':E?f!‘§?!3:fs‘fd§éiiIi’JFZ . ; l! '7 "' "’ ‘ "’ “ ‘ 99‘~.~‘“‘- “x“ \"~‘--‘."-°~'--s~~ * ' "“H'r v: 3-4"OGQQVA-Rt’nus-QWI‘Wvuwas‘l’mmM‘. PERFORMANCE EXPECTANC-Y as A DETERMINANT 0F ACTUAL PERFORMANCE: THE INFLUENCE OF 0mm CHARACTERSWCS AND SUBJECT NEED- ma . APPROVAL Thesis for the Degree of M. A, MiG-REGAN STATE UNIVERSITY _ KAY AUCE SCHLAPPE 1970 n . ‘ l 4.12 11 £le. 41419.04114/fi? J lllllilllllHHIIIHIHIHIHMIIIIHIIlllmlnullmunnm 3 1293 104484 4609 . Michigan State ' University v—wwq-I “any 1 I Wale”; ABSTRACT PERFORMANCE EXPECTANCY AS A DETERMINANT OF ACTUAL PERFORMANCE: THE INFLUENCE OF DEMAND CHARACTERISTICS AND SUBJECT NEED-FOR-APPROVAL By Kay Alice Schlappe A study by Aronson and Carlsmith (1962) confirmed the prediction of cognitive dissonance theory that people pre- fer to confirm a failure expectancy by failing than to disconfirm the expectancy by succeeding. Later studies have attempted to attribute these results to compliance with experimenter demands for consistency. The results of these studies have been inconclusive, because the manipulations of consistency demands have also affected pressures for achievement. The present study investigated the effects of experimenter demands for consistency in the Aronson-Carlsmith experimental situation by separating the experimenter demands for consistency from the pressures for achievement. The relationship between need-for-approval and responses to experimenter demands and achievement pres- sures was also examined. Forty-six introductory psychology students served as subjects for a partial replication of the Aronson-Carlsmith low expectancy condition. For Ss in the Reliability Kay Alice Schlappe Condition (R), the instructions emphasized the experimenter demands for consistency, i.e. a reliable performance. To maximize achievement pressures, instructions in the Per- sonality Evaluation Condition (PE) sensitized $3 to the personality evaluation aspect of the task. Within each of the conditions. Ss were divided into the high performance (HP) and low performance (LP) treatments of the original study. Need-for-approval was measured by the Marlowe- Crowne Social Desirability Scale. It was hypothesized that Ss in R would strive for a consistent (poor) performance, while Ss in PE would strive for a successful (inconsistent) performance. It was also hypothesized that this differential responding would be greater for high than low n-approval 83. The data were analyzed with a 2 x 2 x 2 analysis of variance. unweighted means. Significant interactions were further analyzed by simple effects tests. The results showed a significant performance effect. Subjects pre- ferred a high rather than consistent performance in both R and PE. Comparison of responses across R and PE re- vealed virtually no differential responding to the in- structions in the HP group. In the LP group, hi-n-approval Ss showed significantly more success-seeking in PE than in R, while lo-n-approval Ss showed an opposite (non-significant) response pattern. Kay Alice Schlappe The over-all pattern of responding indicates that only hi-n-approval Ss complied with the experimenter demands for consistency, and then only when the possibility of a high performance seemed out of reach. It was con- cluded that there is no evidence that experimenter demands are responsible for consistency results in the performance expectancy studies. The conflicting results in the performance expectancy literature were discussed in terms of the effect of desires on cognitive consistency processes. Approved Aéfi::;ég::;1iflvL~ -/ Date // 2.3 /70 PERFORMANCE EXPECTANCY AS A DETERMINANT OF ACTUAL PERFORMANCE: THE INFLUENCE OF DEMAND CHARACTERISTICS AND SUBJECT NEED-FOR-APPROVAL BY Kay Alice Schlappe A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1970 (’7’ u: '- C?5 70 ‘2’ /~ ;°r-7l ACKNOWLEDGMENTS I would like to express my appreciation to Dr. James Uleman, chairman of my thesis committee, for his guidance and encouragement during the writing of the thesis, and to Dr. Lawrence Messe’and Dr. William Crano for serving on.the committee. I would also like to thank Dr. Alice Eagly for her help during the planning and execution of the research. 11 TABLE OF CONTENTS LIST OF TABLES . . . . . . LIST OF FIGURES . . . . . . INTRODUCTION . . . o . . . Disconfirmation of performance expectancies . . . Demand characteristics . . Need-for-Approval. . . . METHOD . . . . . . . SUbjeCtso o o o o o 0 Procedure . . . . Experimental conditions. . Dependent variable . . . Reign 0 0 O O O O 0 RES ULTS O O O O O O 0 Performance. . . Need-for-Approval. Instructions . . Sex differences . . Success of the instructions DISCUSSION. . . . . . . . Performance. .. . Demand characteristics . . Need-for-Approval. . . Another look at the dissonance Suggestions for future research . . . . FOOTNOTE 0 O O O O O 0 0 REFERENCES. . . . o . . . iii predictions Page iv Table 1. LIST OF TABLES Page Mean Response Change on Repetition of 5th Trial 0 O O O O O 0 O 0 O O O 22 Analysis of Variance of the Response Change mta I O O O O O O O O O O O O 22 Mean Response Change with N-Approval Groups Equalized by Sex. . . . . . . . . 27 Perception of Experiment by Type of Instructions . . . . . . . . . . 28 Relationship of Expectancy and Desire to Performance in the Aronson-Carlsmith Paradigm . . . . . . . . . . . 36 iv LIST OF FIGURES Figure Page 1. Significance Levels of Differences Between Means-~Simp1e Effects Analyses . . . . . 23 INTRODUCTION Consistency theories have had a major influence on research in social and personality psychology for approxi— mately the last two decades. The basic idea central to all these theories is that a person strives to minimize incon- sistency in his beliefs, attitudes, behavior, or interper- sonal relationships. or the consistency theories, the theory of cognitive dissonance (Festinger, 195?) has probably generated the most research and the most conp troversy. Within dissonance research, one of the most controversial areas has been that cancerning disconfirmation of performance expectancies. The present research in_ vestigates a crucial issue in this controversy: whether the demand characteristics of the experimental situation may account for the results in those experiments finding dissonance effects. The effect of the subjects' need-for- approval on their responses to the demand characteristics is also examined. Disconfirmation of performance expectancies. According to cognitive dissonance theory, dissonance is aroused when. ever two cognitive elements are psychologically incon- sistent, i.e. "if considering these two alone, the obverse of one element would follow from the other (Festinger, 1957, p. 13).“ The dissonance, which is assumed to be psycho- logically uncomfortable, will be reduced by changing one or both elements, or the relationship between them. In addition, the person will actively try to avoid information or situations which would arouse dissonance. One important case where dissonance is predicted by the theory is the disconfirmation of an expectancy.1 According to the theory, if a person eXpects X and Y occurs, dissonance should be aroused, since the cognition 'X will occur” is inconsistent with the cognition "x.did not occur.I The dissonance prediction is not at all surprising when.what is expected is inherently more desirable than.what actually occurs; the same prediction would be made by any hedonic theory. However, dissonance theory predicts that the dis- confirmation of an expectancy will be dissonance-arousing, and therefore uncomfortable, regardless of the pleasurable- ness or painfulness of the expected event. The dissonance prediction that the disconfirmation of both desirable and undesirable expectancies produces negative affect is sup- ported in a study by Carlsmith and Aronson (1963), and partially supported in replications by Sampson and Sibley (1965) and Keisner (1969). The issue of disconfirmed expectancies becomes es— pecially important when the expectancy involved is an expectancy about one's behavior, a performance expectancy. in the case of performance expectancies, dissonance theory predicts that if a person expects to achieve a certain level of performance, i.e. his self-concept includes a belief about his ability in that area, performing at a markedly different level will arouse dissonance, causing him to mini- mize the discrepant performance.2 In such a case, the per- son's cognition of how he performed is inconsistent with his cognition of how well he should have performed. As in the case of disconfirmed expectancies in general, the crucial prediction for disconfirmed performance expectancies is the one involving a negative performance expectancy, i.e. an ex- pectancy of a poor performance or of undesirable behavior. Dissonance theory predicts that the person.who expects to do poorly and does well should experience dissonance in the same way as the person who expects to do well and does poorly. Aronson and Carlsmith (1962) tested the dissonance predictions for disconfirmed performance expectancies by es- tablishing in their subjects an expectancy of either success or failure on a fake social sensitivity test and then, after either confirming or disconfirming the expectancy, giving the subjects the chance to change their performances. Subjects in the Aronson and Carlsmith experiment performed so as to confirm their eXpectancies, i.e. subjects with dis- confirmed eXpectancies changed significantly more responses on the repetition of the performance trial than subjects whose expectancies were confirmed. This effect was found in both the success and failure expectancy conditions. A great deal of controversy has followed the publica- tion of the Aronson and Carlsmith study, since for many social psychologists the finding that subjects will attempt to confirm a failure expectancy is dissonant with their be- lief that peOple try to maximize their achievement. As a result of this controversial nature of their findings, a number of attempts to replicate the Aronson and Carlsmith study have been carried out. Of these, Ellsworth's study (1966), two studies by Brook, Edelman, Edwards, and Schuck 5 (1965), and one group of subjects in Silverman and I Marcantonio's study (1965) have shown the consistency effect obtained by Aronson and Carlsmith. Studies which have found that subjects change in the direction of maximizing achieve- ment rather than consistency include five experiments by Brook et a1. (1965), Jones (1968), two experiments by Lowin and Epstein (1965), one group of Silverman and Marcantonio's subjects (1965), and Ward and Sandvold (1963). A slight but nonsignificant tendency toward success-seeking was found by Marcentonio (1966). The findings of Cottrell (1965) are equivocal; although he found a significant per- formance-eXpectancy interaction, his low expectancy subjects showed more of a success-seeking than a consistency effect. It is obvious from the above summary that the Aronson and Carlsmith study and the replications of that study have not settled the issue of whether a person prefers to confirm a failure expectancy or to perform successfully. Since both consistency and achievement effects have been found using essentially the same experimental situation, it follows that there must be unintended and unidentified differences among the experiments which are significantly influencing the re- sults. One factor which may be contributing to the con- fusion in this area is differential reSponding by the sub— jects to the demand characteristics of the experimental situation. Demand characteristics. Orne (l962)has suggested that subjects in psychological experiments are concerned with making a positive contribution to the research. It is, therefore, important to the subject that he be a "good subject." For the typical subject, being a good subject means that he performs in such a way that the experiment is successful, i.e. the experimenter's hypothesis is validated. The experiment, therefore, is a problem-solving situation for the subject in which he attempts to determine the experi- menter's hypothesis and respond so as to confirm the hypothe- sis. The cues which convey the experimenter's hypothesis to the subject, Orne refers to as "the demand characteristics of the experimental situation (Orne, 1962, p. 779)." Aronson and Carlsmith (1962) note the possibility that their results may have been due to demand characteristics created by their stress on the reliability and validity of the test used to establish the expectancy and performance treatments. However, they dismiss this suggestion because of their subjects' "obvious surprise" on being told the nature of the experiment. A somewhat more rigorous test of the demand characteristics explanation was made by Ward and Sandvold (1963) who attempted to minimize consistency demands by de-emphasizing the validity of the social sensitivity test. Subjects in the Ward and Sandvold study behaved as though they preferred to succeed rather than confirm a failure expectancy. Ward and Sandvold con- cluded that experimenter demands for consistency created by the emphasis on the reliability and validity of the test were responsible for the Aronson and Carlsmith re- sults. However, a study by Silverman and Marcantonio (1965) does not support either a dissonance reduction or demand characteristics explanation. Silverman and Mercantonio found, by including the reliability and valid- ity emphasis for one group and excluding it for another, that the reliability-validity emphasis increased success- seeking behavior at the eXpense of consistency behavior. The research reported above does not offer clear support either for or against the demand characteristics interpretation of the results which support dissonance theory, since consistency and achievement effects have both been obtained both in the presence and absence of the reliability-validity emphasis. However, it is question- able whether emphasizing versus minimizing the reliability- validity of the test is an adequate test of the experi- menter demands hypothesis. Since the demand characteristics explanation assumes that in the absence of demands for con- sistency the subject will maximize his performance, the manipulation to minimize the demands for consistency should not be one which at the same time decreases the importance of a good performance. It is reasonable to assume that the subject's belief about how reliable and valid the test is will affect how important to him a positive evaluation on the test will be.. Silverman and Marcantonio (1965) sug- gested that this is what happened in their study; the re- liability-validity emphasis increased their subjects' ego- involvement in the test score. An additional problem with this approach is that, even if it were able to demonstrate that consistency effects are the result of experimenter demands for consistency, it does not explain the fact that most of the replications of Aronson and Carlsmith which obtained achievement effects also contained the reliability- validity emphasis. Demand characteristics, according to Orne's (1962) analysis, are effective in influencing the subject's be- havior because the subject is concerned with the adequacy of his performance as a subject. However, in the Aronson and Carlsmith study, the subject is faced with the problem of performing well on the test of social sensitivity, in addition to the problem of performing as a "good subject." Only in the high expectancy condition can the subject do well on both. In the low expectancy condition, which is the crucial test of the dissonance prediction, the situa- tion provides a conflict for the subject. If he reSponds to the experimenter demands, he performs poorly on the social sensitivity test; if he tries to perform well on the test, he forsakes the role of a "good subject." A study by Barthel and Crowne (1962) suggests that faced with such a conflict, subjects categorize the situation as calling for one or the other of the alternative types of behavior and respond apprOpriately. In the Barthel and Crowne experi- ment, subjects given a perceptual defense test categorized the task as either a personality test or a perceptual test. Subjects who categorized the task as a personality test were more hesitant to report taboo words. In the Aronson and Carlsmith paradigm, the subjects. may have categorized the situation as either calling for performance as "good subjects" in reaponding to the demand characteristics, or as calling for a successful performance on the social sensitivity test. If this analysis is correct, making the personality aspect of the Aronson and Carlsmith experimental situation especially salient should cause the subjects to strive to maximize achievement. Conversely, making the experimenter demands for consistency salient should cause the subjects to strive for consistency. In the present investigation, the saliency of the demands for consistency and the saliency of the personality evaluation were manipulated by instructions to the subjects. In the personality condition it was emphasized that the test was an accurate measure of social sensitivity and that the experimenter was interested in finding out what the subject's personality was like. Consistency demands were created by stressing that scores on the test were usually consistent (reliable) across testing situations and across different sections of the test and that this re- liability was necessary for the experimenter's analysis to be valid.3 Only the low expectancy conditions of the Aronson and Carlsmith experimental situation were used, since the conflict between consistency demands and pres- sures for a favorable test score does not occur in the high expectancy conditions. Thus the experimental manipulations created four treatment conditions: 1) personality evaluation, low expectancy - low performance; 2) personality evaluation, low expectancy - high performance; 3) consistency demands, low expectancy low performance; A) consistency demands, low expectancy high performance. The following hypotheses were tested: 1) When the personality evaluation is made salient, subjects maximize their performance on the social sensi- tivity test, i.e. low expectancy - low performance subjects change more reSponses on the repetition of the performance trial than low expectancy — high performance subjects. k 10 2) When the demands for consistency are made salient, subjects strive for consistency in their performance on the social sensitivity test, i.e. low expectancy - high performance subjects change more reSponses on the repetition of the performance trial than low expectancy - low performance subjects. 3) Subjects in the personality evaluation condition show more achievement maximization than subjects in the con- sistency demands condition, i.e. low eXpectancy - low performance subjects in the personality condition change E more reSponses on the repetition of the performance trial than low expectancy - low performance subjects in the consistency demands condition, and low expectancy - high performance subjects in the consistency demands condition change more responses on the repetition of the perform- ance trial than low expectancy - high performance subjects in the personality evaluation condition. Nged-for-approvalg_ The concept of need-for-approval is based on the assumption that peOple differ in how im- portant it is to them to be favorably evaluated by other ~people. The Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1964) has been presented as an indirect measure of the need-for-approval. Marlowe and Crowne reasoned that socially desirable responding on a per- sonality test may indicate a high need-for-approval, since the approval of other peOple is usually dependent upon be- having or presenting oneself in socially desirable ways. 11 The Marlowe-Crowne SD Scale, therefore, is composed of items for which the socially desirable reSponse would be untrue of most peOple, e.g. "I'm always willing to admit it when I make a mistake." The person who responds so as to present himself in a favorable (socially desirable) light receives a high need-for-approval score. A person's need-for-approval should affect how he re- sponds both to a personality evaluation situation and to the demand characteristics of eXperimental situations. Since high n-approval subjects are very concerned with presenting themselves in a socially desirable manner, the high n-approval subject should be more interested than the low n-approval subject in maximizing his score on a personality test. It is, in fact, this tendency which is being measured by the M-C SD Scale. The high need-for-approval subject should also be more reSponsive than the low need- for-approval subject to demand characteristics of experi- mental situations. Being a good subject is socially de- sirable, and the good subject is more likely to win the experimenter's approval. Marlowe and Crowne (1961) found that high need-for-approval subjects are more responsive than low need-for-approval subjects to "per- ceived situational demands." Their high n-approval subjects rated the experiment, which employed a very tedious task, less unfavorably on the post-experimental questionnaire. 12 Since for the high npapproval subject, both per- sonality and demand pressures are more important than for the low n-approval subject, the conflict between the two in the Aronson and Carlsmith eXperimental situation should be greater for the high n-approval subject. In the Barthel and Crowne (1962) perceptual defense experiment, dif- ferential responding to the experimental situation on the basis of categorizing it as a personality or perceptual test was greater for hi-n—approval than for lo-n-approval subjects. Differentiating the pressure for a positive per- sonality evaluation and the experimenter demands for con- sistency in the Aronson and Carlsmith situation should result in greater differential reSponding to the two conditions by hi-n-approval than by lo-n-approval subjects. In the present investigation, the following hypotheses were tested: 1) When the personality evaluation is made salient, reSponses of hi-n—approval subjects show more striving for achievement maximization than reSponses of lo-n—approval subjects: a) in a low expectancy - high performance treat- ment group, hi-n—approval subjects change fewer responses on the repetition of the performance trial than low n—approval subjects; and b) in a low expectancy - low performance group, hi-n-approval subjects change more responses than lo-npapproval subjects. 13 2) When the demands for a consistent performance are made salient, responses of hi-n—approval subjects show more consistency-striving than responses of lo-n- approval subjects: a) in a low expectancy - high performance treatment group, hi-npapproval subjects change more reSponses than lo-n-approval subjects; and b) in a low expectancy - low performance treatment group, hi-n—approval subjects change fewer reSponses than lo-n- approval subjects. METHOD Subjects. The subjects were 23 male and 23 female undergraduate students at Michigan State University. The subjects were enrolled in the introductory course in psychology and received research credit for participating in the experiment. Recruitment of subjects was by means of a sign-up sheet on the bulletin board announcing the study as a Personality Judgment Experiment. Procedure,_ Up to four subjects were run at one time. As each subject entered the room, he was seated at one of four positions which were separated by partitions and asked to complete the Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1964). The Scale is presented to subjects as a Personal Reaction Inventory. After all subjects in the group had completed the M-C SD Scale, the experiment was introduced as either a personality or correlational study depending upon whether the subjects were in the personality evaluation or consistency demands condition. The subjects were next asked to take a bogus test which was introduced as a test of social sensitivity. The test consisted of 100 cards on which were pasted photographs of three young men. The subjects were told that one of the photos on each card was that of a 14 I f 15 schizophrenic or mentally ill person and that their task was to separate the schizophrenic from the two normals. They were instructed to use whatever cues they deemed relevant to the judging task. The subjects were told that some peOple do extremely well on the test, getting as many as 85% correct, and that some people do very poorly, getting as few as 20% correct, but that it is very difficult for people to judge their own performances. Actually there were no correct answers to the test. The photographs were taken from a 1963 Colorado State College yearbook. To the best of the experimenter's knowledge, none of the photos were of schiZOphrenics. The test was administered in five sections of 20 cards each. An Opaque projector was used to project the pictures onto the wall in front of the subjects. The subjects were asked to indicate their reSponse to each card by holding the apprOpriate one of three response cards over their shoulders. The reSponse cards were marked L, R, and C to stand for the left, right, or center photo. The experimenter who was standing behind the subjects Operating the projector was thus able to see and record the subjects' responses. After the subjects had completed each of the first four sections of the test, the ex- perimenter pretended to score their answers by comparing them to an answer key. Actually the reSponses to be 'scored correct were selected at random. Each subject 16 was given scores of 5,H,4, and 5 on the first four sections. Following the administration of the fifth section of the test, the eXperimenter pretended to have forgotten to time the section. Subjects were given a copy of the answer sheet and asked to score their responses while the experimenter went through her notes to see what could be done about the timing omission. It was expected that having the subjects score their own reSponses on this trial would help allay suSpicions about their scores, especially for those whose expectancies were disconfirmed on this trial. Both the answer key and set of responses t given to the subjects were false and had been arranged beforehand to give a prearranged score to each subject. After the subjects had finished scoring their re- Sponses, the experimenter collected the answer sheets and answer keys and informed the subjects that since she did need the time measure to complete her records, it would be necessary to have them take the fifth section of the test again. The subjects were asked to respond to the pictures as though they had never seen them before in order to make the timing accurate. After the subjects had responded to the fifth section of the test for the second time, a post-experimental questionnaire asking for the subject's reaction to the experiment was administered. The questionnaire was de- signed to test for the subject's insight into the purpose {Ks .1 78.317 ~ mR-AV‘.» '. 17 of the experiment and to determine if the instructions had successfully established differing perceptions of the nature of the experiment in the two conditions. To accomplish this, two questions were asked: "what do you think is the purpose of this experiment?" and, "in.what way do you think the experimenter will use the test scores?" The subjects were then informed of the real purpose of the {Au experiment and reassured as to the falseness of their social sensitivity scores. 5 Experimental conditions. Two experimental conditions 9 were created by different instructions about the nature of the experiment. The instructions in the Personality Evaluation Condition were designed to make salient the need for a positive personality evaluation. I'm interested in finding out the personality characteristics of Michigan State University students. The test you have just completed is one type of personality test. Now I'd like for you to take another personality test. This test has been.widely used with remarkable success by psychologists for many years. It is an excel- lent measure of how sensitive an individual is to other people, i.e. the peOple who score high on this test are the same peOple who, when in- terviewed, exPress a good deal of understanding and insight into other people. People who score low on this test, on the other hand, tend to express a very superficial understanding of other people when interviewed. The instructions in the Reliability (consistency) Demands Condition were designed to make salient the con- sistency demands of the experimental situation by indi- cating that the experiment's success depended upon the test scores being reliable. 18 I'm interested in correlating scores on two dif- ferent types of psychological tests. You have just taken the first test. Now I'd like for you to take the test to which I wish to compare it. This next test has been used by psychologists for several years as a measure of how sensitive an individual is to other peOple. I decided to use this particular test of social sensitivity because of its high reliability; that is, people who take the test more than once make about the same score each time they take it, and peOple have been found to make just about the same scores on different parts of the test. I need a reliable test in order for the comparison with the first test to be valid. So what I want to find out then, by having you take these two tests, is whether or not people in general make the same scores on this test as they make on the first test I gave you. Both the Personality Evaluation and Reliability De- mands conditions were subdivided into low expectancy - high performance and low expectancy - low performance conditions. The scores on the first four sections of the social sensi- tivity test were considered to be expectancy scores. All subjects were given low expectancies by scores of 5,4,4, and 5 on the first four sections. The score on the fifth section.was the performance score. Half of the subjects were given a score of 17 on the fifth section, and half of the subjects were given a score of 5. The low performance of those subjects receiving a score of five on the fifth section confirmed their low expectancy. The high per- formance of those subjects receiving a score of 17 on the fifth section disconfirmed their low expectancy. 19 Dependent variable. The dependent variable was the number of responses changed on the repeat performance of the fifth trial. The number of responses changed was taken as an indication of whether the subject was striving for a consistent performance or a high performance. Changing no responses would guarantee an identical per— formance, while changing a large number of responses would guarantee a low score if the previous score were high, and virtually guarantee a higher score if the previous score were very low. A large number of responses changed in the high performance conditions would indicate a striving for P— a consistent performance since the eXpectancy was low, while a small number of responses changed would indicate striving for a high performance. In the low performance condition, a large number of responses changed would in~ dicate a striving for a high performance, while a small number of responses changed would indicate striving for a reliable performance. Design. Each group of subjects reporting for the experiment was assigned to either the Personality Evaluation or Reliability Demands condition, with the two conditions being alternated. Within each group, half of the subjects were assigned to the high performance con- dition and half to the low performance condition. 20 Within each of the four resulting instructions-performance combinations, subjects whose scores on the Marlowe- Crowne Social Desirability Scale were above the median were classified as high need-for-approval, while those whose scores were below the median were classified as low need-for-approval. RESULTS Table 1 shows the mean number of reSponses changed according to performance trial, need-for-approval, and type of instructions. The npapproval classification.was obtained by dividing the scores on the Marlowe-Crowne Social Desirability Scale at the median within each performance-instructions group. The data from five sub- jects are not included in the means or the subsequent analyses. Four of these subjects' responses on the post- eXperimental questionnaire indicated a suSpicion that the experimenter was interested in how many answers they changed on the repetition of the fifth trial. The fifth subject whose data were discarded was suSpicious of the false answer sheet. The data were analyzed by a three- way analysis of variance, unweighted means (Winer, 1962). The analysis of variance is summarized in Table 2. Since the second-order interaction of performance, instructions, and need-for-approval was significant, simple effects analyses were also performed. Figure 1 indicates the significance levels of the differences between means ob- tained by the simple effects analyses. 21 22 Table l.--Mean Response Change on Repetition of 5th Trial. Need- Instructions Performance for— Approval Personality Reliability High 5.40 (n:5)* 5.00 (n=4>* High Low 8.00 (n=5)* 8.00 (n=4)** High 13.40 (ns5) 8.00 (n=6) LOW * Low 9.67 (n=6) 11.67 (n=6) *Data from one subject not included because of insight into **the purpose of the experiment. Data from one subject not included because of his discovery of the false answer sheet. Table 2.--Analysis of Variance of the Response Change Data Source df MS F p Performance (A) 1 166.77 24.67 4001 N-Approval (B) 1 19.17 2.84 Instructions (C) 1 9.02 1.33 A X B 1 20.01 2.96 (.10 A X C 1 5.62 .83 B X C 1 38.00 5.62 .<.05 A x B x c 1 30.61 4.53 4.05 Error 33 6.76 23 msoHpOSADmsH apaamsomacm 333333 an) 1.. .T \ .\.\ (item .\\ I) l.\. \\..\ NIH A..- .V \aov L. . \ a. 33 309 no. Ill.m.s Hoc.ll. cosmsaomamm mosmsaouaom 11.no. OH. _ .d.s H swam RE 3 \ x x. \.m.s\ :3 \ LINE \\ 1\ $\\\. .\.\um.s1\ \wmboammdlz :3 swam Hmpoama