Abstract

Four studies demonstrated both the power of group influence in persuasion and people's blindness to it. Even under conditions of effortful processing, attitudes toward a social policy depended almost exclusively upon the stated position of one's political party. This effect overwhelmed the impact of both the policy's objective content and participants' ideological beliefs (Studies 1-3), and it was driven by a shift in the assumed factual qualities of the policy and in its perceived moral connotations (Study 4). Nevertheless, participants denied having been influenced by their political group, although they believed that other individuals, especially their ideological adversaries, would be so influenced. The underappreciated role of social identity in persuasion is discussed. One of the most durable lessons in social psychology is the power of group influence. Yet, that lesson commands relatively modest attention in contemporary research on attitude change. While it is acknowledged that group influence can bias responses to persuasion The interpersonal transmission of meaning is underscored by the finding that as early as 12 months of age human beings interpret objects in light of the emotional expressions of others One aspect of this definitional process involves the factual qualities ascribed to the attitude object Another aspect of this definitional process involves the moral qualities highlighted in the object (see As the foregoing discussion implies, attitudes do not follow from the objective features of the object alone, for these features are, to a large extent, inferred on the basis of in-group judgments and have no intrinsic merit independent of the decision-maker's values He or she will construct the attitude object to fit its assigned social meaning, by ascribing it value-consistent factual qualities (e.g., "teaches important job skills") rather than value-inconsistent ones (e.g., "gives money away"), and by highlighting moral commitments congenial to the party's position (e.g., "we must help people to help themselves") rather than uncongenial (e.g., "we must reduce government spending"). According to this analysis, attitudes derive from the objective features of the attitude object, and their merit in light of preexisting beliefs, primarily when information about ingroup judgments is absent and social meaning must thus be inferred. Attitudes are less responsive to these factors when information about ingroup judgments is available and social meaning directly conveyed. Two conditions facilitate the effect of in-group judgments on attitude change. The first condition is ambiguity in the attitude object. To the extent that the merits or deficiencies of the object are unambiguous, and its moral connotations established (as with abortion and capital punishment), neither its factual qualities nor its moral qualities will be amenable to alternative interpretation and attitude change will thus be limited (see Blindness to Group Influence In spite of the posited large impact of group influence, people may be blind to it, and instead assume that their attitudes follow from an impartial assessment of relevant facts. Several phenomena are relevant to this prediction. First, people are motivated to see themselves as objective and free of bias, as research on the "bias blind spot" Overview of Studies The first objective of the present studies involved testing the postulated effect of group influence on attitude. Political partisans evaluated an object germane to the values of their group-a proposed welfare reform. As involved partisans, they were expected to undertake effortful (i.e., systematic or central-route) processing of the information (see A second objective of this research involved assessing the correspondence between the actual effect of group information and its perceived effect. All four studies thus asked participants to estimate how much they had been influenced by the position of their party. Study 1 Liberal and conservative college students were presented with one of two versions of a welfare policy. One version provided generous benefits, whereas the other version provided stringent benefits. Informal pilot testing confirmed both that self-identified liberals preferred the generous policy to the stringent one and that self-identified conservatives preferred the stringent policy to the generous one (effects more systematically documented in Study 2). The manipulation of policy content was crossed with a manipulation of reference group information. Half the participants were told that Democrats supported the policy; the remaining participants were told that Republicans supported it. The experiment thus featured a 2 (participant ideology: liberal or conservative) ϫ 2 (policy content: generous or stringent) ϫ 2 (reference group information: Democrats favor or Republicans favor) betweensubjects design. Method Participants All but 12 participants were recruited by telephone from a list of students enrolled in an introductory psychology course; they received course credit 809 GROUP INFLUENCE AND POLITICAL BELIEFS for their participation. The other 12 students were recruited by telephone from a separate departmental pool of participants and paid $5. Liberal and conservative students were identified using a mass testing survey administered earlier in the academic term. Students indicated their political affiliation (1 ϭ very much a Democrat, 9 ϭ very much a Republican), their political ideology (1 ϭ very much liberal, 9 ϭ very much conservative), their attitude toward welfare (1 ϭ very much oppose welfare, 9 ϭ very much favor welfare), and how strongly they felt about their attitude toward welfare (1 ϭ not at all strongly, 9 ϭ extremely strongly). Students who indicated that they were extremely Democrat and liberal (a 3 or lower on each of the relevant scales) and that they favored welfare (a 7 or higher on the relevant scale) qualified as prospective liberal participants. Students who indicated that they were extremely Republican (a 7 or higher), very conservative (a 6 or higher), and generally opposed to welfare (a 5 or lower [the bottom quartile of the distribution]) qualified as prospective conservative participants. (Because of the limited number of conservative students in the participant pool, the relevant selection criteria were loosened somewhat.) To help ensure a sample of involved partisans, all participants were also required to have indicated that they felt strongly about welfare (at or above the median response of 5). The selection procedure yielded a sample of 48 liberals (67% women) and 31 conservatives (42% women). 1 Procedure Upon arrival, participants were told that the study concerned "memory of everyday current events." They were informed that they would read a random selection of two newspaper reports and afterward complete a test assessing their recall. They were told that each report was followed by a questionnaire assessing their responses, as these might "be related to your performance on the later memory exercise." While the first report served no purpose except to reinforce the plausibility of the cover story, the second one presented a state welfare proposal. Although it was fabricated, the report was formatted to resemble an authentic newspaper article. Policy content manipulation. Participants were randomly assigned to one of two versions of the welfare policy report. The "generous policy" version offered almost $800 per month to a family with one child, an extra $200 for every additional child, full medical insurance, $2,000 in food stamps, extra subsidies for housing and day care, a job training program, and 2 years of paid tuition at a community college. While it limited benefits to 8 years, it guaranteed a job after benefits ended, and it reinstated aid if the family had another child. By contrast, the "stringent policy" version provided only $250 per month and $50 for each additional child. It offered only partial medical insurance, and imposed a lifetime limit of 1.5 years without the possibility of reinstating aid. In contrast to the generous policy, the stringent one provided no food stamps, housing, day care, job training, paid work, or college tuition. By real-world standards, the contrast between the two policies was stark. No existing program was more generous than the generous one featured here-almost all real-world programs provided less than $700 per month and limited lifetime benefits to 5 years or fewer; none provided all the additional services supplied in the generous policy used in the present research (see State Policy Documentation Project, 2001; Reference group information manipulation. Policy endorsement was also manipulated in the welfare report. In the Democrats favor condition, the report noted that the policy was supported by 95% of House Democrats (and 10% of Republicans). In the Republicans favor condition, these percentages were reversed. The reference group information was buttressed by including policy-relevant rhetoric ostensibly provided by prominent Democrats and Republicans. In the Democrats favor condition, a Democrat was quoted as saying that the policy would "lighten the financial burden of the poor," and accused Republicans of "victim blaming." A Republican was also quoted in dissent, stating that "the program was too costly," and would reward people for "having children [that they] cannot support." By contrast, in the Republicans favor condition, a Democrat was quoted as saying that the policy was "only a band-aid effort." A Republican was again quoted in dissent, stating that the program "provides sufficient coverage . . . without undermining a basic work ethic and sense of personal responsibility." Dependent measures. To assess attitude, participants were asked to indicate their support of the policy on a scale ranging from 1 (extremely opposed) to 7 (extremely in favor). To assess the perceived causal sources of their attitude, they were asked to estimate the extent to which each of the following four factors contributed to their attitude toward the welfare proposal: "The specific details of the proposal"; " [Their] own personal philosophy of the role of government in social issues"; "What the typical Democrat or Republican believes"; and " [Their] own background/experience with people on welfare." Responses were made on four separate scales ranging from 1 (doesn't contribute at all) to 7 (contributes a great deal). Next, memory of the report was assessed, and the responses served as checks on the manipulations. Participants first estimated the percentages of Democrats and Republicans who supported the policy, and then they estimated the monthly payment and the relevant time limit on benefits. Finally, participants were debriefed and thanked. Results and Discussion Preliminary Analyses To ensure that the experimental manipulations were successful, memory of relevant report details was assessed. With a few exceptions, participants accurately remembered both policy content and reference group information. Moreover, accuracy of recall did not vary as a function of experimental condition. Ultimately, however, data from 7 students were discarded prior to analysis-5 students whose memory either of policy content or of reference group information was very inaccurate (i.e., 3 or more standard deviations away from the relevant condition mean) and an additional 2 students who guessed the experimental hypothesis. Because most participants in all studies accurately recalled the content of the manipulations, these checks receive no further discussion. Analyses of Attitude Data Attitude data were examined using a 2 (participant ideology: liberal or conservative) ϫ 2 (policy content: generous or stringent) ϫ 2 (reference group information: Democrats favor or Republicans favor) analysis of variance (ANOVA). As predicted, reference group information was the most influential of the three independent variables. The relevant interaction involving participant ideology and reference group information proved highly significant, F(1, 64) ϭ 120.81, p Ͻ .001. Regardless of whether 1 The participants in the four studies presented in this article were drawn from introductory psychology classes and paid participant pools over several academic terms. According to the selection criteria used, the overall percentage of liberal students completing a given mass testing administration ranged from 7% to 38%, while the percentage of conservative students ranged from 4% to 17%. No study yielded evidence that the overall extremity of participants' ideology (as assessed in mass testing) moderated the effect of group information on attitude. However, there arose an unexpected interaction effect involving policy content and reference group information, F(1, 64) ϭ 14.74, p Ͻ .001. Regardless of participant ideology, a policy was rated more favorably if it appeared contrary to the assumed ideological biases of its supporters. The generous policy was thus rated more positively (M ϭ 4.94) than the stringent one (M ϭ 3.70) in the Republicans favor condition, t(64) ϭ 3.93, p Ͻ .001, and the stringent policy was rated marginally more positively (M ϭ 4.59) than the generous one (M ϭ 3.88) in the Democrats favor condition, t(64) ϭ 1.65, p ϭ .10. Self-Perceived Bases of Attitudes The factors that participants claimed to have affected their attitudes diverged from the factors that actually did. They asserted that the "details of the proposal" (M ϭ 5.71) and their own "philosophy of government" (M ϭ 5.88) contributed to their attitude most, but that "what the typical Democrat or Republican believes" contributed least (M ϭ 3.25), paired t(71) ϭ 9.59, p Ͻ .001; t(71) ϭ 11.00, p Ͻ .001, respectively. Indeed, they rated the position of their party as no more influential than their experience with people on welfare (M ϭ 3.33), t Ͻ 1. Summary For both liberal and conservative participants, the effect of reference group information overrode that of policy content. If their party endorsed it, liberals supported even a harsh welfare program, and conservatives supported even a lavish one. The results are consistent with the contention that people base their attitudes on social meaning. Once the policy was socially defined as liberal or conservative, the persuasive impact of its objective content was reduced to nil. Participants also denied having been influenced by the stated position of Democrats and Republicans, and instead they claimed that their beliefs followed from an apprehension of policy content (guided by their personal philosophy of government). One unpredicted finding was that each party was more persuasive when its actual position differed from its expected one (i.e., when Democrats supported a stringent policy, and when Republicans supported a generous one). This result could reflect people's preference for moderates over extremists (see Study 2 A second study was conducted to replicate and extend the initial findings and to address questions that were not answered in Study 1. One question involved the predicted pattern of results in a condition where reference group information was absent. As discussed in introducing the present studies, under such circumstances people are expected to infer social meaning by evaluating the policy in light of long-held ideological beliefs. Liberals should thus prefer the generous policy to the stringent one, and conservatives should prefer the stringent policy to the generous one. The effortful determination of social meaning would appear to reflect systematic or central-route processing, as one hallmark of such processing is responsiveness to message content The addition of a no-group-information condition would also help to improve upon the test used to assess awareness of group influence. Participants in Study 1 may have been aware of the effect of "other Democrats and Republicans," but they may have reasoned that policy content was more influential because it constituted the very object of judgment. Study 2 addressed this issue by simply assessing whether participants given group information claimed to be more affected by their party's position than participants given no group information. A third question addressed by Study 2 concerned the role of prior knowledge in moderating group influence. On the one hand, people who are knowledgeable about welfare might feel more confident in their own appraisal of the policy and thus be less influenced by the position of their party, especially if they are aware of relevant background information such as cost-of-living expenses. On the other hand, high-knowledge individuals might prove as responsive to group influence as their low-knowledge peers, if they too base their attitudes not on objective content per se, but on subjective and socially conferred meaning. Note. Scale is from 1 (extremely opposed) to 7 (extremely in favor). GROUP INFLUENCE AND POLITICAL BELIEFS Method Design and Participants The selection criteria were identical to those used in Study 1. Because of the limited number of conservative students in the participant pool, only liberal students (37 women, 17 men) took part in Study 2. They were assigned to one of four experimental conditions. In two of the conditions, either the generous policy or the stringent one was presented without further commentary. In the remaining two conditions, either the generous policy or the stringent one was presented with reference group information expected to attenuate the effect of policy content (generous policy/Republicans favor; stringent policy/Democrats favor). The other two conditions used in Study 1-the generous policy/Democrats favor condition and the stringent policy/Republicans favor condition-were conceptually unnecessary; indeed, the effect of group information in these conditions would be limited by ceiling or floor effects. Accordingly, these two conditions were dropped from Study 2. To assess the effect of prior knowledge, the mass testing questionnaire also asked students to rate their knowledge of welfare on a scale ranging from 1 (very little knowledge) to 9 (a great deal of knowledge). 2 Procedure Except for the noted change in the featured experimental conditions, the procedure was generally identical to that used in Study 1. A minor alteration was made to the questionnaire item used to assess awareness of group influence. Rather than ask participants to rate the influence of "the typical Democrat or Republican," the relevant item omitted the pejorative term "typical" and simply referred to "other Democrats and Republicans." Results and Discussion Data Analytic Strategy A median split was performed on participant knowledge, yielding a low knowledge group (with scores ranging from 1 to 5) and a high knowledge group (with scores ranging from 6 to 9). Two omnibus tests were then undertaken to analyze the attitude data. First, a 2 (policy content: generous or stringent) ϫ 2 (reference group information: absent or available) ϫ 2 (participant knowledge: low or high) ANOVA was performed to assess whether the effect of policy content was, as predicted, attenuated by reference group information and whether the effect of group information was, in turn, attenuated by participant knowledge. Because of the nature of the experimental design (i.e., policy content was not fully crossed with group information), a conservative omnibus test was also undertaken-a 4 (condition) ϫ 2 (participant knowledge) ANOVA. After both analyses were performed, follow-up contrasts were conducted to clarify obtained effects. A few participants failed to fill out the questionnaire completely; as a result, degrees of freedom vary slightly for different comparisons. Attitude Both the policy content ϫ group information interaction obtained in the 2 ϫ 2 ϫ 2 ANOVA, F(1, 46) ϭ 27.66, p Ͻ .001, and the overall main effect of condition obtained in the 4 ϫ 2 ANOVA, F(3, 46) ϭ 9.23, p Ͻ .001, yielded a highly significant result. In the absence of group information, participants (all of whom, as noted, were liberal) favored the generous policy more (M ϭ 4.75) than the stringent one (M ϭ 3.31), t(46) ϭ 3.51, p Ͻ .002. However, when group information was available, these preferences reversed, such that participants favored the Democratsupported stringent policy (M ϭ 5.00) more than the Republicansupported generous policy (M ϭ 3.20), t(46) ϭ 3.92, p Ͻ .001. Neither analysis yielded main effects or interactions involving participant knowledge, all Fs Ͻ 2. (Using participant knowledge as a continuous variable in a regression model yielded the same null result, all Fs Ͻ 1.) Self-Perceived Bases of Attitudes As in Study 1, participants attributed their attitude to policy content (M ϭ 5.65) and to their philosophy of government (M ϭ 5.41) more than to the views of "Democrats and Republicans" (M ϭ 3.15), paired t(53) ϭ 9.38, p Ͻ .001; t(53) ϭ 8.54, p Ͻ .001, respectively. Indeed, they felt that their experience with people on welfare was marginally more important (M ϭ 3.87) than the stated position of their party, t(53) ϭ 1.82, p ϭ .07. Moreover, they did not rate the position of their party as more influential in the two group information conditions (M ϭ 3.46) than in the two nogroup-information conditions (M ϭ 2.91), t(50) ϭ 1.24, p Ͼ .22. Summary In the absence of reference group information, liberal participants favored the generous policy more than the stringent one.