Determinants of confirming and disconfirming responses to negative social labels

Determinants of confirming and disconfirming responses to negative social labels

JOURNAL OF EXPERIMENTAL Determinants SOCIAL 14, 31-42 (1978) PSYCHOLOGY of Confirming to Negative and Disconfirming Social Labels SHARON B. GU...

1MB Sizes 0 Downloads 42 Views

JOURNAL

OF EXPERIMENTAL

Determinants

SOCIAL

14, 31-42 (1978)

PSYCHOLOGY

of Confirming to Negative

and Disconfirming Social Labels

SHARON B. GURWITZ

Response

AND BR~CETOPOL

Northwestern

University

Received September 21, 1976 Two experiments explored the question of when people will respond to negative labels by confirming them and when they will respond by disconfirming them. In a field experiment subjects were accused ofnot taking advantage ofthe opportunities available in a nearby city, and in a laboratory experiment they were accused of having low self-confidence. After these accusations, subjects were given au opportunity to behave in ways that varied in the degree to which they were consistent with the experimenter’s labels. For some subjects the experimenter’s accusations included mention of the fact that the negative label also applied to many other members of their group, while for other subjects this group factor was not mentioned. The reference to the group either increased or decreased subjects’ label-confirming behavior, depending on whether the subject had provided the experimenter with evidence consistent with the label before the accusations were made. When the experimenter had this evidence, subjects subsequently confirmed the label more if the fact that the negative label applied to many group members was mentioned than if it was not mentioned. When the experimenter accused the subject without having any evidence, subjects disconfirmed the label more ifthe group was mentioned than if it was not.

Several lines of research indicate that people often behave in ways that confirm their own and others’ expectations. Rosenthal’s (1966) work on expectancy effects has demonstrated, in contexts ranging from psychological experiments with rats to classroom situations with children, that behavior confirms expectations. Research in a dissonance theory tradition has found that people will sometimes behave so as to make their performance confirm their expectations, even when these expectations are for poor performance (Aronson & Carlsmith, 1962). More recent researc has demonstrated that informing people that certain characteristics have been attributed to them can increase the likelihood that those characteristics will occur. Miller, Brickman, and Bolen (1975) found that children’s behavior changed more in response to an attribution which established an This research was supported by National Institute of Mental Health Grant No. MIX25503 to the first author. We are grateful to Philip Brickman for his comments on an earlier draft of the manuscript. Requests for reprints should be sent to Sharon Gurwitz, Department of Psychology, Northwestern University, Evanston, Illinois 60201. 31

0022-1031/78/0141-0031$02.00/O Copyright All rights

0 1978 by Academic Press, Inc. of reproduction in any form reserved.

32

GURWITZ

AND

TOPOL

expectancy for similar behavior in the future (e.g., “This classroom is one of the cleanest in the building”; “ This class is very neat”) than in response to persuasive attempts to establish the same type of behavior. People’s behavior is also often determined by the behavioral expectations that accompany their social roles. For example, Stein, Pohly, and Mueller (1971) found that children spent more time on a task if it was labeled as appropriate for members of their own sex than if it was labeled as appropriate for members of the opposite sex. Sociologists who advocate a labeling approach for understanding deviant behavior (e.g., Becker, 1963; Schur, 1971) also maintain that people sometimes behave as they do in order to fulfill label-based expectancies that others and themselves hold. For example, suppose someone had been caught stealing something. Because people label him as a thief, this person who has committed theft once will come to think of himself in terms of this label and will subsequently behave so as to make his behavior consistent with the label. In other words, labeling someone as a thief might have the unintended consequence of increasing the likelihood of his stealing in the future. In an experimental test of this social labeling hypotheses, Kraut (1973) told some subjects who gave to charity that they were charitable, and he did not label others; similarly, he told some subjects who refused to give that they were uncharitable, and he did not label others. Subjects were later asked to contribute to a second charity. As predicted, there was an interaction between whether a label was given and the nature of the label: Subjects who had given the first time gave more the second time if they had been called charitable than if they had not been labeled; while subjects who had not given the first time gave less the second time if they had been called uncharitable then if they had not been labeled. Despite all this evidence in support of the notion that people behave so as to confirm their own and others’ expectations, it seems unlikely that this pattern is universal. Everyday experiences indicate that there are times when people strive to disconfirm expectations and to demonstrate that their behavior is inconsistent with the label which has been assigned to them. This rejection of a label is most likely to occur when the label is negative. There are many times when knowledge that one’s past performance was poor will increase subsequent effort rather than make one feel resigned to mediocrity (e.g., Brickman, Linsenmeier, & McCareins, 1976). There are also many times when people will attempt to disprove negative expectations when they are based on stereotypes about groups to which they belong (e.g., Lambert, Libman, & Poser, 1960). For example, many women will respond to the allegation that they are doing a “man’s job” by trying harder to do it well. These observations raise the interesting and important question of when people will respond to negative expectations by confirming them and when they will respond by disconfirming them. The present research was designed to begin investigating this issue.

SOCIAL LABELS

33

While countless studies have demonstrated that people confirm expectations, much less research has addressed the question of when people behave in ways that are inconsistent with the labels that have been applied to them. In one recent paper, Steele (1975) demonstrated that subjects who were “called names” that were negative reacted wit behavior that was inconsistent with those names. In this research, an experimenter called housewives on the telephone and berated them for their lack of concern with cooperative efforts for the betterment of their fellow man. When they later had an opportunity to help out w community project, the housewives who had experienced the ne name calling complied with the request to help more than those who had not received the derogatory phone call. The studies by Kraut (1973) and by Steele (1975) provide conflicting answers to the question of whether people confirm or disconfirm a negative label: Kraut found that people labeled as uncharitable tended to be 1 charitable in the future than those not labeled, while Steele found t those who had been labeled as noncooperative were more cooperative i the future than those who had not been labeled. Since these studies were similar in many ways, it seemed that an analysis of the differences between them would provide a useful beginning for understanding when labels produce confirming behavior and when they produce disco behavior. A careful comparison of the studies revealed two po important dimensions on which they differed, and these dimensions became the independent variables of the present research. One difference between the Kraut (1973) study and the Steele (1975) study is that Kraut’s subjects provided evidence consistent with t uncharitable label before the label was applied to them, while Steele’s subjects did not provide the experimenter with any label-relevant evidence before he began berating them. One possibility, then, is that the subjects in Kraut’s study found if difficult to deny that the label applied to them an thus adopted it as a guide for their future behavior, while the subjects i Steele’s study rejected the label psychologically and thus attempted to reject it behaviorally. A second difference between the two experiments was that the label in the Kraut study was directed toward the subject as an individual, while the label in the Steele study was directe subject as a member of her community. This difference may also have m it easier for Steele’s subjects to reject the experimenter’s label inapplicable to them personally. Experiment 1 was designed to test the effects of these variables on subjects’ responses to a negative label. The two independent variabl (Evidence/No Evidence and Individual/Group label) were crosse factorially, resulting in one condition which most closely resemble Kraut study (Evidence/Individual label), one condition which most closely resembled the Steele study (No Evidence/Group label), and two other conditions that were similar to each of these studies in one way. It was

34

GURWITZ

AND TOPOL

predicted that subjects’ behavior would be most consistent with the label in the condition similar to Kraut’s study and most inconsistent with the label in the condition similar to Steele’s study. If this prediction were confirmed, it was hoped that the results for the other two conditions would indicate whether either or both independent variables produced the effects. EXPERIMENT

1

Northwestern University, where the experiment was conducted, is located in a suburb adjacent to Chicago. Despite the university’s proximity to the city, many students do not go into Chicago very frequently, even though they often have intentions to do so. Therefore, the accusation directed against the subject by the experimenter was that he or she did not take advantage of what Chicago has to offer, and that this is very bad. Method Subjects. Subjects were Northwestern University undergraduates living in a large dormitory which only contains single rooms, so that the experimenter could be reasonably sure that the same individual was reached on two occasions. For the first stage of the experiment, which involved a telephone call to the subject, subjects’ names were randomly selected from a list of dormitory residents. All students whose families lived in or around Chicago were excluded from the list, and then equal numbers of males and females were randomly assigned to each of five experimental conditions. A total of 94 subjects (9 or 10 members of each sex per condition) completed the experiment. An additional 10 subjects received the telephone manipulation but could not be located several days later to respond to the dependent measures. Procedure. Subjects in each of the four experimental cells were contacted by telephone by a male experimenter who identified himself as a student from another local university who was working on a sociology class project on college students’ use of the facilities available in Chicago. He said he had selected the subjects’s name randomly from the Northwestern Student Directory. If the subject was in the Individual Evidence condition, the experimenter continued by saying, “I’d like to know how often you have gone into Chicago to use its various resources and facilities this year. About how many times have you been to Chicago since the beginning of the school year in September. 7” If the subject was in the Group Evidence condition, the words, “I’dlike toknow how often you have gone into Chicago” were replaced with the words, “I’d like to know how often Northwestern students such as yourself have gone into Chicago.” The modal response to this question was zero, even though the study was conducted during the winter. A total of eight subjects in the two evidence conditions indicated that they went to the city very often (one or more times a week), and so the experimenter’s accusations would have been inappropriate for them. Therefore, dependent measures were not collected from these subjects, In the No Evidence conditions, the experimeter did not ask the subject how often he went to Chicago. Rather he followed his introductory comments with his accusations. The experimenter’s accusations were carefully prepared with several goals in mind: It was important that subjects see the remarks as having some validity, and that they realize that the comments were negative, but that they not find them terribly offensive. The comments made by the experimenter in Steele’s (1975) study were used as a model. When a draft of the experimenter’s speech was completed, he called a number of students, made his speech, and immediately debriefed them and asked for their reactions and comments. These preparations resulted in the following set of accusatory remarks, made to subjects in the Individual label conditions:

35

SOCIAL LABELS You know, there’s an awful lot to do in Chicago, and it’s really wrong of you not to make proper use of these opportunities. Isn’t it true that you spend most of your free time just sitting around in Evanston? Really, wouldn’t you say that you are timid and afraid to explore? I mean, you are being so lazy and apathetic. I can’t believe you care so little about what’s going on around you. You may as well not go to school near a city if you’re not going to take advantage of these facilities.

After these remarks, the experimenter concluded by saying that be would be sending the subject a questionnaire about how he takes advantage of what Chicago has to offer, and be thanked the subject for his time. In the Group label condition, the experimenter changed two of his remarks. The first sentence stated that it’s really wrong that Northwestern students don’t make proper use of the available opportunities, and the next-to-last sentence stated that the experimenter couldn’t believe people from Northwestern care so little about what’s going on around them. The rest of his comments were the same in the two conditions. The experimenter memorized all his comments so that he could say them in as natural a way as possible. The dependent measures were obtained by a female experimenter who knocked on the subject’s door and said that she was from the Norris Center (the student union) Programming Council, agroup which organizes activities for members of the Northwestern community. She said that they were interested in assessing the popularity of a number of activities they were considering organizing for the next few academic quarters, and she asked the subject TV fill out a brief questionnaire indicating whether he or she would want to participate in each of the activities listed. All subjects who were asked to fill out the questionnaire agreed to do so. The ten activities listed included trips to sports events, theaters, restaurants, etc., and, with the exception of a trip to the suburban Brookheld Zoo, all the activities would take place in Chicago. The subject rated each activity on a 5-point scale, where 1 meant “No, 1 definitely would not participate in this activity,” 2 meant “I probably would not participate in this activity,” 3 meant “I might or might not participate in this activity,” 4 meant “I probably would participate in this activity,” and 5 meant “Yes, I definitely would participate in this activity.” A large number of l’s and 2’s, and a low total score for the 10 activities, would be interpreted as conhrming the first experimenter’s accusations, while a large number of 4’s and 5’s, and a high total score for the ten activities, would be interpreted as disconfirming the accusations. TABLE SUBJECTS’

INTEREST

1

IN THE TEN ACTIVITIES

Mean rating for the ten activities

IN EXPERIMENT

1

Number of activities rated 1 or 2 (Confirming the label)

Number of activities rated 4 or 5 (I&confirming the label)

Individual label

Group label

Individual label

Group label

Individual label

Group label

Evidence

3.24 (?I = 20)

3.07 (n = 18)

2.85

3.50

4.70

3.94

No evidence

2.68 (n = 19)

3.23 (n = 18)

4.90

2.78

3.36

4.6i

Control

3.06 (n = 19)

3.58

3.89

36

GURWITZ

AND TOPOL

The first experimenter called subjects on Sunday and Monday evenings. At the end of each evening, he prepared a list of names and room numbers of the people the second experimenter should contact. This list contained the names of the people who had been called plus the names of control subjects who were not called. The three women playing the role of the second experimenter knew nothing about the study other than their own roles, and so they did not know that subjects had received different treatments or which condition a particular subject was in. The second experimenter first attempted to contact the subjects who had been called on Sunday the following Wednesday and to contact the subjects who had been called on Monday the following Thursday. Most subjects were, in fact, contacted on the intended day. If a subject could not be reached then, the experimenter continued to attempt to reach him or her until about noon on Saturday, after which the subject was dropped from the experiment. The experiment was conducted for 3 consecutive weeks. Immediately after it was completed, a debriefing letter explaining the methods and purposes of the study was sent to all participants, and the person who had served as the first experimenter held adiscussion session in the experimental dormitory for anyone who wished further information about the study. It appeared that no subjects suspected that there was a connection between the phone call and the questionnaire.

Results Each subject’s ratings of his or her interest in attending each of the ten events were coded to produce the following three dependent measures: mean interest score for the ten events; the number of events receiving a rating of 1 or 2 (indicating that the subject was not interested in the event); and the number of events receiving a rating of 4 or 5 (indicating that the subject was interested in the event). The mean scores for these measures for each of the five conditions are shown in Table 1. These data were analyzed using analyses of variance with four contrasts. Three of these contrasts were used to test for the two main effects and the interaction for the four experimental cells, and the fourth was used to compare the four experimental cells with the control cell. Sex of subject was also included as a factor in the analyses, but it had no significant effects on the results and thus will not be discussed further. The analyses of variance revealed that the interaction between the Evidence/No Evidence and the Individual/Group label variables was the only significant effect for each of the three measures. Subjects gave higher ratings to the ten activities in the Individual/Evidence and the Group/No Evidence conditions than in the Group/Evidence and the Individual/No Evidence conditions, (F( 1,84) = 5.35, p < .03). They assigned more ratings of 1 and 2 (indicating low interest in the activity) in the Group/Evidence and the Individual/No Evidence conditions than in the other two experimental conditions (F(1,84) = 8.13, p < .005), and they assigned more ratings of 4 and 5 (indicating high interest in the activity) in the Individual/Evidence and the Group/No Evidence conditions than in the other two experimental conditions (F( 1,84) = 4.23, p < .05). The contrast comparing the four experimental cells with the control cell was not significant. It had been anticipated that subjects in the Group/No Evidence condition

37

SOCIAL LABELS

would show relatively high interest in the ten activities. This was the condition most similar to Steele’s (1975) study, where subjects behavedin a way that disconkmed the experimenter’s label. Subjects in this condition did, indeed, give high ratings to the activities, indicating that the label did not apply to them. Their mean interest level was high, they gave scores of and 5 (indicating probable or definite interest in the activity) to nearly the activities, and they gave scores of 1 and 2 to relatively few activi The results for the Individual/Evidence condition, however, are contrary to prediction. This condition was designed to be most similar to Kraut’s (1973) study, where subjects given a negative label tended to behave consistently with that label in the future. In the present study, however, subjects in the Individual/Evidence condition behaved very similarly to subjects in the Group/No Evidence condition. They gave relatively high ratings to the ten activities, they expressed a positive interest in a relatively large number of the activities, and they expressed a negative interest in relatively few activities. It was in the other two conditions, where subjects had provided evidence for the label that was directed at them as members of their group and where subjects provided no evidence for the label that was directed at them as individuals, that subjects showed relatively lo in the activities, thus behaving in a manner that was consistent label. Because the anticipated difference between the condition modeled after Kraut’s (1973) study and the condition modeled after Steele’s (1975) stud did not occur, it seemed desirable to see whether the obtained results wou replicate in a second study using different procedures. Therefor Experiment 2 was designed to be conceptually similar to, but procedurally different from, Experiment 1. The label was changed, the depende measures were changed, and the setting was changed from the field to the laboratory. EXPERIMENT

2

Method Subjects. Subjects were 39 male and 42 female students in an introductory psychology class at Northwestern. Participation in the study fulfilled part of a course requirement. Subjects’ names were drawn from a list of students who had previously participated in a “group testing session,” where they filled out a variety of questionnaires. They were called on the telephone and asked to take part in the study, but the group testing session was not mentioned when they were called. Procedure. The labeling in the present study involved focusing the subject on his feelings of insecurity and his lack of self-confidence. Subjects came to the laboratory individually and were greeted by a male experimenter. He began by telling the subject that the research was part of alarge project dealing with people’s feelings of insecurity and lack of self-confidence. He said that the researchers were now attempting to identify the types of people who lack self-confidence and the types of situations that make people feel insecure, and that they were trying to learn more about these areas by developing and administering questionnaires and by interviewing people. To increase the subject’s perceptions ofthe value of the research project,

38

GURWITZ

AND TOPOL

he said that the ultimate goal was to be able to understand the determinants of low self-confidence and to develop therapeutic techniques to help people who suffer from this problem. If the subject was in the No Evidence treatment, the experimenter said that they were now calling students like the subject, selected at random from the subject pool, and asking them to fill out a detailed questionnaire dealing with the topic of insecurity and self-confidence. If the subject was in the Evidence treatment, the experimenter instead said that the research team already knew that the subject was someone “who experiences feelings of insecurity and low self-confidence.” He reminded the subject about the questionnaires he had filled out during the group testing session, and said that, although there was no special questionnaire in that booklet dealing with insecurity and lack of self-confidence, there were some questions throughout the booklet that were relevant to these characteristics. Because the subject’s personality profile on these questions indicated that he experienced feelings of insecurity and low self-confidence, the researchers were interested in having him fill out their more detailed questionnaire dealing with these topics. The experimenter continued by saying that he would like the subject to spend a few moments thinking about his own feelings of insecurity and his own lack of self-confidence. He then recited a short speech intended to remind the subject of the relevance of the label to his behavior. This speech was carefully constructed so that it would seem to have personal relevance for just about anyone who heard it. In the Individual label conditions, the experimenter said: Think about how you feel when you are having trouble making a decision. Think about how you feel when you do make a decision but then start thinking that an alternative that you didn’t choose would probably be better than the alternative that you did choose. Think about the times when you say something and then later either regret what you said or think of a better way you could have said it. These are all times when you are displaying insecurity and poor self-confidence. When things get tough and you wish that someone else would come along and help out or take over, you are demonstrating poor self-confidence. When you feel anxious when first meeting someone on whom you want to make a favorable impression, this too is a reflection of your insecurity and low self-confidence. Think of the times when you worry about other people’s opinions of you. These worries indicate insecurity on your part. Since these are the types of feelings that you experience so often, we hope that you will appreciate the importance of doing research on this topic. In the Group label conditions, there were two additions to the experimenter’s accusatory speech. First, he began by saying that past research has indicated that the problems of insecurity and lack of self-confidence are extremely common among Northwestern students, and that comparisons among universities indicate that Northwestern students have more problems in these areas than students at other comparable schools. Second, the final sentence was changed to state, “Since these are the types of feelings that you and so many other Northwestern students experience so often, we hope that you will appreciate the importance of doing research on this topic.” In addition to the four experimental conditions, there was a control condition which did not receive the experimenter’s labeling speech. For these subjects, the experimenter simply made his opening remarks about the purpose of the research and said that the subject had been selected at random from the subject pool to fill out the questionnaire. There were from 15 to 17 subjects in each of the five conditions, with between seven and nine members of each sex in each group. After making his remarks, the experimenter handed the subject the questionnaire, which was entitled the Z & SC ZnventorpVersion ZZZ.The questionnaire consisted of 13 questions related to the subject’s experiences with feelings of insecurity and low self-confidence. The first two questions were the primary dependent measures of interest. The first one asked,

39

SOCIAL LABELS

“How would you rate your own general level of self-confidence?“, and the subject was to respond by circling a number from 1 (“extremely low”) to 7 (“extremely high”). The second question asked, “How would you rate your own general level of insecurity/security?“, and the subject was to circle anumber from 1 (“extremely insecure”) to 7 (“extremely secure”). The next four questions were also fairly general (e.g., “Relative to other students at this university, how would you rate your own level of self-confidence?“), and the final seven questions were more specific (e.g., “When you are with a group of strangers in a social situation, do you talk less than you do when you are with friends?“). The questionnaire was made as long as possible (and was even retyped to fill three pages ratker than two) so that the subject, upon receiving it, would believe that it was indeed a research instrument of substantial importance. When the subject completed the questionnaire, the experimenter led him to an adjacent room and told him that the experimenter in there would interview him. The second experimenter’s interview consisted of a very thorough debriefing session in which the purposes, procedures, and descriptions of the experiment were fully described to the subject. Special care was taken to insure that subjects in the Evidence conditions understood that in fact they had been selected for the study at random from the list of those who participated in the group testing session, and that the experimenters had not inspected their responses to the group testing session questionnaires. The woman playing the role of the second experimenter was a very articulate and understanding person who took great care to see that the subject left the laboratory feeling positively. By having a separate experimenter conduct the debriefing sessions, the first experimenter could be kept blind to the experimental hypotheses. Although he anticipated that the experimental subjects would respond differently on the questionnaire than the control subjects, he was not informed of the anticipated differences among the experimental conditions, nor did he generate predictions which were related to the results.

Results Table 2 shows the mean responses by subjects in each condition to the “How would you rate your own general level of selfconfidence?” Higher numbers indicate higher self-confidence and thus are in the direction of discotirming the experimenter’s label. As the table shows, the pattern ofresults replicates that of Experiment 1: Subjects in the Individual/Evidence and Group/No Evidence conditions confirmed the experimenter’s remarks less than subjects in the Group/Evidence and Individual/No Evidence conditions. The analysis of variance indicated ahat

question,

TABLE RESPONSES

TO THE YOUROWN

Evidence No evidence Control

QUESTION

2

IN EXPERIMENT

2,“How

WOULD you RATE

GENERALLEVELOFSELF-CONFIDENCE?"

Individual label

Group label

4.82 (n = 17)

(n = 15)

4.67 (n = 15)

5.00 (n = 17)

3.93

4.94 (n = 17)

40

GURWITZ

AND

TOPOL

the only significant effect for this measure was the interaction between the two independent variables, F(1,71) = 4.08, p < .05. The differences among the conditions in response to the question, “How would you rate your own general level of insecurity/security?” were not statistically significant. The pattern of results for that and the remaining four general measures showed the lowest scores (in all cases meaning more confirmation of the experimenter’s label) for the subjects in the Group/Evidence condition. As in Experiment 1, sex of subject did not affect the results, nor was the contrast comparing the four experimental cells with the control cell significant. DISCUSSION

The results for the two experiments are consistent with each other, but they are not consistent with predictions. It had been anticipated that subjects in the Individual/Evidence condition, which was modeled after Kraut’s (1973) study, would confirm the experimenter’s label to a greater extent than subjects in the Group/No Evidence condition, which was modeled after Steele’s (1975) study. Instead, subjects in these two groups behaved very similarly, and both groups disconfirmed the label relative to the other two experimental groups. The arguments proposed earlier in support of the original predictions were fairly straightforward: It was reasoned that those conditions that would make the subject perceive the label as relevant to himself (e.g., his having provided evidence consistent with the label, and his being addressed as an individual) would result in behavior that confirmed the label, while those conditions which would make the subject perceive the label as inapplicable to himself (e.g., his having provided the accuser with no label-relevant evidence, and his being addressed as a representative of his group) would result in behavior which disconfirmed the label. The obtained pattern of results requires rethinking the ways in which the independent variables may have affected subjects’ implicit responses and ultimately their responses to the dependent measures. It is useful to begin by considering ways in which people might respond to the Group label manipulation, where the subject was informed that the negative quality in question was particularly characteristic of a group to which he belongs. There are several types of reactions to this group label that might result in attempts to discotirm it. One such response might be, “I’m different from the other members of this group, and the next chance I get I’m going to prove it.” A different type of response, also likely to result in disconfirming behavior, is, “The group is not really like that, and the next chance I get, I, as a group member, am going to prove it.” On the other hand, the group label might at times elicit other responses that increase the likelihood of label-consistent behavior. One likely possibility is that the social support provided by the knowledge that other group members have the problem might make the negative behavior more acceptable, so that a

SOCIAL LABELS

-41

subject might think, “I guess it’s not so bad to be like this, since many other too.” Therefore, there are a number of ways in which mention of the group is likely to make the subject either more likely or less likely to confirm the negative label, relative to the case where the group is not mentioned. While these various responses to the Group label manipulation all sound plausible, we must be able to predict when confirming responses and when disconfirming responses are likely to occur. The Evidence manipulation provides a basis for making these predictions. If the subject is aware that he has already provided behavioral evidence consistent with the label, then he should find it difficult to deny the relevance of the label to his behavior. If he also learns that many other members of his group share the same problems, the social support provided by this knowledge should make the negative characteristic seem more acceptable. This should increase the likelihood that subjects in the Group/Evidence condition will confirm the label more than subjects in the Individual/Evidence condition. Also, the impact Q Evidence manipulation may be increased by the experimenter’s claim many members of the subject’s group have the same problem, since this implies that the subject, as a group member, will have it as well. The combination of direct evidence and indirect evidence by implication may also increase the likelihood of label-confirming behavior by the Group/ Evidence subjects. The effect of the Group Label may be quite different, however, when the subject has not provided the experimenter with label-consistent evidence. If the subject accepts the experimenter’s claim that the negative quality characterizes his group, then he may seek to dissociate himself from the group by disconhrming label-based expectations. If he does not accept the experimenter’s claims about the group, then he should be eager to disconhrm the label as a representative of his group. Motivations of this sort, which are likely to be provoked when the experimenter has presented no evidence that the subject’s behavior fits the label, should increase the likelihood that Group/No Evidence subjects will disconfirm the label. In sum, the Group label can elicit cognitions that either increase or decrease label-confirming responses, relative to those that occur when one’s group is not mentioned. The Evidence manipulation is one determinant of which type of effect the Group label will have. It should be noted that the effects of the independent variables must have been less superficial than a simple desire to please or displease the person doing the labeling, since, in Experiment 1, there was no apparent connection between the experimenter who administered the independent variables and the experimenter who collected the dependent measures. The present interpretation and the original line of reasoning use designing the studies make different predictions primarily for Individual/Evidence condition. The original line of reasoning was bas part on the research by Kraut (1973), and the results from the present two experiments forced a reconceptualization of the issues because of their people who are similar to me have the problem

42

GURWITZ

AND TOPOL

inconsistency with Kraut’s results. Kraut’s study was selected as an empirical base for the reasoning that led to the present studies both because of its apparent procedural similarities but outcome dissimilarities to Steele’s (1975) research, and because it represented the diverse body of studies finding expectancy-confirming results. It should be noted, however, that the support for the negative labeling effect within the Kraut study was weak: Although the interaction between whether a label was given and whether it was positive (charitable) or negative (uncharitable) was significant, the comparison between the Uncharitable label and the Uncharitable No label conditions was not significant. Thus, the Kraut study provided a very useful springboard for the present line of research. The failure to replicate it, however, may in part be due to the weakness of the original findings. While the interpretation of the data that we have proposed is admittedly post hoc, it does account for the results and suggest hopefully fruitful conceptual issues for subsequent research. Because the two experiments reported here used such different procedures, we can have confidence that the effects are in fact reliable ones that are not unique to one labeling situation nor due to specific experimental quirks. The data indicate that negative labels do not have a standard effect on behavior, but that there is situation-specific variability in people’s responses to them. This area of investigation is an interesting and important one both theoretically and practically, and so it is hoped that subsequent research will clarify and expand our knowledge of the determinants of confirming and disconfirming responses to negative labels. REFERENCES Aronson, E., & Carlsmith, J. M. Performance expectancy as a determinant of actual performance. Journal of Abnormal and Social Psychology, 1962, 65, 178- 182. Becker, H. S. Outsiders. New York: The Free Press, 1963. Brickman, P., Linsenmeier, J. A. W., & McCareins, A. G. Performance enhancement by relevant success and irrelevant failure. Journal of Personality and Social Psychology, 1976, 33, 149-160. Kraut, R. E. Effects of social labeling on giving to charity. Journal of Experimental Social Psychology, 1973, 9, 551-562. Lambert, W. E., Libman, E., & Poser, E. G. The effect of increased salience of a membership group on pain tolerance. Journal of Personality, 1960, 28, 350-357. Miller, R. L., Brickman, P., & Bolen, D. Attribution versus persuasion as a means for modifying behavior. Journal of Personality and Social Psychology, 1975,31,430-441. Rosenthal, R. Experimenter effects in behavioral research. New York: AppletonCentury-Crofts, 1966. Schur, E. M. Labeling deviant behavior: Its sociological implications. New York: Harper & Row, 1971. Steele, C. M. Name-calling and compliance. Journal of Personality and Social Psychology, 1975, 31, 361-369. Stein, A. H., Pohly, S. R., & Mueller, E. The influence of masculine, feminine, and neutral tasks on children’s achievement behavior, expectancies of success, and attainment values. Child Development, 1971, 42, 195-207.