Journal «/ Personality and Social Psychology

1968, Vol. 9, No. 2, 121-127

DISHONEST BEHAVIOR AS A FUNCTION OF DIFFERENTIAL

LEVELS OF INDUCED SELF-ESTEEM1
ELLIOT ARONSON AND DAVID R. METTEE 2
University oj Texas

After taking a personality test, Ss were given false feedback aimed at tempo-
rarily inducing either an increase in self-esteem, a decrease in self-esteem, or
no change in their self-esteem. They were then allowed to participate in a
game of cards, in the course of which they were provided with opportunities
to cheat under circumstances which made it appear impossible to be detected.
Significantly more people cheated in the low self-esteem condition than in the
high self-esteem condition. A chi-square evaluating cheater frequency among
the high self-esteem, the no information (no change in self-esteem), and the
low self-esteem conditions was significant at the .OS level. The results are
discussed in terms of cognitive consistency theory.

Recent theorizing and experimentation have
suggested that a person's expectancies may be
an important determinant of his behavior.
Working within the framework of the theory
of cognitive dissonance, Aronson and Carl-
smith (1962) conducted an experiment in
which subjects were led to develop an expect-
ancy of poor performance on a "social sensi-
tivity" test. The subjects then proceeded to
perform beautifully. Aronson and Carlsmith
found that these subjects subsequently
changed their superior performance to an
inferior one when retested over the same
material. Similarly, Wilson (196S) found that
subjects were significantly more attracted to
a negative evaluator than to a positive evalu-
ator if the negative evaluations were in accord
with a strong performance expectancy which
had led the subjects to withdraw from a com-
petitive event. Consistency theory thus has
received some support in regard to specific
expectancies and performance directly related
to these expectancies.
But what about more pervasive expect-
ancies such as those about the self? Bramel
(1962) showed some evidence for the impact
of self-esteem on subsequent behavior. In his
study he temporarily raised or lowered the
'•This experiment was supported by a National
Science Foundation Graduate Fellowship (NSF-26-
1140-3971) to David R. Mettee and by grants from
the National Science Foundation (NSF GS 750) and
the National Institute of Mental Health (MH
12357-01) to Elliot Aronson. Authors are listed in
alphabetical order.
2
Now at Yale University.
121
subjects' self-esteem by providing them with
positive or negative information about their
personalities. He then allowed them to dis-
cover irrefutable negative information about
themselves. The individuals who held low
self-concepts were more willing to accept this
information; that is, they were not prone as
people who had been induced toward high
self-esteem to project this specific negative
attribute onto others. These results are con-
sistent with the work of Rogers (1951), who
argued that negative or maladaptive responses
occur as the result of being consistent with a
negative self-concept, and that such responses
can be altered only by first changing the
self-concept in a direction consistent with
adaptive responses.
The prediction being tested in the present
experiment is in accord with the experiments
cited above. In addition, it carries our inter-
est in the self-concept one step further in the
direction Rogers has taken—toward greater
generalization. What Aronson and Carlsmith
showed is that people try to behave in a
highly specific manner which will coincide
with a highly specific self-expectancy; that
is, people who believe that they are poor in
a "social sensitivity" test will take action
aimed at performing poorly on that test. But
does this generalize? If we feel low and worth-
less on one or two dimensions do we behave
generally in low and worthless ways—even if
the behavior is not directly and specifically
related to the low aspects of the self-concept?
For example, if a person is jilted by his
122 ELLIOT ARONSON AND DAVID R. METTEE
girlfriend (and thus feels unloved), is he more
apt to go out and rob a bank, kick a dog,
or wear mismatched pajamas?
In the present experimental situation we
are predicting just that. Concretely, indi-
viduals who are provided with self-relevant
information which temporarily causes them to
lower their self-esteem (but does not specifi-
cally make them feel immoral or dishonest)
are more apt to cheat than those who are
made to raise their self-esteem—or those who
are given no self-relevant information at all
(control condition). Similarly, people who
are induced to raise their self-esteem will be
less likely to cheat than the controls. This
hypothesis is based upon the assumption that
high self-esteem acts as a barrier against
dishonest behavior because such behavior is
inconsistent. In short, if a person is tempted
to cheat, it will be easier for him to yield
to this temptation if his self-esteem is low
than if it is high. Cheating is not incon-
sistent with generally low self-esteem; it is
inconsistent with generally high self-esteem.
METHOD
General Procedure
The subjects were led to believe that they were
participating in a study concerned with the correla-
tion between personality test scores and extrasensory
perception (ESP). They were told that their person-
alities would be evaluated with the self-esteem scales
of the California Personality Inventory (CPI) and
that their ESP ability would be ascertained with
the aid of a modified game of blackjack. Before
participating in the blackjack game, subjects took the
personality test and received false feedback (either
positive, negative, or neutral) about their person-
alities. During the blackjack game subjects were
faced with the dilemma of either cheating and
winning or not cheating and losing in a situation in
which they were led to believe (erroneously) that
cheating was impossible to detect. The opportunity
to cheat occurred when the subjects were "acci-
dentally" dealt two cards at once instead of one.
The rightful card put the subject over 21 and ensured
defeat, whereas the mistakenly dealt extra card, if
kept, provided the subject with a point total that
virtually assured victory. The card which the subject
kept constituted the dependent variable.
Subjects
The subjects were 45 females taken from intro-
ductory psychology classes at the University of
Texas, who were randomly assigned to one of three
self-esteem conditions: high, low, and neutral. In
actuality, SO subjects were run; the results of five
subjects were discarded because of suspicion. Three
of these were in the low self-esteem condition; two
were in the high self-esteem condition. The criteria
for elimination were determined a priori and were
followed rigidly throughout the experiment. It was
made explicit that the experimenter had no precon-
ceptions as to how personality traits might be
related to ESP ability, but simply wanted to deter-
mine whether or not, for example, people who are
easily angered have more ESP than calm people.
Personality Test
All subjects came to the first session together and
were given the self-esteem scales of the CPI. The
CPI was administered by a person who introduced
himself as a member of the University Counseling
Center staff. Subjects were told at this session that
the experiment was concerned with ESP and per-
sonality characteristics. They were informed that
this session was to determine the personality traits
of the subjects, with ESP ability to be measured
in the second session.
A shortened version of the CPI was used to
evaluate the personalities of the subjects. This version
contained only the six scales related to self-esteem
and, for our purposes, constituted a measure of the
subjects' chronic self-esteem. However, the primary
experimental purpose of this test was merely to
provide the opportunity and rationale for situation-
ally manipulating the subjects' self-esteem via pre-
programmed feedback regarding subjects' personality
test results.
In order to separate the experimenter as much as
possible from the personality evaluation aspects of
the experiment, subjects were told by the experi-
menter that Miss Jacobs,3 a member of the Uni-
versity of Texas Counseling Center staff, had kindly
consented to administer the personality tests. It was
emphasized that she would score the personality
inventories and that the experimenter's access to their
scores would not be on a name basis but via a
complicating coding process. It was indicated that,
as a matter of convenience, subjects would be given
feedback regarding their personality tests when they
came for the second session of the experiment. Fol-
lowing this, the CPI's were distributed, subjects
completed them, and before leaving were assigned a
time to return for the second session with three
subjects assigned to each specific time slot.
Personality Score
In the second session subjects were tested in groups
of three. Upon arrival for the second session, sub-
jects were greeted by Miss Jacobs and told to be
seated in an outer office which provided access to
three adjoining offices. After all three subjects had
arrived, Miss Jacobs handed each subject a manila
envelope bearing appropriate Counseling Center in-
signia and assigned each subject to a different office
3
The authors would like to express their apprecia-
tion to Sylvia Jacobs for her assistance in running
the study.
 DISHONEST BEHAVIOR AS A FUNCTION OF INDUCED SELF-ESTEEM
where she was to go to read the results of her
personality test. In addition, Miss Jacobs told the
subjects that she had been given a sheet of instruc-
tions to deliver to them. The sheet of instructions
was handed to the subjects along with the manila
envelope.
The personality test results consisted of three
standard feedbacks unrelated to subjects' perform-
ance on the CPI. Each of the three subjects present
at any one specific time was randomly assigned feed-
back of either high self-esteem (HSE = positive),
no self-esteem (NSE = neutral), or low self-esteem
(LSE = negative) in content.
The HSE and LSE personality reports were paral-
lel in content except, of course, for the nature of
the evaluation. For example, a portion of the HSE
report stated:
The subject's profile indicates she has a stable
personality and is not given to pronounced mood
fluctuations of excitement or depression. Her
stableness does not seem to reflect compulsive tend-
encies, but rather an ability to remain calm and
level-headed in almost any circumstance. Her pro-
file does suggest she might be rather impulsive
concerning small details and unimportant decisions.
This impulsive tendency is probably reflected in a
lack of concern with material things. In addition,
it appears that material things are important to
the subject only insofar as they enable her to
express her generosity, good nature, and zest for
living. . . . [she] is intellectually very mature for
her age.
The corresponding portion of the LSE report stated:
The subject's profile indicates that she has a rather
unstable personality and is given to pronounced
mood fluctuations of excitement or depression. Her
instability seems to reflect compulsive tendencies
and relative inability to remain calm and level-
headed in circumstances which involve tension and
pressure. Her profile does suggest she might be
rather meticulous and careful concerning small
details and when making unimportant decisions.
In addition, it appears that material things are
very important to the subject as an end in them-
selves. She appears to be a very selfish person
who clings to material things as a source of
personal gratification and as an emotional crutch.
In the NSE condition the subject was told that
her report had not yet been evaluated due to a
heavy backlog of work at the Counseling Center.
Instead, she was presented with a sample profile
which was described as a typical or average CPI
profile such as one might find in a psychological
textbook. The comments in this NSE report paral-
leled those of the HSE and LSE reports (e.g.,
". . . fairly stable personality . . . occasionally ex-
periences mood fluctuations . . ."). The contents of
the rest of the reports was designed to be global in
nature and contained comments concerning the per-
son's ability to make friends, general impact of per-
sonality, and depth of thought. Since the dependent
123
variable involved honesty, special care was taken to
refrain from mentioning anything directly involving
honesty. Similarly, nothing whatever was mentioned,
either explicitly or implicitly about the person's
moral behavior or "goodness" of conduct.
Warm-up Instructions
After reading their personality test results, subjects
turned to their page of instructions. The instructions
stated:
The purpose of this experiment is to correlate
extrasensory perception ability with personality
characteristics. In order to get a true measure of a
person's ESP ability, it is necessary that one's mind
be primed for thinking. In order to accomplish this,
I am having you engage in a period of cerebral
warm-up. It's not important what you think about
•—anything will do—but the crucial point is that
you are to use your mind, warm it up by thinking.
You need not concentrate intensely, just keep your
mind active and filled with thought. A few minutes
of this will suffice to prepare you for the ESP
experiment.
The purpose of this "warm-up" period was to pro-
vide an opportunity for the impact of our manipula-
tions to sink in. Since the subjects had just received
an evaluation of their personalities we were quite
confident that they would be thinking about this
material.
Following the warm-up period, the subjects were
sent, one at a time, at intervals of approximately
30 seconds, from the second floor of the psychology
building to a room on the fourth floor. Here they
were met by the experimenter who was unaware of
which subject had received which type of personality
feedback. The subjects were placed in one of three
cubicles which isolated them from each other. Thus
it was impossible for subjects to converse with one
another between the time they received their per-
sonality results and the time the dependent measure
was collected.
Apparatus
The apparatus consisted of four booths or cubicles;
the front panels of three of the cubicles bounded an
area 1 foot square. Access to this area was available
via the fourth cubicle which had no front panel.
The experimenter, when sitting in the fourth cubicle,
was thus able to receive and dispense playing cards
to the subjects in the other three cubicles via slots
in their front panels. Subjects in their cubicles had
no means of communicating with their fellow sub-
jects nor could they see the experimenter due to
plywood panels on both sides and at the top of
their booths.
Inside each booth there were two slots and two
toggle switches. One switch was designated to be
turned, on to indicate a "no" answer and the other
to indicate a "yes" answer. One of the slots was
horizontal with the base of the booth, and was
situated in the middle of the front panel i inch
above the cubicle base. The other slot was vertical
124 ELLIOT ARONSON AND DAVID R. METTEE

and was the slot through which subjects returned "in order to insure against possible interference with
cards to the experimenter. your ESP due to another person handling the cards."
The experimenter's cubicle contained three inclined In actuality, the machine was used in order to pro-
sheet-metal slides leading down to the horizontal vide an opportunity for the subjects to cheat, and
slots of the subjects' booths, three vertical slots fromto make it easier for subjects to cheat since the
each subject's booth which were shielded so that machine apparently removed the experimenter from
subjects could not see into the experimenter's com- the situation; this will be described below. According
partment, three scoreboards (one for each subject), to the experimenter's description, the machine auto-
six small light bulbs, each connected to one of the matically dealt them a card on each round. In order
subjects' "yes" and "no" switches. The experimenter's to "stand pat," subjects had to switch on their
booth also contained a small electric motor. The "no" light so that the experimenter could divert
motor was functionless except for sound effects, and the dealing machine from giving them a card on the
was activated periodically during the experimental next round. The use of light signals rather than
session. verbal ones to inform the experimenter that the
subject wanted to stand pat was justified by "the
Experimental Materials necessity of keeping talking at a minimum in order
The experimenter had two decks of cards. One not to interfere with your ESP." The modifications
deck was used for the general game of cards to be and rules of the blackjack game were as follows:
played; the other deck was divided into three stacks Each subject was provided with $5 by the psy-
of cards (one stack for each subject) with a pre- chology department as a stake with which to play
arranged sequence. These were the crucial hands the game. Each subject was to "bet 50$, no more
which a subject would be dealt on those occasions or no less," on the outcome of each hand. Subjects
when she would be given the opportunity to cheat. were informed that at the conclusion of the experi-
The prearranged stacks consisted of four hands of ment they would be allowed to keep all winnings
blackjack with four cards in each hand. The first over $5, but that subjects having less than $5 at
two cards in all hands totaled approximately 11-13 the game's conclusion would not be required or
points with the third card sending the point total obligated to make up the deficit out of their own
over 21. The fourth card, if substituted for the pockets; this made it impossible for any subject
third, always brought the point total to between 19 to "lose" any of her own money.
and 21 points. We presented the following cover story regarding
Each subject's cubicle contained, in addition to how ESP ability would be measured: The subjects
the two toggle switches, a IS-watt light bulb and were told to "concentrate for approximately 5 to 10
10 fifty-cent pieces (or $5). The half-dollars were seconds before each hand," that is, to think about
used as "chips" and potential reward in the experi- whether they would win or not win the upcoming
mental task. hand. If they thought or felt they were going to
win, they were to push the "yes" switch in front
Experimental Instructions of them to the "on" position for S seconds, whereas
if subjects felt they were not going to win the
After the subjects were seated, the experimenter upcoming hand, they were to push the "no" switch
recited the instructions. He described the results of to the "on" position for S seconds. The number of
some previous experiments concerning ESP. Rhine's correct guesses above chance supposedly constituted
conclusion that some persons do indeed possess ESP a subject's ESP score. Prior to each guess, all sub-
was presented; however, no indication was given jects deposited a fifty-cent piece in the vertical slot,
that certain types of persons possess ESP while and it was explained to the subjects that on every
others do not. In order to make the cover story hand they were trying to win a kitty with $1.50
appear more credible, subjects were asked to read in it.
a recent newspaper article taped to a side panel A thorough explanation of the objectives of black-
of their cubicles, which told of how a young girl had jack was given in order to equalize differential card
been winning an astonishing number of raffle con- game experience between the subjects. It was empha-
tests which, according to a research institute at Duke sized that point totals of 14-15 made the choice of
University, was perhaps due to her ESP powers. This whether or not to take another card especially diffi-
article indicated that having ESP could be materially cult because of the high probability of going over
valuable, but no mention was made either in the 21 if another card was taken, and the low proba-
article or in the instructions that having ESP was bility of winning by standing pat. The point totals
an intrinsically positive or valued trait to possess. of the various cards were also emphasized with the
Following this, subjects were told that their ESP Ace counting only 11 points rather than 1 or 11
ability was going to be evaluated in the context points, to simplify the experiment; also, all face
of a modified game of blackjack. It was clearly cards were worth 10 points, and all numbered cards
explained that the game was to be played among the were worth their face values.
three subjects and that the experimenter was not It was explained to subjects that:
participating in the game as a contestant. The sub-
jects were then informed of the presence and neces- The card-dealing machine is a "home-made" affair.
sity of a "card-dealing machine." The experimenter It has not as yet been perfected and at times
said that the cards wer? to be dealt by a machine makes mistakes. So far we've been able to iron
 DISHONEST BEHAVIOR AS A FUNCTION OF INDUCED SELF-ESTEEM
out most of the mistakes, but occasionally the
machine will deal two cards at a time instead of
one. The machine is set up so that it deals from
the top of the deck, as in a normal game of
cards; when a person receives two cards, the top
card is actually the card he should have gotten
and the bottom card is the mistaken card. Now,
the machine is fairly good, but at times will make
this mistake. If it ever happens that you are
dealt two cards, pick up the two cards and im-
mediately return the bottom card to me by slipping
it through the vertical slot. Remember, the top
card is your card, the bottom card should be
returned to me through the vertical slot.
Subjects were told that when all "no" switches
were turned on, the experimenter would say "Game,"
which was a signal to subjects that they were to
push all their cards through the vertical slot to the
experimenter. Again, no talking was allowed under
the pretext of not interfering with ESP. The experi-
menter then examined each player's hand, determined
the winner, collected the three fifty-cent pieces and
slid them down the slide of the winner. When the
experimenter said "Begin," indicating that the next
game was to start, the two subjects not having
received any money were thus informed of their
losing status.
Resume of Procedural Instructions to the
Subjects
The word "begin" was a signal to subjects that
the previous hand was finished and that a new hand
was to commence immediately. Following the word
"begin," subjects deposited their fifty-cent pieces
through the vertical slot, concentrated for S seconds,
made a choice of the "yes" or "no" switches regard-
ing their outcome expectancy on the upcoming hand,
and the game began. The hand continued until all
subjects had indicated they no longer wanted any
more cards, at which time the experimenter said
"game" and subjects then pushed all their cards
through the vertical slot to the experimenter. This
same procedure was repeated 35 times for all sub-
jects. However, subjects were not aware as to ex-
actly how many hands had been played at any given
point in the experiment, nor did they know precisely
when the game would end.
Dependent Measure
The dependent measure was the number of times
during the experimental session that subjects kept
the card they should have returned to the experi-
menter, thus enabling them to win the hand. The
experiment was designed so as to present each sub-
ject with four opportunities to cheat during a ses-
sion; thus, as far as each subject was concerned, the
machine had mistakenly dealt two cards at once only
four times in 35 hands. In reality, the experimenter
dealt all cards but synchronized his dealing with the
onset and termination of the machine-generated
sound effects.
125
The method for providing subjects with a cheating
opportunity was quite simple. However, complica-
tions arose in determining precisely when this oppor-
tunity should present itself. We decided to present
a subject with an opportunity to cheat when the
following conditions could be satisfied: (a) If she
(the subject) had guessed "yes"; (6) when she pos-
sessed the same amount of money as the other
subjects; (c) if she had not previously received more
opportunities to cheat than the other subjects;
(d) when her ESP hit rate was near chance level.
The separate stacks of "cheating cards" were pre-
arranged so that the first two cards totaled 11-14
points, the third put the total over 21, and the
fourth, if substituted for the third, brought the
total to 19-21 points. Thus, when subjects received
two cards at once they were faced with a dilemma:
if they did not cheat, they would lose the hand;
if they did cheat, they would almost certainly win.
Cheating, therefore, was made relatively safe from
exposure by having the cards dealt by a machine
and also enabled subjects to net $1. The behavioral
measure of cheating was whether or not subjects
returned to the experimenter the card that was
actually theirs and kept the bottom card which
enabled them to win. The remaining two subjects
in such a game were dealt cards from the general
deck. If the subject had cheated, she was always
declared the winner on that hand and was given
the three fifty-cent pieces. This subject, of course,
lost if she did not cheat. The other two subjects,
in this hand, were always dealt a hand less than 21
or at times one of them was dealt a hand exceeding
21. If a subject did not cheat, the other subject with
the score closest to and under 21 was declared the
winner. The experimenter used scoreboards to keep
track of how often a subject had won a hand, how
often each had been given a cheating opportunity,
and whether or not a subject had cheated. The
Scoreboard also provided for an evaluation of trial
effects.
Following the card game, all subjects were asked
to be seated at a table in the same room. They
then filled out a questionnaire consisting of a check
on the experimental manipulations and several filler
items. Subjects were then debriefed completely. The
purpose of the experiment was explained and sub-
jects were assured that the results of the personality
test were preprogrammed and had no relationship
to their actual test scores.
RESULTS AND DISCUSSION
Analyses of variance on continuous data
were all nonsignificant, although the mean
differences were of the order hypothe-
sized (mean cheats: LSE =1.87, NSE = 1.S4,
HSE = 1.07). Frequency analyses, however,
produced significant chi-squares. Subjects
were divided according to whether they never
cheated or cheated on at least one occasion.
Table I shows a 2 X 3 contingency table chi-
126 ELLIOT ARONSON AND DAVID R. METTEE
TABLE 1
NUMBER or PEOPLE CHEATING AT LEAST ONCE AS A
FUNCTION or SELF-ESTEEM
Condition Cheat Never cheat
LSE 13 2 curred subsequent to the card game, this dif-
NSE 9 6
HSE 6 9
Note.— X^H,, = 7.00, d/ = 2,p < .05. x2LH = 5.17. d/ - 1,
p < .03.
square with dj — 2. Note that 13 people who
were given negative feedback cheated at least
once, while there were only 6 cheaters among
the positive-feedback subjects. The chi-square
proved to be significant at beyond the .05
level (x2 = 7.00, p < .05). Another chi-square
was computed to evaluate the cheating dif-
ferences between just the high and low self-
esteem groups. This chi-square with dj = 1
also proved to be significant after the correc-
tion for continuity had been made (x2 = 5.17,
p < .03).
Taken as a whole, the data indicate that
whether or not an individual cheats is influ-
enced by the nature of the self-relevant
feedback he received. People who learned un-
complimentary information about themselves
showed a far greater tendency to cheat (on
at least one occasion) than individuals who
received positive information about them-
selves. This suggests that individuals with
low self-esteem are more prone to commit
immoral acts than individuals with high self-
esteem, at least when the immoral act is
instrumental in producing immediate material
gain. Moreover, the number of people cheat-
ing in each of the above groups fell on either
side of the neutral condition. Although neither
experimental condition was significantly dif-
ferent from the control, it is important to
note that in the high self-esteem condition
there was a greater trend toward honest be-
havior than in the control, whereas in the
low self-esteem condition there was a greater
trend toward cheating than in the control.
Our interpretation of the results hinges
upon our contention that the manipulation
employed in the experiment made subjects
feel good about themselves or bad about
themselves. In short, we contend that the
feedback the subjects received had some im-
pact (however temporary) upon their level
of self-esteem. This interpretation is bolstered
by our check on the manipulation which indi-
cated that LSE subjects felt worse about
themselves than either NSE or HSE subjects.
However, since the manipulation check oc-
ference might be due to the cheating behavior
rather than the self-esteem manipulation.
There is some additional evidence which is
also consistent with our interpretation. This
involves the cheating behavior of subjects in
terms of their chronic levels of self-esteem.
Although it was not our intention to measure
chronic self-esteem, one can extract a rough
measure by looking at the self-concept scores
on the CPI, which all subjects filled out as
part of the cover story of the experiment.
According to these measures, people of high,
medium, and low self-esteem had been almost
equally distributed among experimental con-
ditions. The following data emerge: In the
no feedback (NSE) condition, slightly more
"low chronics" cheat than "high chronics."
Also, as one might expect, the greatest per-
centage of cheaters falls among the low
chronics who were given negative feedback;
the smallest percentage of cheaters falls
among the high chronics who were given posi-
tive feedback. The small number of subjects
in each cell makes statistical analysis of these
data unfeasible. The most one can say about
these results is that they are consistent with
our interpretation of the overall data on
the basis of the experimental treatments
themselves.
It should be emphasized that one of the
unique aspects of this study is the nonspecific
nature of the self-concept manipulation. The
subjects were not told anything about them-
selves which would lead them to infer that
they were moral-honest or immoral-dishonest
people. Rather, they were told things designed
to reduce their self-esteem in general. The
social implications of these findings may be
of some importance. Our results suggest that
people who have a high opinion of themselves
are less prone to perform any activities which
are generally dissonant with their opinion.
Similarly, it may be easier for a person with
a low self-concept to commit acts of a crimi-
nal nature. Moreover, it may be that a
common thread running through the complex
 DISHONEST BEHAVIOR AS A FUNCTION OF INDUCED SELF-ESTEEM
variables involved in successful socialization
(Sears, Maccoby, & Levin, 1957) is that of
different development of self-esteem. Granted
that most children become aware of what
behavior is approved (moral) or disapproved
(immoral), the development of high self-
esteem in the individual may be crucial in his
choosing a moral rather than an immoral
mode of behaving.
This discussion is highly speculative to say
the least. Further experimentation is necessary
before the validity of our reasoning can be
determined. One reason for this note of
caution is the fact that it is difficult to per-
form an experiment involving complex human
cognitions, emotions, and behavior which
leaves us with a single, untarnished explana-
tion for the results. This experiment is no
exception. One conceivable alternative inter-
pretation concerns aggressiveness: the sub-
jects in the LSE condition, because they
received a negative evaluation, may have been
angry at the evaluator and, consequently, may
have cheated as a way of punishing him. We
attempted to eliminate this possibility in two
ways: (a) We separated the experimenter
who ran the ESP experiment from the evalu-
ator (a member of the Counseling Center).
Toward this end, the experimenter appeared
ignorant of and uninterested in these evalua-
tions; it was made to seem to be strictly be-
tween the subject and the Counseling Center;
(b) the subjects were playing against other
subjects rather than against "the house."
Thus, when an individual cheated, she was
clearly not hurting the experimenter; rather,
she was unjustly taking money from a fellow
college student.
Perhaps a more compelling explanation in-
volves compensation. The subjects in the low
self-esteem situation may, in effect, be saying,
"Well, I may not have done well on that
personality test, but at least I'm going to
see to it that I win some money." This
explanation is quite different from the one
that holds that it is easier to cheat because
such behavior is consistent with feelings of
low self-esteem. Note, however, that the "com-
pensation" explanation applies only to the
subjects in the low self-esteem condition.
Thus, this alternative explanation would have
been weakened if the subjects in the high
127
self-esteem condition had cheated less than
those in the control condition. But as the
reader will recall, although these data were
in a direction favoring consistency theory,
they were not statistically significant. Thus,
until further research is performed on this
problem, compensation remains a possible
explanation.
A final piece of unclarity should be men-
tioned. We predicted that people in the LSE
condition would cheat and people in the HSE
condition would not cheat because we felt
that such actions would reflect a consistency
between self-esteem and behavior. But cheat-
ing is merely one of many ways in which a
person's behavior could show consistency with
low or high self-esteem. For example, in the
present situation a LSE subject could try to
lose; being a loser might be considered as
consistent with a low self-esteem. It should
be noted that the experimenters took special
pains to assure the subjects that they could
not lose money. In this situation it seems
reasonable to assume that being a loser simply
means having bad luck—not being a bad
person or even a poor person (financially).
We selected cheating as our dependent vari-
able because we felt that it is an unambigu-
ously unethical piece of behavior. Although
losing at cards is not pleasant, it does not
seem as bad, especially since a loss of
money is not involved. The data indicate
that our reasoning was correct—that if a
sizable number of LSE subjects had sought
to lose, our data would have failed to reach
significance.
REFERENCES
ARONSON, E., & CARLSMITH, J. M. Performance ex-
pectancy as a determinant of actual performance.
Journal of Abnormal and Social Psychology, 1962,
65, 178-182.
BRAMEL, D. A dissonance theory approach to de-
fensive projection. Journal of Abnormal and Social
Psychology, 1962, 64, 121-129.
ROGERS, C. R. Client-centered therapy. Boston:
Houghton-Mifflin, 1951.
SEARS, R. R. MACCOBY, E. E., & LEVIN, H. Patterns
of child rearing. Evanston, 111.: Row-Peterson,
19S7.
WILSON, D. T. Ability evaluation, postdecision dis-
sonance, and co-worker attractiveness. Journal of
Personality and Social Psychology, 1965, 1, 486-
489.
(Received May 11, 1967)