact that
he called this a delay-of-reward study indicates how closely allied are the
two fields, delay of reinforcement and knowledge of performance. W a r -
rick (54) systematically introduced lags into the transmission of control
movements in a compensatory tracking situation. H e found that time-on-
target was lowered by the introduction of any lag. Angel1 (6) experimented
with chemistry quiz results. W h e r e the students were given immediate
information about right and wrong answers, they did better than where they
were not told until the next class meeting. I n Keller’s (31, 3 2 ) work with
code reception, subjects who found out after each word whether o r not they
were right learned much more rapidly than those where knowledge of
performance was given only after much larger units. Houston ( 2 8 ) de-
Downloaded by [McMaster University] at 10:58 14 October 2014

signed a study of “firing” with the Pedestal Sight Manipulation T e s t so

that it was possible to compare trainees who were informed immediately
about their performance by seeing a red filter drop over the picture of the
target plane with trainees who were only given their scores after blocks of
trials. Those given immediate knowledge of performance by means of the
red filter were much superior in performance.
T h r e e studies which seem to indicate a failure of knowledge of perform-
ance to have an effect possibly throw some light on this generalization. Crafts
and Gilbert ( 2 0 ) had their subjects learn mazes, and found that those who
were given complex error and time scores after each trial did no better than
those who were not. It has already been mentioned in discussing Generaliza-
tion 5 that the subjects were probably confused by the complexity of the
knowledge (scores) given. Perhaps, since the delay from turning into a
given alley and being given scores a t the end of the trial was quite appreciable,
immediate knowledge of performance of another kind was affecting the
groups. T h e complex scores were ineffective, and both groups were receiv-
ing what amounted to immediate knowledge of performance with respect
to rate a t a given moment and entries into blind alleys.
An experiment by Lorge and Thorndike ( 3 4 ) called for subjects to throw
balls over their shoulders a t a target without looking. T h e n , after delays
up to 6 sec., the subject was told where the ball had landed. Lorge and
Thorndike reported that delays of 4 and 6 sec. seemed to lessen the improve-
ment somewhat as compared to shorter delays ( 1 and 2 sec.). T h e y sug-
gested that the failure of a steep gradient to develop was due to the fact
that improvement depended on postural changes which are slow and may
take up to 6 sec. T h e real cue is that when the results of a throw were
 R. B. A M M O N S 289

announced to a subject only after one intervening throw, there was no

appreciable improvement. From this, we can see how closely delay of
knowledge and interference from intervening activity are related.
T h e third study is one by Alexander (1). H e had subjects throw darts
over a barrier in a darkened enclosure; he then informed the subjects after
various intervals (0, 2, 4, 8, or 16 secs.) of where their darts had hit. There
was no statistically significant difference among the groups. However,
two things were observed. First, the subjects with shorter delays were bet-
ter able to guess where their darts had hit. Second, the subjects were receiv-
ing immediate knowledge of performance from the look of the dart going
over the barrier or the “feel in their arms” as they threw. Alexander pointed
out that what the subjects were apparently learning was not how to throw
the darts but the response which was being immediately rewarded; i.e.,
Downloaded by [McMaster University] at 10:58 14 October 2014

guessing where they would fall.

It is quite possible that the learner may not be able to use information
given more than 15 or 20 seconds after the response, so he uses what is im-
mediately available in the way of information and this may be irrelevant
or even misleading. W e might suggest with Brown (19) that knowledge
of performance serves as a reward, and add that what is rewarded is the on-
going behavior at the time of reward. When we delay knowledge we sim-
ply decrease the possibility that the behavior which we intend to reward is
actually rewarded, and increase the possibility that some relatively irrelevant
response is rewarded. W e must remember that the subject can modify the
operation of this delay gradient by developing an ability to reinstate sym-
bolically both the stimulus situation and the response. T o the extent that
he is able to do this, he can make quite satisfactory use of information given
after rather considerable delays.
It is probable that there is an optimum delay for every task and every
stage of learning. If the knowledge comes too soon, it often cannot be
used. Hick (27) found that time to change a response when more force is
called for is about 0.3 second in the case of a hand and arm movement. T h e
spectacle of a batter in baseball swinging at a slow ball even when he
“knows” he is too early is a common one. Even if the knowledge can be
used, this use may retard the development of the subject’s ability to make
use of other more permanent or stable cues. In Houston’s study (28) of
gunnery, the use of the red filter to indicate hits probably slowed down the
subject in learning to judge his own hits from relevant cues of size and
direction of movement of target. H e simply came to rely on the red filter
for information. I n many instances, delay of knowledge allows the per-
 290 J O U R N A L OF G E N E R A L PSYCHOLOGY

former to make a better over-all evaluation of his performance when he is

finally given the information. Delay might well be useful to a person who is
following directions in carrying out a complex task. Some time later he may
gain much by making a verbal comparison of his activities and the “correct”
ones.
7. Generalization 7
In the case o f discontinuous tasks where knowledge of performance is given,
small intervals between trials are generally better for learning than are longer
ones. T h i s generalization is based directly on only one study. Macpherson,
Dees, and Grindley ( 3 6 ) had subjects push a lever up to a given pressure
against a weight. T h e subjects preferred intervals of only a few seconds
because these short intervals allowed them to “hold the feel” of the correct
Downloaded by [McMaster University] at 10:58 14 October 2014

movement. Objectively, performance measure favored the shorter intervals

with this and several simple tasks. T h e authors speculated that there might
also be an additional effect from the short intervals in that the subjects had
less time to let their attention wander. I n any case, a large loss in perform-
ance was usually found over a 24-hour period.
Superficially this generalization seems to be in conflict with the results
from distribution of practice studies. However, not all such studies report
benefit from distribution, as can be seen in any review of the problem. T h e
present author has found, for example, that in simple perception studies with
Rorschach cards there is often better performance with intervals of less than
15 sec. between trials where the task is discrete (see 9 for description of ex-
perimental technique). T h e subject is apparently better able to “keep in
mind” his previous response. Early in practice the optimum interval between
trials may be somewhat longer than it is later on. Macpherson, Dees, and
Grindley ( 3 6 ) observed that, when subjects had not yet mastered the opera-
tion of their apparatus, they tended to become confused if the interval be-
tween trials was too short.

8. Generalization 8
W h e n knowledge of performnnce is decreased, performance drops. Whether
the drop would bring the performance back to the same level as that if this
lesser degree of knowledge of performance had obtained all along cannot
be ascertained in most of the following studies. It would have been necessary
to give a great many trials in the new conditions, and seldom was this done.
T h e following studies seem to have a bearing on the generalization.
a. W i t h subjects throwing a ball over their shoulders a t a target, Lorge
 R. B. AMMONS 291
and Thorndike (34) found a definite decrease in performance when they
stopped telling the subjects where the ball hit.
b. Seashore, Underwood, Houston, and Berks (47;53, pp. 414-415)
observed a definite drop in ranging scores by trainees on the Pedestal Sight
Manipulation Test when they stopped using the red filter arrangement to
indicate firing hits on test trials.
c. Houston (28) also used the Pedestal Sight Manipulation Test with
the red filter over the target image to indicate hits. H e found a sharp drop
in ranging scores when he stopped using the red filter. Although these people
did not drop all the way to the level of the group which had never been
given knowledge of performance, the difference in the final levels was nor
statistically significant.
d . In another study with the Pedestal Sight Manipulation Test, Morin
Downloaded by [McMaster University] at 10:58 14 October 2014

and Gagne (40) also found that ranging performance dropped to a level
not appreciably higher than that of a control group when the red filter was
no longer used to give knowledge of performance.
e. Elwell and Grindley (23) reported a big drop in the performance
of a two-hand coordination test when they stopped giving obvious knowledge
of results.
f. In Thorndike’s classic experiment (49) with line drawing, his sub-
jects showed a large decrease in performance on their last (test) trial dur-
ing which trial no regular knowledge of performance was given.
T h e finding of a drop when knowledge of performance is removed is
by no means universal. In the Tufts College study (52) with a gunnery
range and height finder, five of six subjects dropped appreciably when the
instructor stopped sounding a buzzer when they were off-target. T h e
sixth subject did not drop, however. In a study with the Pedestal Sight
Manipulation Test, Bilodeau ( 14) found no final statistically significant
differences when the red filter was no longer being used among groups given
varying amounts of knowledge of performance with the filter during train-
ing. However, as shown quite clearly in his Figure 2, the means (all 361)
range themselves consistently with the hypothesis that there is a residual
and that it is proportional to the amount of training with the filter. T h e
control group is lowest, the 50 per cent red filter next, and the 100 per cent
group highest with the difference being quite appreciable. Perhaps the sta-
tistics were not sensitive enough to detect the difference. I t is also possible
that there would be more residual if more training were given. Biel, et al.
(10) found no decrease at all in mean performance of men pointing guns
when they were no longer given knowledge of performance in the form of
 292 J O U R N A L OF GENERAL PSYCHOLOGY

a buzzer sounded by the instructor. However, mean improvement by the

group was observed only where the buzzer was being used.
Some hint as to a possible basis for the drop when knowledge of perform-
ance is withdrawn may be found in the work of Macpherson, Dees, and
Grindley (35). These investigators observed that a t least in the case of
simple tasks the problem was primarily motivational. Their subjects showed
a drop on the first trial without knowledge of performance, not before
knowledge had actually been withheld, but after an announcement that no
information would be given after the next trial. Although this “motiva-
tional” drop was not observed for some skills, it was found with line draw-
ing and estimation of time interval. Ross (42) had school children make
tally marks as rapidly as possible. T h e y did not decline in performance
when he stopped giving them knowledge of performance in the form of scores,
Downloaded by [McMaster University] at 10:58 14 October 2014

and he speculated that this might well be due to the presence of a group
motivational phenomenon.

9. Generalization 9
W h e n knowledge of performance is decreased, performance drops more
rapidly when trials are relatively massed. Macpherson, Dees, and Grindley
(36), using simple motor tasks, demonstrated this in the case of knowledge
of performance. I t should be noted that this generalization depends on only
one study. No other pertinent data are available.

10. Generalization 10
W h e r e subjects are not being given supplementary knowledge of per-
formance by the experimenter any longer, the ones w h o maintain their per-
formance level probably have developed some substitute knowledge o f per-
formance. I t should be clear that there is often no way to test this gen-
eralization. I t is clearly supported, however, by the results in an ergographic
study by Arps (7, 8 ) . T h e subjects reported developing an imagery which
substituted for actual knowledge of performance. Elwell and Grindley
(23) proposed that those subjects who succeeded in maintaining a high
level of performance for any length of time after knowledge of results had
been removed were very probably using, consciously, some clue, such as the
kinaesthetic sensations from the wrists a t the end of each movement, as a
check on that movement, and as a guide to the nature of the correction re-
quired in the next movement. T h i s development of substitutes for supple-
mentary knowledge of performance could have very great practical signifi-
cance.
 R. B. AMMONS 293
1 1. Generalization 11
W h e n direct (supplementary) knowledge of Performance is removed, sys-
tematic “undershooting” or “overshooting” may appear in performance. Eaton
( 2 2 ) had his subjects make a rotary movement which produced a circle
on a blank piece of paper, with overlapping or failure to close it measured as
an error. Although no direct evidence for this generalization can be found
in the study, it was noted that there was definite overshooting (overlapping)
to begin with and that much of the improvement consisted of eliminating
this.
Direct evidence is found in a study reported by Dees and Grindley (21).
When subjects were not given their scores as usual, there was systematic
overshooting with knob-turning and lever-pushing against pressure, but
none with holding a telegraph key closed for a given interval. There is a
Downloaded by [McMaster University] at 10:58 14 October 2014

possibility that interval between trials and amount of overshooting or under-

shooting are related. It was noted by Dees and Grindley that there was
systematically greater overshooting with knob-turning and lever-pushing
where there was a short interval (1.8 sec.) between trials than where there
was a longer interval (33 sec.).
These 11 generalizations summarize the information which is currently
available in the field of knowledge of performance.
C. DISCUSSION
W h a t is meant by “giving knowledge of performance?” Does it mean
just to make the stimuli available to the subject? Or does it imply some-
thing about training him to use the available information? With these
thoughts in mind, plus the problem of transfer from training to test situa-
tion, the author has made a brief outline of statements which might form
the basis for a theoretical formulation of the general problem of how learn-
ing and performance are affected by knowledge of performance.8 Since,
however, the generality of these statements is far beyond that of the present
paper, no attempt has been made to marshal evidence for them. Where an
illustration occurred, it was included.
1. Statement I
T h e performer can learn t o recognize an object correctly in a given situa-
tion. Often such recognition is the necessary first step in any successful
SA systematic series of statements dealing with the knowledge of performance
variables in a discrete task situation is presented elsewhere ( 5 ) . Parts of it are so
highly speculative that no attempt has been made to develop the system here.
 294 JOURNAL OF GENERAL PSYCHOLOGY

or satisfactory behavior sequence. T o fire a t a target effectively, one must

first be able to recognize this target. T h e recognition of camouflaged gun
emplacements would be this kind of behavior. A study by Atkinson and
Ammons ( 9 ) has shown how recognition behavior develops. W h e n the
subject first looks a t unstructured or meaningless material (an ink blot, in
this case), he sees nothing special. However, with experience he comes to
be able to recognize it quickly as some particular thing with a familiar con-
tour and (often) internal organization. Knowledge of performance pro-
vides reward for this recognition behavior and thus speeds its acquisition ( 4 ) .

2. Statement 2
T h e performer can learn t o discriminate along a sensory continuum in a
Downloaded by [McMaster University] at 10:58 14 October 2014

given situation. T h i s discrimination can be along a simple continuum, such

as pitch of sound or brightness on a radar scope. Usually, however, it is on
a much more complex continuum, with many sensory dimensions combined
o r acting together. W e know little about how ability to discriminate along
this kind of continuum transfers from a training situation to any other.
Several examples of this type of learning come to mind. Gibson and Smith
( 2 4 ) , in a study of depth perception, found that subjects learn to discrim-
inate distance of objects from specially prepared photographs with other
cues minimized. Improvement by subjects in estimating the length of a
line was reported by Smith (48). Hamilton’s subjects (25) improved a t
the task of discriminating lengths. Knowledge of performance is basic to
this discrimination learning.

3 . Statement 3
T h e performer can learn to discriminate between various similar acts by
himself in a given situation. It can be readily seen that Statements 1, 2,
and 3 give a rough indication of the source and kind of “information” which
can become available to the subject in the knowledge of performance situation.

4. Statement 4
T h e performer can establish discriminating control over his various acts
in a given situation. I t should be pointed out that the performer’s ability
to discriminate is definitely limited, and that probably for most purposes it
is not very fine. T h u s the flexible gunner may well be unable to discrim-
inate between many of the pairs of paths over which he would move the
gunnery apparatus in following an attacking plane.
 R. B. AMMONS 295

5. Statement 5
T h e potential to make a more nearly “correct” (physically) response is
increased as the performer learns to make finer and finer (more precise)
response discriminations of the kind mentioned in Statement 4. These dis-
criminations con be transferred from task t o task. A t the start of flexible
gunnery training, the trainee often feels lucky to point the “gun” in a gen-
erally correct direction and “fire.” Improvement consists in part of learning
to aim more precisely, fire only a t certain angles, lead by so much, etc. T h a t
is, the trainee learns to make finer adjustments within the dimension of aim-
ing and firing. Two studies of rotary pursuit by Ammons (2, 3) give evi-
dence on this point rather clearly. I n both cases it was found that the over-
all improvement of rotary pursuit performance is accompanied by a decrease
in the average duration of “misses” or times off target. I t is apparent that
Downloaded by [McMaster University] at 10:58 14 October 2014

the performer is learning to make finer adjustments of his responses to the

speed of the target and the shape of the path followed by the target.
6. Statement 6
T h e more similar a “new” situation is t o an rrold’’ (i.e., familiar) situa-
tion, the greater the transfer of fine response adjustments (discriminations)
f r o m the old t o the new situation. Abilities to put accurately varying pres-
sures on a trigger mechanism will transfer more to a new but somewhat
similar trigger arrangement than to a knob-turning control. Here, again,
reference is to the kind of discriminating motor control mentioned in State-
ment 4, based on knowledge of performance.
7. Statement 7
T h e more similar a “new” situation is t o an “old” (i.e., familiar) situa-
tion, the greater the transfer o f stimulus discriminations f r o m the old t o
the new. This transfer problem always comes up when trainees are shifting
from a preliminary training method giving much information about per-
formance to an operational situation giving much less information about
performance. A simple illustration is found in the work of Keller (31, 32)
on code learning. H e observed that subjects had trouble shifting from a
code receiving method giving quick knowledge of performance after each
letter to a situation calling for larger units of material to be received before
knowledge of performance was given.
8. Statement 8
T h e ability t o behawe more adequately (response more nearly ‘rcorrect”)
will increase as the trainee learns t o discriminate and recognize the finer waria-
 296 J O U R N A L OF GENERAL PSYCHOLOGY

tions in the stirrruIus situation. I t is possible that in a thoroughly worked

out system this might be a deduction. In any case, it is apparent that as
the performer comes to differentiate the stimulus situation, he is better able
to “apply” the appropriate response. I t may be that this could be the major
function of the giving of “artificial” or supplementary knowledge of per-
formance early in training-the performer is motivated to learn to make the
necessary stimulus discriminations.
These eight statements are included only to provoke thought about the
basic processes involved in the change of performance due to knowledge of
performance. From these statements, most of the findings can be predicted,
a t least loosely. It is apparent that much more experimentation must be
done before a comprehensive learning theory, accounting for the operation
of knowledge of performance, can be developed.
Downloaded by [McMaster University] at 10:58 14 October 2014

D. SUMMARY
Knowledge of various kinds which the performer receives about his per-
formance affects his behavior. These effects have been summarized in the
form of 11 generalizations, each followed b y the available supporting evi-
dence from research studies. A theoretical framework is suggested for an
organized, systematic approach to the processes underlying the phenomena of
knowledge of performance.
REFERENCES
1. ALEXANDER, L. T. Knowledge of results and the temporal gradient of rein-
forcement. Unpublished doctor’s dissertation, Ohio State Univ., 1950.
2. AMMONS,R. B. Effect of distribution of practice on rotary pursuit “hits.” J.
Exper. Psychol., 1951, 41, 17-22.
3. .
- An analysis of “hits” in continuous rotary pursuit before and after
a single rest. J . G e n . Psychol., 1953, 48, 3-10.
4. ~ . Experiential factors in visual form perception: I. Review and for-
mutation of problems. J . Genet. Psychol., 1954, 84, 3-25.
5. .
- Knowledge of performance: Survey of literature, some possible
applications, and suggested experimentation. Wright-Patterson A i r Force Base,
Ohio: W r i g h t Air Development Center, Aero Medical Laboratory, February,
1954. (WAUC Technical Report 54-14.)
6. ANGELL, G. W. T h e effect of immediate knowledge of quiz results on final
examination scores in freshman chemistry. J . Educ. Res., 1949, 42, 391-394.
7. ARPS,G. F. A preliminary report on work with knowledge versus work with-
out knowledge of results. Psychol. R e v . , 1917, 24, 449-455.
8. ~ . W o r k with knowledge of results versus work without knowledge
of results. Psychol. Monog., 1920, 28, No. 125.
9. ATKINSON,R. C., & AMMONS, R. B. Experiential factors in visual form percep-
tion: 11. Latency as a function of repetition. J. Enper. PJychol., 1952, 4,
173-178.
 R. B. AMMONS 297

10. BIEL, w. C., BROWN,G. E., GOTTSDANKER, R. M., & HALL,R. c. The effective-
ness of a check sight technique for training 40-mm. gun pointers. Tufts Col-
lege, 1944. (OSRD Report 4054.)
11. BILODEAU, E. A. A further study of the effects of target size and goal attain-
ment upon the development of response accuracy. San Antonio, Texas: Human
Resources Research Center, Lackland Air Force Base, February 1952. (Re-
search Bulletin 52-7.)
12. . A preliminary study of the effects of reporting goals as a function
of different degrees of response accuracy. San Antonio, Texas: Human Re-
sources Research Center, Lackland Air Force Base, January 1952. (Research
Bulletin 52-4.)
13. . Some effects of modification of information about a previous re-
sponse upon the acquisition of two lever positioning habits. San Antonio.
Texas: Human Resources Research Center, Lackland Air Force Base, Jan-
uary 1952. (Research Bulletin 52-4.)
14. . Some effects of various degrees of supplemental information given
a t two levels of practice upon the acquisition of a complex motor skill. San
Antonio, Texas: Human Resources Research Center, Lackland Air Force Base,
Downloaded by [McMaster University] at 10:58 14 October 2014

April 1952. (Research Bulletin 52-15.)

15. .- Speed of acquiring a simple motor response as a function of sys-
tematic transformations of knowledge of results. San Antonio, Texas: Human
Resources Research Center, Lackland Air Force Base, November 1952. (Re-
search Bulletin 52-33.)
16. BILODEAU, E. A., & MORIN,R. E. A preliminary investigation of the effect of
removal of the pipper from the projected reticle pattern of the Pedestal Sight
Manipulation Test. San Antonio, Texas: Human Resources Research Cen-
ter, Lackland Air Force Base, February 1951. (Research Note 51-2.)
17. BILODE,AU, E. A., & MORIN,R. E. Proficiency on the Pedestal Sight Manipula-
tion Test with and without the tracking pipper. San Antonio, Texas: Human
Resources Research Center, Lackland Air Force Base, December 1951. (Re-
search Bulletin 51-27.)
18. BOOK,W. F., & NORVELLE, L. The will to learn: an experimental study of incen-
tives in learning. Pedagog. Sem., 1922, 29, 305-362.
19. BROWN,J. S. A proposed program of research on psychological feedback
(knowledge of results) in the performance of psychomotor tasks. In Re-
search planning conference on perceptual and motor skills. San Antonio,
Texas: Human Resources Research Center, Lackland Air Force Base, Novem-
ber 1949. (Conference Report 49-2.) Pp. 81-87.
20. CRAFTS,L. W., & GILBERT,R. W. T h e effect of knowledge of results on maze
learning and retention. J . Educ. Psychol., 1935, 26, 177-187.
21. DEES, V., & GRINDLEY, G. C. T h e effect of knowledge of results on learning
and performance: IV. T h e direction of the error in very simple skills. Quart.
J . Exper. Psychol., 1951, 8, 36-42.
22. EATON,M. T. A study of latent learning. J . Exper. Pzychol., 1935, 18, 683-707.
23. ELWELL,J. L., & GRINDLEY, G. C. T h e effect of knowledge of results on learn-
ing and performance: I. A coiirdinated movement of the two hands. Brit. J .
PJychol., 1938-9, 39, 39-54.
24. GFBSON, E. J., & SMITH, J. T h e effect of training in distance on the judgment
of size-at-a-distance. San Antonio, Texas : Human Resources Research Cen-
ter, December 1952. (Research Bulletin 52-39.)
25. HAMILT~N, H. C. T h e effect of incentives on accuracy of discrimination meas-
ured on the Galton bar. A r c h . of Piychol., 1929, No. 103.
 298 J O U R N A L O F GENERAL PSYCHOLOGY

26. HELMSTADTER, G. C., & ELLIS, D. S. Rate of manipulative learning a s a func-

tion of goal-setting techniques. J . Exper. Psychol., 1952, 43, 125-129.
27. HICK, W. E. Reaction time for the amendment of a response. Quart. J . E x -
per. psycfio!., 1948-9, 1, 175-179.
28. HOUSTON,R. C. T h e function of knowledge of results in learning a complex
motor skill. Unpublished master’s thesis, Northwestern Univ., 1947.
29. JOHANSON, A. M. Influence of incentive and punishment on reaction time.
A r c h . of Psychol., 1922, 8, No. 54.
30. JUDD, C. H. Practice without knowledge of results. Psychol. R e v . M o n o g .
SUPPI., 1905-6, 7, 185-198.
31. KELLER,F. S. Studies in International Morse Code. I. A new method of teach-
ing code reception. J . A p p l . Psycfiol., 1943, 8, 407-415.
32. . T h e Radio Code Research Project: Final report of Project SC-88.
New York: Psychological Corporation, 1945. (OSRD Report 5379.)
33. LINDAHL,L. G. Movement analysis as a n industrial training method. J. Appl.
Psychol., 1945, b,420-436.
LORGE, I., Sr THORNDIKE, E. L. T h e influence of delay in the after-effect of a
Downloaded by [McMaster University] at 10:58 14 October 2014

34.
connection. J. Exper. Psychof., 1935, 18, 186-194.
35. MACPHERSON, S. J., UEES,V., & GRINDLEY, G. C. T h e effect of knowledge of re-
sults on learning and performance: 11. Some characteristics of very simple
skills. Quart. J . Exper. Psychol., 1948-9, 1, 68-78.
36. .- T h e effect of knowledge of results on learning and performance.
111. T h e influence of the time interval between trials. Quart. J. Exper. Psy-
chol., 1948-9, 1, 167-174.
37. MELTON,A. W. (Ed.) Apparatus Tests. A A F Aviation Psychology Program Re-
search Report No. 4 , 1947.
38. MORGAN, C. L., & MORGAN, L. V. Effects of immediate awareness of success and
failure upon objective examination scores. J. Exper. Educ., 1935, 4, 63-66.
39. MORIN,R. E. T h e functions of informative feedback and rewarding feedback
in acquisition of a lever positioning habit. San Antonio, T e x a s : Human Re-
sources Research Center, Laekland A i r Force Base, December 1951. (Re-
search Note 51-11.)
40. MORIN,R. E., & GAGNE, R. M. Pedestal Sight Manipulation T e s t performance
a s influenced by variations in type and amount of psychological feedback.
San Antonio, T e x a s : H u m a n Resources Research Center, Lackland A i r Force
Base, October 1951. (Research Note 51-7.)
41. PRESSEY, S. L. Development and appraisal of devices providing immediate
automatic scoring of objective tests and concomitant self-instruction. J. of
Psychol., 1950, 29, 417-447.
42. Ross, C. C. A n experiment in motivation. J . Educ. Psychol., 1927, 18, 337-346.
43. . T h e influence upon achievement of a knowledge of progress. J.
Educ. Psycho[., 1933, %, 609-619.
44. S A C K ER. ~ , S. T h e effect of knowledge of scores on learning a simulated track-
i n g problem. A m e r . Psychol., 1947, 2, 299. (Abstract.)
45. SALTZMAN, I. J. Delay of r e w a r d a n d human verbal learning. J . Exper. Psy-
cho[., 1951, 41, 437-439.
46. SEASHORE, H., & BAVELAS,A. T h e functioning of knowledge of results in
Thorndike’s line-drawing experiment. Psycho/. Rev., 1941, 48, 155-164.
47. SEASHORE, R. H., UNDERWOOD, B. J., HOUSTON,R., & BERKS,L. T h e influence
of knowledge of results on performance. Unpublished manuscript. (See Un-
derwood ( 5 3 ) ).
 R. B. AMMONS 299

48. SMITH,F. 0. Repetition without knowledge of results as a factor in learning.

Psycho!. Bull., 1933, 30, 673-674. (Abstract.)
49. THORNDIKE, E. L. T h e law of effect. Amer. J . Psychol., 1927, 39, 212-222.
50. THORNDIKE, E. L. Human Learning. New York: Century, 1931.
51. TROWBRIDGE, M. H., & CASON,H. An experimental study of Thorndike’s theory
of learning. J . Gen. Psychol., 1932, 7 , 245-260.
52. Tufts College. Knowledge of results training in ranging on a moving target.
Report to the National Defense Research Committee, Project 10, Section D-2.
53. UNDERWOOD, B. J. Experimental Psychology. New York: Appleton-Century-
Crofts, 1949.
54. WARRICK, M. J. Effect of transmission-type control lags on tracking accuracy.
Dayton: Wright-Patterson Air Force Base, 1949. (USAF Tech. Rep. No.
5916.)
55. WATERS,
R. H. T h e specificity of knowledge of results and improvement. Psy-
chol. Bull., 1933, 30, 673. (Abstract.)
56. WRIGHT, W. R. Some effects of incentives on work and fatigue. Psychol. Rev.,
1906, 13, 23-24.
Downloaded by [McMaster University] at 10:58 14 October 2014

Department of Psychology
University af Louisville
Louisville 8, Kentucky