The Motivational Effects of Knowledge of Results:

The Influence of Goal-setting

Abstract

Research on knowledge of results (KR) has generally not controlled

for motivational effects resulting from differential goal-setting. The

present experiment was carried out to separate the effects of KR and

goal-setting using a ZxZ fixed model design; the variables were KR
versus No KR; and specific hard goals versus "do best" goals. The

goals (manipulated by instructions) were representative of the goals

typically assigned (explicitly or implicitly) to KR and no KR Ss in previous
studies, respectively. No difference was found between KR and no KR
groups, but a significant goal effect was found in favor of Ss given specific hard
goals. The results indicated that effects previously attributed to differen-
tial KR were actually due to different levels of motivation produced by the
different goals.

:' Journal of Applied Psychology, 1967 (in press).

C , 1

• •,16
ISCLAIMEI NOTICE

THIS DOCUMENT IS BEST

QUALITY AVAILABLE. THE COPY
FURNISHED TO DTIC CONTAINED
A SIGNIFICANT NUMBER OF
PAGES WHICH DO NOT
REPRODUCE LEGIBLY.
 The Motivational Effects of Knowledge of Results:
The Influence of Goal-settingI

Edwin A. Locke

American Institutes for Research

Washington Office

The positive influence of knowledge of results (KR) on learning and

performance is one of the best established findings in the research literature
(Ammons, 1956; Bilodeau & Bilodeau, 1961, etc.) Research on KR has not

been overly concerned with theoretical issues, but the fact that occasional
studies fail to find significant KR effects on performance (e.g. , Chapanis,
1964) emphasizes the need to integrate and systematize findings in this area.

Several investigators (e. g. , Vroom, 1964) have distinguished between

the information or cueing function of KR and the motivational function. The

former refers to information given to Ss regarding the nature and locus of

errors and the nature and type of the correct response (epistemic KR). Most

studies of KR or "feedback" have been of this type (e. g. , Bilodeau & Bilodeau,
1961). Given a constant motivational state, the more information given to S
about the task or about how to correct errors, the better his performance
level or learning.

The manner in which knowledge functions to niotivate S is much mor-

complicated. One possibility for differentiating be twe'n episterni- .and
motivational KR is to give S knowledge that cannot be uscd for correcting
errors or for changing the timing or locus of the respunsc. Examples ot

KR that do not cue S as to a better method of performing would be: total

time-on-target on a pursuit rotor task; total score after several trials on a
computation task; total score after several throws on a dart throwing task.

-I-

Lr
Knowledge of score on simple motor twsks such as reaction time and weight
lifting also falls into this category, since there are no right or wrong
responses on such tasks. Any effects on performance of this type of KR may

be attributed to motivational factors since presumal ly only S's level of effort

is influenced.1

In the "classic" studies dealing with motivational KR such as those of

Book and Norvell (1922), Crawley (1926), Mackworth (1950), and Ross (1927),

Ss given knowledge of score were usually told to try to improve their perform-
ance or were given explicit goals to reach, whereas the No KR Ss were told
not to think about trying to improve their scores or were told simply to "do
their best. " Although the KR groups performed better than the No KR groups
in these experiments, the effects of goal-setting were confounded with the
effects of knowledge of score, making it impossible to determine which was
the critical variable or the interdependence, if any, between them. Similar

criticisms can be made of recent studies, e.g. , Ch,'ch ano Camp (1965);

McCormack, Binding, and Chylinski (1962); McCormack, Binding, and

McElheran (1963); and Payne and Hauty (1955). In these studies goals were

not manipulated explicitly, but KR was given in relation to standards (such

as S's best previous score) which clearly must have suggested goals to the Ss.

In five other studies (Arps, 1920; Johanson, 1922, Mace, 1935, Manzer,

1935; Gibbs & Brown, 1955) positive results were obtained for KR without
explicit manipulation of Ss' goals; however, no attempt was made to deter-
mine if different goals were set by KR and No KR groups (although Mace
argued that the effect of KR was to suggest a performance standard to the KR
Ss). In contrast, Chapanis (1964), attempting to replicate the finding of Gibbs
and Brown (1955), did two things to eliminate "demand characteristics" (Orne,
1962) that may have affected previous results: 1) he "hired" Ss as employees

doing a job for pay (punching digits onto a tape) rather than as experimental
Ss; and 2) he ran Ss individually rather than in groups (as was the case in most

previous experiments in this area). In the absence of the implicit goal setting

-2-
demands (e.g. , "improvement") inherent in the typical experimental situa-
tion and the possible effects of implicit competition, Chapanis found no
effect of KR on performance. Further, Locke and Bryan (1966b) found no
overall effect of KR on performance on a complex c(mputation task. But
when Ss were re-grouped according to their a posteriori performance goal
descriptions, a significant relationship of goals to pcrformance was found,
suggesting that differences in Ss' goals had more effect on performance

than differences in KR.

The above findings suggest that the effects of KR should be separated

from the effects of goal-setting in order to determine whether KR automati-

cally influences performance level or whether its effects are mediated by

goal-setting activity. The purpose of the present experiment was to test the
hypothesis that effects previously attributed to differential KR were actually
due to different performance goals associated with the different KR conditions.

The major premise underlying this hypothesis is that level of effort on a task
is determined largely by S's conscious performance goals. This emphasis
on a cognitive approach to motivation is supported by recent theoretical
developments (e. g. , Dulany, 1962; Miller, Galanter, & Pribram, 1960;
Ryan, 1958, 1964; Spielberger, 1965) and by the findings of a number of
recent studies (e.g., DeNike, 1965; Dulany, 1962; Locke, 1966a, 1966b;
Locke & Bryan, 1966a, 1966b; Spielberger, Berger. & Howard, 1963; Spiel-
berger, Bernstein, & Ratliff, 1966; Spielberger, Levin, & Shepard. 1962).

Method

Task. The task was simple addition. Each problem consisted of three
two-digit numbers and was presented on a separate 3x5 index card. The
cards were placed consecutively in boxes holding 720 cards each. The Ss
wrote their answers on sheets containing space for 90 answers. As each
answer sheet was completed, S was instructed to insert it in a slot underneath
a one-way mirror through which he was observed by E rduring the experin•nt.

-3-
The S worked 1'L" oMe. hour at the task, which was divided into five trials
separated by short rest pertods. The trials were alternately 10 and 15 min-
utes in length (i.e. , 10'; 15'; 10'; 15'; and 10'). All Ss were told the length
of each tria; in advance.

Subjects. The Ss were 24 male and 12 female naid college (under-

graduate) voluntec,-s selected from a larger pool of 62 volunteers. All Ss
in the original pool were given a pre-test consiE ting of three I-minute trials
on the addition task and asked to make three attitude ratings indicating their
liking for and interest in "tasks like this. " From this pool four matched
groups of nine Ss each were selected as follows: four Ss were chosen who
were approximately equal in ability and equal on the three attitude ratings;
each of these Ss was then assigned at random to one of the four experimental
conditions. This procedure was repeated until there were nine Ss in a cell.
Thus the cells had almost identical means and distributions on the four
matching variables. No attempt was rnade to match for sex; but as it turned
out, the proportion of females to males was approximately equal for the

main effects.

Design and Proc •cui-e

The experiment was introduced as a study of the development of attitudes

toward the task. The structure of the task was explained. The Ss were told
not to spend a lot of time checking answers since the total number correct
was more important than the percent correct. While S worked, E stayed in
the adjacent observation room. The S was made aware of the one-way mirror
and was told E would remain there during each trial so as not to disturb S.

Communication was possible through an intercom systen.

The design was a 2xZ fixed model. The fixed variables were knowledge
of score (KR) and type of goal.

KR Condition. The KR Ss were told the number of problems they had

gotten e'jrrect at the end of each trial.

-4-
 No KR Condition. The No KR Ss were not given their scores. Since

Ss had to hand in each answer sheet after completion and since they did not
know how many of their answers were wrong, these Ss could not easily keep
track of their score except in a very general way.

Do Best Goal. "Do Best" Ss were told to "do their best" on each trial.

(This was the goal typically given to NoKR Ss in the experiments discussed
earlier.) Do Bez.t Ss in the KR condition could not easily use their scores
to set specific personal goals due to the alternating trial lengths. In addition,

Do Best-KR Ss were not allowed to perform any computations between trials

that might enable them to determine their rate per minute (a pre-condition
for setting a goal such as "improvement").

Hard Goal. Goals were set for the Hard Goal Ss on the basis of the
scores attained by the matched "Do Best" Ss. On the first trial a given S's
goal was set about 10 percent higher than the score achieved on the same

trial by a matched Do Best S. Then E adjusted S's goal before each succeeding
trial depending upon how well S had done on the previous trial. If S did not
get near the goal, the next goal was lowered slightly; if S reached or exceeded
the goal, the succeeding goal was raised.

The goals were marked by means of a colored 3x5 index card placed
vertically at the appropriate point in the box of problem cards. This card

represented the point they had to reach by the end of the trial in order to
reach their goal (if they got all problems correct). The Ss were told to try
and surpass this point since they were bound to get sorvie wrong. They were

told at the end of each trial whether or not they had blawil) the ga-l but were'
not given their actual scores unless they were in the KR condition.

On the average the goals of the Hard Goal Ss were set from 0% to 32%
(mean and median = 11%) above the scores attained by the matched Do Best
Ss. Hard Goal Ss wereable to reach or beat their goals on 16% oi the trials.

The Ss had approximately two minutes' rest between trials during

which E corrected their answer sheets (if they were in the KR condition) and
determined th'ir new goals (if in the Hard Goal condition). To set the new
goal, E: came in the experimental roomr• and placed the colored card Mt the
appropriate place in the card box. No Ss were told how maray problems werte
nieeded to reach the goal, and Lhey could getya roLugh idea of this by seving

where the colored card was placed in the card box.

At the end of the experiment Ss were asked to look at a set of cards,

each of which described a possible goal they could have had (e. g. , "I tried
to do my best"; "I tried for the assigned goal") and were asked to pick out
the card that best represented their performance goal during the experiment.

Results

Three different performance measures were used: (a) deterioration

scores defined as the difference between the mean number of problems cor-
rect per minute on the three pre-test trials and the mean number of problems
correct per minute during the five experimental trials; (b) linear slope scores
calculated from the number of problems correct per minute on the five experi-
mental trials only; and (c) percent error on the five experimental trials.

The performance curves for each experimental sub-group, in terms o)f

total problems correct per minute, are shown graphically in Figure 1.i It is
evident that both Hard Goal groups did progressively better than the Do Best
groups, the difference between them reaching a peak in the last (10') trial

period. On the other hand, the KR and No KR curves did not diverge except
for the Hard Goal group in the last 10-minute period.

Insert Figure 1 about here.

The results of analyses of variance (using a standard Zx2 fixed model

design) for the three performance measures are shown in Table 1. The goal
effec't was significant using both deterioration scores-, F (1. 32) = 4. 8M,
p < .Or', and linear slope scores. F (1, 3Z) = 12.87., p < .01, with theI lHard

Goal group bcing superior to the Do Best group in both catses. A mantched

-6-
groups test comparing the Hard Goal and D)o B~est groups (performed by
matching each Hard Goal S with a Do Dest S of equal initial ability) yielded
a t-ratio for mean deterioraticn of 2.80 (17, p ( .02), and a similar test

using slope scores yielded a t of 3. 18 (17, p ( .01). There was no overall

effect of KR on any of the performance measures, nor waF there evidence of
any interaction between KR and type of goal.

Insert Table I about here.

There were no significant main effects nor any interaction for the
percent error criterion, indicating that the higher performance level attained
by the Hard Goal group was not achieved at the expense of a relatively greater
number of errors in relation to problems attempted. Both the Hard Goal and

Do Best groups averaged about 5.6% errors during the experimental trials.

The results of the post-experimental questioning indicated that all Hard

Goal Ss were either trying explicitly for the goals or had them in mind as
they worked along. Eleven of the Do Best Ss said they were trying to "do
their best"; four said they were trying to go at a "reasonably fast pace"; two
tried to go at a "normal" or "slow" pace; and one S reported that he tried to
complete one box of (720) problem cards by the end of the hour. The latter
S was the only S in the Do Best condition who worked at a faster pace (in
terms of problems correct per minute) during the e'xperimental than during
the three pre-test trials.

Discussion

The results of the present experiment support the hypothesis that moti-
vational effects previously attributed to differential knowledge of results were
actually a function of differential performance goals associated with the KR
conditions. When differential goal setting by KR and No KR Ss was controlled,
no overall effect of KR on performance was found. Howc\.er, when the effccts

of the goals typically set by KR and No KR S: in previou.s vxp.riments (e. g.

-7-
specifi Hard GoQ'% versus L, Besýj were compared, a significant goal effect
was found. These results were consistent with previous findings by Locke
and Bryan (1966b) where goals were measured by post-experimental inter-

views rather than manipulated by instructions.

Although the findings of the present study provide support for the initial
hypothesis, there are a number of issues still open. It may be observed from
Figure 1 that there was an effect of KR condition between trials 4 and 5. The
KR groups increased their performance rate on trial 5, whereas the No KR
groups decreased. The F for the difference scores (1, 32) was 8.83 (p < .01).

It would be useful to determine whether this effect was due to implicit goal-
setting on the part of the KR Ss, who by this time may have been able to get
some idea of their rate, or to some other effect.

Only two goal classes were used in the present study. In one sense the

Hard Goals in the present experiment were harder than the Do Best goals,
since the Hard Goals were set above the performance level of the Do Best Ss.

However, just "how much" harder is not known. A greater variety of per-
formance goals should be utilized in subsequent studies and some attempts
made to scale the goals as to difficulty or motivational level.

In more general terms, the results of the present study suggest one
mechanism by which incentives of all types might work (e. g. , KR, money,
instructions, participation, praise, reproof, verbal "reinforcement, " etc.);
it is possible that such incentives are effective only to the degree that they

affect S's goals or intentions. Research in the area of verbal learning

'DeNike, 1965; Dulany, 196Z; Spielberger et al, 1962) suggests thait a verbal
reinforcement (given that the response-reinforcement contingency is known)
is effective in changing behavior to the precise extent that S desires or intends
to get the reinforcement. In view of the seemingly inconsistent and often
unpredictable effects of incentives on performance, it seems probable that

the use of goals and intentions as mediating variables will be necessary if

fully adequate explanations cf the effects of incentives are to be achieved.

-8-
 The present results suggest that in order to predict the effect of

knowledge of score on performance level, ;t is nut enough to know that

the individual has such 'nowledge. It is also necessary to know what he
does with it ,i.e. how he evaluates it and what goals he sets in response
to it.

-9-
 Footnotes

1. This research was supported by contract Njnr 4792(00) between the

Office of Naval Research and the American Institutes for Research. The
author would like to thank Miss Judith Bryan of A. I. R. for analyzing the
data for this study and Dr. Charles D. Spielberger of the National Institutes
of Mental Health for his many helpful comments on this paper.

2. While it is true that scores of this type could be taken as a cue or

signal to change one's method of performing the task or to change one's

strategy, such scores do not yield information about what one did wrong
or where the mistake(s) occured. Nor does such knowledge tell one what
kinds of corrections to make.

3. It is evident that in general the Ss showed deterioration (rather than

improvement) in their rate of performance from the pre-test to the

experimental trials. This can be attributed to two factors. First, little
or no learning occurs over a short period of time with mature Ss on a task
like addition. Secondly, the slightest lapse in concentration will slow one's
addition rate. Since over a longer time span some attention lapses are
inevitable, this produces a slower rate on the longer trials. The experimental
trials in this case were 10 to 15 times as long as the (three one-minute)
practice trials.

- 9a -
 References

Ammons, R. B. Effects of knowledge of performance: a survey and

tentative theoretical formulation. Journal of General Psychology,
1956, 54, 279-Z99.

Arps, G. F. Work with knowledge of results versus work without knowledge

of results. Psychological Review Monographs, Supplement, 1920, 28,
No. 3.

Bilodeau, E. A., & Bilodeau, I. McD. Motor-sl.ills learning, in Annual

Review of Psychology, 1961, 12, 243-280.

Book, W. F. , & Norvell, L. The will to learn: an experimental study of

incentives in learning. Pedagogical Seminary, 1922, 29, 305-362.

Chapanis, A. Knowledge of performance as an incentive in repetitive,

monotonous tasks. Journal of Applied Psychology, 1964, 48, 263-267.

Church, R. M. , & Camp, D. S. Change in reaction time as a function of

knowlcdge of results. American Journal of Psychology, 1965, 78, 102-106.

Crawley, S. L. An experimental investgation of recovery from work.

Archives of Psychology, 1926, No. 85

DeNike, L. D. Recall of reinforcement and conative activity in verbal

conditioning. Psychological Reports, 1965, 16, 341--346.

Dulany, D. E., Jr. The place of hypotheses and intentions: an analysis

of verbal control in verbal conditioning. In C. W. Eriksen (ed.)
Behavior and awareness. Durhan-, North Carolina: Duke University

Press, 19ý2, 102-129.

Gibbs, C. B. , & Brown, I. D. Increased production from the information

incentive in a repetitive task. Medical Research Council, Applied
Psychology Research Unit, (Great Britain), 1916, No. 230.
-10-
Johanson, A. M. Influence of incentive and punishment on reaction time.
Archives of Psychology, 192Z, 8, No. 54.

Locke, E. A. The relationship of intentions to level of performance.

Journal of Applied Psychology, 1966, 50, 60-66. (a)

Locke, E. A. A closer look at level of aspiration as a training procedure:

a re-analysis of Fryer's data. Journal of Applied Psychology, 1966
(in press). (b)

Locke, E. A. , & Bryan, J. F. Cognitive aspects of psychomotor performance:

the effects of performance goals on level of performance. Journal of

Applied Psychology, 1966,5roj-a9/. (a)

Locke, E. A. , & Bryan, J. F. The effects of goal-setting, rule-learning,

and knowledge of score on performance. American Journal of Psychology,

1966, (in press). (b)

McCormack, P. D., Binding, F. R. S., & Chylinski, J. Effects on reaction-

t'Al
time of knowledge of results of performance. Perceptual, Motor Skills,
1962, 14, 367-372.

McCormack, P. D., Binding, F. R. S., & McElheran, W. G. Effects on

react ion-time of partial knowledge of results of performance.
Perceptual and Motor Skills, 1963, 17, 279-281.

Mace, C. A. Incentives: some experimental studies. Industrial Health

Research Board (Great Britain), 1935, Report No. 7U.

Mackworth, N. I-. Researches on the measurement of human performantc,.

Medical Research Council Special Report Series (Great Britain),

1950, No. 268.

Manzer, C. W. The effect of knowledge of output on muscular work.

Journal of Experimental Psychology, 1935, 18, 80-90.

Miller, G. A. , Galanter, E. , & Pribran., K. H. Plans ar-i the structure of

behavior. New York: Henry Holt, 1960.

-11-
Orne, M. T. On the social psychology of the psychological experiment with
particular reference to demand characteristics. American Psychologist,
1962, 17, 776-783.

Payne, R. B., & Hauty, G. T. Effect of psychologicz 1 feedback upon work

decrement. Journal of Experimental Psychology, 1955, 50, 342-351.

Ross, C. C. An experiment in motivation. Journal of Educational Psychology,

1927, 18, 337-346.

Ryan, T. A. Drives, tasks, and the initiation of behavior. American Journal

of Psychology, 1958, 71, 74-93.

Ryan, T. A. Chapter I: Explaining behavior; Chapter II: Explanatory concepts;

Chapter V: Experiments on intention, task, and set; Chapter VI: Intentional

learning; Chapter VII: Unintentional learning. Cornell University,
Department of Psychology, (unpublished mimeos), 1964.

Spielberger, C. D. Theoretical and epistemological issues in verbal condition-

ing. In S. Rosenberg (ed.) Directions in psycholinguistics. New York:

MacMillan, 1965, 149-ZOO.

Spielberger, C. D., Berger, A., & Howard, K. Conditioning of veroal

behavior as a function of awareness, need for social approval, and
motivation to receive reinforcement. Journal of Abnormal and Social
Psychology, 1963, 67, 241-?46.

Spielberger, C. D., Bernstein, I. H., & Ratliff, R. G. Information and

incentive value of the rel-dorcing stimulus in verbal conditioning.
Journal of Experimental Psychology, 1966, 71, 26-31.

Spielberger, C. D., Levin, S. M., & Shepard, M. The effects of awareness

and attitude toward the reinforcement on the operant conditioning of
verbal behavior. Journal of Personality, 1962, 30, 106-121.

Vroom, V. H. Work and motivation. New York: John Wiley, 1964.

-12-
 Table I

Analysis of Variance Results for Three Performance Criteria

Deterioration

Source df F

Goals 1 4.83*
KR 1 ( 1
GoalsxKR I CI
Within (mean square) 3z (8,695.28)

Linear Slope (Trials 1-5)

Goals 1 12. 87**

KR 1 1.36
GoalsxKR I 1I
Within (mean square) 32 (797. 78)

Percent Error (No. Wrong/No. Attempted)

Goals I (1
KR 1 1. 32
GoalsxKR 1 41
Within (mean square) 32 (9. 29)

*p - .05
S*p K .01

-13-
 01. .
Isz
-z Ix
0 Z-Ze
z a,
0 00
o~4 4%) s)
% /
0 -%n
4.4
4 F- -0
C:4 E-)
0
00
C-)
0 00 00
04fluiW, J~d 4:)OJJO:) swalq0Jdj 10401