JOURNAL
OF EXPERIMENTAL
SOCIAL
Self-Persuasion,
PSYCHOLOGY
Social
4,
143-152 (1968)
Approval,
Determinants
of
impression
and
Self-Esteem
Task
Success
as
following
Management
HARRY S. UPSHAW AND LAWRENCE A. YATES thiversity
of IEEhois
at Chicago
Circle
Subjects were instructed to present themselves in the best or worst possible light in response to an “objective” personality inventory. Half of the subjects received a socially desirable profile in return, and half a socially undesirable profile. Subsequently, self-esteem was assessed. Three hypotheses derived from the literature on ingratiation were tested : self-persuasion, direct approval, and task success. The data support the task-success hypothesis which relates self-esteem in the “managed impression” situation to success in creating a false impression.
In his recent book, hyratiation, Jones (1964) reviewed the evidence of several studies supporting the proposition that “when confronted with approving feedback, people will tend to extract from it signifying information concerning their own basic worth, even though they have distorted their self-picture in the attempt to gain this approval” (Jones, 1964, pp. 57-58). Particularly relevant to this proposition are studies by Jones, Gergen, and Davis (1962) and by Gergen (1965). In the first of these studies subjects were instructed either to present themselves as they truly were or t’o make a maximally good impression upon an interviewer. Half of the subjects in each experimental group were suhsequent’ly informed that the impression made upon the interviewer was, in fact, a favorable one, and half were informed that it was unfavorable. The subjects were then asked to what extent the information given to the interviewer described the subject as he truly wae:. Analyzing these data, the investigators found a main effect due to ingratiation VS. accuracy instructions, which served to validate t,he experimental induction. Of greater interest to the present discussion was the fact that they also observed a significant main effect due to positive vs. negative feedback. Thus, whether or not the subject knowingly misrepresented himself in a favorable direction, he claimed the information to be more accurate when it was followed by favorable feedback than when it was followed by unfavorable feedback. 143
144
UPSHAW
AYD /
YATES
In the study by Gergen two groups of subjects were instructed to present themselves honestly, and one group was instructed to make the best possible impression on an interviewer. In the ingratiation condition and in one of the accuracy conditions, t,he interviewer, by prearrangement, provided reinforcement for every comment t,hat a subject made praising himself. In the other accuracy condition the interviewer was impassive and nonreinforcing. Records were kept of the frequency of self-enhancing statements during the interviews. These records showed clearly that reinforcement under accuracy instructions was associated with an increase in self-enhancing statements and that reinforcement under ingratiation instruct,ions resulted in even more self-enhancement. Following the interview, subjects were induced to make statements about themselves in order to supply (‘normative data.” The frequency of selfenhancing statements during this period was the critical dependent variable as regards the present issue. It was found that the subjects who had been reinforced for self-praise, whether under accuracy or ingratiation instructions, made more self-praising responses than did the non-reinforced subjects who had been interviewed under accuracy instructions. Furthermore, the frequencies of self-enhancement in the ingratiation and reinforced accuracy conditions were not significantly different,. The data-collection instrument that Gergen used, t#he Self-Valuation Triads test, consists of sets of three items referring to values, attitudes, int.erests, etc. The items for each triad were selected on the basis of their successive intervals scale values on a continuum of favorability (Jones, 1964, p. 59). Self-enhancement in the experiment,, therefore, corresponded to the subject’s presenting himself in a more socially desirable light than presumably otherwise would have been the case. In t,he ingratiation conditions of both the Jones et al. and Gergen studies, the subject was instructed to create a maximally good impression for an interviewer, even if the performance of that t,ask required the falsification of information presented to the interviewer about oneself. In both studies the ingratiating subject was given the task of managing a false (favorable) impression. In response to the attempted impression management, some subjects were led to believe that they had succeeded: they received approving feedback agrecin g with their exaggerated claims of personal worth. Others were led to believe that they had failed to convey the intended impression: they received disapproving feedback which did not agree with their claims. Jones int,erpreted the resu1t.s of the two studies as evidence of a human failing whereby approving feedback is easily credited and accepted as evidence of one’s real worth, and disapproving feedback is attributed t,o other causes. Gergen related this view to psychological theory in t.he
DETERMINANTS
OF
SELF-ESTEEM
145
following way: “. . . reinforcement, or feedback, seems to be continuously influential in moulding the self-concept in at, least two ways: (a) From the standpoint of operant conditioning a person can be reinforced by others for providing certain behavior at certain times. Since such responses should then occur with increasing frequency, the person should come to feel that such responses are truly representative of the self. (b) From the viewpoint of classical learning a person can come to associate his overt behavior with descriptions communicated to him by others. . . . According to this view, a person’s self-definition is primarily a reflection of the views of significant others about him” (1965, p. 413). The process accounting for the increased self-esteem of the successful ingratiator is, according to Jones and to Gergen, one of self-persuasion which is presumably mediated by social reinforcement in the form of agreement expressed by an attractive other person. Thus, with reinforcement one comes to believe that he actually possesses the socially desirable traits that he has claimed. This explanation has the virtue of both fitting the data and being consistent with the reinforcement folklore of psychology. There are, however, at least two alternatives to the selfpersuasion hypothesis which fit the data equally well and are equally consistent with reinforcement principles. One of the alternatives views the link between self-esteem and social approval as direct, as opposed to the self-persuasion hypothesis which views the relationship as being mediated by a change in beliefs about oneself. We shall refer to the former hypothesis as the “direct approval” interpretation. According to this interpretation a person’s self-esteem is expected to reflect the amount of social approval that he has obtained, without regard to the nature of the behavior upon which that approval was based. Another explanation of the increased self-esteem of the successful ingratiator is the “task success” hypothesis. According to this interpretation, a person’s self-esteem is increased by the fact of being successful at any task. Thus, the feedback in the earlier studies may be viewed as evidence that the subject is being successful in the task he has undertaken. For the ingratiating subject, the approving feedback may signify the success of his deception, whereas the same feedback to the subject who is striving to portray himself accurately may be viewed as evidence that he and the interviewer are truly communicating. The three interpretations of the Jones et al. and Gergen results differ in the meaning attributed to the feedback to the ingratiating subject. For the self-persuasion hypothesis the feedback serves to inform the subject that the target of his managed impression agrees or disagrees with the claims that were made. The same feedback conveys to the subject in-
146
UPSHAW
AND
YATES
formation concerning the target’s approval or disapproval; and this is the critical feature with respect to the direct-approval hypothesis. Finally, the feedback also provides the subject with information about the success of his efforts to portray himself as a paragon of desirable traits. This aspect of feedback is critical for the task-success hypothesis. The concept of reinforcement in modern psychology is broad enough to encompass all of these features of feedback in this type of experimental situation. The present study is one in which the three alternative hypotheses discussed above lead to different predictions. The design of the study consists of four experimental conditions. Subjects were instructed either to make the best impression possible (I+), or the worst (I-). Half of the subjects in each of these two groups received favorable feedback (F+), and half received unfavorable feedback (F-). According to the self-persuasion hypothesis the individuals showing the highest self-esteem at the time of testing are those in t.he (I+ Ff) condition. The Jones propoeit,ion regarding self-persuasion states that unfavorable information about a person tends not to be believed. If t,his is true, then t,he three groups U+ F-), (IF-), (IF+) should not differ in self-esteem. Thus, this version of the self-persuasion hypothesis implies the following prediction (where (‘>I’ means “has greater self-est’eem than”) : (I+ F+) > (I+ F-) = (I- F-) = (I- Ff). Another version of the selfpersuasion hypothesis is that one tends to believe that information about himself which he states, and with which an attractive other person expresses agreement. Under this version of the hypothesis the expectation is: (I+ F+) > (I+ F-) = (IF+) > (IF-1. The direct-approval hypot,hesis states that self-esteem is elevated in proport’ion to the amount of social approval, regardless of the reason for the approval. In terms of t’he present design, this hypothesis predict#s a main effect, due t,o feedback, i.e.: (I+ F+) = (I- F+) > (I+ F-) = (IF-). The task-success hypothesis states t,hat self-esteem is a function of successful performance. In the present design all subjects were instructed to manage an impression. The feedback provided success experiences for half of the subjects and failure experiences for the other half. Therefore, under t’his hypothesis the predict,ion is: (I+ F+) = (IF-) > (I+ F-) = (IFf). METHOD
two
Thirty-two social
volunteer fraternities
at
subjects from the Chicago
an introductory Circle Campus
psychology cowse of the University
and from of Illinois
DETERMINANTS
OF
SELF-ESTEEM
147
participated in the study. The subjects met in one of two sessions where they were told the following: “Recently a new personality inventory has been developed for use in personnel selection in industry. This instrument, the Instant Personality Survey (IPS), requires only a few minutes for administration and scoring, a computer program being available which prints a unique analysis of the personality of each applicant in terms of trait descriptions. This feature of the IPS provides the personnel manager with the kind of information that formerly required a great deal of additional time during which he consulted with psychiatrists or clinical psychologists. Now he can have the equivalent of professional consultation within minutes after the applicant completes the survey. Although preliminary research on the instrument has shown it to be unusually reliable and valid, inadequate research has been done to determine whether it is sensitive to deliberate misrepresentation by a person who is trying to put himself in a good or bad light.” All of the 16 subjects in one session were asked to respond to the IPS in a self-enhancing way, and the 16 subjects in the other session were asked to respond to it in a self-degrading way. Specifically, they were instructed to play the role of a person in an employment setting who tries to impress the personnel manager in a maximally favorable or unfavorable way. In the course of the experiment, two forms of the “IPS” were used. The first form was, in reality, the first 19 items of the Edwards Social Desirability Scale, and the second consisted of the second 19 items of the SD Scale. Thus, the first 38 of t,he 39 MMPI items that Edwards (1957) has keyed for socially desirable responses were used. That the arbitrary partition of the SD Scale does not destroy its usefulness as a measure of striving to manage a socially desirable impression is supported t,v the relatively high internal consistency reliability coefficient (82) reported by Edwards ( 1964). The subjects first completed the IPS feigning either a maximally good or a maximally bad impression. The experimenter collected their machine-scorablc answer sheets and sent them to the “computer.” After 10 minutes each subject was given a computer printout (headed by his identification number) which listed nine personality descriptions attributed by the “computer” to the subjects. The statements were actually selected from those used by Forer (19491, from an astrology book, and from a well-known text in the psychology of personality. Although the subjects did not know it, each received either a standard favorable or a standard unfavorabie form. The following are examples of the statements included in the descriptions given to aribjects in the favorable-feedback condition. “COMP. X323-This person tends to like a certain amount of variety and becomes dissatisfied when hemmed in by arbitrary restrictions and limitat.ions.” “COMP. XII%-This person tends to take pride in his indepcndcnlec of thought and to require satisfactory proof of others’ statements before accepting them.” Examples of statements from the unfavorable-feedback condition are : “COMP. Z646-This person tends often to blame others when things go wrong.” “COMP. Z525-This person tends to disregard the feelings of others.” ilfter reading the computer statements about them, the subjects were asked to evaluate the (,xteut to which the IPS had succeeded in detecting their true pcrsonalities. despite lln,ir attempts at impression management, by rating the personality desGpt,ion on a S-point sea!c (1 = “~mor,” 5 = "good"). Following this rating, the suhjccts were asked to complete a second form of the IPS, this time responding to
148
UP,SHAW
AND
YATES
each item as truthfully as possible. These “accuracy” instructions were justified as a means of objectively determining the extent to which the subjects had succeeded in “beating” the test whrn it was first administered. The answ-er sheets for the “accuracy” condition were then collected, and the subjects were instructed again t,o respond to the second form of the IPS, this time attempting to make the best possible impression. This final attrmpt at impression management was included as a control by which to insure that the feedback from the first attempt did not serve to change differentially the views of the subjects concerning the social desirability of the responses to ittms on the scale. Without this control it would not be possible to distinguish changt,s of srlf-csterm following impression management from efforts by the subject to use the information obtained from feedback to perfect whatever impression he tried to make, whether good or bad. The justification offered the subjects for the Iast task was that they now had experience which might improvl t,heir ability to manage an impression. At the end of the session the rationale of the experiment was explained, apologies for deception were made, and gratitude for the subjects’ participation was expressed. To summarize, the chronology of experimental events was as follow;-.: 1. The subjects responded to the first form of the IPS in a way that, was calcIlIated to make a maximally good or bad impression. 2. Half of the subjects in each experimental condition read computer descriptions of their personalities (presumably based upon their falsified data) which wsre tlattering, and half read unflattering descriptions. 3. The subjects rated the accuracy of the computer statement as a description of their real personalities. 4. A second form of the IPS was responded to under instructions tu be as accurate as possible. 5. The subjects again responded to the second form of the IPS: this time all subjects were asked to try to make the best possible impression. 6. The subjects were debriefed. RESULTS
AND
DISCUSSION
The responsesmade by subjects under instructions to create either a favorable or an unfavorable impression were assigned scores based on the number of times the keyed socially desirable alternative was Ihosen. An analysis of variance based on a two-by-two, fixed-effects, factorial design was done on these data for the variables of good versus bad impression and favorable versus unfavorable feedback. As expected, only the impresTABLE MEAN
1
SOCIAL-DESIRABILITY SCORES OBTAINED INSTRUCTIONS T O MANAGE AN IMPRESSION
UNDER
Impression attempted Feedback
Favorable
Unfavorable
Favorable Unfavorable
17.8 18.5
4.0 10
DETERMINANTS
OF
149
SELF-ESTEEM
sion main effect was significant (F = 6.13, 1 and 28 df, p < .05). Table 1 presents the means on this initial measure. The analysis establishes that the subjects who were instructed to make a good impression made higher social-desirability scores than did those instructed to make a bad impression. Table 2 presents the means and Table 3 the summary of the analysis of variance of t.he rat,ings of the accuracy of the feedback which followed TABLE MEAN
RATINGS
2
OF ACCURACY
OF FEEDBACK
Impression Feedback
Favorable
Favorable Unfavorable
attempted Unfavorable
3.87
3.74
2.18
1.68
the subject’s attempt at impression management. These dat,a indicat,e a main effect, such that the favorable feedback is said to be more accurate tha,n the unfavorable feedback. The central problem of t,he present study is the effect of favorable and unfavorable feedback on the self-esteem of subject.s following an attempt to manipulat’e an impression. Thus, the critical analyses are concerned with the socia.1 desirability of the responses made by subjects when they TABLE ANALYSIS
OF VARIANCE
Source Impression Feedhack EXF ErrOr *p
OF RATINGS
df 11) (Fj
1 1
1 28
3 OF FEEDBACK
MS 0.78 “8.12
0.28 0.68
ACCURACY
F 1.15 41.35*
-
< .Ol.
were asked to respond truthfully. Before presenting these data, however it is appropriat,e to consider additional control data which were collected. It was anticipated that subjects might respond to the feedback by changing their criteria for good and bad impressions. In other words, a subject who feigned a, good impression, but received unfavorable feedback, might account for the feedback on the basis that his deception was detected. Accordingly, he might decide that the feedback was based on his true personality or. perhaps, on the personality characteristics of one who would try t,o deceive. Alternatively, he might decide that the norms of the
150
UPSHAW
AND
YATES
personality test imply standards of social desirability which are at odds with those he holds. It will be recalled that the subjects were instructed just before the end of the experiment to attempt once again to manage an impression: this time everyone was instructed to make the best possible impression. These data were analyzed by means of a 2 X 2 (impression, feedback) factorial, fixed-effects analysis of variance. None of the effects was significant, suggesting that. all subjects held a common standard of social desirability at the close of the experiment,. TABLE 4 SELF-ESTEEM SCORES FOLLOWIXG IMPRESSION MANAGEMENT
MEAN
FEEDBACK
Impression Feedback
Favorable
Favorable Unfavorable
13.9 13.0
FOR
attempted Unfavorable 11.8 16.0
Table 4 presents the means of the social-desirability scores for the four experimental conditions obtained under the instruction to be truthful, following feedback for the original attempted good or bad impression. These are the critical data for distinguishing among the three hypotheses of the study. The hypotheses were tested by means of one or more TABLE 5 OF COMPARISONS TESTING ALTERNATIVE CONCERNING SELF-ESTEEM
ANALYSTS
HYPOTHESES
Comparison Hypothesis Self-persuasion I
+3
Self-persuasion IIa IIh
+1
Direct Task a b c
approval
I+F-
I+F+
I -
--I
F+
-1
0 0
0
1-I
-1
+l
-1
+1
fl +1
-1
1 -I?-
31 s
-1
..il
/<‘?I
< I 00
1X.06 (Ii.*>5 -.
Y.15 1.09
-1
‘32 .7s
3 9X
+1 -1
$2 ,353 18.06 6.25
9 17* 3 15 1 09
-1
0
success -1 0 0
0
+1
a The error MS is for all comparisons * p < .Ol, double-tailed.
-1
MS
error:
0
5.73,
df = 28.
DETERMIKAA-TS
OF
SELF-ESTEEM
1.51
single-degree-of-freedom comparisons, each of which is logically equivalent to a t test in which weighted combinations of cell means are compared. Table 5 lists the weights for the comparisons for each hypothesi,;, along with the mean square and F value. Two versions of the self-persuasion hypothesis were discussed. The first of these specifies that the self-esteem of the (I+F+) group is expected to be higher than that of the other groups. The comparison rorresponding to this hypothesis is described in Table 5 as “Self-pcrsuesion, I.” The da,ta do not support this hypothesis. The second version of t,he self-persuasion hypothesis specifies that the highest self-cstcem would be found in the (I+F+) group, the lowest, in the (I-F-) group, with the other two groups being of intermediate value. The “Self-persuasion, II” hypothesis is expressed in Table 5 as two comparisons. The first of these is between the two groups that are expect,ed to be highest and lowest in self-esteem. These two groups do not differ significantly. Furthermore, the direction of their difference is opposite to that hypothe&ed. The second comparison is likewise insignificant. However, insignificance in this case is predicted by the hypothesis. Thus, the only support found in this et#udy for either version of the self-persuasion hypothesis ia the confirmation of an expected null hypothesis. The direct-approval hypothesis specifies what is essentially n feedback main effect. whereby self-esteem is elevated by social approval. The single comparison corresponding to that hypothesis not only failed to attain a significant va,lur, but the direct.ion of the observed difference was opposite to that, hypot’hesized. Therefore, it seems appropriate to dismiss this hypothttsi!: from further considerat.ion, at least with respect to the present c1:it:i. The task-success hypothesis specifies that the average self-esteem oi the (I+F+ I and (I-F-) groups should be higher than the average of the (I+F1 and (I-F+) groups. The comparison corresponding to that, hypothesis was found to be highly significant. Two auxiliary comparisons deriving from the task-success hypothesis were expected not to be significant, and they were not. It, can be observed in Table 5, however, that the comparison between the (IfFf) and (I-F-) groups was nearly significant, with the (I-F-) group displaying higher self-esteem. If this difference is, in fact, a reliable one the t.ask-success hypothesis would require as an explanation that the subjects of the (I-F-) group experienced greater success than those in the (I+F+) group. It will be recalled that t’he data reported in Tables 2 and 3 indicated that the subjects who received unfavorable feedback considered the computergenerated profiles to be less descriptive of their true personalities than those who rcceiycd favorable profiles. Thus, it may be that those sul)jects
152
UPSHAW
AND
YATES
who succeeded in “beating” the test with an unfavorable impression did, in fact, experience greater task success. The data of the present study rather clearly support the view that a person’s self-esteem is increased by success at managing an impression. It seems reasonable to generalize this conclusion to the point of suggesting that success with any interactional goal elevates self-esteem. If this generalization is permitted, the task-success hypothesis will account for the present data in which all subjects managed a false impression, as well as earlier studies in which self-esteem was observed to increase when a social agent agreed with self-flattering statements made un+r instructions to be accurate. One feature of the present experiment that may strike the reader as less than optimal is that the same instrument was used to assess a subject’s dissimulation in one phase and to measure his self-esteem in the other. The earlier studies by Jones et al. and by Gergen, which prompted the present research, were similar in that a subject’s degree of self-esteem was inferred from information concerning the extent to which he privately believed socially desirable statements about himself, ‘The apparent inability of investigators to differentiate operationally betweell lying to produce a favorable impression and honestly expressing high regard for oneself is a cause for some chagrin. If the line of research inaugurated by Jones and his colleagues is to continue and flourish, early eff0rt.s toward resolving these conceptual probIems must be forthcoming. The ultimate interpretation to be made of these results of the present study depends upon how those problems are resolved. The dual use of the social-desirability measure in this design is troublesome, but the trouble lies in theory, not method. REFERENCES EDWARDS, A. L. The social desirability variable in personality ass~sment and rcsearch. New York: Dryden, 1957. EDWARDS, A. L. Social desirability and performanceon the MMPI. Psychometrikrc, 1964, 29, 295-308. FORER, B. R. The fallacy of personal validations: A classroom dr:monstrat,lon of gullibility. Journal of Abnormal and Social Psychology, 1949, 44, 118-123. GERGEN, K. J. The effects of interaction goals and personalistic feedback on the presentation of self. Journal of Personality and Social Psychology, 1965, 1, 41% 424. JONES, E. E. Ingratiation. New York: Appleton, 1964. JONES, E. E., GEFCGEN, K. J., AND DAVIS, D. E. Some determinants of rractlons to being approved or disapproved as a person. Psychological Monoqraph,.s. 1962. 76 (Whole No. 521). (Received
April
3, 1967)