When are intergroup attitudes based on perceived consensus information?

Abstract

Two studies examined the extent to which individuals' attitudes toward familiar and unfamiliar social groups are differentially related to perceptions of the attitudes held by other people about those groups. In Study 1, participants indicated their own attitudes, as well as their perceptions of the attitudes of relevant ingroup members, toward nine familiar and nine unfamiliar groups. The relationship between participants' perceptions of their own attitudes and the attitudes of others was stronger for groups that were unfamiliar to them than for groups that were familiar to them. In Study 2, we manipulated perceived group familiarity and again found that individuals were more likely to desire and use consensus information about unfamiliar than familiar groups. We also found that individuals were more confident in their intergroup attitudes after receiving consensus information and that attitude confidence mediated the relationship between group familiarity and the effects of consensus on intergroup attitudes. Implications of the research for the stability and change of intergroup attitudes, as well as for the processes of social influence more generally, are discussed.

Keywords:

• Stereotypes
• prejudice
• consensus
• attitudes
• familiarity
• groups

Information about the beliefs of other people, social consensus information, has a profound influence on individuals' intergroup attitudes and behaviors (Haslam et al., 1996; Sechrist & Stangor, 2001; Sechrist, Stangor, & Killen, 2005; Stangor, Sechrist, & Jost, 2001a, 2001b; Wittenbrink & Henly, 1996). However, it is likely that consensus information influences the beliefs about some social groups more than others. That is, individuals may be more likely to depend on the social norms provided by others when making judgments about some social groups as compared to other groups. For instance, when individuals have little experience with or knowledge of a particular group, they may be especially likely to turn to others to have their attitudes validated. In the present research we tested the hypothesis that social consensus information has a greater influence on beliefs about social groups with which individuals have (or perceive to have) little experience or knowledge (that is, for unfamiliar groups) in comparison to groups which are more familiar.

CONSENSUS INFORMATION AND INTERGROUP ATTITUDES

Previous research demonstrates that perceptions of the attitudes of other people have an important influence on the development and change of stereotypes and prejudice (Crandall, Eshleman, & O'Brien, 2002; Haslam et al., 1996; Schaller & Conway, 1999; Sechrist & Stangor, 2001; Sechrist et al., 2005; Stangor & Jost, 1997; Stangor et al., 2001a, 2001b; Thompson, Judd, & Park, 2000; Wittenbrink & Henly, 1996). For instance, Stangor et al. (2001a) demonstrated that altering people's perceptions about the extent to which stereotypes are shared with others significantly changed racial stereotypes. Participants became significantly more positive in their attitudes toward African Americans when they learned that other people held more favorable stereotypes than they had originally estimated, and they became more negative toward African Americans when they learned that others held less favorable stereotypes than they had originally assumed. Furthermore, Sechrist and Stangor (2001) found that these changes constituted real acceptance of the attitudes of other people, because information that validated one's prejudices made people more likely to act prejudicially, when behavior was assessed in an unobtrusive manner, and information validating one's stereotypes made those attitudes more cognitively accessible, as assessed on a non‐reactive reaction time measure.

Additional studies support the notion that stereotypes and prejudice are based, in some measure, on social consensus. For example, Wittenbrink and Henly (1996, Experiment 3) demonstrated that high‐prejudiced participants expressed more favorable attitudes toward African Americans after they had been provided with positive, as opposed to negative, feedback about the beliefs of others. Haslam et al. (1996) found that people changed their stereotypes of national groups to be more similar to the beliefs allegedly held by members of a desirable ingroup (other nonprejudiced students at one's college), and they changed their stereotypes away from those allegedly held by an undesirable outgroup (prejudiced people). In addition, Crandall et al. (2002) demonstrated that participants followed social norms not only in their expression of prejudice, but also in their evaluations of discrimination scenarios and reactions to hostile jokes. They demonstrated that prejudiced‐based jokes were tolerated only when consistent with social norms (Study 3). If norms to suppress the expression of prejudice were salient, then such jokes were considered offensive. Taken together, these experiments provide strong and consistent support for the hypothesis that intergroup attitudes are affected by perceptions about the extent to which those attitudes are shared with others.

GROUP FAMILIARITY

However, it is not clear if information about the beliefs of other people influences individuals' attitudes toward some social groups more than their attitudes toward other social groups. It is expected that individuals' attitudes toward groups with which individuals are unfamiliar may be more influenced by social consensus information. Familiarity has been previously defined as experience with or knowledge of someone from a particular group (see Corrigan, Edwards, Green, Diwan, & Penn, 2001a; Corrigan, Green, Lundin, Kubiak, & Penn, 2001b; Corrigan et al., 2005; Holmes, Corrigan, Williams, Canar, & Kubiak, 1999). Familiarity with a social group may vary in intensity from having no experience or knowledge of members of the group, to seeing television portrayals of members of the group, to having a friend who is a member of a group, to being a member of the social group. In addition, individuals do not have to actually interact with members of a group to indicate being familiar with the group, but they must perceive that they have experience with members of group; that is, they must have knowledge of the group. For example, a person may never have had direct contact with a member of a particular social group, but still consider him‐ or herself to be highly familiar with the group based on his or her experiences with friends, television, documentaries, etc., which provided information about the group. Thus, familiar groups are groups with which people have experience or perceive they have experience.

We expect that individuals' attitudes toward familiar and unfamiliar social groups are differentially related to perceptions of the attitudes held by other people about those groups. Attitudes based on personal or direct experience tend to be particularly strong, such that they are resistant to change attempts and are predictive of behavior (Doll & Ajzen, 1992; Eagly & Chaiken, 1998; Fazio, 1989). Furthermore, attitudes toward familiar, as compared to unfamiliar, groups are expected to be held with greater confidence. Research on consumer decision making and choices suggests that confidence increases with level of familiarity with consumer products (Park & Lessig, 1981). Thus, as familiarity with a social group increases, so may one's confidence in their attitude about that group. Furthermore, because direct or perceived experience is likely to be seen as a valid source of information, the perceptions of others may not play an important role in perceptions of these groups.

However, when individuals have very little or no prior knowledge or experience with a social group, individuals have less information on which to base their attitudes and may be less confident in their judgments. In fact, research demonstrates that when faced with uncertainty, individuals turn to others to have their beliefs validated (Berscheid, 1966; Deutsch & Gerard, 1955; Haslam, 1997; Haslam et al., 1996; McGarty, Turner, Oakes, & Haslam, 1993; Sherif, 1936; Turner, 1991). Although the present research does not examine uncertainty as a moderator of consensus effects, we do expect that attitudes about familiar groups are held with greater confidence than attitudes about unfamiliar groups. Under circumstances where they are faced with making decisions about unfamiliar groups, individuals may turn to their perceptions of the attitudes held by others as a primary source of their intergroup attitudes.

In other words, when asked an opinion or attitude of a familiar group, an individual is likely to be able to perform this task without too much difficulty because he or she is likely to have prior experiences and/or direct interactions upon which to base his or her opinion. However, when asked an opinion or attitude of an unfamiliar group, individuals have less information in memory to guide them, so they will likely use social consensus information as a source of information upon which to base their opinion. Thus, individuals' intergroup attitudes may be more influenced by social consensus information when groups are unfamiliar, rather than familiar, to them.

Previous research demonstrates that familiarity with members of a group influences prejudicial attitudes. For example, familiarity with persons with mental illness has been shown to be inversely associated with prejudicial attitudes toward mental illness (Holmes et al., 1999). Furthermore, familiarity with mental illness decreased the likelihood that individuals stigmatized persons with mental illness, in terms of being stereotyped as dangerous, and discriminated against those with mental illness, in terms of avoiding them (Corrigan et al., 2001b; see also Corrigan et al., 2005). The approach of this research is that contact or experience with the group will reduce prejudice and discrimination. However, this is not always the case. In fact, Corrigan and colleagues (Corrigan et al., 2005) found that for adolescents, familiarity with persons with mental illness led to an increase in discrimination. Our approach is not to address whether familiarity with a group reduces prejudice and discrimination, but rather to examine whether social consensus information is more influential in changing individuals' attitudes toward some groups (unfamiliar groups) rather than others (familiar groups). Nevertheless, this previous research is important because it suggests that familiarity with members of a social group is highly associated with attitudes and behavior toward the group (Corrigan et al., 2001a; Corrigan et al., 2001b; Holmes et al., 1999; Link & Cullen, 1986), which is consistent with our notion that attitudes toward familiar groups are quite strong.

The present research also has important practical implications regarding stereotype change attempts. Intergroup attitudes may be the result of direct personal experience with group members (intergroup contact), as well as perceptions about the extent to which those attitudes are shared with others. Most models of stereotype formation and change (Eagly & Kite, 1987; Hamilton & Gifford, 1976; Hewstone & Brown, 1986; Rothbart & John, 1993), as well as current reviews of the stereotyping and prejudice literature (Brewer & Brown, 1998; Fiske, 1998; Hamilton & Sherman, 1994), have primarily focused on the influence of direct intergroup contact on group attitudes. Furthermore, the prevailing assumption within the literature is that stereotypes and prejudice are best changed as a result of intergroup contact, even though it is acknowledged that this change is not always that easy to create (Aronson, Blaney, Stephan, Sikes, & Snapp, 1978; Hewstone & Brown, 1986; Pettigrew & Tropp, 2000; Rothbart & John, 1985).

Although prior research has demonstrated that both pathways to stereotype development are possible, no research that we are aware of has attempted to determine the variables that influence which group attitudes are the result of which learning process. It would be expected that some group attitudes are more likely to be a function of experience, whereas other group attitudes would more likely be a function of the perceived beliefs of others. According to the present research, if individuals lack experience with or knowledge of a particular group, then information about the beliefs of others may be especially influential. Thus, attempts to change such individuals' stereotypes should focus on the beliefs of other people (specifically, providing people with information about the favorable beliefs of others). However, if individuals have direct or perceived experience with members of the social groups, then information about the beliefs of others may not be as influential. Instead, stereotype change attempts directed at these individuals may benefit from focusing on intergroup contact approaches. Thus, this research would suggest different approaches to stereotype change for different individuals and for different social groups.

It is important to acknowledge that social consensus information involves both informational and normative psychological processes (for a review see Pratkanis, 2007). Social consensus is informational such that it provides evidence about the nature of reality and one accepts this information. In other words, social consensus provides individuals with social proof about how to behave and what to think, and these behaviors and thoughts are accepted as the correct view of reality. In support of this idea, Reingen (1982) demonstrated, across five experiments, that showing targets a list of other compliers and then asking them to comply to a request increased both money and blood donations. Social consensus information is normative such that it creates a social pressure to agree or go along with the group. In other words, people conform to the expectations of others because they want to please, gain acceptance from, or avoid being rejected by those others. For example, in his classic experiments, Asch (1951, 1952, 1956) showed that individuals will publicly conform to others on an unambiguous task (i.e., simple perceptual problem), so as to not appear foolish or different. In the present studies, we focused more on the former process. That is, our research examined the informational or social proof component of social consensus information. We provided participants with information about the beliefs of other individuals. Although some participants may have felt pressure to conform to the group, we did not intentionally set up a situation in which they should have felt they had to or were forced to comply. All responses were confidential. This is not to say, however, that the normative nature of consensus is not important. We return to this point in the General Discussion.

THE PRESENT RESEARCH

This general idea that individuals' attitudes toward familiar and unfamiliar social groups are differentially related to perceptions of the attitudes held by other people about those groups leads to a number of hypotheses, which can be tested using both correlational and experimental designs. First, attitudes about existing groups that are more familiar, because they are based on experience with or knowledge of the group, should therefore be relatively more independent of the perceived beliefs of relevant others. On the other hand, attitudes about groups that are low in familiarity, because there is little direct or perceived experience, are therefore more likely based on perceived social consensus. Thus, we predicted that the relationship between one's own attitudes and the perceived attitudes of others would be stronger for unfamiliar groups than for familiar groups, and we tested this hypothesis in Study 1.

In Study 2, in order to directly test the causal hypothesis that consensus information is more important for groups with which people have little experience or knowledge (in comparison to more familiar groups), we manipulated perceived group familiarity and then assessed the extent to which individuals (a) wanted to view and (b) were influenced by consensus information about the groups. We expected that individuals would be more likely to want to learn about the relevant opinions of others when they were unfamiliar, and thus unsure, about the group characteristics. We then provided participants with either positive or negative consensus information regarding the beliefs of other people and measured their attitudes toward the familiar and unfamiliar groups following this consensus information. We expected that participants would be more influenced by consensus information about unfamiliar social groups than by consensus information about familiar social groups. In addition, we examined whether attitudes toward familiar groups are held with greater confidence than attitudes toward unfamiliar groups and whether individuals' confidence in their attitudes mediated the relationship between group familiarity and the effect of consensus information on their attitudes toward the social groups.

STUDY 1

In Study 1, familiarity was measured directly by asking participants to indicate how familiar they were with various social groups, based on their estimates of their own daily interaction with others. We assessed individuals' own attitudes, as well as their perceptions of the attitudes of others, about 18 social groups. The “others” in this case were members of a relevant ingroup, because prior research has demonstrated that the attitudes of ingroup members are more influential than the attitudes of outgroup members (Abrams, Wetherell, Cochrane, & Hogg, 1990; Haslam et al., 1996; Stangor et al., 2001a; Tajfel & Turner, 1979; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987; van Knippenberg & Wilke, 1988). To help assure generalization, a broad variety of social groups (including occupational, gender, ethnic, and religious groups) were chosen for the study. Furthermore, the groups were selected such that our participants (who were college students) would be likely to be more familiar with some of the groups than others. We expected the perceived attitudes of others to be more strongly related to personal attitudes for unfamiliar, as compared to familiar, social groups.

Method

Participants

A total of 26 male and 46 female students at the University of Maryland participated in the study in exchange for course credit in an introductory psychology course.

Procedure

Participants were told that it is common for people to have positive and negative attitudes about groups, and that the researchers were interested in these attitudes. In addition, they were told that there were no right or wrong answers, and that their responses would be completely anonymous.

To assess their personal attitudes, participants were presented with a list of 18 social groups (see Table 1), and were asked to indicate how favorable they were toward each group (1 = “Not at all favorable”, 10 = “Extremely favorable”). To assess perceptions about the attitudes of others, participants were presented with the same groups and response scales, but asked to indicate how they thought the average person in their classroom would respond to the questions, regardless of how they personally felt. Attitude order was counterbalanced, such that one‐half of the participants made the self‐ratings before the other‐ratings, whereas the other half made the other‐ratings first. Furthermore, for half of the participants the social groups were presented in alphabetical order, whereas for the other half they were presented in reverse alphabetical order. To assess perceived familiarity, participants also indicated how frequently they interacted with members of each of the groups (1 = “Interaction never occurs”, 6 = “Daily interaction”). Finally, participants were given the list of 18 social groups and asked to indicate of which groups they were members.

Table 1. Mean familiarity ratings, standard deviations, and correlations between participants' own attitudes and their perceptions of others' attitudes for each group, Study 1.

Group	Mean	SD	r
Unfamiliar
Pilots	2.42	0.93	.83**
Policemen	2.90	1.17	.62**
Construction workers	3.00	1.46	.41**
Ministers	3.04	1.50	.50**
Nurses	3.10	1.00	.46**
Mailmen	4.22	1.30	.61**
Senior citizens	4.53	0.96	.59**
Protestants	4.62	1.68	.57**
Republicans	5.07	1.34	.48**
Familiar
Asians	5.34	0.81	.54**
Catholics	5.51	1.03	.49**
Democrats	5.54	0.87	.30*
Jews	5.54	0.93	.59**
Blacks	5.72	0.59	.48**
Whites	5.96	0.20	.43**
Men	5.98	0.12	.56**
Women	5.99	0.12	.58**
College students	5.99	0.12	.55**
* = the correlation is significant at p<.05; ** = the correlation is significant at p<.001.

Results

The analyses of interest involved individual judgments nested within group order (alphabetical, reverse), attitude order (own attitudes first, others' attitudes first) and gender, and thus are best analyzed with Hierarchical Linear Modeling (HLM 5; Raudenbush, Bryk, & Congdon, 2000). A series of Hierarchical Linear Modeling (HLM) equations was constructed to examine the relationships among the variables. Within these analyses, personal attitudes was the Level 1 (within subjects) outcome variable, and participants' perceptions of the attitudes of others, participants' group familiarity ratings, and the interaction between the attitudes of others and group familiarity were the Level 1 predictor variables. Gender, group order, and attitude order were the Level 2 (between subjects) variables. The Leve1 1 equation for personal attitudes was as follows:

where Y indicates the participant's personal attitudes; B0 refers to the intercept for the participant; B1 and B2 represent the slope between personal attitudes and perceptions of others' attitudes and familiarity ratings, respectively; B3 refers to the slope between personal attitudes and the others' attitudes×familiarity ratings interaction term; and R represents error. In HLM, the parameters in the Level 1 model become the outcome variables in the Level 2 model. The Level 2 model equation for B0 was:

where G00 represents the intercept; G01 indicates the effect of gender; G02 refers to effect of attitude order (own attitudes first, others' attitudes first); G03 indicates the effect of group order (alphabetical, reverse); and U represents residual variances of the intercepts between persons (a random effect). The other Level 2 model equations were:

where G represents the intercept and U represents residual variances of the intercepts between persons (a random effect).

We found a significant positive relationship between participants' perceptions of the attitudes of others and their personal attitudes, B = .73 (.09), p<.001.1 In addition, we found that participants' group familiarity ratings predicted their personal attitudes, B = .58 (.15), p<.001, such that familiar groups were given more favorable personal attitude ratings than unfamiliar groups. More importantly, and as predicted, the group familiarity×perceptions of others' attitudes interaction also predicted personal attitudes, B = −.05 (.02), p<.01. This finding indicates that the relationship between perceptions of the attitudes of others and personal attitudes decreases as the groups become more familiar. There were no significant effects of the Level 2 variables—gender, group order, or attitude order. In other words, gender, group order, and attitude order did not predict personal attitudes.

To allow comparison of the magnitude of the parameters for familiar and unfamiliar groups, we examined the mean group familiarity rating for each of the groups. The results are shown in Table 1. On the basis of these familiarity ratings, and as shown in Table 1, nine groups were considered to be familiar groups and nine were considered to be unfamiliar groups. Table 1 also includes the correlations between personal attitudes and perceived attitudes of others for each of the 18 groups. We used HLM to test whether the perceived attitudes of others were correlated with personal attitudes for familiar and unfamiliar groups separately. Within these analyses, personal attitudes was the Level 1 outcome variable, and the perceived attitude of others was the Level 1 predictor variable. Gender, group order, and attitude order were the Level 2 variables.

We found a significant positive relationship between participants' perceptions of the attitudes of others and their own attitudes for both unfamiliar, B = .64 (.05), p<.001, and familiar groups, B = .38 (.05), p<.001. However, and validating the prior analysis, the association between own and others' attitudes was significantly higher for unfamiliar groups than for familiar groups, z = 3.47, p<.001. There were no significant relationships between gender of the participants, group order, or attitude order and personal attitudes.

Follow‐up analyses

Because the stimuli involved different types of social groups (e.g., gender, ethnic, political, religious, and occupational), it is possible that characteristics of these groups may have varied with familiarity judgments. Indeed, an inspection of Table 1 reveals that the familiar groups were more likely to be ascribed social categories (such as gender, political party, religion, and ethnicity) whereas the unfamiliar groups were more likely to be occupational categories. Therefore, it seemed important to attempt to control for these differences. We first conducted an analysis that controlled for the type of group (social categories versus occupations). To do so, we repeated the HLM analysis separately for the occupational groups (college students, construction workers, mailmen, ministers, nurses, pilots, and policemen) and the social categories (Asians, Blacks, Catholics, Democrats, Jews, Men, Protestants, Republicans, Whites, and Women). In these analyses personal attitudes was again the Level 1 outcome variable, and the attitude of others, group familiarity ratings, and the interaction between others' attitudes and group familiarity ratings were the Level 1 predictor variables. Gender, group order, and attitude order were Level 2 variables. Results showed that the expected perceptions of others' attitudes×group familiarity interaction was significant for both the occupational groups, B = −.07 (.02), p<.01, as well as for the social categories, B = −.14 (.02), p<.01. Again, gender, group order, and attitude order did not influence personal attitudes.

Another potential alternative explanation for our results is that participants are more likely to be members of the familiar groups (e.g., gender, religion, political party, and race), but not the unfamiliar groups (e.g., occupational categories). To rule out the possibility that ingroup status may have been producing the differences, we examined the demographics of participants (pertaining to their race, gender, religion, and political party) and calculated new variables based on the ingroup–outgroup status of the participants for each of these categories. Then we conducted two HLM analyses, one on the groups for which the participants were members, and one on the groups for which the participants were not members. Personal attitudes was again the Level 1 outcome variable, and the attitudes of others, group familiarity ratings, and the interaction between the attitudes of others and group familiarity ratings were the Level 1 predictor variables. Gender, group order, and attitude order were the Level 2 variables. Results showed that the expected perceptions of others' attitudes×group familiarity ratings interaction was significant for outgroups, B = −.08 (.02), p<.001, but not for ingroups, B = −.02 (.10), ns. Consensus information may not be as relevant for ingroups, because individuals are highly familiar with these groups. In fact, the mean familiarity rating for ingroups was a 5.85 on a 6‐point scale. Personal attitudes were not predicted by the Level 2 variables.

Discussion of Study 1

Study 1 provided evidence that attitudes toward groups that are relatively unfamiliar are more strongly related to perceptions of others' attitudes about those groups in comparison to attitudes about familiar groups. Specifically, the relationship between individuals' own attitudes and their perceptions of others' attitudes was stronger for unfamiliar groups than for familiar groups.

However, in Study 1, we used a correlational design, and the fact that group familiarity was measured rather than manipulated in this research naturally poses some interpretive difficulties. For example, in Study 1, familiarity was operationalized in terms of self‐report of how often participants interacted with members of different social groups. Although this seems a straightforward method of assessing familiarity, this approach does leave open an alternative explanation of our results in terms of false consensus or pluralistic ignorance. False consensus research demonstrates that people tend to overestimate how many other people hold the same attitudes and make the same judgments as themselves (Ross, Greene, & House, 1977). For example, in a study on smoking prevalence, Sherman, Presson, Chassin, Corty, and Olshavsky (1983) found that both adult and adolescent smokers estimated significantly higher numbers of smokers in the actual population (see also Marks, Graham, & Hansen, 1992; for a review see Marks & Miller, 1987). Pluralistic ignorance is a psychological state in which people believe that their private beliefs are different from those of other people, even though their public behavior is identical to that of other people (Miller & McFarland, 1991; Miller & Prentice, 1996). In a study on pluralistic ignorance and alcohol use, Prentice and Miller (1993) found that participants, in comparing themselves to the average student at their university and to friends, reported higher levels of discomfort with alcohol practices on campus.

According to a false consensus or pluralistic ignorance interpretation, it is not that social consensus leads to a more certain or stronger attitude about a social group, but that one's own attitude about a social group is used to estimate the amount of consensus about that social group. In other words, it is possible that the causal arrow is reversed in this study, such that one's own attitude is used as a heuristic to infer attitudes of other people, and that the ability to do so was stronger for familiar versus unfamiliar groups. This interpretation is also in many ways consistent with our overall expectations about the development and change of group beliefs. If people suffering from pluralistic ignorance are unaware about the extent to which others hold similar beliefs or about the social norm of beliefs, they may believe that other people hold more negative beliefs toward outgroups than they actually do, despite the fact that they exhibit similar behaviors and in actuality hold similar beliefs. However, the pluralistic ignorance and false consensus literature does not address what happens to individuals or individuals' attitudes when they find out that others hold similar or different attitudes than individuals initially perceived.

To rule out an interpretation in terms of false consensus or pluralistic ignorance and to address other issues resulting from the use of a correlational design in Study 1, we conducted a second study in which we manipulated group familiarity by having participants think about and list their prior experiences or interactions with social groups. Again, we expected that consensus information is more important for groups with which people have little contact or knowledge than familiar groups.

STUDY 2

In Study 2, we manipulated familiarity of the group and then assessed the extent to which individuals wanted to view consensus information about the groups. We also provided participants with either positive or negative consensus information regarding the beliefs of others and measured their attitudes toward the groups following this information to examine the influence of consensus information on individuals' beliefs about familiar and unfamiliar groups.

As illustrated in Study 1, we expected that consensus information would be perceived as more useful and more influential in forming attitudes about unfamiliar groups than familiar groups. Specifically, we predicted that participants would be more likely to desire consensus information about unfamiliar than familiar groups. In addition, we expected that participants would be more influenced by positive and negative consensus information for unfamiliar than familiar groups, such that they would be more likely to change their beliefs in the direction of this information when indicating their attitudes toward unfamiliar groups than toward familiar groups.

In addition, we measured participants' confidence in their beliefs prior to and following the consensus information feedback. The purpose of the former rating was to provide further support for our group familiarity manipulation, such that beliefs about familiar groups should be held with greater confidence than beliefs about unfamiliar groups. In addition, this rating was included to examine whether individuals' confidence in their attitudes mediated the relationship between group familiarity and the effects of consensus information on their attitudes toward the social groups. The latter rating was included to test the impact of consensus information, such that if participants internalize the consensus information then they should be more confident in their beliefs after learning about the beliefs of other ingroup members (Baron et al., 1996; Goethals & Nelson, 1973; Orive, 1988).

Method

Participants

A total of 142 (81 male and 61 female) students at the University at Buffalo, The State University of New York, participated in the experiment in exchange for course credit in an introductory psychology course.

Design

The experimental design of this study was a 2 (Group Familiarity: familiar, unfamiliar)×2 (Social Group Domain: ethnic, occupational)×2 (Consensus Information: positive, negative) with group familiarity as a within‐subjects factor.

Procedure

Participants were informed that the researchers were interested in people's beliefs about different social groups. Participants were given a packet of questionnaires and asked to complete them page by page without returning to a previous page or flipping forward. Participants first completed the group familiarity manipulation. To manipulate low familiarity, participants were asked only to indicate how familiar they were with a particular social group. To induce high familiarity, participants were provided with four blank lines and asked to think about all the interactions they had with members of a particular social group and to list some of these interactions.2 That is, participants were reminded of their prior experiences with the social group and asked to recall their perceptions of these experiences. They then were asked to indicate how familiar they were with the group.

The social groups that were rated consisted of either ethnic (African Americans and Asians) or occupational (nurses and mailmen) groups. Participants rated either both ethnic groups or both occupational groups. The familiarity manipulation was counterbalanced, such that sometimes African Americans were the familiar group and Asians the unfamiliar group, and vice versa, and sometimes nurses were the familiar group and mailmen were the unfamiliar group and vice versa. Familiarity ratings were made on a 1 (Not at all familiar) to 7 (Extremely familiar) Likert scale.

Participants then were told that it was possible to receive information about how other University at Buffalo students rated the two social groups they were going to rate. Participants were then asked to choose which group they wanted to learn about (either African Americans or Asians for some participants or nurses and mailmen for other participants). After they made the choice of which group they wanted to receive information about, participants were asked to indicate how much they would like to view other students' beliefs for each group on 1 (Not at all) to 7 (Very much) Likert scales. Participants also indicated how confident they were in their opinions about each group on 1 (Not all confident) to 7 (Extremely confident) Likert scales.

Participants were then provided with the beliefs of other University at Buffalo students about each group (regardless of which group they had chosen), so they could see how their beliefs compared to others. They were told that this information was based on extensive testing at the university. Participants were provided with information that other students' beliefs were either very favorable toward both groups or very unfavorable toward both groups. Specifically, in the positive consensus information condition, participants were informed that according to data collected at the university, the average student at the University at Buffalo ranked the social group as a 9.4 (or 9.6) out of a possible 10 on the favorability scale. The scale values were counterbalanced across groups. In the negative consensus information condition, participants were informed that the average student ranked the social group as a 4.4 (or 4.6) on the favorability scale. Again, the scale values were counterbalanced across groups. Participants were provided with favorable (or unfavorable) consensus information about two social groups. In order to decrease the likelihood that participants would guess the hypotheses, participants were informed that the average student ranked one social group as a 9.4 and the other group as a 9.6 (4.4 and 4.6 in the negative consensus information condition) on the favorability scale, instead of 9.5 (or 4.5) in both cases. Information was provided to participants such that other students' beliefs about the groups were on separate pages.

Participants then were asked to make several ratings regarding their beliefs about each of the two groups for which they had been provided information. First, participants were asked to indicate how favorable they were toward the group on a 1 (Not at all favorable) to 10 (Extremely favorable) Likert scale. They then were asked to indicate how confident they were in this rating on a 1 (Not all confident) to 7 (Extremely confident) Likert scale. To provide an additional measure of group perceptions, participants also indicated the extent to which they thought the group was intelligent, hardworking, and friendly on 1 (Not at all) to 7 (Very much) Likert scales. Participants made all of these ratings for one of the groups and then all of the ratings for the second group. Participants also indicated their gender and other demographic information. They then were given a suspicion check and were fully debriefed.

Results and discussion of Study 2

No participants reported being suspicious of the hypotheses or experimental conditions and thus data from all participants were analyzed. Furthermore, there were no differences pertaining to race or gender of the participants, so these variables were omitted from analyses. No differences were found regarding the type of group (occupational versus ethnic) or order of the group ratings, and therefore these variables also were deleted from subsequent analyses.

Manipulation check

Following the familiarity manipulation, participants were asked to indicate how familiar they were with each group. Results of a paired‐samples t‐test demonstrated that the manipulation of familiarity was successful, such that participants perceived the manipulated familiar group as more familiar (M = 4.78) than the manipulated unfamiliar group (M = 3.88), t(141) = 4.90, p<.001.

Desire for consensus information

Participants were asked to indicate for which group they would like to receive consensus information and to rate how much they would like to view consensus information about both the unfamiliar and familiar group. A chi‐square analysis demonstrated that, as predicted, given the choice between the unfamiliar and familiar group, a greater percentage of participants chose to view information about the unfamiliar group (M = .63) than the familiar group (M = .37), χ²(1, N = 142) = 9.71, p<.001. Furthermore, participants indicated that they were more interested in obtaining consensus information about the unfamiliar group (M = 2.94) than consensus information about the familiar group (M = 2.69), F(1, 141) = 5.37, p<.05, η = .19.

Attitudes toward social groups

After receiving the positive or negative consensus information, participants also were asked to indicate how favorable they were toward the unfamiliar and familiar groups. A 2 (Consensus Information: positive, negative)×2 (Group Familiarity: familiar, unfamiliar) ANOVA with repeated measures on the latter factor was conducted. We found a main effect for consensus information, such that participants' beliefs about the social groups were more favorable in the positive (M = 7.81) than in the negative consensus information conditions (M = 6.51), F(1, 140) = 25.22, p<.001, η = .39. Results also demonstrated a significant consensus information×group familiarity interaction, such that participants were more influenced by the consensus information when rating the unfamiliar group than the familiar group, F(1,140) = 17.55, p<.001, η = .33 (see Figure 1). Specifically, in rating the unfamiliar group, participants' favorability ratings were more favorable following positive consensus information (M = 7.99) than negative consensus information, (M = 6.01), F(1, 140) = 39.69, p<.001, η = .47. In rating the familiar group, the difference was also significant, such that participants' favorability ratings were more favorable following positive (M = 7.63) than negative (M = 7.01) consensus information, F(1, 140) = 4.39, p<.05, η = .17, but not to the extent that it was in the unfamiliar condition. Furthermore, following negative consensus information, participants' favorability ratings were more unfavorable when rating unfamiliar (M = 6.01) than familiar (M = 7.01) groups, F(1, 140) = 15.70, p<.001, η = .32, and following positive consensus information, participants' favorability ratings were marginally more favorable when rating unfamiliar (M = 7.99) than familiar (M = 7.63) groups, F(1, 140) = 3.27, p = .07, η = .15.

Figure 1. Mean favorability attitudes toward social groups as a function of group familiarity and consensus information, Study 2.

In addition, participants were asked to indicate the extent to which they thought members of the social groups were intelligent, hardworking, and friendly. These three items formed a reliable measure (Cronbach's alpha = .77), and thus analyses were conducted on the mean of the three items. Consistent with the favorability results, we found a main effect for consensus information, such that participants were more likely to assign these positive traits to the social groups in the positive (M = 5.74) than in the negative consensus information conditions (M = 5.23), F(1, 140) = 12.26, p = .001, η = .28. Results also demonstrated a significant consensus information×group familiarity interaction, such that participants were more influenced by the consensus information when rating the unfamiliar group than the familiar group, F(1, 140) = 6.21, p<.05, η = .21 (see Figure 2). Specifically, for unfamiliar groups, participants' trait ratings were more favorable following positive consensus information (M = 5.84) than negative consensus information, (M = 5.12), F(1, 140) = 18.23, p<.001, η = .34. However, for familiar groups, this difference was not significant, F(1, 140) = 2.90, ns. In addition, following negative consensus information, participants' trait ratings were marginally more unfavorable when rating unfamiliar (M = 5.12) than familiar (M = 5.35) groups, F(1, 140) = 3.62, p<.06, η = .16, and following positive consensus information, participants' trait ratings were marginally more favorable when rating unfamiliar (M = 5.84) than familiar (M = 5.63) groups, F(1, 140) = 3.17, p<.08, η = .15.

Figure 2. Mean trait ratings of social groups as a function of group familiarity and consensus information, Study 2.

Confidence ratings

We also assessed participants' confidence in their beliefs both before and after receiving the perceived consensus information. In support of our familiarity manipulation, after receiving the manipulation but prior to viewing the consensus information, participants indicated that they were more confident in their beliefs about the familiar group (M = 4.81) than the unfamiliar group (M = 4.42), F(1, 141) = 6.04, p<.05, η = .20.

A 2 (Group Familiarity: familiar, unfamiliar)×2 (Confidence: time 1, time 2) ANOVA with repeated measures on both factors was conducted to examine if participants internalized the consensus information, and thus became more confident in their beliefs. Results demonstrated a main effect for group familiarity, such that overall participants were more confident in their opinions about familiar groups (M = 5.23) than unfamiliar groups (M = 4.93), F(1, 141) = 8.15, p<.01, η = .23. Furthermore, we found a significant main effect for confidence, such that participants were more confident in their beliefs after receiving consensus information (M = 4.62 vs M = 5.54), regardless of feedback or group judged, F(1, 141) = 77.01, p<.001, η = .59. The interaction between group familiarity and confidence was not significant, F = 1.20, ns, indicating that confidence increased over time for both groups.

We also examined whether individuals' confidence in their attitudes mediated the relationship between group familiarity and the effects of consensus on favorability and trait ratings. To conduct this analysis, we followed Baron and Kenny's (1986) procedures for conducting mediational analyses. According to these procedures, four effects are necessary to support a mediational model. First, group familiarity needs to predict the outcome measure (effects of consensus information on attitudes toward social groups). Second, group familiarity needs to predict the mediator, attitude confidence. Third, attitude confidence needs to predict attitudes toward social groups controlling for group familiarity. And fourth, when the mediator, attitude confidence, is included in the analysis, the direct effect of group familiarity on the outcome measure needs to be significantly reduced.

Separate mediational analyses were conducted for the two measures of attitudes toward social groups: favorability ratings and trait ratings. Regarding favorability attitude ratings, group familiarity accounted for significant variation in the effects of consensus on attitudes, β = .36, t = 6.41, p<.001, and attitude confidence, β = .51, t = 9.86, p<.001. When attitude confidence was simultaneously regressed on both group familiarity and the effects of consensus on attitudes, the relation between group familiarity and effects of consensus on attitudes was reduced, β = .19, t = 3.11, p<.01. Testifying to the partial mediational role of confidence, the Sobel (1982) test was significant, z = 6.11, p<.001.

Regarding trait attitude ratings, group familiarity again accounted for significant variation in the effects of consensus on attitudes, β = .36, t = 6.53, p<.001, and attitude confidence, β = .51, t = 9.86, p<.001. When attitude confidence was simultaneously regressed on both group familiarity and the effects of consensus on attitudes, the relation between group familiarity and effects of consensus on attitudes was reduced to non‐significant levels, β = .09, t = 1.36, ns. Demonstrating the mediational role of confidence, the Sobel (1982) test was significant, z = 6.11, p<.001.

GENERAL DISCUSSION

The results of the present research suggest that when individuals are unfamiliar with a particular social group, they are especially susceptible to perceived social consensus information. In other words, group familiarity is an important moderator of the influence of perceived social consensus information on intergroup attitudes. Specifically, Study 1 demonstrated that attitudes toward groups that are relatively unfamiliar are more strongly related to perceptions of others' attitudes about those groups in comparison to attitudes about familiar groups. Study 2 showed that individuals are more likely to desire and depend on consensus information when groups are unfamiliar than familiar. These studies have both theoretical as well as applied implications.

We defined familiarity as experience with or knowledge of members of the social group. In Study 1, familiarity was operationalized in terms of self‐report of how often participants interacted with members of different social groups. In Study 2, we manipulated group familiarity in a very different way—by having participants think about and list their prior experiences or interactions with social groups. We believe that by using these different operational definitions, we have strong evidence demonstrating the role of group familiarity. However, it is nevertheless worthwhile noting that each manipulation has some limitations. As previously discussed, the measure of familiarity used in Study 1 does leave open an alternative explanation of our results in terms of false consensus or pluralistic ignorance. The manipulation used in Study 2 rules out an interpretation in terms of false consensus or pluralistic ignorance; however, it does allow an interpretation in terms of cognitive accessibility. That is, it could be suggested that our manipulation, in which participants were asked to think about all the interactions they had with members of a particular social group and to list some of these interactions, increased the accessibility of those exemplars. In support of our interpretation, our manipulation check showed that participants perceived the group manipulated to be more familiar as more familiar than the group manipulated to be less familiar. However, there is a close link between accessibility and familiarity, as is well known from the availability bias (constructs that are accessible are judged as frequent or familiar). Although, this interpretation does not apply to Study 1, in which familiarity with the social groups was measured.

We also predicted that attitudes about familiar groups are held with greater confidence, making them less susceptible to consensus, whereas attitudes about unfamiliar groups are held with less confidence, making them more likely based on consensus information. The results of Study 2 showed that after receiving the manipulation, but prior to viewing the consensus information, participants indicated that they were more confident in their beliefs about the familiar group than the unfamiliar group. In addition, when asked to list interactions with members of a particular social group, most of the interactions participants described were positive or neutral. In fact, only five participants described negative interactions with members of the social groups, and these participants also provided descriptions of positive interactions. Thus, our hypothesis that attitudes about familiar groups are held with greater confidence, and thus consensus information is less relevant, does not seem to be hampered by our manipulation because participants did not seem to freely recall episodes that would generate uncertainty; that is, overall they did not generate both positive and negative events. In addition, showing support for our hypothesis, attitude confidence mediated the relationship between group familiarity and the effects of consensus on attitudes toward the social group.

Because beliefs about unfamiliar groups were more easily changed, this finding is consistent with previous research demonstrating that strongly held attitudes are more difficult to change (Eagly & Chaiken, 1998). For example, across several issues (e.g., hiring quotas for women, banning pornography), Bassili (1996) found that attitudes held with certainty were less likely to change following a brief counter‐attitudinal message, compared with attitudes held with less certainty (see also Rucker & Petty, 2004; Tormala & Petty, 2002; Wu & Shaffer, 1987). Thus, relative to attitudes low in certainty, attitudes held with high certainty are not only more resistant to social consensus information, but to all kinds of information (for reviews see Gross, Holtz & Miller, 1995; Petty, Haugtvedt, & Smith, 1995). We suggest that familiarity influences confidence, as a result of direct or perceived experiences. However, familiarity could also lead to confidence by reducing the competition of conflicting responses, by increasing ease and fluency, by activating positive meanings, or through any number of other mechanisms (e.g., Brinol, Petty, & Tormala, 2006; see Koriat & Levy‐Sadot, 1999, for a review). Therefore, future research might want to investigate other mechanisms by which familiarity influences attitude confidence. The confidence results found in Study 2 are also consistent with previous research on the consequences of consensus information, such that following the consensus information participants indicated they were more confident in their beliefs about both familiar and unfamiliar social groups (Baron et al., 1996; Orive, 1988). Thus, the present studies suggest that attitude confidence or certainty is a key mechanism.

In addition to its theoretical interest, understanding the sources of intergroup attitudes also has practical importance. For some groups, such as those that are low in familiarity, negative intergroup attitudes may be most effectively changed by providing individuals with information about the favorable beliefs of other ingroup members. On the other hand, for other groups, such as those with which individuals have experience or knowledge, stereotype change attempts may focus on creating positive intergroup contact. However, it should be pointed out that these implications refer to the informational component of social consensus information. Other effects of social consensus may emerge if the social pressure or normative component of social consensus is engaged. For example, Pettigrew (1958) found that as White southern men entered the army, where the social norms were less discriminatory than they were used to, they became less prejudiced against Blacks. In addition, Minard (1952) found that White and Black coal miners in West Virginia were integrated while at work below ground, but when the workers returned to their homes and communities, there was almost no interaction between the two groups in their social lives, and attitudes and behavior were more negative. In both of these examples, individuals are being influenced by social norms regarding familiar groups. However, in these situations there were also clear social pressures to conform involved. Thus, social consensus as a normative process can be very important for change when the group is familiar. Future research would benefit from further examining the informational versus normative processes involved in social consensus.

Concluding comments

The present research suggests that group familiarity is one important moderator of consensus effects. The relationship between individuals' own attitudes and their perceptions of others' attitudes was stronger for unfamiliar groups than for familiar groups. Furthermore, individuals were more likely to desire and use social consensus information about social groups with which they were unfamiliar as compared to groups with which they perceived to be familiar. These studies may help explain why and how individuals develop strong negative attitudes about members of social groups for which they have little or no contact (Katz & Braly, 1933; Maio, Esses, & Bell, 1994). By providing individuals with information about the favorable intergroup attitudes of other people, interventions designed to reduce prejudice may be particularly effective.

Notes

1. Standard errors are included in parentheses.

2. The majority of the interactions participants described were positive or neutral. The five participants who described negative interactions with members of the social groups also provided a description of positive interactions.