In-group variability and motivation to reduce subjective uncertainty
In-group variability and motivation to reduce subjective uncertainty.
Jetten, Jolanda; Hogg, Michael A.; Mullin, Barbara-Ann. Group Dynamics: Theory, Research, and Practice4.2 (Jun 2000): 184-198.
Building on the subjective uncertainty reduction model of social identity processes (M. A. Hogg, in press-b; M. A. Hogg & B.-A. Mullin, 1999), it was hypothesized that homogeneous groups would be better than heterogeneous groups at reducing uncertainty and therefore that people would show more intergroup differentiation and turn more to a homogeneous in-group under conditions of uncertainty rather than certainty. Two minimal-group studies were conducted to investigate this idea. As predicted, when groups were homogeneous, more intergroup differentiation was observed under low certainty than under high certainty (Study 1). Study 2 provided evidence that group members turn actively to their group to reduce low certainty when the group is homogeneous. Low certainty led group members to request more feedback about other group members' responses, and they were more likely to change their initial responses when the in-group was perceived as homogeneous. These findings confirm predictions derived from the uncertainty model proposing that uncertainty reduction is a motivation for group behavior. These findings also highlight the importance of social structural factors such as perceived group variability in the uncertainty reduction process.
The importance of motivations to reduce uncertainty when interacting or communicating with, for instance, strangers (Berger, 1987), in intercultural encounters (Gudykunst & Hammer, 1988), or when establishing a romantic relationship (Parks & Adelman, 1983) is generally acknowledged in interpersonal relations research. It is assumed that processes relating to understanding the other person and understanding the self are directed at reducing uncertainty and that uncertainty reduction increases perceptions of attractiveness of the other person (Berger, 1987; Berger & Calabrese, 1975). More generally, there is a cluster of human motivations to reduce subjective uncertainty, and to meaningfully structure the social field, that social psychologists (stretching at least as far back as Bartlett, 1932) consider to be critical and fundamental to human existence (see Banaji & Prentice, 1994; Baumgardner, 1990; J. D. Campbell, 1990; Kruglanski, 1989; Sedikides & Strube, 1995; Sorrentino, Holmes, Hanna, & Sharp, 1995; Sorrentino & Short, 1986; Stevens & Fiske, 1995; Swann & Schroeder, 1995).
In line with recent theorizing based on group process theories such as social identity theory (Tajfel & Turner, 1986), the present research extends to the intergroup domain the idea that people are motived by uncertainty reduction. Hogg and his colleagues (Hogg, in press-a, in press-b; Hogg & Abrams, 1993; Hogg & Mullin, 1999) recently proposed that uncertainty reduction is a fundamental motivation for self-categorization as a group member and an important drive for individuals to join groups or turn to groups. A number of studies provide support for this uncertainty reduction model of social identity processes (e.g., Grieve & Hogg, 1999; Hogg & Grieve, 1999; Hogg & Mullin, 1998; Mullin & Hogg, 1998, 1999). The aim of the present research was to develop this analysis by focusing on social structural properties of intergroup contexts that may make some groups better than others at satisfying the uncertainty reduction motive. Specifically, when people are uncertain they are more likely to turn to groups that they believe are better equipped to reduce uncertainty. We propose that perceived group variability is critical to this process because group variability consists of information about the extent to which groups can provide clearly focused and consensual prototypes that are best suited to uncertainty reduction through self-categorization and social identification.
Social Identity and Self-Categorization
Social categorization is a fundamental social–cognitive process that structures the world around us in meaningful ways (e.g., Hogg & Abrams, 1988; Medin, 1989; Rosch, 1978; Tajfel, 1981; Tajfel & Wilkes, 1963; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987), and social categorization is the foundation stone of intergroup behavior (e.g., group identification, intergroup discrimination, stereotyping). From social cognition theories, intergroup behavior is explained with reference to the limited cognitive capacity of human beings that causes biased information processing and reliance on categories instead of individual information (Fiske & Neuberg, 1990; Fiske & Taylor, 1984; Macrae, Milne, & Bodenhausen, 1994). While fully acknowledging the critical role of social categorization in structuring the social world and one's place within it, social identity researchers have traditionally placed a greater emphasis on how the process is guided by self-evaluative motives (Tajfel & Turner, 1986). Because evaluative self-definition is contingent on category membership, individuals and groups strive for evaluatively positive social identity and intergroup distinctiveness and thus, ultimately, positive self-esteem. Much traditional social identity research on intergroup relations has mainly focused on the social comparison and enhancement aspects of the theory, with relatively little attention given to the categorization process itself (see Hogg & Mullin, 1999).
This emphasis is rebalanced by self-categorization theory (Turner, 1985; Turner et al., 1987), which can be viewed as an extension of social identity theory. Self-categorization theory elaborates the operation of the categorization process as the generative mechanism for social identification and group behavior. On the basis of considerations such as metacontrast, normative fit, and comparative fit, particular social categorizations become contextually salient (see Oakes, 1987; Oakes, Haslam, & Turner, 1994), so that cognition, perception, affect, and behavior are framed by well-differentiated in-group and out-group prototypes. Self-categorization transforms the self so that it is governed by an in-group prototype that describes and prescribes behavior and thus produces typical group and intergroup behaviors (e.g., stereotyping, cohesion, ethnocentrism, in-group bias, normative behavior). This more cognitive emphasis in self-categorization theory implicitly refocuses attention on motivations associated with the categorization process to processes such as perceptual simplification, structural clarification, maximization of meaning, and uncertainty reduction.
Uncertainty Reduction Model
The uncertainty reduction model (Hogg, in press-a, in press-b; Hogg & Abrams, 1993; Hogg & Mullin, 1999) suggests that the underlying motive for self-categorization is the reduction of subjective uncertainty. This argument grew out of the general observation that social identity research did not lend much support to the role of self-esteem as the fundamental drive for intergroup comparisons (Hogg & Abrams, 1990; Long & Spears, 1997; Rubin & Hewstone, 1998). Hogg (in press-a, in press-b; see also Hogg & Abrams, 1993; Hogg & Mullin, 1999) has argued that one reason for these disappointing results with respect to the self-esteem hypothesis might be that other motivations may operate that are fundamental to social identity processes. According to the uncertainty reduction model, subjective uncertainty reduction is just such an alternative motivation, and it is a motivation that is very well satisfied by the process of self-categorization that generates social identity phenomena. Although group-derived self-esteem is still seen as an important group motivation, it is argued that uncertainty reduction is at least as important and in some contexts may be more important.
Because human survival rests substantially on being able to construct a meaningfully predictable world, uncertainty about one's attitudes, beliefs, feelings, and perceptions, as well as about oneself and other people, is generally an aversive state associated with feelings ranging from unease to fear (e.g., Lopes, 1987). It is predicted from the uncertainty reduction model that people strive to minimize or reduce uncertainty by means of self-categorization and assimilation of the self to the in-group prototype. Category prototypes of self-inclusive categories prescribe behavior, locate us in the social world, regulate our expectations and perceptions of self and others, and furnish consensual validation for who we are, what we believe, and what we do. Uncertainty is produced by contextual factors that challenge people's certainty about their cognitions, attitudes, perceptions, feelings, and behaviors. However, it is predicted that not all uncertainty, but only uncertainty about subjectively important issues that have implications for the self-concept, will motivate group identification. Furthermore, in a given context some groups are more relevant than others for uncertainty reduction. This is largely dependent on the extent to which the prototypical properties of the group are relevant to the dimension of uncertainty (Hogg & Mullin, 1998).
Previous research has already provided evidence that people identify with groups and define themselves as group members to reduce subjective uncertainty. Hogg and Grieve (1999) conducted a traditional minimal-group experiment in which participants were randomly categorized into X group, Y group, or were not categorized. It was assumed that categorization under these normal minimal-group conditions embodied relatively high uncertainty. Hogg and Grieve created low uncertainty by giving practice trials on point allocation matrixes (matrixes were also used to measure intergroup bias). They found that intergroup bias was significant only among participants categorized under uncertainty. Grieve and Hogg (1999, Experiment 1) replicated this finding in a more extensive experiment also containing self-report measures of group identification. In this study, group identification was higher, and intergroup bias occurred only among participants who were categorized under uncertainty. A similar result was obtained in a second experiment (Grieve & Hogg, 1999, Experiment 2). Again, it was found that only participants who were categorized under uncertainty identified with the category and exhibited group behavior. Mullin and Hogg (1998), in another minimal-group experiment, disentangled situational uncertainty from task uncertainty and examined their independent effects on intergroup discrimination. Situational uncertainty was defined as uncertainty arising from the relationship between the self and other people in the social setting. It was manipulated by selecting participants who had no experience as a participant in psychological research (high situational uncertainty) versus participants who had participated at least five times in previous research (low situational uncertainty). As in Hogg and Grieve's research, task uncertainty was manipulated by procedural practice on the allocation task. In the low task certainty condition, participants were given six practice trials and were told to complete as many trials as they needed to make sure they felt completely certain about the task. For both dimensions of uncertainty, significant group identification and in-group bias were found only when participants were categorized under uncertainty. Finally, Hogg and Mullin (1998) reported a set of three minimal-group experiments in which they provided some evidence that category relevance is important. Participants who were categorized under uncertainty about something important identified significantly more strongly with the category if the category was self-conceptually relevant to the dimension of uncertainty.
Together, these studies provide support for the basic tenets of the uncertainty reduction model (see Hogg & Mullin, 1999, for an overview). The last set of studies (Hogg & Mullin, 1998) also suggests that properties of the social category, in this case its relevance to the dimension of uncertainty, may influence whether people self-categorize. It is this idea that we took up in the present research. What types ofgroups are contextually appropriate to reduce uncertainty? Why do we seek out groups, join groups, and self-define ourselves as members of particular groups and not others when we are uncertain?
Group Variability and Uncertainty Reduction
According to social comparison theory, people generally seek out similar other individuals to validate their beliefs (e.g., Festinger, 1954a, 1954b; Kruglanski & Mayseless, 1990; Schachter, 1959; Suls & Wills, 1991). This theory rests on the assumption that people are driven to know that their opinions are correct and to know what they can and cannot do. Only when physical reality checks are not possible do people make social comparisons with others, and then with others who are similar to them on relevant dimensions. Festinger was quite explicit that such comparisons were interpersonal. In his historical commentary, Wheeler (1991) stated that “comparison theory emphasized individuals using others to fulfill their own need to know. ... The shift ... was clearly toward an individualistic focus” (p. 3), and the idea that people construct a group norm against which to compare themselves was not elaborated—the group and intergroup aspect of social comparison was not explored (see Hogg, in press-a). Although self-evaluation, possibly uncertainty reduction, is a motive for interpersonal social comparison, subsequent research on social comparison focused much more on the self-enhancement motive (see Wills & Suls, 1991). Contemporary social comparison research is thus largely about people making interpersonal comparisons for self-enhancement reasons.
The uncertainty reduction model explains how people find certainty and validate their perceptions, attitudes, feelings, and behaviors through identification with an in-group (see also McGarty, Turner, Oakes, & Haslam, 1993; Tajfel, 1978). However, as Hogg and Mullin's (1998) studies indicated, not all in-groups are functionally equivalent in reducing subjective uncertainty, and the ability of a group to reduce uncertainty may be affected by its social structural characteristics. Some attention has been paid to social structural characteristics of the in-group in research conducted by McGarty et al. (1993; Experiments 1 and 2), who investigated uncertainty reduction when the reference group was similar or dissimilar to the self. However, in this research no evidence was found that confidence about responses on a perceptual task was more influenced by feedback from a similar reference group (students) than by feedback from a dissimilar reference group (a group of visually impaired people). The focus of the present research is slightly different in that the effect of group variability information on the effectiveness of turning to a group to reduce uncertainty is examined.
Groups are homogeneous when there is marked similarity among members and heterogeneous when there are substantial individual differences within the group (Simon, 1992; Simon & Brown, 1987). The perception of group variability comprises information about similarity among group members and the “groupness,” or entitativity, of a group (D. T. Campbell, 1958). For instance, research has demonstrated that in-group homogeneity increases the social distinctiveness and social identity of a group and that in-group homogeneity is positively related to in-group favoritism (Hogg, 1992; Simon, 1992; Simon, Kulla, & Zobel, 1995; Vanbeselaere, 1991). In line with previous research, it is predicted that perceived agreement and similarity among in-group members (Wilder, 1981, 1986), degree of in-group normative consensuality (Grieve & Hogg, 1999), perceived group cohesion (Hogg, 1992), perceived groupness or entitativity (D. T. Campbell, 1958; Sherman, Hamilton, & Lewis, 1999; Yzerbyt, Rogier, & Fiske, 1998), and perceived intergroup distinctiveness (Jetten, Spears, & Manstead, 1997, 1998) would be higher in homogeneous groups than in heterogeneous groups.
Homogeneous groups thus provide more distinctive, consensual, and clearly defined prototypes than heterogeneous groups. We predict that homogeneous groups are therefore particularly well suited to uncertainty reduction through self-categorization. Group heterogeneity represents diversity and disagreement among members, there is considerable variation within the group, the prototype of the group is less clearly defined, and it is therefore less clear what the group as a whole stands for (McGarty et al., 1993). Self-categorization–based assimilation of self to such a prototype would do little to resolve uncertainty. It is even possible that extreme heterogeneity undermines feelings of groupness altogether and that the group is better described as an aggregate of individuals. We predicted that people would be more likelyto self-define with and turn to homogeneous groups under conditions of low certainty than under high certainty. Certainty versus uncertainty should not affect heterogeneous groups differently, because group heterogeneity limits the effective and functional reduction of uncertainty.
We examined the interactive effect of uncertainty reduction and perceived group variability in a minimal-group setting. We manipulated group variability by providing participants with a homogeneous or heterogeneous distribution of the in-group. Uncertainty was manipulated to be low certainty about the correctness of social attitudes. We examined our prediction—that attempts to reduce low certainty affect the extent to which group members categorize and identify with their group—by using degree of intergroup differentiation as the main dependent measure. Intergroup differentiation in the present research was used as a measure of group identification and group processes and was measured in terms of the difference between in-group and out-group evaluations on traits stereotypical for these groups. Other important dependent measures were social identification and the degree to which group members turn to their own group to validate their attitudes in order to reduce subjective uncertainty. We predicted that homogeneous groups are better suited to, and more effective at, reducing high versus low subjective uncertainty and that uncertainty would have no effect when the in-group is heterogeneous.
Design and participants
This experiment had a 2 (certainty: low vs. high) × 2 (group variability: homogeneous vs. heterogeneous) factorial design, with random allocation of participants to conditions. Students (36 women and 20 men, distributed evenly across conditions) at a large Australian university participated in partial fulfillment of course requirements.
Participants arrived at the laboratory in groups varying in size from 10 to 20 and sat at individual tables on which answer booklets had been placed. The female experimenter introduced the study as an investigation into “attitudes and perception” and informed participants that the study involved three tasks.
The first task, taken from Mullin and Hogg (1999), had participants indicate how personally important they felt each of 10 “serious and important issues” was to them (e.g., having a group of good friends, being physically fit and healthy). After completing this task, participants in the high-certainty condition were informed that
The next two tasks are related to these important attitudes and beliefs. Please note that there are no objectively right or wrong answers to these questions. There may be some variation in people's attitudes and beliefs regarding such important issues—that is OK. Your own attitudes and beliefs are just as valid as anyone else's.
Participants in the low-certainty condition were told:
The next two tasks are related to these important attitudes and beliefs. Previous research has shown, however, that some attitudes and beliefs are actually more valid than others because they are grounded in factual information. So, some people actually have more correct views on these issues than other people. [ 1 ]
Next, participants completed manipulation checks for the certainty manipulation (see Dependent variables section, below) and continued with the second task, for which they were categorized into two groups on the basis of a visual perception task. Participants were led to believe that previous research had shown that attitudes and perception are related and that it is useful to distinguish between two kinds of perceivers: detailed and global (Jetten, Spears, & Manstead, 1996). To measure their own perceptual style participants were given a “dot estimation task” in which they were presented for 5 s with seven separate overhead sheets filled with dots. They were asked to estimate the number of dots (which ranged from 40 to 100), after which the experimenter asked them to calculate the total scores for these seven estimations. The experimenter inspected each participant's score and selected one of two booklets to hand out. The booklet contained information about the participant's perceptual style. All participants were categorized as detailed perceivers. [ 2 ]
Hereafter participants were informed that previous studies had demonstrated that there are variations in the scores within the group of detailed perceivers. It was explained that it is possible to locate the score of detailed perceivers on a continuum that ranges from moderate in detailed perceiving (0) to very high in detailed perceiving (100). The booklet contained a frequency distribution of scores of detailed perceivers who had participated in a previous study. In all conditions, the in-group mean was 50. The intragroup range (i.e., the difference between the scores of the least and most detailed perceivers) was 30 in the homogeneous in-group condition and 70 in the heterogeneous in-group condition.
All ratings were made on 9-point scales ranging from 1 (not at all) to 9 (very much). Certainty was checked with three items taken from Mullin and Hogg (1999) that measured how confident and certain participants felt about their responses on the attitude–belief scale and how confident they felt as a participant in this experiment (α = .86). Two items checked whether the in-group was seen as more variable in the heterogeneous condition compared to the homogeneous condition (“According to the histogram detailed perceivers are very different from each other in the way they perceive” and “According to the histogram detailed perceivers are very similar in the way they perceive”). We averaged these two items after we recoded the first item (α = .91). Higher means indicate a higher perception of in-group homogeneity.
To assess the extent to which group members turn to their own group for attitude validation, participants were told that they would receive samples of other detailed perceivers' responses to the attitude–belief scale after the final task. Two questions followed: “How much would you like to find out the attitudes/beliefs of the other detailed perceivers in your group?” and “How much would you like to discuss these attitudes/beliefs with the other detailed perceivers in your group?” (α = .93). Social identification was then measured with three items adapted from Jetten et al. (1996): “I identify with detailed perceivers,” “I see myself as part of the group of detailed perceivers,” and “I feel strong ties with the group detailed perceivers” (α = .87).
In-group and out-group evaluations were assessed with semantic differentials. On the basis of a pilot study, traits were selected that were stereotypical for detailed and global perceivers. Participants had to indicate on a 9-point scale where they judged the average in-group member and the average out-group member to be. One endpoint always represented stereotypical traits for detailed perceivers (accurate, profound, rigid, very organized), and the other endpoint represented stereotypical traits for global perceivers (slovenly, superficial, flexible, not at all organized). Analysis of the difference scores (in-group evaluation minus out-group evaluation) revealed that the reliability of the scale was acceptable (α = .78) after deleting the in-group negative and the out-group positive trait “rigid vs. flexible.” The difference scores between in-group and out-group evaluations were averaged, with higher scores indicating higher positive intergroup differentiation.
We conducted 2 (certainty) × 2 (group variability) analyses of variance (ANOVAs) on manipulation checks, which revealed only the expected main effect of certainty on the certainty check, F(1, 52) = 110.70, p < .001. Participants were less certain in the low-certainty condition (M = 5.84) than in the high-certainty condition (M = 7.83). The expected main effect of group variability (and no other effects) was found on the variability check, F(1, 52) = 130.40, p < .001. Participants in the heterogeneous condition perceived the group as less homogeneous (M = 2.64) than participants in the homogeneous condition (M = 7.73). No other effects were significant on these measures.
Attitude validation and group identification
A two-way ANOVA on the attitude validation measure revealed a main effect of certainty, F(1, 52) = 141.47, p < .001. Participants in the low-certainty conditions turned more to their group for attitude validation (M=7.66) than did participants in the high-certainty condition (M = 3.86). The interaction between group variability and certainty was not significant (F < 1). On the identification scale there was only a significant main effect of certainty, F(1, 52) = 15.77, p < .001. Identification was higher for participants in the low-certainty condition (M = 5.71) than in the high-certainty condition (M = 4.27). Moreover, although overall group identification was not significantly different from the midpoint of the scale (M = 4.99, t < 1), identification was significantly higher than the midpoint of the scale for participants in the low-certainty condition, t(27) = 3.91, p < .001, and significantly lower than the midpoint of the scale for those in the high-certainty condition, t(27) = 2.33, p < .05.
A two-way ANOVA on the scale measuring positive intergroup differentiation revealed an interaction between certainty and group variability, F(1, 52) = 6.93, p < .01. As predicted, low-certainty participants whose in-group was homogeneous engaged in greater positive differentiation of the in-group from the out-group than did other participants (see Table 1). Simple main effects analysis revealed, as predicted, that greater differentiation occurred in the homogeneous–low-certainty condition than in the homogeneous–high-certainty condition, F(1, 53) = 8.38, p < .01. There was also a tendency for differentiation to be greater in the homogeneous–low-certainty condition than in the heterogeneous–low-certainty condition, F(1, 53) = 2.73, p = .11. Compared to the homogeneous–high-certainty condition, differentiation was greater in the heterogeneous–high-certainty condition, F(1, 53) = 4.14, p < .05.
Enlarge this Image.
Study 1: Intergroup Differentiation (In-Group–Out-Group Evaluative Difference) on In-Group–Favoring Dimensions as a Function of Certainty and Group Variability
The results of Study 1 replicate the findings of earlier studies that subjective uncertainty increases identification with an available social category and that participants who experience low certainty turn to the in-group to reduce this uncertainty (e.g., Grieve & Hogg, 1999; Hogg & Grieve, 1999; see Hogg & Mullin, 1999). However, on these measures, this effect was not influenced by the extent to which participants perceived their own group as homogeneous or heterogeneous.
Support for the role of group variability in uncertainty reduction was found on the intergroup differentiation measure. When groups were homogeneous, differentiation between the in-group and out-group on dimensions where the in-group compared favorably to the out-group was greater for participants who were uncertain compared to those who were certain. Differences between in-group and out-group evaluations did not differ as a function of uncertainty for participants who had a heterogeneous in-group. It is interesting that homogeneity not only led to higher differentiation under uncertainty but also to lower differentiation in the high-certainty condition (compared to the heterogeneous conditions). This suggests that for participants who already were high in certainty, not only might the group be less important, but also group homogeneity might be perceived as threatening. That is, imposed homogeneity might threaten distinctiveness needs for group members who do not need to rely on the group for uncertainty reduction (Brewer, 1991).
One possible reason why in Study 1 no effects were observed for group variability on the extent to which group members turn to other group members to validate the correctness of their attitudes to resolve uncertainty is that it may have been unclear how the group could actually provide guidance in reducing low certainty. Although the group was more clearly defined when it was homogeneous than when it was heterogeneous, perhaps it was still not clear to participants what the group stood for and how the group might effectively reduce or resolve low certainty. Specifically, it is possible that there were differences in intention to reduce low certainty as a function of group variability and certainty but that no opportunity was available to actually receive information from a homogeneous group under low versus high certainty. We would expect that group variability would become more important if the group actually provided information that could be attended to and could be influential. We felt it important to demonstrate that uncertainty reduction occurs through a process of actively using group-based information and that effects might be more powerful if such an opportunity is given compared to the more passive setting of Study 1. In Study 2 we tested whether people turn to groups to reduce uncertainty, by examining to what extent group members are interested in other group members' responses and to what extent the group exerts social influence on the decision-making process. We also made changes to the nature of the uncertainty manipulation. Decision certainty was manipulated, and participants were given the opportunity to ask for information about other in-group members' decisions.
As in Study 1, we predicted that in-group homogeneity in particular would allow group members to effectively reduce low certainty. As in Study 1, the experimental setting involved minimal groups. The main dependent variables were the extent to which participants were interested in responses of other in-group members and the extent to which feedback about other group members' responses would motivate participants to change their own initial decisions.
Design and participants
The independent variables of certainty (low vs. high) and group variability (homogeneous vs. heterogeneous) were manipulated in a 2 × 2 factorial design. University of Amsterdam students (39 women and 37 men, distributed evenly over conditions) received 10 Dutch guilders (approximately $5 U.S.) for their participation in the study.
Participants were seated in separate cubicles and completed the experimental tasks on personal computers. They received brief instructions on how to work with the computer and then began the experiment, which was introduced as an investigation of the way people make decisions. The experiment consisted of three phases: the certainty manipulation, followed by a categorization phase, and then the group variability manipulation. The critical dependent variable was the extent to which group members were interested in receiving information about decisions made by other in-group members and the extent to which they would like to change their initial decision after having had the chance to view other group members' decisions.
We manipulated certainty by providing participants with five personnel decision cases (adapted from Doosje, Ellemers, & Spears, 1995; Ellemers, Spears, & Doosje, 1997). For each case participants had to decide which of two options was the best solution to the personnel problem. For instance, participants were told:
There is a vacancy at the administration department for a new supervisor. Two candidates have been short-listed for this position. Candidate 1 started 5 years ago as a secretary at the administration department. A few years ago she was promoted and now she is the assistant of the supervisor that is leaving. She knows exactly what is going on in the department, but she has no management experience. Candidate 2 is currently supervisor of a production line. She would prefer a job in the office but does not have experience with the administration department. Which candidate is in your view best qualified for the job, 1 or 2?
Participants in the high-certainty condition were told that they “had to keep in mind that there are no objectively right or wrong decisions and that your decision is just as valid as anyone else's.” Participants in the low-certainty condition were told that they
had to keep in mind that one of the two decisions was clearly more valid than the other one. In other words, one option is more correct than the other option. The decisions differ in their effectiveness to solve the problem.
After participants had given their answers to each of the five cases, the certainty manipulation checks were administered (see Dependent measures section).
As in Study 1, the second part of the experiment was designed to categorize participants. As before, they were told that previous research had shown that there are two kinds of perceivers: detailed and global. They completed the dot estimation task on the computer to measure their own perceptual style. Again, all participants received feedback that they were detailed perceivers. In the third part of the experiment participants were provided with information (by means of histograms) about the distribution of scores of detailed perceivers. Homogeneity and heterogeneity were manipulated in the same way as before. After participants had studied the in-group distribution histogram for 20 s, they were provided with a written interpretation of the results (projected onto the computer screen together with the histogram).
The manipulation checks for certainty and group variability were similar to Study 1. We averaged the three items checking for certainty and the two items checking for group variability (αs = .80 and .88, respectively). Group identification was measured with the same three items as in Study 1 (α = .85).
The main dependent measure consisted of a task in which participants were given the opportunity to access other group members' decisions on each of the five personnel decision cases. First, participants were re-presented with the case and were asked if they were interested in the decision of another group member. If they pressed the “yes” key, the decision of an ostensibly randomly chosen other in-group member was displayed on the screen. This feedback was randomized so that only in half of the requests for other decisions did the other in-group member seem to have made the same decision as the participant. For each case, participants were given the option to ask for more decisions (pressing the “yes” key), up to a maximum of five, or continuing with the next personnel case (pressing the “no” key). The feedback was constructed so that on average 2 or 3 out of 5 other in-group members would respond similarly to the participant. We computed a 6-point scale by counting the number of times participants had requested the decision of other in-group members for each case. This scale ranged from 0 (no decision of other group members required) to 5 (decisions of all group members required). Scores were averaged across the five cases.
After being given the opportunity to request the decisions of other in-group members, participants were asked if they would like to change their initial decisions. If they pressed the “yes” key, they were given the opportunity to type in their new decision. We computed a scale by counting the number of times each participant changed his or her initial decision. This scale ranged from 0 (no initial decisions changed) to 5 (all initial decisions changed). To assess the extent to which participants would turn to an expert to reduce uncertainty, participants were asked to rate the extent to which they would like to know the decision of organizational psychologists for each of the cases (the scale ranged from 1 [not at all] to 9 [very much]). This question was asked twice: right after the manipulations of group variability and certainty as part of the cover story (Time 1), and then after the task in which participants could request the decisions of other in-group members (Time 2).
A two-way ANOVA with the factors of certainty and group variability on the certainty check revealed only a main effect of certainty, F(1, 72) = 6.12, p < .05. Participants in the high-certainty condition were more certain of their decisions (M = 6.92) than participants in the low-certainty condition (M = 6.26). Analysis of the group variability check revealed a main effect of group variability, F(1, 72) = 389.73, p < .001, and no other effects. The group was perceived as more variable in the heterogeneous condition (M = 7.51) than in the homogeneous condition (M = 2.23).
In contrast to Study 1, no significant main or interaction effects were found on group identification. Overall, mean group identification was 3.89, which is significantly lower than the midpoint of the scale, t(75) = 4.96, p < .001.
Request for in-group members' decisions
An ANOVA revealed a marginally significant interaction between certainty and group variability on participants' requests for in-group members' decisions, F(1, 72) = 3.68, p = .06. As predicted, participants requested more in-group members' decisions in the homogeneous–low-certainty condition than in other conditions (see Table 2). Simple main effects analysis revealed that the request for information was higher in the homogeneous–low-certainty condition compared to the homogeneous–high-certainty condition, F(1, 72) = 3.81, p < .05.
Study 2: Requests for In-Group Members' Decisions as a Function of Certainty and Group Variability
Change of initial decisions
Few participants changed their initial decisions at all, let alone more than once. Therefore, we created a dichotomous variable differentiating between those who “did not change any initial decisions” (0) and those who “changed initial decisions” (1). Analysis revealed also a marginally significant interaction between certainty and group variability, F(1, 72) = 2.95, p = .09. There was a tendency for participants to change their initial decisions more in the homogeneous–low-certainty condition and in the heterogeneous–high-certainty condition (see Table 3). Simple main effects analysis revealed no significant differences.
Study 2: Desire to Change One's Initial Decision on the Personnel Decision Task as a Function of Certainty and Group Variability
Finally, we analyzed the extent to which group members wished to know the decisions of organizational psychologists for each of these cases with a three-way ANOVA with time as a repeated measure. Means for Time 1, Time 2, and change over time are displayed in Table 4. This analysis revealed a three-way interaction among group variability, certainty, and time, F(1, 72) = 4.44, p < .05. Simple main effects analysis revealed two significant effects. Participants in the homogeneous–low-certainty condition became more interested in the decisions made by organizational psychologists over time (M = .37) than participants in the homogeneous–high-certainty condition (M = −.50), F(1, 72) = 4.56, p < .05. Furthermore, the interest in these decisions was also higher over time in the homogeneous–low-certainty condition than in the heterogeneous–low-certainty condition (M = .05), F(1, 72) = 4.56, p < .05.
Study 2: Interest in the Decisions Made by Organizational Psychologists as a Function of Group Variability, Certainty, and Time
The focus of Study 2 was the extent to which group members turn to other group members to reduce uncertainty, as a function of subjective uncertainty and group variability. Against a background of successfully manipulated independent variables, the results were generally in line with the intergroup differentiation finding of Study 1. Although there were no significant effects of group identification (see General Discussion section), effects on other measures supported predictions about when participants who were given the possibility to actively reduce their uncertainty would do so. Results revealed that when the group was homogeneous, not only did group members low in certainty request access to more decisions made by other group members than those who were high in certainty, but there was also a tendency for them to be more likely to change their own initial decision and to increase their interest in decisions made by experts (organizational psychologists).
As in Study 1, it should be noted that information search, changing of initial decision, and interest over time in decisions of organizational psychologists also were consistently higher in the heterogeneous–high-certainty condition. As discussed more fully below, it is possible that group members who were already certain found heterogeneity to satisfy a distinctiveness need and thus were curious about decisions made by fellow group members (Brewer, 1991, 1993a, 1993b).
The results of two studies suggest that perception of in-group variability by group members has implications for the reduction of subjective uncertainty. Homogeneous groups were more relevant and effective than heterogeneous groups as foci of social identity-contingent uncertainty reduction. Participants differentiated between the in-group and out-group on stereotypical dimensions more strongly, and thus showed more positive intergroup differentiation, if they were subjectively low in certainty and the in-group was relatively homogeneous (Study 1). In Study 2 we examined the extent to which group members actively turn to other members to reduce low certainty. Consistent with Study 1, participants requested more information from fellow in-group members when they were subjectively low in certainty and the group was homogeneous. In addition, there was evidence that in this condition participants were more influenced by group members to change their initial decisions than in other conditions. Manipulation checks confirmed that the manipulation of uncertainty was successful in both studies. There is of course the possibility that the manipulation may have had other effects. For example, high-uncertainty participants might have also felt insecure and under threat. Indeed, uncertainty is aversive, and this quite probably involves feelings of insecurity and threat. The precise relationship among feeling uncertain, feeling insecure, feeling threatened, and so forth needs to be disentangled in future research.
It is reasonable to infer that people turn to homogeneous groups to reduce uncertainty because their members are prototypically similar, and such groups are more likely to furnish clearly focused, consistent, and consensual information (Hogg & Mullin, 1999; see also McGarty et al., 1993). Heterogeneous groups are less helpful for uncertainty reduction because when coherence and unity within the group are low, there is an attitudinal and behavioral dissensus, and the prototype is less clearly defined, making it difficult to predict how the group might actually resolve one's uncertainty.
In sum, our findings support the notion that uncertainty motivates group behavior and that social structural characteristics of the group such as perceived group variability (homogeneity vs. heterogeneity) are important in this process. However, although our results indicate that these group processes are stronger when groups are homogeneous than when they are heterogeneous, there may be occasions when in-group heterogeneity has an advantage in the uncertainty-reduction process. One such occasion may be when group members do not identify very strongly with the group. Under these conditions, the diversity within heterogeneous groups that is usually considered a disadvantage may become an advantage. With greater diversity comes an increased probability that people will always be able to find at least some members who validate their own behavior. However, decreasing identification is associated with a gradual shift from prototype-based social attraction to interpersonal-relationship–based personal attraction (e.g., Hogg, 1992), and thus interpersonal comparison processes (e.g., Festinger, 1954a; Schachter, 1959) may also be more prevalent when identification is low. Heterogeneous groups may also be appropriate for reducing relatively low levels of uncertainty or uncertainty about less self-conceptually critical matters. Under these circumstances, homogeneous groups might be viewed negatively as being overly controlling.
Another occasion when heterogeneity may be better suited than homogeneity to uncertainty reduction is when uncertainty is related to factual knowledge rather than to matters of opinion or perspective. The diversity of heterogeneous groups increases the probability that there is someone in the group with the correct information or relevant expertise, whereas the uniformity of homogeneous groups increases the consensuality of social support. It might be argued that Study 1 was mainly concerned with judgmental (attitudinal, opinion, and value) uncertainty and Study 2 with intellective uncertainty (i.e., a demonstrably correct solution was possible). When uncertainty is experienced on judgmental dimensions in which expert knowledge is not fundamental to reduce uncertainty, it can be predicted that heterogeneous groups do not provide the best group structure to reduce uncertainty. However, availability of expert knowledge would probably be important and beneficial for participants who were asked to make personnel decisions. The effect found in Study 2 that uncertainty led to more interest in decisions made by organizational psychologists might point to that. [ 3 ]
The present research can also be related to Brewer's (1991, 1993a, 1993b) optimal distinctiveness theory. Brewer has argued that people are motivated by opposing needs for assimilation and for distinctiveness or uniqueness (Snyder & Fromkin, 1980) and that they strive for an optimal balance between satisfaction of these needs. If assimilation needs are overly satisfied people seek distinctiveness, and if distinctiveness needs are overly satisfied they seek assimilation. Homogeneous groups satisfy assimilation needs relatively more than distinctiveness needs, and heterogeneous groups satisfy distinctiveness needs relatively more than assimilation needs (Brewer & Miller, 1996). From an uncertainty perspective, we argue that under conditions of high uncertainty people are drawn to homogeneous groups because prototype clarity and consensuality satisfy uncertainty. However, when uncertainty is low, people have little need to identify with groups at all, or they may identify relatively weakly with heterogeneous groups. One prediction is that people appear to strive for relative distinctiveness or uniqueness when groups are heterogeneous. The possibility exists that whether people strive for assimilation or distinctiveness may be determined by their level of uncertainty motivation. With increasing uncertainty people become increasingly motivated to assimilate and less motivated to be distinct and unique. There is some support for this idea from the high levels of intergroup differentiation in Study 1 and the high decision-request rates and subsequent changes of initial decisions among low-uncertainty participants in heterogeneous groups in Study 2. It is possible that these effects were not so much the result of uncertainty reduction processes as the effect of greater adherence to the group when distinctiveness needs are preserved. As discussed before, homogeneous groups might even represent a threat under low uncertainty because people's need to turn to groups is lower when certainty is high. Group homogeneity might in that case be more threatening to the identity than heterogeneity, because the group can be perceived as overly controlling. Further research is required to investigate in greater detail the interplay between distinctiveness needs resulting from group variability feedback and uncertainty reduction motivations, in particular when group members are relatively certain.
It was unexpected that no effects for group variability or uncertainty were found on the group identification measure in Study 2. This is not only inconsistent with previous uncertainty research (e.g., Grieve & Hogg, 1999; Mullin & Hogg, 1998, 1999), but it also weakens the key argument that group members categorize themselves more as group members and identify more strongly with the group when they are uncertain. However, as described in the Results section of Study 2, overall identification with the detailed-perceivers group was weak and even significantly below the midpoint of the scale. It might be that mentioning the availability of experts (organizational psychologists) in the context of an intellective task (see discussion above) made the experts a contextually more relevant comparison group than the in-group consisting of other students and that this thus depressed in-group identification. It was probably less clear to group members how other detailed perceivers could reduce their task-specific uncertainty than experts in this field. This argument is quite consistent with the notion that people who are uncertain are more likely to identify with groups that are more relevant rather than less relevant to uncertainty reduction (Hogg & Mullin, 1998; McGarty et al., 1993).
Our studies confirm basic tenets of the uncertainty reduction model of social identity processes and provide evidence for the importance of structural characteristics of the group that might influence the extent to which group members turn to their group when uncertain. The results of both studies show that group members used information about in-group variability when they searched for a group to help them reduce uncertainty. Generally, when certainty was low rather than high, people turned more to homogeneous groups, presumably because homogeneity ensures a clear and consensual prototype that is effective at reducing uncertainty. Group variability is only one structural feature of groups that may influence uncertainty reduction through group identification. Other characteristics of groups, such as their size, status, power, or distinctiveness from other groups, might also be taken into account when group members judge the suitability of groups to reduce uncertainty. We might predict that minority groups, high-status groups, powerful groups, and groups that are clearly distinctive from other groups would be considered most capable of reducing or resolving uncertainty. These are promising directions for future research on subjective uncertainty reduction as a motivation for social identity processes.
Banaji, M. R., & Prentice, D. A. (1994). Annual Review of Psychology.
Bartlett, F. C. (1932). Remembering. Cambridge, England: Cambridge University Press.
Baumgardner, A. H. (1990). Journal of Personality and Social Psychology.
Berger, C. R., Rolotte, M. G., & Miller, G. R. (Eds.) (1987). Interpersonal processes: New directions in communication research. Newbury Park, CA: Sage.
Berger, C. R., & Calabrese, R. J. (1975). Human Communication Research.
Brewer, M. B. (1991). Personality and Social Psychology Bulletin.
Brewer, M. B., Hogg, M. A., & Abrams, D. (Eds.) (1993a). Group motivation: Social psychological perspectives. London: Harvester Wheatsheaf.
Brewer, M. B. (1993b). Social Cognition.
Brewer, M. B., & Miller, N. (1996). Intergroup relations. Buckingham, England: Open University Press.Campbell, D. T. (1958). Behavioral Sciences.
Hogg, M. A. (1992). The social psychology of group cohesiveness: From attraction to social identity. Hemel Hempstead, England: Harvester Wheatsheaf.
Hogg, M. A., Suls, J., & Wheeler, L. (Eds.) (). Handbook of social comparison: Theory and research. New York: Plenum.
Hogg, M. A. (). European Review of Social Psychology.
Hogg, M. A., & Abrams, D. (1988). Social identifications: A social psychology of intergroup relations and group processes. New York: Routledge.
Hogg, M. A., Abrams, D., Abrams, D., & Hogg, M. A. (Eds.) (1990). Social identity theory: Constructive and critical advances. London: Harvester Wheatsheaf.
Hogg, M. A., Abrams, D., Hogg, M. A., & Abrams, D. (Eds.) (1993). Group motivation. Hemel Hempstead, England: Harvester Wheatsheaf.
Hogg, M. A., & Grieve, P. (1999). Asian Journal of Social Psychology.
Hogg, M. A., & Mullin, B.-A. (1998). Reducing subjective uncertainty by group identification: The role of group relevance.
Hogg, M. A., Mullin, B.-A., Abrams, D., & Hogg, M. A. (Eds.) (1999). Social identity and social cognition. Oxford, England: Blackwell.
Tajfel, H., & Tajfel, H. (Ed.) (1978). Differentiation between social groups. London: Academic Press.
Tajfel, H., Turner, J. C., & Giles, H. (Eds.) (1981). Intergroup behaviour. Oxford, England: Blackwell.
Tajfel, H., Turner, J. C., Worchel, S., & Austin, W. G. (Eds.) (1986). Psychology of intergroup relations. Chicago: Nelson Hall.
Tajfel, H., & Wilkes, A. L. (1963). British Journal of Psychology.