Isenberg’s (1986) review found evidence consistent with the PAT explanation of the choice shift. The first type of evidence comes from studies that have coded the arguments presented in group discussion, and then examined whether the number of risky (cautious) arguments correlates positively with the riskiness (cautiousness) of the final decision. Studies of this type generally find that there is a correlation between the number of risky (cautious) arguments presented during discussion, and the extent of a risky (cautious) shift. Of course, strong evidence for causality cannot be provided by correlational data, as a third factor could be the cause of both the type of arguments presented in discussion and the type of risky shift observed.
More robust evidence for PAT comes from studies that measure individual posttest positions and compare them to the final group position. If persuasion did occur and is the mechanism that explains the shift, then the final group position is expected to equal the mean of the individual group members’ posttest positions. Studies that have examined both variables (Boster, Fryrear, Mongeau, & Hunter, 1982; Boster, Mayer, Hunter, & Hale, 1980) have found that there is a very high correlation between them (exceeding .80.)
Another type of evidence discussed by Isenberg (1986) involves experiments that control the number of risky versus cautious arguments received by group members. When subjects receive more risky (cautious) arguments than cautious (risky) arguments, a risky (cautious) shift typically emerges, which is the expected outcome if the choice shift is the result of persuasion, not normative pressures.
Although the data indicate that process of persuasion explains choice shifts accurately, there is disagreement as to how that processes is manifested (Meyers & Seibold, 1990). A number of alternative explanations of the persuasion process have been proposed, including Hoffman’s (1961) group valence model (GVM; see also Hoffman & Kleinman, 1994), McPhee, Poole, and Seibold’s (1982) distribution of valence model (DVM; see also McPhee, 1994), and Boster and colleagues’ (1980, 1982) linear discrepancy model (LDM).
GVM proposes that the most important factor in determining the adoption of a decision choice is not the attitudes of individual group members toward each decision choice, but the group’s perception of a decision choice’s valence. A study testing the GVM would code the arguments provided in group discussion for each decision option, and calculate a valence score for each option by subtracting the number of negatively valenced comments from the number of positively valenced comments. The decision option the theory predicts the group will choose is the option that has the largest positive differential score.
The DVM proposes, on the other hand, that individual judgments do matter in determining the group outcome. The group decision choice predicted by the DVM is the choice that has the highest valence for the most group members. An experiment testing the DVM would assess the valence of comments made by individual group members during discussion, and calculate which option was described more positively by each group member. A decision criterion, such as majority rule, is then applied, such that the decision option that was most positive for the most group members would be the decision option predicted to be chosen by the group.
In many cases both models make the same predictions, but there are cases where the models differ. An example would be when a minority makes a number of strong arguments for a given position, and a majority argues less strongly for a different position. The GVM would predict that the minority position would prevail, as its decision option would have had the highest differential between positive and negative arguments presented. Alternatively, DVM would predict that the majority position would prevail, as the largest number of group members would have spoke in favor of the majority position during discussion. Meyers and Brashers (1998) is one of the few studies to test competing predictions from the two models, and found the data to be more consistent with the DVM than the GVM for cases in which the models made competing predictions.
Although both the GVM and DVM attend carefully to the types or arguments offered in discussion, they specify neither the operative form of social influence nor the process used by group members to influence each other. A model that attempts to address this issue is the LDM (Boster et al., 1980, 1982, 1991), a version of French’s (1956) theory of communication discrepancy. The LDM proposes that all group members are persuaded by the arguments advanced by their fellow group members throughout the group discussion process until a consensual position is reached. Individual group members hear the arguments of other group members, compare the argument position to their own attitude position, and adjust their attitude in the direction of the position advocated in the argument. If each group member were to speak with equal frequency, LDM would predict that the final group decision would be equal to the mean of individual group members’ attitudes before discussion. If group members do not speak with equal frequency, the final position reached by the group will be a weighted mean of individual prediscussion attitudes, with frequency of speech being the weighting variable.
Tests of the LDM have provided data consistent with the model (Boster et al., 1980, 1982, 1991). When group members with risky initial attitudes talk more than group members with cautious initial attitudes, a risky shift is observed; the opposite occurs when group members with cautious initial attitudes dominate discussion. For choice decision items that consistently produce a risky (cautious) shift, LDM posits (as does PAT) that there exists more risky (cautious) arguments available and accessible to group members on these topics and that they are expressed during discussion.
The LDM does not propose that extreme members are disproportionately influential unless they speak more frequently than other group members. Research on extreme group members has generally found they are not more influential than more moderate group members (Myers & Murdoch, 1972). Nevertheless, there is some evidence that extreme group members play a disproportionate role in group discussions. Van Swol (2009) found evidence of extreme group members talking more, both in terms of the proportion of total words spoken and the proportion of conversational turns. But, although groups were polarized after discussion, groups with an extreme member did not polarize to a greater extent than groups without extreme members.
Van Swol (2009) suggests that an explanation for why the presence of extreme members does not lead to increased polarization is that the conversational dominance of extreme group members prevents others group members from hearing the positions that might be espoused by all of their fellow group members, as well as from having the opportunity to be polarized by the repetition of their own positions. Brauer, Judd, and Gliner (1995) found that the repeated expression of one’s opinion (and defense of that opinion) in group discussion leads to believing that opinion more strongly. Employing an experimental design that induced independently the number of times subjects repeated their position and the number of times each subject heard another subject’s position (Study 1), they found that the number of times subjects repeated their position was a substantial predictor of group polarization. Thus, informational influence’s role in the group polarization effect might be due not just to group members persuading each other to be more extreme, but group members persuading themselves as well.
Both the normative and informational influence camps can point to a large body of research consistent with their position. In an extensive literature review Isenberg (1986) speculated that both normative and informational influence have independent effects on the choice shift. Other alternatives are possible, of course, such as one form of influence leading to the other, or normative and informational influence combining nonadditively to affect group decisions.
An experiment by Boster and Mayer (1984) was done to test such an alternative possibility. They designed an experiment that varied majority position and argument strength, to test the effects of normative and informational influence, respectively. The majority position induction worked correctly, in that subjects were aware of the position held by a majority of group members. The argument strength induct
ion did not work, however. Instead, the authors’ report that participants perceived the arguments of the majority to be more persuasive than those of the minority. Causal analysis found that the perception of argument strength mediated the relationship between majority position and choice shift. Put differently, normative and informational influences did not operate independently to affect group decisions, but instead were linked in a causal sequence in which the perceived quality of arguments was determined by the number of persons advocating them.
Other studies report evidence of a more sophisticated relationship. Lee (2008) found that in a computer-mediated context, deindividuated subjects failed to distinguish between strong and weak arguments, whereas individuated subjects did. Additionally, individuated subjects were more persuaded by strong arguments than weak arguments, whereas deindividuated subjects were not. Examining individuated and deindividuated subjects separately across two decision tasks, Lee found that for individuated subjects, perceived argument quality, and not group identity, was an important predictor of conformity. Alternatively, for deindividuated subjects both perceived argument quality and group identity were statistically significant predictors of conformity for the first decision task, and only group identity for the second task.
Thus, deindividuation and group identity inductions influenced whether normative or informational cues affected decisions. Similar results were found by Boster and Hale (1989), who tested whether varying the ambiguity of the experimental materials affected whether normative or informational influence would dominate. When the experimental materials were ambiguous, SCT (but not PAT) predicted the choice shift; when the experimental materials were not ambiguous, PAT (but not SCT) predicted the choice shift. These findings point to a more nuanced picture of normative and informational influence in choice shifts, and are consistent with previous research on the ability of contextual features to alter the meaning of a persuasive message (Asch, 1948).
The Influence of Minorities
* * *
A large body of research has emerged on the ability of minority group members to influence majority group members, beginning with the seminal work of Moscovici, Lage, and Naffrechoux (1969). The Moscovici et al. experiment was motivated by the fact that a substantial corpus existed on the processes and effects of majority influence; whereas little research had been done examining whether and how minorities can influence majorities. In the Moscovici et al. (1969) experiment subjects were shown blue slides and in the presence of others had to report what color they saw. Of the six people in the group, four were subjects and two were confederates instructed to answer “green” instead of “blue.” In one condition, the confederates reported the slide as green on all 36 trials; in a second condition, the confederates reported the slide as green on 24 trials and blue on the other 12 trials. In a control condition, subjects answered without hearing the minority group members’ judgments. The consistent minority condition resulted in subjects giving an answer of green 8.42%. The influence of the minority decreased dramatically when it behaved inconsistently, resulting in wrong answers only 1.25% of the time. In the control condition only 0.25% of the subjects judged the slides to be green. Although the size of the effect certainly was not large in comparison to research on majority influence, it was still sizable, and especially so given that correct answer was obvious.
A variety of theories with divergent experimental procedures has emerged to explain the effect. One of the earliest was Moscovici’s (1980, 1985) dual process model, which posits that minorities influence majorities using informational influence, whereas majorities influence minorities via normative influence. More precisely, majorities encourage a comparison process in minorities, whereby minorities compare their position to the majority and focus on the relationships they have (and want to maintain) with the majority. The result is superficial processing that, if it results in a change in position, will produce compliance without internal acceptance. In contrast, minorities encourage the majority to engage in a validation process, whereby majority members try to understand the minority position. This process is thought to entail deeper processing, and thus result in more attitude change (if not public acceptance). Other dual process models (e.g., Nemeth, 1986) that have been proposed focus instead on how the relative positions of each group encourage convergent thinking (minorities focus on the majority position) or divergent thinking (majorities exposed to minority positions focus not just on the minority position, but other related ideas as well). Other attempts at explaining the phenomena include single process models (Latane & Wolf, 1981; Tanford & Penrod, 1984), as well as more general theories (Turner, 1991).
Given the variety of theories proposed on the topic, as well as the different ways in which tests of the various models have been conducted, it is difficult to draw firm, specific conclusions (Wood et al., 1994). Nonetheless, some broad conclusions can be drawn. The first is that although there is evidence that minority influence occurs, and this evidence is consistent with the dual process model (Wood et al., 1994), majority influence is both more powerful and more prevalent.
Second, minority members are more likely to influence majority members if they are consistent, both in terms of comments made throughout the course of the discussion, and in terms of consistency with other minority group members (Wood et al., 1994; for an example, see Meyers, Brashers, & Hanner, 2000). Note that the confederates in Moscovici et al. (1969) experiment met both of these behavioral criteria—in the condition in which the highest rate of influence was obtained, both confederates claimed that the slides were green on all 36 trials. Recent research has uncovered an additional way in which minorities can demonstrate consistency and increase their ability to influence—via the use of abstract language (Sigall, Mucchi-Faina, & Mosso, 2006). Linguists have found that when action is described abstractly, it tends to convey permanence and immutability to situational context, whereas concrete language conveys the opposite. Sigall et al. found that minorities using abstract language versus concrete language in group discussion were viewed as more consistent, and also had more success in exerting indirect influence. The influence of majorities, however, was not affected by whether they used abstract or concrete language.
Other behavioral characteristics of minority group members increase their persuasiveness as well. Persuasiveness of minorities is increased if the minority has agreed with the majority previously (Hollander, 1958), or has convinced majority members to defect (Clark, 1998). Whether the minority is an active versus a passive minority is another contributing factor (Kerr, 2002). An active minority is one that is aware that they are in the minority, is interdependent with others, and expects to interact with group members in the future. Kerr (2002) found that for low-relevance topics, passive minorities were not persuasive, regardless of whether they proffered strong or weak arguments. Active minorities, however, were more persuasive when presenting strong arguments compared to weak arguments, indicating that only active minorities instigate systematic processing among majority group members. Similar to the notion of an active minority, Baron and Bellman (2007) found data consistent with a “courage hypothesis;” minorities harassed by the majority in group discussions were more persuasive on multiple persuasion measures than minorities who were not harassed. These data indicate that minorities who defy majorities despite potential negative social sanctions are more likely to influence.
Another factor that affects the extent to which minority influence is likely to occur is the discussion task. Intellective tasks, or tasks that have demonstrably correct answers, result in more minority influence than judgmental tasks, or tasks without demonstrably correct answers. As Crutchfield (1955) demonstrated, normative influence increases when ambiguity increases, and it is possible that the reduction in normative influence on intellective tasks minority persuasion to occur. Intellective tasks likely also allow for stronger arguments to be made, and given that minority influence occurs primarily through informational influence, intellective tasks present
a more fertile setting for minority influence to take root.
When attitude change does occur as a result of minority influence, investigators have also found that the nature of that attitude change differs from the attitude change observed as a result of majority influence. Martin, Hewstone, and Martin (2003, 2008) found that minority messages employing strong arguments were more resistant to counterpersuasion than majority messages, an effect found to result from the minority message instigating systematic processing. Consequently, the attitudes changed via minority influence were more highly correlated with behavioral intentions than attitudes changed via majority influence, again suggesting minorities instigate systematic processing (Martin, Martin, Smith, & Hewstone, 2007). Even when minority influence is unsuccessful immediately, it can have the effect of reducing the certainty with which majority members hold their opinion, when those majority members recognize that, they rejected the persuasive message because it was delivered by the minority, and when they recognize that this reason for rejecting these arguments is illegitimate. Additionally, those who viewed the reason as illegitimate (versus those who did not) were more likely to be persuaded by a subsequent message (Tormala, DeSensi, & Petty, 2007).
The SAGE Handbook of Persuasion Page 73