Internalization versus compliance: Differential ... - Wiley Online Library

18 downloads 50760 Views 1MB Size Report
Mar 2, 2013 - Internalization versus compliance: differential processes underlying minority influence and conformity. ANNE MAASS. Psychologisches Institute ...
European Journal of Social Psychology, Vol. 13, 197-21.5 (1983)

Internalization versus compliance: differential processes underlying minority influence and conformity ANNE MAASS Psychologisches Institute, Universitat Kiel. Olshausenstrasse 40160,2300.Kiel, West Germany

RUSSELL D. CLARK, Ill Florida State University at Tallahassee, Florida, U.S.A.

Abstract Two experiments investigated whether minority influence and conformity operate by the same or by different processes. It was predicted that subjects who were simultaneously exposed to a majority and a minority opinion would move towards the minority in private but towards the majority in public. The results of Experiment I supported this hypothesis. Experiment 2 investigated three hypotheses predicting that ( 1 ) the above interaction would be replicated, (2) minorities would trigger more arguments and counter-arguments, and (3) cognitive activity would mediate internalization but not compliance. Hypotheses I and 3 were supported. The second hypothesis was not supported. However, minorities were found to trigger more arguments and fewer counter-arguments than majorities. The results were interpreted as supporting the dual process model. INTRODUCTION Since Sherif’s (1935) and Asch’s (195 1) early work on conformity, it has become a social psychological truism that individuals tend to yield to a majority position even when that position is clearly incorrect. For more than 30 years, the field of social influence has been dominated by conformity research that investigated the impact of groups on the attitudes and behaviour of individuals. Since Moscovici, Lage, and Naffrechoux’s first formulation of minority influence theory in 1969, a parallel line of research has focused on the reversed influence process from the minority to the majority. A controversial issue that has recently received a great deal of attention is the question whether the social influence phenomena of conformity and minority influence are based on distinctly different processes o r whether one common influence process underlies both phenomena. Traditionally, different mediating 0046--2772/83/030197- 19$01.90 0 1983 by John Wiley & Sons, Ltd.

Received 20 September 1982 Revised I 9 November 1982

198

A. Maass and R. D. Clark, III

variables have been stressed to account for the two modes of social influence. Conformity is generally believed to be a function of dependence and power. By virtue of its size, the majority is the more powerful agent of social influence particularly where the minority is dependent upon the majority either for information or for other benefits such as social acceptance and approval (cf. Jones and Gerard, 1967; LatanC and Wolf, 1981). In contrast, minority influence is generally believed to be a function of the minority’s behavioural style (cf. Moscovici, 1976). According to Moscovici’s theory, the key to minority influence is the inference made by the majority during conflict. An initially powerless minority can induce a conflict with the majority by insisting on an alternative position. If the minority adopts certain behavioural styles, notably consistent support of the minority position, the majority will perceive the minority as certain and confident. More recently the idea of two distinctly different processes has been challenged by Doms (in press), Doms and Van Avermaet (1980b), and LatanC and Wolf (1981) who suggest that one common influence process underlies both conformity and minority influence. Although a series of studies support this notion (for summaries see Doms, in press; LatanC and Wolf, 1981), the authors acknowledge that empirical evidence is limited to studies that assess publicly expressed attitude change. In contrast, Moscovici (1980) has argued in his notion of conversion behaviour that the two modes of social influence produce qualitatively different effects, indicative of differential underlying processes. People confronted with a majority opinion in the conformity paradigm tend to comply publicly to the majority’s position without privately accepting it. Consistent minorities, in contrast, are likely to produce private acceptance even when no public compliance occurs. In line with Moscovici’s (1980) notion of conversion behaviour, a number of minority influence studies found that (a) publicly expressed changes in opinions generalized to private assessments (Nemeth and Wachtler, 1973), (b) attitude changes on reactive measures generalized to non-reactive assessment (Mugny, Rilliet and Papastamou, 1981), and (c) opinion changes on a target issue generalized to related issues (Moscovici et al., 1969; Nemeth and Wachtler, 1974; Wolf, 1979). These observations suggest that compliance in the minority influence paradigm are paralleled by private acceptance (Deutsch and Gerard, 1955) or internalization (Kelman, 1958). The conformity paradigm, in contrast, has frequently been found to produce only superficial compliance (e.g. Asch, 1951;Nemeth and Markoswki, 1972), although lasting attitude changes have been reported in cases where the majority consisted of an important reference group (e.g. Newcomb’s Bennington study, 1943; Pearlin, 1959) or where the stimulus was highly ambiguous (Sherif, 1935). More informative are those studies that have experimentally compared both influence phenomena and that have frequently supported Moscovici’s notion. Of the eight published studies comparing the two paradigms, three (Moscovici and Lage, 1976; Mugny, 1974-75, 1976) found more public compliance in the conformity paradigm than in the minority influence paradigm while the remaining five studies obtained no differences between the two paradigms (Doms and Van Avermaet, 1980; Moscovici, Mugny and Papastamou, reported in Moscovici, 1980; Moscovici and Personnaz, 1980; Personnaz, 1981; Sorrentino, King and Leo, 1980). The opposite was observed for internalization. Here, six studies reported

Internalization versm compliance

199

that minorities had greater impact than majorities (Moscovici and Lage, 1976; Moscovici et al. reported in Moscovici, 1980; for delayed but not for immediate assessment of internalized attitude change; Moscovici and Personnaz, 1980; Mugny, 1974-75, 1976; Personnaz, 1981), while two studies observed no differences between the two paradigms (Doms and Van Avermaet, 1980a; Sorrentino et al., 1980). These studies further suggest that public compliance in the conformity paradigm is generally not accompanied by internalization. While three studies obtained neither public compliance nor private acceptance (Moscovici and Personnaz, 1980; Personnaz, 1981; Sorrentino et al., 1980), four of the remaining five studies reported public compliance without internalization (Moscovici and Lage, 1976; Moscovici et al. reported in Moscovici, 1980; Mugny, 1974-75, 1976). Thus, with the exception of Doms and Van Avermaet’s (1980a) results, people in the conformity paradigm seem to comply superficially without internalizing the publicly expressed attitude change. Turning to the minority influence paradigm, the comparison studies suggest with only one exception (Moscovici et al. reporting in Moscovici, 1980 for immediate assessment) that people not only comply but also internalize the minority’s position. Not considering the two studies that failed to obtain either public compliance or private acceptance (Mugny, 1976, Experiment 2; Sorrentino et al., 1980) all of the remaining studies indicated that subjects in the minority influence paradigm showed internalized attitude changes (Doms and Van Avermaet, 1980a; Moscovici and Lage, 1976; Moscovici et al. reported in Moscovici, 1980, for delayed assessment; Moscovici and Personnaz, 1980; Mugny, 1974-75; Personnaz, 1981). The majority of these studies further found that people internalized the minority position even when they did not show public compliance (Doms and Van Avermaet, 1980a; Moscovici and Lage, 1976; Moscovici and Personnaz, 1980; Mugny, 1974-75; Personnaz, 1981) and one study (Moscovici and Lage, 1976) reported that internalized attitude change was equally likely for subjects who had previously expressed public attitude changes than for those who had not (for similar results see Moscovici et al., 1969). Although the empirical support is impressive, the idea of internalization without public compliance should be treated with caution for two reasons. First, all studies reporting internalization without public compliance employed different dependent measures for the internalization measure than for the compliance assessment. Most of these studies assessed internalization in a non-reactive way while public compliance is by definition reactive. Thus, changes that were detected by very sensitive internalization measures might not have been detectable with less sensitive public compliance measures (e.g. Doms and Van Avermaet, 1980a). It is interesting to note that the only two studies that employed identical measures for internalization and compliance, and that analysed these differences statistically’, found either no differences between compliance and internalization (Nemeth and Wachtler, 1973) or detected differences in the opposite direction (Kiesler and Pallak, 1975). Second, all but two studies (Moscovici et al. reported in Moscovici, 1980; Mugny et al., 1981) employed designs in which internalization was preceded ‘Note that two additional studies (Moscovici et al. reported in Moscovici, 1980; Mugny et al., 1981) analysed differences between the two levels of influence statistically. Since direct and indirect items were nonequivalent, however, the same criticism can be applied that was outlined above.

200

A. Maass and R. D. Clark, 111

by public compliance. Thus, despite the great number of studies reporting internalization without compliance, it still remains to be demonstrated that internalization in the minority influence paradigm will be greater than public compliance when (a) assessed in the same manner, and (b) when no prior public compliance had occurred. The answer to this question, however, has some important theoretical implications. Petty and Cacioppo (1981) have recently proposed an ‘elaboration-likelihood model’ of attitude change that distinguishes two routes of information processing, a central and a peripheral one. Assume that a person is confronted with a counter-attitudinal communication. If this person is (a) motivated and (b) able to process the counter-attitudinal message, s/he will generate arguments (favourable thoughts) or counter-arguments (unfavourable thoughts) that will lead to an enduring internalized attitude change. If the person is either not motivated or not able to process the communication, s/he will not engage in active issue-relevant thinking. Consequently, s/he will most likely show a temporary attitude shift that will endure only as long as the persuasion cues (e.g. powerful or attractive influence source) are present. In some instances, however, temporary compliance may induce dissonance which in turn may lead to permanent attitude changes. Thus, the peripheral route of information processing may eventually lead to the same outcome as the central route: a permanent change of attitudes. Translated to the social influence situation, two explanations can be offered for the observed differences between minority influence and conformity. First, a dual process model can be proposed according to which minorities (minority influence paradigm) and majorities (conformity paradigm) trigger different cognitive processes. Minorities will induce active thinking leading to permanent attitude changes while majorities will trigger peripheral information processing leading only to public compliance. This application of Petty and Cacioppo’s model closely resembles Moscovici’s (1980) notion of conversion behaviour. According to Moscovici, a consistent minority will trigger a validation process in which the subject raises arguments and counter-arguments leading to internalized attitude changes. The majority in the conformity paradigm, in contrast, will trigger a comparison process in which the subject simply compares the contradicting opinions without paying further attention to the issue in question. Here, the conflict of opinions will be resolved by public compliance while the person will pertain his/her initial attitudes in private. Second, one may propose a single process model according to which a minority communication will be processed in an identical manner as a majority communication (see e.g. Doms, in press; LatanC and Wolf, 1981). Both common sense and empirical data (Nemeth and Wachtler, in press) suggest that a person will experience more social pressure to comply in the conformity than in the minority influence paradigm. According to both cognitive dissonance theory (Festinger and Carlsmith, 1959; Frey, 1978) and self perception theory (Bem, 1967, 1972) internalized attitude change will be more likely to follow public compliance when external pressure is low than when it is high. It follows that persons in the conformity paradigm can be expected to return to their initial attitude following temporary public compliance while persons in the minority influence paradigm will be more likely to show an enduring attitude change following prior public

Internalization versus compliance

20 1

compliance. Thus, without recourse to differential cognitive processes, the observed differences in internalization can readily be explained on the basis of cognitive dissonance or self perception theory-if internalization is associated with prior public compliance. If internalized attitude change can be demonstrated to occur without prior public compliance, however, the single-process model can no longer sufficiently account for the observed differences between minority influence and conformity. Previous research has not yet provided unequivocal evidence that internalized attitude change may occur without prior public compliance. Studies have either suffered from a lack of comparability (employing differential dependent measures) or have not supported the idea of internalization without public compliance (Nemeth and Wachtler, 1973; Kiesler and Pallak, 1975). It was therefore the purpose of the first experiment to demonstrate that a minority position will be accepted in private even when no prior public compliance has occurred. In order to test the hypothesis that internalization can occur without prior compliance, subjects in Experiment 1 were exposed simultaneousIy to a minority and to a majority opinion both of which differed from the subject’s own opinion2. Gay rights were selected as target issue. In order to avoid confoundings due to the direction of argumentation, the minority was presented as favouring gay rights and the majority as opposing gay rights in half the experimental conditions, while roles were reversed for the remaining experimental conditions. Subsequently, half of the subjects were asked to express their opinion in public while the other half indicated their opinion under conditions of privacy and confidentiality without having stated their public opinion at any point during the experiment. In line with previous research, subjects were expected to move towards the majority position when asked to announce their own attitudes in public while shifting towards the minority position when expressing their attitudes in private. Hence, the following interaction was predicted: In the condition in which the minority favoured and the majority opposed gay rights, subjects were expected to become more favourable towards gay rights when asked to respond in private, but more unfavourable towards gay rights when asked to respond in public. In the condition in which the minority opposed and the majority favoured gay rights, the opposite prediction was made. Here, subjects were expected to express more unfavourable attitudes towards gay rights in private while expressing more favourable attitudes in public. No differences were expected between public and private responses in the control condition in which subjects were not exposed to an influence source. EXPERIMF,NT 1 Method Topic of discussion

‘Gay rights’ were selected as topic of discussion for the following reasons. (a) At the time when the study was conducted, the issue of gay rights received considerable *It should be noted that this design differs from both the traditional conformity and minority influence paradigm since the subject in the present experiment is neither part of the minority nor of the majority group. As an outsider, the subject is exposed simultaneously to both types of social influence (see also Latane and Wolf, 1981).

202

A . Maass and R. D. Clark, I l l

attention in Florida due to the ‘Trask-Bush Amendmenf3. Since the amendment was expected to have an immediate impact on Florida’s university system, the issue can be assumed to have some relevance to the subjects. This appears important in view of Petty and Cacioppo’s (1981) observation that a central route of attitude change is unlikely to occur when the issue is considered irrelevant. (b) A pilot study on 178 female undergraduates (conducted in October 1981) revealed that the majority of the respondents (56 per cent) expressed moderate attitudes towards gay rights. Thus, both a strongly favourable and a strongly unfavourable position on gay rights can be expected to be counter-attitudinal from the perspective of the majority of subjects. Pilot study

A four-item attitude scale on gay rights was constructed and pretested on a sample of 178 female undergraduates. The seven-point attitude scale assessed the subjects’ attitudes towards gay rights in general and towards job-related legal protection, the right to homosexual marriages, and adoption rights of homosexuals. The single item scores were averaged in order to obtain a single attitude score for each subject ranging from 1 (strongly in favour of gay rights) to 7 (strongly opposed to gay rights). The pilot study revealed that the majority of respondents (56 per cent) expressed moderate or neutral attitudes towards gay rights (slightly agree-undecided, or slightly disagree) while 12 per cent expressed favourable (strongly agree, agree) and 32 per cent unfavourable attitudes (disagree, strongly disagree). The internal consistency of the scale yielded 0.86 (Cronbach’s Alpha). Subjects

An initial attitude assessment was conducted on 326 female and 96 male undergraduates in order to select subjects with moderate attitudes towards gay rights (scoring between 2.75 and 5.25). Eighty-four female undergraduates with moderate attitudes (slightly agree to slightly disagree) towards gay rights served as subjects for Experiment 1. Procedure

In a 3 (influence condition) x 2 (public versus private response) factorial design, subjects were randomly assigned to four experimental and two control conditions. Subjects were run in groups of 1 to 4 persons per group. Communication between subjects was minimized by seating subjects at a distance of 6 feet from each other and by prohibiting communication during the experiment. Majority-conlminority-pro condition. Subjects were told that they were participating in a study on human relations and communication. They were asked to read the summary of an ostensible 60-minute group discussion on gay rights involving five college age females, supposedly freshmen of the same school. One of 3The Trask-Bush Amendment mandated the withdrawal of all public funds to institutions that provide facilities or other support to groups advocating premarital sex. The Amendment is generally considered an indirect attack on gay groups; its constitutionality was being discussed in the Florida Supreme Court during the months in which the study was conducted. The Court’s ruling was that the Amendment was unconstitutional.

Internalization versus compliance

203

the five participants in the summary was described as consistently presenting a minority position in favour of gay rights while the remaining four persons were consistently defending an anti-gay rights position. Thus, subjects were exposed simultaneously to minority and majority influence. The minority presented eight arguments in favour of gay rights and the majority presented eight arguments opposing gay rights. These arguments were selected from a pilot study in which 38 undergraduates were asked to present arguments for and against gay rights, The eight most frequently presented arguments were selected for the present study‘. Since people tend to expect minorities to yield to a majority and since uncompromising consistency is considered to be the key to minority influence (e.g. Moscovici, 1976), a one sentence statement at the end of the summary explained that the minority individual did not change her opinion during the course of the discussion. Maiority-pro/minority-con condition. The same procedure was followed in the second experimental condition. Here, however, the minority took the anti-gay rights position while the majority defended a pro-gay rights position. The number and content of the arguments was the same as in the majority-con/minority-pro condition. After reading the summary, half of the subjects were instructed to indicate their own attitudes towards gay rights in public while the other half performed the same task in private. Control condition. Subjects in the control condition were asked to express their attitudes towards gay rights in public or private without having been exposed to the summary of the group discussion. Private acceptance versus public compliance. Half of the subjects in all three conditions were asked to respond to the attitude questionnaire in private while the other half was instructed to announce their opinion in public. In the public compliance condition, the experimenter (with a clearly visible calendar in hand) gave the following instruction: ‘The group discussion you just read about was the first session of this group that is meeting several times a week for an ongoing discussion of gay rights. As I told you, this is an experiment on human relations and communication. In particular, we are interested in the effect of new group members on an already established group. We therefore would like each of you to participate in one of the group sessions. Each of you is going to meet this group at a different time for a further short discussion of gay rights. Before arranging a schedule that is convenient for you, I want you to fill in this attitude questionnaire for the following reason. You have now a very good idea of who is taking which position in this group and I would like them to have the same information about you. So, please, indicate on this questionnaire what position you are going to take in the group discussion. I will then pass your response on to the group. In this way, they will be informed what position you intend to take and will be able to prepare for the group discussion’.5 4The list of arguments may be obtained from the authors upon request. One may argue that the manipulation of public compliance versus internalizationwas confounded by the anticipation of a group interaction in the public response condition. With reference to Petty and Cacioppo’s (1981) elaboration-likelihood model, a public response may be defined as a response that is expressed in the presence of the influence source. For obvious logistic reasons, a public response was defined in the present experiment as a response in the onficipated presence of the influence source. Hence, rather than confounding public compliance and internalization, the anticipation of the group interaction was a central component of their operationalization.

A. Maass and R. D. Clark, III

204

In the private acceptance condition, the subjects were instructed to indicate their ‘true’ opinions on gay rights under purportedly confidential conditions. ‘Before we move on to the next step in this experiment, I would like you to fill in your true personal opinion about gay rights in this attitude questionnaire. For reasons of confidentiality, I do not want you to put your name or social security number or any other identification on this questionnaire. Make sure that you do not identify yourself in any way. After you have answered all questions on the questionnaire, please, put your questionnaire in this ballot box (pointing to a large ballot box) so your response will remain anonymous’. The questionnaires, although otherwise identical had slightly different margins that allowed the experimenter to identify individual response sheets. Subjects were then carefully debriefed, sworn to secrecy, and thanked for their participation. Dependent measure The post-test attitude scores served as dependent measure and the pre-test score as covariate. This procedure was given preference over the calculation of difference scores because of the small sample size (N = 14 per cell) and possible randomization problems associated with small samples. Further, random assignment to public versus private response conditions was somewhat limited due to the fact that subjects who were run individually (N = 9) had to be assigned to the public response conditions, i.e. the experimental manipulation of confidentiality in the private response conditions was credible only when more than one subject was present.

Results Attitude change A 3 (majority-con/minority-pro versus majority-pro/minority-con versus control condition) x 2 (private versus public) ANACOVA was performed on the post-test attitude scores using the pre-test attitude scores as the covariate. As predicted, neither influence condition nor private versus public response yielded reliable main effects, while the two variables were found to interact significantly with each other, F (2,77) = 4.49, p < 0.02, see Table 1.

Table I . Adjusted means of post-test attitude scores in Experiment 1* Influence condition Maj-con/ min-pro

Control

Maj-pro/ min-con

Private

3.70

3.94

4.38

Public

4.22

4.14

3.48

Types o f response

‘Means represent attitude scores on a seven-point scale with higher scores indicating stronger opposition t o gay rights.

Internalization versus cornplinnce

205

A priori determined, follow-up f-tests compared private with public responses in the control and both experimental conditions. No differences between private and public responses were observed in the control group, t (77) = 0.58, n.s. As predicted, subjects in the majority-pro/minority-con condition expressed significantly more favourable attitudes towards gay rights in public than in private, t (77) = -2.63, p < 0.005; subjects in the majority-con/minority-pro condition expressed more favourable opinions towards gay rights in private than in public, although this difference did not reach customary levels of significance, t (77) = 1 . 4 6 , < ~ 0.08. Further, private responses in the majority-con/minority-pro condition were significantly more favourable towards gay rights than were responses in the majority-pro/minority-con condition, f (77) = 1.89, p < 0.05.

EXPERIMENT 2 The second experiment pursued two objectives. First, a replication of the first experiment seemed warranted, particularly since the results may seem intuitively surprising. Second, it was the objective of this experiment to investigate the premises of the dual process model in more detail. Moscovici (1980) has argued that minorities are more likely than majorities to trigger the generation of arguments and counter-arguments. Further, both Petty and Cacioppo (1981) and Moscovici (1980) have proposed that internalized attitude change is achieved by the generation of issue-relevant arguments and counter-arguments while public compliance does not require this type of cognitive activity. Hence, three hypotheses were formulated. Hypothesis I predicted that the interaction between influence condition and private versus public responses observed in Experiment 1 would be replicated. Hypothesis I1 predicted that a minority source of influence would be more likely than a majority source to trigger the generation of arguments and counter-arguments. Hypothesis I11 predicted that private acceptance but not public compliance would be mediated by the generation of arguments and counter-arguments. In order to test these hypotheses, subjects were exposed to a similar experimental procedure as in the previous experiment. In addition, the subjects’ cognitive effort was assessed following a procedure commonly employed in research on self-generated attitude change (e.g. Petty and Cacioppo, 1977; Petty, Harkins and Williams, 1980). Method

Subjects Thirty-one6 female and eighteen male undergraduates with moderate attitude towards gay rights (slightly agree to slightly disagree) were selected on the basis of the initial attitude assessment (see Experiment 1).

6 0 n e female subject was excluded since she reported to have revised her attitude towards gays dramatically since the pretest due to a new job that allowed daily interaction with gay co-workers.

206

A. Maass and R. D. Clark, 111

Procedure The procedure was similar as in Experiment 1. The design, however, was extended to a 2 (majority-con/minority-pro versus majority-pro/minority-con) x 2 (minority versus majority source of influence) x 2 (order of presentation) x 2 (public versus private assessment) factorial design in which the first, third, and last factor represent between subjects variables while the second factor is varied within subjects’. Subjects were randomly assigned to one of eight experimental conditions. Subjects were run in groups of one to four students; each group consisted either of male or of female students only. Majority-conlminority-pro versus majority-prolminority-con condition. As in Experiment 1, subjects in the majority-con/minority-pro condition were instructed to read the summary of an ostensible group discussion in which the majority opposed and the minority favoured gay rights while the distribution of opinions was reversed in the majority-pro/minority-con condition. For male students, the experimental material was modified so that the ostensible group was described as consisting of five male rather than female students. Minority versus majority source of influence. After reading the summary, subjects in both conditions were given 15 minutes to list their own arguments to the position presented by the minority (minority source of influence) and to that presented by the majority (majority source of influence). Two sets of six ‘idea spaces’ were provided for this task. It was explicitly pointed out to the subjects that they did not need to use all ‘idea spaces’, nor were they required to fill an equal number of spaces in each set. After 15 minutes subjects were interrupted by the experimenter and asked to complete the thought they were currently recording, without starting a new one. After completion of their last thought, subjects were asked to rate each thought as either in favour of, neutral towards, or opposing gay rights. Order ofpresentation. In an attempt to avoid an order bias, half of the subjects in both influence conditions received a form on which the majority-related idea spaces were located on the left and the minority-related idea spaces on the right side while the order was reversed for the remaining half of the subjects. In order to minimize confusion, the summaries of the ostensible group discussions were organized in the same manner as the idea spaces (either the majority arguments left and the minority arguments right, or vice versa). Private versus public response. Immediately before responding to the post-test, subjects were reminded of the distribution of opinions in the ostensible group (four favouring and one opposing gay rights or vice versa). This procedure was considered necessary since an extensive pilot study had revealed that subjects were frequently unable to recall the distribution of opinions accurately after the 15 minute delay’. After recalling the distribution of opinions, half of the subjects in each cell were asked to indicate their opinion on gay rights in public while the remaining subjects responded in private, using the same manipulation as in Experiment 1. ’Note that public versus private response was introduced in the experiment after the assessment of cognitive activity. It was included in the present analysis only in order to assure that no bias had occurred in the assignment of subjects to the private and public response conditions. *Particularly in the majority-pro/minority-concondition, the subjects’ memory may have been altered by their perceptions of the opinion distributionoutside the laboratory. Note that a comparable sample of 30 undergraduates with moderate attitudes towards gay rights estimated that 70.2 per cent of the American public opposed gay rights.

Internalization versus compliance

207

At the end of the experimental session, subjects were carefully debriefed, sworn to secrecy, and thanked for their participation. Dependent variables Cognitive activity. The subjects’ cognitive activity (generation of arguments and counter-arguments) was assessed by a procedure commonly used in studies on self-generated attitude change (e.g. Petty and Cacioppo, 1977, 1979; Petty, Cacioppo and Heesacker, 1981). Subjects were given 15 minutes to list their thoughts to both the minority’s and the majority’s arguments. Twelve 4 x 1.25 inch (10.16 cm x 3.16 cm) boxes created an ‘idea space’ providing for a maximum of six spaces for reactions to the minority and six spaces for reactions to the majority position. Subjects were instructed to write their ideas or thoughts down, one per idea space. After recording their thoughts, subjects were asked to rate each of their thoughts as either supporting gay rights (+), as neutral (0), or as opposing gay rights (-). Since previous research has indicated high correlations between the subject’s own ratings and those of independent raters (e.g. Petty et al., 1981), no additional judges were used. Three sets of cognitive activity scores were calculated for each subject. First, the total number of thoughts generated in the minority and in the majority idea space (between 0 and 6 respectively) represented the two summary scores of cognitive activity. Second, separate argumentation and counter-argumentation scores of cognitive activity for the source of influence were calculated for both, the thoughts generated in the majority and those generated in the minority idea spaces. The argument scores consisted of the number of thoughts that supported the viewpoint of the influence source, the counter-argument scores of those that opposed the viewpoint of the influence source (majority or minority, respectively). These two sets of scores, the summary scores of cognitive activity and the argumentation/counter-argumentationscores, served to test differences between the majority’s and the minority’s ability to trigger cognitive activity (Hypothesis 11). A third set of scores, direction scores ofcognitive activity, were calculated for the majority and the minority spaces combined. The three direction scores consisted of the total number of arguments supporting gay rights, the total number of neutral arguments, and the total number of arguments opposing gay rights. The latter scores served to assess the mediating function of cognitive activity (Hypothesis 111).

Attitude change

In order to test the first hypothesis (replication of interaction pattern observed in Experiment l ) , the data were analysed in the same manner as in Experiment 1. A 2 (majority-con/minority-pro versus majority-prolminority con) x 2 (private versus public response) x 2 (order) ANACOVA was performed on the post-test attitude scores with pretest attitude scores and sex of subject as covariates. Sex of subject was included as covariate in all analyses since an all-female sample was used in the first experiment while both males and females served as subjects in the second experiment. As in Experiment 1, a significant interaction between influence condition (majority-con/minority pro versus majority-pro/minority-con) and type

208

A . Maass and R. D. Clark, 111 Table 2. Adjusted means of post-test attitude scores in Experiment 2' Influence condition Maj-con/ min-pro

Maj-pro/ min-con

Private

3.68

4.33

Public

4.24

3.34

Types of response *Means represent attitude scores on a seven-point scale with higher scores indicating stronger opposition to gay rights.

of response (private versus public) was observed, F (1,38) = 8.74, p < 0.01, see Table 2. None of the remaining interactions or main effects yielded significant results. Further, sex of subject was not found to effect the dependent variable in a systematic way while the pretest scores, not surprisingly, were found to be a reliable < 0.001. predictor of the post-test score, F (1,38) = 2 9 . 1 7 , ~ A priori determined follow-up t-tests on the post-test scores between private and public responses in both influence conditions revealed that subjects in the majority-pro/minority-con influence condition expressed significantly more favourable attitudes towards gay rights when asked to respond in public than when asked to respond in private, t (38) = - 2 . 6 5 , ~ < 0.01. In contrast, subjects in the majority-con/minority-pro influence condition tended to show more favourable attitudes when asked to respond in private than when asked to respond in public, although this comparison did not reach customary levels of significance, t (38) = 1 S 0 , p < 0.08. Further, private responses in the majority-con/minority-pro condition were significantly more favourable towards gay rights than private responses in the majority-pro/minority-con condition, t (38) = 1.72, p .c 0.05. Cognitive activity Amount of cognitive activity. The second hypothesis predicted that a minority would trigger more cognitive activity (generation of arguments and counter-arguments) than a majority. In order to test the second hypothesis, the summary scores of cognitive activity (number of majority- and minority-induced arguments) were subjected to a 2 (majority-con/minority-pro versus majority-pro/minority-con) x 2 (private versus public response) x 2 (order) x 2 (majority versus minority source of influence) ANACOVA with repeated measures on the last variable and sex of subject as the covariate. A main effect was expected for majority versus minority source of influence. Contrary to this expectation, none of the above factors yielded significant results. In particular, subjects were not found to generate more arguments when the source of the influence was a minority (X = 3.33) than when it was a majority (X= 3.70). Direction of cognitive activity. Cognitive activity is generally conceptualized as embracing two components, argumentation (generation of thoughts that support the viewpoint of the influence source) and counter-argumentation (generation of thoughts that oppose the viewpoint of the influence source). In view of this distinction, it is conceivable that minorities may trigger more arguments than

Internalization versus compliance

209

majorities while majorities may trigger more counter-arguments than minorities. In order to explore this possibility, a 2 (majority-con/minority-pro versus majority-prolminority-con) x 2 (order) x 2 (majority versus minority source of influence) x 2 (direction of argumentation: arguments versus counter-arguments) ANACOVA was performed on the argument and counter-argument scores with repeated measures on the last two variables and sex of subject as the covariate. The results of the analysis revealed a main effect for direction of argumentation, F (1,44) = 17.93,p < 0.001, indicating that subjects were more likely to generate arguments (supporting the viewpoint of the influence source, X = 3.58) than counter-arguments (opposing the influence source, X = 1.81). Further, an interaction between influence condition and source of influence, F (1,44) = 4.73, p < 0.05, indicated that more arguments were generated in reaction to the majority source in the majority-con/minority-pro condition (X = 3.20) than in any other condition (X = 2.42 for minority source in majority-con/minority-pro condition; 2 = 2.46 for majority and X = 2.70 for minority source in majoritypro/minority-con condition). Most importantly, a significant interaction was found between influence source and direction of argumentation, F (1,44) = 4.11, p < 0.05. Minorities were slightly more likely than majorities to trigger the generation of arguments although the follow-up t-test was non-significant (X= 1.90 versus X = 1.64, respectively). Majorities, however, were significantly more likely than minorities to evoke counter-argumentation (X = 1.15 versus X = 0.66, respectively), t (44) = 2.04, p < 0.05. In summary, the resuits of the above analyses indicate that the two influence sources do not trigger differential amounts of cognitive activity. When considering the direction of argumentation, however, minorities appear somewhat (though not significantly) more likely than majorities to trigger arguments (thoughts that support their viewpoint) while they are significantly less likely than majorities to trigger counter-arguments (thoughts that oppose their points of view).

Mediating function of cognitive activity The last, and most critical, hypothesis predicted that the degree and direction of cognitive activity would mediate internalization but not public compliance. In order to test this hypothesis, an hierarchical multiple regression procedure was employed (e.g. Batson, 1975) in which the primary predictor variable (influence condition) was initially entered into the regression equation before the variables that are believed to mediate (number of arguments supporting, neutral towards, or opposing gay rights). In a second analysis, the primary predictor variable (influence condition) was entered into the regression equation after the mediator variables. This procedure was followed separately for private and public responses, using attitude change scores (pretest minus post-test scores) as the dependent variable. In the private response condition it was predicted that influence condition (majority-con/minority-pro versus majority-pro/minority-con) would account for overlapping portions of the variance in attitude change. Thus, any effect of influence condition on attitude change was expected to be wiped out if the mediator variables (number of arguments supporting, neutral towards, and opposing gay rights) were entered into the regression model before the influence condition. In

210

A. Maass and R. D. Clark, III

Table 3. Multiple regression analysis of cognitive activity scores as mediators between influence condition and attitude change Order of entry of cognitive activity scores After influence condition Response condition Private response

Public response

Source

df

MS

F

p

4.00