Scholarly article on topic 'Measuring the Measures: A Meta-Analytic Investigation of the Measures of Outgroup Homogeneity'

Measuring the Measures: A Meta-Analytic Investigation of the Measures of Outgroup Homogeneity Academic research paper on "Psychology"

0
0
Share paper
OECD Field of science
Keywords
{""}

Academic research paper on topic "Measuring the Measures: A Meta-Analytic Investigation of the Measures of Outgroup Homogeneity"

Measuring the Measures: A Meta-Analytic Investigation of the Measures of Outgroup Homogeneity

Group Processes & -p* Intergroup Relations 1

2007 Vol 10(2) 157-178 j

Jennifer G. Boldry

Montana State University

Lowell Gaertner

University of Tennessee

Jeff Quinn

Duke University

We meta-analytically synthesized the intergroup variability literature (177 effect sizes, from 173 independent samples, and 12,078 participants) to test the potential moderating effect of 11 measures of perceived variability. Aggregating across the measures, we detected a small but reliable tendency to perceive more variability among ingroup than outgroup members and such outgroup homogeneity was stronger among non-minimal than minimal groups. Furthermore, analyses that distinguished among the 11 measures revealed systematic discrepancies among the patterns of perception detected by those measures. Those systematic discrepancies further varied across social contexts defined by relative group status, with some measures yielding ingroup homogeneity and others outgroup homogeneity. We discuss the possibility that the measures of variability require different mental activities that interact with contextually induced cognitive and motivational processes to yield disparate intergroup perceptions.

keywords group perception, group variability, heterogeneity, ingroup homogeneity, intergroup perception, intergroup status, meta-analysis, outgroup homogeneity, relative group status

Reflecting the stereotype that members of other groups 'are all alike', research on intergroup perception initially identified a tendency for persons to perceive more variability among ingroup members than among outgroup members (e.g. Judd & Park, 1988; Linville, Fischer, & Salovey, 1989; Park & Rothbart, 1982). However, this outgroup homogeneity effect is not always

the rule. Sometimes persons perceive less variability among ingroup than outgroup members

Author's note

Address correspondence to Jennifer G. Boldry, National Analysts Inc., 1835 Market Street, Philadelphia PA, 19103, USA [email: jennifer_boldry@yahoo.com].

Copyright © 2007 SAGE Publications (Los Angeles, London, New Delhi and Singapore) 10:2; 157-178; DOI: 10.1177/1368430207075153

(i.e. ingroup homogeneity; e.g. Simon, 1992; Simon & Brown, 1987; Simon & Mummendey, 1990) and other times they perceive the ingroup and outgroup as equally variable (e.g. Boldry & Kashy, 1999; Linville, et al., 1989). In light of these flip-flopping patterns, research questions have evolved to a focus on socio-structural factors that moderate perceptions of intergroup variability (e.g. Mullen & Hu, 1989; Ostrom & Sedikides, 1992).

In the current research, we follow the evolving path toward the identification of socio-structural moderators, but do so as a backdrop for the exploration of a methodological moderator that has received surprisingly little attention: the measurement of perceived variability. A count of the intergroup relations literature reveals at least 11 measures of perceived variability. While methodological diversity is desirable (e.g. Campbell & Fiske, 1959), such diversity can accrue a cost in that nuances among measures can contribute to variation, even inconsistency, in a research literature. Park and Rothbart (1982), for example, assessed perceived variability between gender groups with a percent estimate measure and detected perceptions of outgroup homogeneity, whereas Linville et al. (1989) employed a probability of differentiation measure and detected equivalent perceptions of variability among gender groups. To assess the possibility of systematic discrepancies among measures, we empirically examined the intergroup variability literature with a meta-analytic technique. As we subsequently explain, we go beyond the basic question of an omnibus difference among measures and explore the more interesting possibility that those differences vary systematically with socio-structural factors.

Measures of perceived variability

The 11 measures of perceived variability are derived from six distinct tasks required of participants. We briefly describe those tasks and their derived measures.

Distribution task

This task requires participants to construct an imagined histogram of the perceived distribution

of group members across a trait dimension. Participants, for example, indicate the frequency of 100 randomly selected group members (or the percentage of group members) described by each interval of a Likert-type scale for a given trait (e.g.Judd & Park, 1988; Linville et al., 1989; Park &Judd, 1990). Similarly, the visually oriented dot distribution task requires participants to place dots of varying size along the intervals of a Likert-type scale to indicate the perceived frequency of group members described by a given interval (e.g. Park & Judd, 1990). From those distribution tasks researchers derive measures of standard deviation ( SD; or variance) and probability of differentiation (Pd; Linville et al., 1989). SD (and variance) assesses the average distance (or squared distance) of group members from the mean. Pd assesses the probability of distinguishing between two group members. Lower scores on those measures reflect perceptions of lesser variability.

Memory task

For this task, participants review information about group members (e.g. traits, behaviors, faces) and, subsequently, attempt to match the information to the particular member with whom the information was initially associated (e.g. 'who said what?'). Summing the number of within-group errors (i.e. the number of times information associated with one member is incorrectly attributed to another member of the same group) provides an assessment of perceived variability, with more errors reflecting a perception of lesser variability.

The memory task comes in two varieties: recognition and recall. Recognition tasks represent the previously presented information and participants must match that information to the corresponding group member; such tasks provide a measure of recognition error (e.g. Cabecinhas & Amancio, 1999; Howard & Rothbart, 1980; Lorenzi-Cioldi, 1998). Recall tasks do not re-present information and participants must recall the previously presented information and match it to the corresponding group member; such tasks provide a measure of recall error (e.g. Ostrom, Carpenter, Sedikides, & Li, 1993; Sedikides, 1997; Stewart & Vassar, 2000).

A sizable literature utilizes a face-recognition task, which is a subset of the recognition task, in which participants identify from an array of faces those faces that were previously presented; such a task provides a measure of face-recognition error (Brigham & Barkowitz, 1978; Chance, Goldstein, & McBride, 1975; Feinman & Entwisle, 1976). To the extent that recall and recognition involve different memory processes (e.g. Garcia-Marques & Hamilton, 1996), recall and recognition measures might be better conceptualized as being derived from disparate tasks.

Range task

This task requires participants to identify the two points along a Likert-type scale or semantic-differential scale that best describe the two most extreme group members. The absolute difference between those extreme scores provides the range measure, with a smaller range reflecting a perception of lesser variability (e.g. Jones, Wood, & Quattrone, 1981; Simon & Brown, 1987). This task is somewhat similar to the distribution task in that participants ostensibly consider the distribution of group members across a trait dimension. The range task, however, requires participants to consider only the two most extreme members.

Similarity task

This task requires participants to rate on a Likert-type scale the extent to which they perceive group members as being similar (or dissimilar). Anchoring thejudgment to a particular attribute (e.g. 'how similar are men in regard to the trait nurturing?') provides a measure of trait similarity (e.g. Boldry & Gaertner, 2006; Park &Judd, 1990). Freeing the judgment from a particular attribute (e.g. 'how similar are men?') provides a measure of global similarity (Park & Judd, 1990; Park & Rothbart, 1982; Quattrone & Jones, 1980). Higher values of similarity reflect perceptions of lesser variability.

Stereotyping task

This task requires participants to consider the members of a group in regard to stereotypic attributes and counterstereotypic attributes. The percent estimate measure requires participants to

estimate the percentage of group members who possess stereotypic traits and counterstereotypic traits, and the latter estimate is subtracted from the former (e.g. Pickett & Brewer, 2001; Ryan & Judd, 1992; Ryan, Judd, & Park, 1996). The mean measure requires participants to rate on Likert-type scales the degree to which stereotypic traits and counterstereotypic traits describe the group and the latter rating is subtracted from the former (e.g. Park & Judd, 1990; Ryan, Judd, & Park, 1996). These measures assess the extent to which group members are perceived as homogeneously conforming to the group's stereotype, with larger values reflecting perceptions of lesser variability.

Subgroup generation task

This task requires participants to list subgroups that potentially exist within a group (e.g. African-Americans and European-Americans might be subgroups of Americans; e.g. Linville, Fischer, & Yoon, 1996; Park, Ryan, & Judd, 1992). The number of listed subgroups provides the subgroup generation measure, with fewer subgroups reflecting a perception of lesser variability.

Measure as mechanism

Despite the number and variety of measures, a modicum of research has empirically compared the measures. Obviously, variation can arise across measures due to issues of reliability and validity. Assessment of relative reliability and validity requires the inclusion of multiple measures in a given study and an appropriate analysis of relative reliability/validity. To date, only one published study meets such a requirement. Park and Judd (1990) employed a structural equation model that explored patterns of association among 6 of the 11 measures. The data consisted of male and female perceptions of gender groups (i.e. men and women). Results suggested that perceived variability comprises two dimensions: dispersion and stereotypicality. Dispersion reflects the extent to which group members are perceived to be dispersed around their group's mean position on a trait dimension and was indexed uniquely by the measures of range, SD, and Pd. Stereotypicality reflects the

extent to which group members are perceived to uniformly fit the group's stereotype and was indexed uniquely by the measures of percent estimate and mean. A measure of global similarity indexed both dimensions.

Park and Judd's (1990) structural model is an important step toward an examination of the multiple measures. Nonetheless, the analysis has its limitations. Time and attention span prohibit the inclusion of numerous measures in a research project and Park and Judd did not include the measures of subgroup generation, recall, recognition, and face recognition. Consequently, it is not clear how those measures align with the included measures. Further, because several measures are derived from common tasks (e.g. standard deviation and probability of differentiation are both derived from a distribution task), it is plausible that characteristics of the task contribute to variation in perception. Such shared variation might be more than a methodological nuisance and might be meaningful conceptually. The mental processes required of participants by a particular task might actively contribute to the perception of intergroup variability.

Although links between mental processes and particular tasks (or measures) are admittedly speculative, hints of such links can be glimpsed in theoretical accounts of the mental activities regulating perceptions of variability. Exemplar-based accounts, for example, suggest that perceived variability fluctuates as a function of the number and diversity of the particular group members (i.e. exemplars) brought to mind at the point ofjudgment (e.g. Linville, Salovey, & Fischer, 1986; Linville et al., 1989). Abstraction-based accounts, on the other hand, suggest that perceived variability fluctuates both as a function of the particular exemplars that come to mind and stored abstractions about the group as a whole (e.g. Park & Judd, 1990; Park, Judd, & Ryan, 1991). Interestingly, research on those theoretical accounts often employs different tasks (i.e.measures). Research consistent with the exemplar-based account often employs Pd or SD measures derived from the distribution task, which requires participants to consider individual

group members. Research consistent with the abstraction-based account often employs the percent estimate and means measures derived from the stereotyping task, which requires participants to consider group-level information in the form of a stereotype. Consequently, in the current research we examine discrepancies at the level of both measure and task.

However, a less obvious and more insidious limitation of Park andJudd's (1990) work, which is endemic to any single study, is that the associations among measures and their relative patterns occur within a particular constellation of socio-structural factors. In the case of Park and Judd, the obtained data are situated in the dynamics of male-female relations, in which the gender-based groups differ in status (e.g. Eagly, Wood, & Diekman, 2000). To the extent to which such dynamics fluctuate across intergroup relations, so might the obtained pattern of data. Of course, researchers methodically pursue such moderating variables within and across studies. However, the non-obvious difficulty arises when different dependent variables are included across studies. To the extent to which measures of perceived variability are differentially sensitive to different moderators, substantial heterogeneity (and apparent inconsistencies) accumulates in a research literature. The issue of differential sensitivity can be conceptualized statistically as an interaction among measure and moderator.

An interaction among measures and socio-structural moderators of perceived intergroup variability suggests that heterogeneous patterns across measures are not simply a function of the differential psychometrics of the measures (i.e. validity and reliability), and, instead, enables the interesting possibility that the measures actively contribute in conjunction with particular socio-structural moderators to perception. In particular, different tasks (and measures) plausibly require different mental activities. For example, some measures, such as similarity, require participants to make judgments about the group as a whole, whereas the range focuses attention on individual group members. Likewise, different socio-structural moderators differentially elicit social processes. Membership in

low status groups, for example, activates social identity concerns that motivate positively distinct perceptions of the ingroup relative to the outgroup (Boldry & Gaertner, 2006; Ellemers, Van Rijswijk, Roefs, & Simons, 1997; Tajfel & Turner, 1979) to which tasks/measures that focus attention at the group (vs. individual) level might be acutely sensitive. Stated otherwise, systematic patterns of perceptions might arise when the particular mental activity required by a measure synchs with the particular process elicited by a socio-structural moderator.

Any single study cannot sufficiently address the issue of differential sensitivity due, in part, to the previously discussed limitations of time and attention span and the inherent complexity of a design that manipulates multiple moderators. A meta-analysis, on the other hand, capitalizes on the existing data collected across numerous studies that vary in regard to moderators and included measures. Consequently, a meta-analysis represents a desirable tool for assessing the multiple measures of perceived variability and we employ such a tool in the current research. We have three objectives for the current metaanalysis. First, we meta-analytically assess a possible omnibus difference among the 11 measures of perceived intergroup variability. Second, we meta-analytically assess the potential moderating effects of socio-structural factors identified in previous literature reviews or empirical studies. Third, and most importantly, we meta-analytically assess potential interactions between the measures of perceived variability and the socio-structural moderators to determine whether the measures evidence differential sensitivity to patterns of intergroup perception produced by the potential socio-structural moderators. When exploring the moderating effect of measurement, we examine potential differences at the levels of both measure and task.

Socio-structural moderators of perceived intergroup variability

We consider two socio-structural variables that: (a) occurred with sufficient regularity in the intergroup variability literature and, (b) co-occurred with a sufficient subset of the measures

of variability as to make a meta-analysis feasible. Those moderators are relative group status and minimal vs. non-minimal groupings.1

Relative group status

Literature reviews (Rubin, Hewstone, Crisp, Voci, & Richards, 2002; Sedikides & Ostrom, 1993; Voci, 2000) consistently identify relative group status as a moderator of perceived group variability. With few exceptions (e.g. Brigham, Maas, Snyder, & Spaulding, 1982; Judd, Ryan, & Park, 1991), the majority of empirical studies indicate that members of high status groups perceive outgroup homogeneity (e.g., Boldry & Kashy, 1999; Lorenzi-Cioldi, 1998; Sedikides, 1997). However, results of those studies are heterogeneous in regard to members oflow status groups. Some studies suggest that members of low status groups perceive ingroup homogeneity (e.g. Brown & Smith, 1989; Cabecinhas & Amancio, 1999; Lorenzi-Cioldi, Deaux, & Dafflon, 1998) and others suggest that they perceive equivalent variability among ingroup and outgroup members (Boldry & Kashy, 1999; Hewstone, Islam, & Judd, 1993; Lorenzi-Cioldi, 1998). Such discrepancies could be attributed to the particular measures employed. Among members of low status groups, perceptions of ingroup homogeneity have been detected by recall error (Lorenzi-Cioldi, 1998) and range (Brown & Smith, 1989) and perceptions of equivalent intergroup variability have been detected by recognition error (Cabecinhas & Amancio, 1999) and similarity (Hewstone et al., 1993).

Minimal vs. non-minimal gpoup

Some studies fostered novel groups in the laboratory using a variant of the minimal group paradigm (e.g. Howard & Rothbart, 1980; Lorenzi-Cioldi, 1998; Simon & Brown, 1987). Other studies investigated pre-existing or nonminimal groups, such as those based on gender (e.g. Bardach & Park, 1996; Lorenzi-Cioldi, Eagly, & Stewart, 1995), ethnicity (e.g. Cabecinhas & Amancio, 1999), religion (e.g. Hewstone et al., 1993), and college class (e.g. freshman, sophomore; Boldry & Kashy, 1999; Lorenzi-Cioldi et al., 1998). A meta-analysis (Mullen & Hu, 1989) and literature

review (Ostrom & Sedikides, 1992) suggest that perceptions of outgroup homogeneity are stronger among non-minimal than minimal groups. Consistent with the potential for interactions among measures and moderators, the literature review suggested that the range is more reliable among non minimal groups than is SD or Pd.

Method

Literature searches

We conducted computer searches on PsychInfo from 1887 to January 2002, Sociological Abstracts from 1963 to January 2002, and Digital Dissertations from 1861 to January 2002, using as key words: homogeneity, heterogeneity, variability, intergroup perception, outgroup (and, in all searches, the hyphenated out-group) homogeneity, outgroup heterogeneity, outgroup variability, ingroup (and, in all searches, the hyphenated in-group) homogeneity, ingroup heterogeneity, ingroup variability, and face recognition. Additionally, we searched PsychInfo with the names of researchers active in the area of group variability: B.A. Bettencourt, R.Y. Bourhis, J.C. Brigham, R. Brown, S. Carpenter, J.E. Chance, B. Doosje, N. Ellemers, G.W. Fischer, A.G. Goldstein, A. Guinote, S.A. Haslam, M.A. Hogg, J. Jetten, C.M. Judd, P.W. Linville, F. Lorenzi-Cioldi, D.M. Mackie, R.S. Malpass, P. Oakes, T. Ostrom, B.P. Park, M. Rothbart, C.S. Ryan, I. Sachdev, P. Salovey, C. Sedikides, J. Shepard, B. Simon, R. Spears, S.E. Taylor, A. van Knippenberg, and D.A. Wilder. To avoid the exclusion of recently published studies not yet added to electronic databases we searched psychology journals from 1998 to 2002 that routinely publish research on group variability: Journalof Personality and Social Psychology, Journalof Applied Social Psychology, Journal ofSocial Psychology, Journal of Experimental Social Psychology, Personality and Social Psychology Bulletin, Personality and Social Psychology Review, Psychology Review, and Psychology Bulletin. We searched the reference sections of all obtained articles to identify potential studies that our previous searches did not identify. Finally, we posted email requests via the list-serves for the Society for Personality and Social

Psychology and the Society for Experimental Social Psychology.

Criteria for inclusion

We included studies that contrasted the perceived variability of the ingroup with that of the outgroup using any of the 11 measures of group variability (e.g. range, Pd, SD/ variance, percent estimate, mean, global-similarity, trait-similarity, recognition error, face-recognition error, recall error, subgroup generation). Because we were interested in perceived variability of the ingroup relative to the outgroup we necessarily excluded studies that: (a) assessed the perceived variability of only the ingroup or only the outgroup (e.g. Kashima & Kashima, 1993; Park & Hastie, 1987), or (b) assessed perception of the ingroup vs. outgroup between-subjects (e.g. Bardach & Park, 1996; Linville et al., 1996).2 We also excluded computer simulated data (e.g. Fiedler, Kemmelmeier, & Freytag, 1999; Linville, et al., Study 5, 1989). Finally, 50 studies were necessarily excluded because the published reports did not provide sufficient information to compute an effect size and the authors could not provide us with the necessary information (e.g. Lee, 1993; Rothgerber, 1997).

Coding scheme

We coded studies in regard to type of measure, relative group status, and minimal vs. nonminimal group. We coded the relative status of a group either on the basis of a manipulation in the primary study or on the basis of research that attests to perceived status differences among social groups. For example, several studies investigated intergroup perceptions among members of White vs. Black ethnic groups or male vs. female gender groups. Whites are recognized (by Blacks and Whites) as having higher social status than Blacks in the United States (Sidanius, Levin, & Pratto, 1996; Sidanius, Liu, Shaw, & Pratto, 1994; Sidanius, Pratto, & Rabinowitz, 1994) and men are recognized (by men and women) as having higher status than women (e.g. Carli & Eagly, 1999; Eagly et al., 2000). In addition, we coded groups as having 'unknown' status in instances in which the relative status of the groups could not be

determined. It is important to note that groups included in this catch-all category of unknown may indeed have a value for relative group status. For example, group members, whose group's status is not manipulated or lacks clear social consensus, may perceive their group to be of higher status. In the absence of an explicit manipulation or clear social consensus, however, such perceived status remains unknown to an external judge. Finally, there were no instances in which the groups were defined explicitly as having equal status.

Two independent raters coded each study. We assessed interrater reliability of the coded moderators with Cohen's kappa. Kappas ranged from .98 to 1.00, indicating strong interrater reliability. We assessed interrater reliability for the calculated effect sizes with Pearson correlations. Correlations ranged from .99 to 1.00, indicating high reliability calculating the effect sizes. The few discrepancies were discussed and resolved prior to analysis.

Calculation of effect sizes

All included effects were derived from within-subjects designs, consequently we calculated effect sizes as the mean perception of the ingroup minus the mean perception of the outgroup divided by the standard deviation of that difference score (Rosenthal, 1994). Because dependent measures were recoded such that higher numbers reflect less variability (i.e. more homogeneous), a positive effect size indicates that the ingroup was perceived to be less variable than the outgroup (i.e. ingroup homogeneity), a negative effect size indicates that the outgroup was perceived to be less variable than the ingroup (i.e. outgroup homogeneity), and an effect size of zero indicates perceptions of equivalent ingroup vs. outgroup variability. We transformed all effect sizes from Glass's g to Cohen's d to correct for the over-estimation of population effect sizes typically present in Cohen's g estimates (Hedges, 1981).

Analytic issues

When approaching a meta-analysis, the researcher has a choice between two statistical models: random-effects or fixed-effects (Field,

2001; Hedges & Vevea, 1998; Hunter & Schmidt, 2000). Fixed-effects models can be more powerful when their homogeneity assumption (i.e. all effect sizes estimate a common population effect) is satisfied. However, when the latter assumption is not satisfied, fixed-effect models underestimate standard errors of parameter estimates and inflate the Type I error rate (i.e. underestimate confident intervals). Monte Carlo simulations, for example, suggest that the Type I error rate in heterogeneous fixed-effects models ranges between .43 and .80, which is dramatically higher than the nominal .05 level (Field, 2003). In all of the analyses, we initially tested the homogeneity assumption of the fixed-effects model (which is equivalent to the test of random-effects variance). The homogeneity assumption was violated in all but two instances (as subsequently noted) and we employed the more appropriate random-effects models. To determine whether aggregated effect sizes differed between moderator categories, we used the x2 distributed QB statistic.

Thirty of the 173 independent samples (17%) included more than one of the 11 dependent measures. Because those 30 samples contributed more than one effect size, we constructed two data sets to deal with the introduction of non-independence. One data set aggregated across the single or multiple measures of a given sample so that each sample contributed a single effect size. We used the latter data set for analyses that ignored potential discrepancies across measures such as analyses that examined the average ingroup vs. outgroup difference, the average moderating effect of minimal vs. nonminimal group, and the average moderating effect of relative group status. The second data set consisted of effect sizes computed from each dependent measure included in a given sample so that we could compare potential differences among the multiple measures. Consequently, the 30 samples that included more than one measure contributed multiple effect sizes and introduced a small degree of non-independence. We used this data set for analyses that examined the average difference among measures and interactions between the measures and the socio-structural moderators.

Results

Average ingpoup vs. outgpoup perception

Analysis of the average perception of the variability of the ingroup vs. outgroup was based on 177 effect sizes derived from 173 independent samples of 12,078 participants.3 Consistent with Mullen and Hu's (1989) meta-analysis, there was a small but reliable outgroup homogeneity effect (d = -.20, 95% confidence interval (CI) -.26 / -.13). Averaging across the 11 measures and socio-structural moderators, persons perceived less variability among outgroup than ingroup members.

Discrepancies among measures and tasks

When examining potential discrepancies among the measures of intergroup variability, we conducted analyses at the level of the 11 measures and at the level of the 6 tasks from which the measures were derived. Although those levels are somewhat redundant, they are both informative. The level of the measure is informative because researchers typically interpret results derived from particular measures. The level

of the task is informative for two reasons. The task level aggregates across measures derived from a common task and such aggregates are more reliable because they are based on a larger number of effect sizes. Additionally, the mental processes required by a particular task could contribute to perceptions of intergroup variability.

Measures Comparison of the effect sizes for the 11 measures indicated that perceived intergroup variability fluctuated across measures (QB(10) = 31.38, p < .01). Table 1 displays the estimated effect size and the number of effects upon which the estimate was based for each measure. Reliable outgroup homogeneity was detected by percent estimate (d = -.31, 95%CI -.57/-.05), and face recognition (d = -.37, 95% CI -.49/-.26). The remaining measures did not detect differences in the perceived variability of the ingroup vs. outgroup.

Tasks As indicated in Table 1, the range task consists only of the range measure, the distribution task aggregates across Pd and SD (variance),

Table 1. Intergroup variability as a function of task and measure

Task Measure ka d 95% CI QB Powerc

Range Range 31 -.14 -.30/.02 - -

Distribution 10 .12 -.17/.40 .00 .49

Pd 5 .13 -.26/.52

SD/variance 5 .10 -.32/.51

Stereotyping 13 -.32* -.56/-.08 .02 .57

Percent estimate 11 -.31* -.57/-.05

Mean 2 -.38 -1.00/.24

Similarity 26 .12 -.05/.29 .78 .81

Global 21 .16 -.03/.34

Trait 5 -.04 -.41/.34

Subgroup Subgroup 5 -.21 -.61/.19 - -

Memory 120 -.25* -.33/-.16 10.30* 1.00

Face recognition 64 -.37* -.49/-.26

Recall errors 13 -.14 -.38/.10

Recognition errors 43 -.11 -.24/.03

*p<.05.

Note. Perceived variability varied across the 11 measures (QB(10) = 31.38, p < .01) and the 6 tasks (QB(5) = 20.32, p < .01). aIndicates the number of effect sizes.

b QB tests for differences across measures within a given task, with d/equal to 1 less than the number of measures (e.g. df = 1 for the distribution task, which consists of Pd and SD). cIndicates the estimated statistical power of the corresponding QB test. CI = confidence interval.

the stereotyping task aggregates across percent estimate and mean, the similarity task aggregates across global similarity and trait similarity, the subgroup task consists only of the subgroup measure, and the memory task aggregates across face recognition, recall errors, and recognition errors. Comparison of the effect sizes for the six tasks indicated that perceived intergroup variability fluctuated across the tasks (QB(5) = 20.32, p < .01). Reliable outgroup homogeneity was detected by stereotyping (d = -.32, 95%CI -.56/-.08), and memory (d = -.25, 95%CI -.33/-.16). The remaining tasks did not detect differences in the perceived variability of the ingroup vs. outgroup.

Comparisons among measures within a given task indicated, with one exception, that measures derived from a common task did not differ. For the distribution task, the estimated effect sizes for Pd and SD (variance) did not differ (QB(1) = .00, ns). For the stereotyping task, the estimated effect sizes for percent estimate and mean did not differ (QB(1) = .02, ns). For the similarity task, the estimated effect sizes for global and trait similarity did not differ (QB(1) = .78, ns). On the other hand, differences were detected among the three measures from the memory task (QB(2) = 10.30, p < .05). Among recall errors, recognition errors, and face recognition, only the latter two measures differed, such that the outgroup homogeneity effect for face recognition was stronger than the null effect for recognition errors (d = -.27, 95%CI -.46/-.07).

Minimal vs. non-minimal groups

Consistent with Mullen and Hu's (1989) meta-analysis, the minimal vs. non-minimal distinction moderated perceptions of intergroup variability (QB(1) = 4.24, p < .05). Non-minimal groups evidenced a reliable outgroup homogeneity effect (d = -.23, 95%CI -.30/-.16). Minimal groups, however, did not evidence a reliable difference between the perceived variability of the ingroup vs. outgroup (d = -.04, 95%CI -.20/.12).

We subsequently tested whether measure or task interacted with minimal vs. non-minimal distinction. Because the included minimal-group studies did not assess percent estimate, mean, subgroup generation, or face recognition, and the included non-minimal-group studies did

not assess trait similarity, those measures could not be included in the analyses. Analyses based on range, Pd, SD (variance), global similarity, recall errors, and recognition errors indicated that minimal vs. non-minimal distinction did not interact with measure (QB(5) = 1.06, ns; estimated power = .99) or task type (QB(3) = 3.66, ns; estimated power = .99).

Relative group status

The status distinction (low, high, or unknown) did not moderate perceptions of intergroup variability (QB(2) = 4.08, ns). We subsequently tested whether measure or task interacted with relative group status. Because the included low status and high status studies did not assess SD (variance), the included low status studies did not assess Pd, and the included unknown status studies did not assess mean, those measures could not be included in the analyses. Table 2 displays the estimated effect size and the number of effects upon which the estimate was based as a function of measure and task within levels of status.

Measures Relative group status and measure interacted to affect perceived intergroup variability (QB(14) = 38.68, p < .01). We decomposed the interaction to explore whether each measure detected an effect of status on perceived intergroup variability. As indicated in Table 2, the presence and nature of the status effect varied across measure. Subgroup generation and recall errors did not detect status effects. For the sake of comprehensiveness, we review in turn the measures that detected status effects.

Range This measure detected a status effect (QB(2) = 7.87, p < .05). Members of high status groups (d = -.19, 95%CI -.37/-.01), and unknown status groups (d = -.32, 95%CI -.51/-.13) perceived outgroup homogeneity, whereas members of low status groups did not evidence a difference in the perceived variability of the ingroup vs. outgroup (d = .07, 95% CI -.13/.26).

Percent estimate This measure detected a status effect (QB(2) = 10.14, p < .01). Members of low status groups (d = -.65, 95%CI -1.30/-.01) and

Table 2. Intergroup variability as a function of task, measure, and status

Task Measure Status QBa kb D 95%CI Powerc

Range Range 7.87* .86

High 11 -.19* -.37/-.01

Low 10 .07 -.13/.26

Unk 10 -.32* -.51/-.13

Distribution11 - -

High 1 .52 -.36/1.40

Low 0 - -

Unk 9 .05 -.13/.23

Pdd - -

High 1 .52 -.36/1.40

Low 0 - -

Unk 4 .05 -.31/.41

SD/Varianced - -

High 0 - -

Low 0 - -

Unk 5 .08 -.10/.26

Stereotyping 11.02** .57

High 4 .43 -.10/.96

Low 4 -.67* -1.20/-.14

Unk 5 -.61* -1.07/-.16

Percent estimate 10.14** .52

High 3 .59 -.05/1.24

Low 3 -.65* -1.30/-.01

Unk 5 -.61* -1.10/-.13

Meand - -

High 1 -.05 -.44/.33

Low 1 -.71* -1.11/-.32

Unk 0 - -

Similarity 7.21* .81

High 6 .25 -.06/.56

Low 9 -.16 -.42/.10

Unk 11 .29* .05/.52

Global 6.49* .73

High 4 .31 -.11/.72

Low 7 -.18 -.49/.14

Unk 10 .34* .07/.60

Traite 6.15* .32

High 2 .15 -.02/.31

Low 2 -.10 -.27/.07

Unk 1 -.34 -.83/.15

Subgroup Subgroup 1.27 .31

High 2 -.19 -.91/.53

Low 2 .02 -.70/.73

Unk 1 -.69 -1.70/.31

Memory 9.11* 1.00

High 45 -.32* -.45/-.19

Low 44 -.08 -.22/.05

Unk 31 -.36* -.51/-.21

(continued)

(continued)

Task Measure Status QBa kb D 95% CI Powerc

Face recognition 7.16* .99

High 25 -.33* -.49/-.17

Low 26 -.27* -.43/-.11

Unk 13 -.61 -.83/-.41

Recall errorse 1.89 .57

High 2 -.10 -.48/.28

Low 2 .13 -.25/.51

Unk 9 -.16 -.31/.00

Recognition errors 8.92* .95

High 18 -.32* -.54/-.10

Low 16 .16 -.07/.40

Unk 9 -.15 -.45/. 16

*p < .05; **p < .01.

Notes. Perceived variability varied as a function of Status x Measure (QB(14) = 38.68, p < .01), and Status x Task (QB(8) = 30.48, p < .01).

aQB provides a 2 df test of status for task and measure, respectively. bIndicates the number of effect sizes.

cIndicates the estimated statistical power of the corresponding QB test.

dThis level of task (or measure) was necessarily excluded from the interaction with Status when there were no effect sizes in one or more levels of status.

eWe used a fixed-effects analysis because the random-effects variance did not differ from zero. CI = confidence interval.

unknown status groups (d = -.61, 95%CI —1.10/— .13) perceived outgroup homogeneity. Members of high status groups evidenced a nonsignificant trend toward ingroup homogeneity (d = .59, 95%CI —.05/1.24).

Global similarity This measure detected a status effect (QB(2) = 6.49, p < .05). Members of unknown status groups perceived ingroup homogeneity (d = .34, 95%CI .07/.60). Members of high status groups evidenced a nonsignificant trend toward ingroup homogeneity (d = .31, 95%CI —.11/.72). Members of low status groups evidenced a nonsignificant trend toward outgroup homogeneity (d = —.18, 95%CI —.49/.14).

Trait similarity This measure detected a status effect (QB(2) = 6.15, p < .05). Although the inter-group perceptions (ingroup vs. outgroup) for members of high, low, and unknown status groups differed from one another, none of the effects were individually significant. We suspect that the effects did not achieve statistical significance because they were based on a limited number

of samples (i.e. k = 2 for low and high status respectively and k = 1 for unknown status groups). Both unknown status groups (d = —.34, 95% CI —.83/.15) and low status groups (d = —.10, 95%CI —.27/.07) evidenced a trend toward outgroup homogeneity. Members of high status groups evidenced a trend toward ingroup homogeneity (d = .15, 95%CI —.02/.31).

Face recognition. This measure detected a status effect (QB(2) = 7.16, p < .05). Members of unknown status groups (d = —.61, 95% CI —.83/—.41), high status groups (d = —.33, 95% CI —.49/—.17), and low status groups (d = —.27, 95% CI —.43/—.11) evidenced outgroup homogeneity, which was stronger among unknown status groups than low status groups.

Recognition errors. This measure detected a status effect (QB(2) = 8.92, p < .05). Members of high status groups perceived outgroup homogeneity (d = —.32, 95%CI —.54/—.10). Members of low status groups evidenced a nonsignificant trend toward ingroup homogeneity (d = .16,

95%CI -.07/.40). Members of unknown status groups evidenced a nonsignificant trend toward outgroup homogeneity (d = -.15, 95% CI -.45/.16).

Tasks Relative group status and task interacted to affect perceived intergroup variability (QB(8) = 30.48, p < .01). We decomposed the interaction to explore whether each task detected an effect of status on perceived intergroup variability. As indicated in Table 2, a status effect was detected by each task, with the exception of subgroup generation (QB(2) = 1.27, ns). The nature of the status effect, however, varied across the tasks.

Descriptively similar status effects were detected by range (QB(2) = 7.87, p < .05) and memory (QB(2) = 9.11, p < .05). Both tasks indicated that members of high status groups perceived outgroup homogeneity (d = -.19, 95%CI -.37/-.01 and d = -.32, 95%CI-.45/-

memory '

.19), as did members of unknown status groups (d = -.32, 95%CI -.51/-.13 and d =

v range ' ' memory

-.36, 95%CI-.51/-.21). Members of low status groups did not evidence a reliable difference between the perceived variability of the ingroup vs. outgroup (d = .07, 95%CI -.13/.26; d

range memory

= -.08, 95%CI -.22/.05).

Descriptively different from the previous effects, descriptively similar status effects were detected by stereotyping (QB(2) = 11.02, p < .01), and similarity (QB(2) = 7.21, p < .05). Members of high status groups perceived a nonsignificant trend toward ingroup homogeneity (d, „„. = .43, 95%CI -.10/.96) and

o J \ stereotyping ' ' '

d. .„ = .25, 95%CI -.06/.56). Members of low

similarity ' ' '

status groups perceived outgroup homogeneity (d . = -.67, 95%CI -1.20/-.14 and d. . =

v stereotyping ' similarity

-.16, 95%CI -.42/.10), though nonsignificant on the latter measure. The two tasks differed, however, with respect to members of unknown status groups for which stereotyping detected perceptions of outgroup homogeneity (d = -.61, 95%CI -1.07/-.16) and similarity detected perceptions of ingroup homogeneity (d = .29, 95%CI .05/.52).

Comparisons among measures within a given task indicated that measures derived from a common task did not differ. For the stereotyping

task, the estimated effect sizes for percent estimate and mean did not differ (QB(1) = 0.27, ns). For the similarity task, the estimated effect sizes for global and trait similarity did not differ (QB(2) = 1.82, ns). For the memory task, the estimated effect sizes for face recognition, recall errors, and recognitions errors did not differ (QB(4) = 6.18, ns). In summary, analyses at the level of task indicated that the tasks of range and memory detected similar perceptions within each level of status which differed from the relatively similar patterns detected within levels of status by the stereotyping and similarity tasks.

Exploring potential confounds of the status x task effect

The Status x Task effect is intriguing because it implies that the tasks (or measures) of perceived variability are differentially sensitive to (evidence different patterns of) the effect of relative group status. The random-effects analysis certainly justifies a broader generalization beyond the contexts defined by the specific characteristics of the included studies. Nonetheless, the validity of the Status x Task effect remains suspect to the extent to which additional study characteristics systematically varied with the effect. Consequently, we coded seven additional characteristics to explore their role as potential confounds. Three characteristics, in addition to group-type (i.e. minimal vs. non-minimal), assess aspects of the social groups: (a) permeable vs. impermeable boundaries (i.e. can members assume membership in the contrasting group?),

(b) history of conflict between groups, and

(c) relative group size (i.e. is a group a numerical minority or is the ingroup and outgroup composed of an equal number of members?). Four characteristics assess aspects of the procedure:

(a) trait valence (i.e. are the traits of the variability measures positively or negatively valenced?),

(b) trait typicality (i.e. are the traits typical or relevant to the group's stereotype?), (c) order in which participants rated the ingroup and outgroup, and (d) scale metric (i.e. did participants respond on scales of a shorter vs. longer metric, e.g. 4 vs. 21 points?).

We examined the proportionate distribution of each characteristic across the levels of status

and task. Inconsistent with their potential role as confounds, none of the study characteristics were distributed in a manner that tracked the patterns of perceived variability evidenced by the Status x Task effect. Consequently, those characteristics do not constitute threats to the validity of the latter effect. Although, this does not eliminate the role of a yet-to-be identified study characteristic that varies systematically with the pattern of the Task x Status interaction, we attempted to explore conceivable confounds.4

We addressed a second potential confound introduced uniquely by the inclusion of effects from minimal group studies. The intergroup variability effect size for minimal groups did not differ from zero (indicating a perception of equivalent ingroup vs. outgroup variability) and those effects were distributed primarily among the tasks of range, similarity, and to a lesser degree, memory. Consequently, introducing effects from minimal group studies into the analysis of Task x Status (and separate analyses of task and status, respectively) could have attenuated estimates of intergroup variability—particularly, for those tasks with higher concentrations of minimal group effects. To examine the extent of this potential confound, we repeated the previous analyses including only effects from non-minimal groups. With two exceptions, all conclusions based on p values and direction of effects remained unchanged.

The first exception is that the previous nonsignificant status main effect became significant (QB(2) = 6.44, p < .05) when non-minimal effects were eliminated. Members of high and unknown status groups perceived outgroup homogeneity (dhigh = —.24, 95%CI —.37/—.11; d t = —.35, 95%CI —.49/—.21) and members

unkown ' '

of low status groups perceived equivalent ingroup vs. outgroup variability (dlow = —.11, 95% CI —.24/02). The second exception is that in the context of the still significant Task x Status effect the similarity task no longer evidenced a significant status effect. It is important to mention, however, that this revised analysis of the similarity task is based on a pool of effects from high status groups that dropped from six to one effect size. In which case, it is not clear whether the now nonsignificant effect is due

to the removal of minimal group effect sizes or severely reduced statistical power. In any event, it is clear that the obtained patterns of the Task x Status interaction were not driven primarily by effect sizes from minimal groups—the same patterns were evidenced among effects from non-minimal groups.

Discussion

We meta-analytically synthesized the intergroup variability literature to examine whether the multiple measures (and tasks from which the measures are derived) moderate perceptions of variability. Furthermore, we examined the possibility that the patterns evidenced by particular measures (and tasks) vary or interact with socio-structural factors. When ignoring the distinction among the measures (and tasks), the meta-analysis revealed a small but reliable tendency to perceive more variability among ingroup than outgroup members and for such an outgroup homogeneity effect to be stronger in non-minimal than minimal groups. However, the meta-analysis further indicated that the distinction among the measures (and tasks) should not be ignored. Perceptions of intergroup variability systematically varied as a function of measure (task) and those discrepancies further varied as a function of intergroup status.

Summarizing the patterns

In the following sections, we briefly summarize patterns of intergroup perception across the tasks (and measures) and as a function of the socio-structural moderators. We subsequently evaluate speculative accounts as to why tasks (and measures) yield discrepant patterns.

Omnibus patterns across tasks and measures

When averaging across (i.e. ignoring) the socio-structural variables, the stereotyping and memory tasks detected significant perceptions of outgroup homogeneity. The constituent measures of those tasks consistently produced effects in the same direction, but those individual effects were not always significant. The percent estimate measure of the stereotyping task and the face recognition measure of the memory task

detected significant outgroup homogeneity. Likewise, the range and subgroup generation tasks/measures detected a nonsignificant trend toward outgroup homogeneity. In contrast, the similarity task (and the global similarity measure) detected a nonsignificant trend toward ingroup homogeneity. Finally, the distribution task (and the constituent measures of Pd and SD/variance) detected near zero effect sizes (i.e. perceptions of equivalent ingroup vs. outgroup variability).

Minimal vs. non-minimal grouping as moderator

When accounting for the moderating effect of minimal vs. non-minimal grouping, the (available) tasks and measures consistently detected the same pattern of perception. Members of non-minimal groups perceived outgroup homogeneity and members of minimal groups perceived equivalent ingroup vs. outgroup variability.

Relative group status as moderator In contrast to the previously described omnibus patterns, we observed a shuffling among the clusters of tasks (and measures) that detected similar perceptions when accounting for the moderating effect of relative group status. Among the omnibus patterns, range, stereotyping, and memory yielded consistent effects (though the effect detected by range was nonsignificant). In the context of relative group status, a different clustering among tasks (and measurers) emerged.

The range and memory tasks evidenced similar patterns of perception which differed from the similar patterns evidenced by the stereotyping and similarity tasks. In particular, the range and memory tasks detected outgroup homogeneity among members of high and unknown status groups and equivalent perceptions of ingroup vs. outgroup variability among members of low status groups. In contrast, stereotyping and similarity detected a trend toward ingroup homogeneity among members of high status groups and outgroup homogeneity among members of low status groups (with the pattern being significant for stereotyping and nonsignificant for similarity). Stereotyping and similarity differed only in regard to the perceptions evidenced by members of unknown status groups, with

stereotyping detecting outgroup homogeneity and similarity detecting ingroup homogeneity.

The subgroup generation task/measure, which yielded relatively large confidence intervals, detected neither a status effect nor differential perceptions of ingroup vs. outgroup variability. We could not examine the distribution task (and measures) because of an insufficient number of effect sizes in the literature. Readers might note descriptive differences among measures within a given task. However, analyses at the level of measure frequently involved effect estimates based on a small number of samples. Consequently, we are more comfortable interpreting analyses at the level of task, which aggregated across related measures and benefited from greater reliability.

Accounting for discrepancies Measurement as mechanism Issues of reliability and validity might be offered as explanations of the discrepancies across tasks (and measures). We suggest, however, that such measurement accounts provide insufficient explanations. The reliability account assumes that all measures tap the same construct, but do so with differing degrees of error. That tasks (and measures) interacted with group status is not fully consistent with a strict reliability account—particularly because the patterns of perception detected by the various tasks (and measures) varied in direction, not just magnitude, across levels of status.

The validity account assumes that the tasks (and measures—or constellations of tasks and measures) tap different constructs. While perceived variability is likely a multidimensional phenomenon (Park & Judd, 1990), a simple validity explanation cannot account sufficiently for the discrepancies among tasks (and measures). Based on a strict validity account we would anticipate that tasks (and measures) tapping a common construct would yield consistent patterns across socio-structural moderators. That the multiple tasks and measures detected similar patterns across the minimal vs. nonminimal variable but different patterns across the relative group status variable is inconsistent with a strict validity account. Of course, we are not suggesting that issues of reliability and validity

do not contribute to the observed discrepancies. Instead, we are suggesting that such measurement issues do not provide exhaustive accounts and a more complex explanation is necessary.

Measure as mechanism We do not pretend to have a complete explanation for the observed discrepancies and, instead, offer glimpses of potential explanations. As suggested in the introduction, the tasks (and measures) potentially require or prompt different mental activities. Some tasks, such as range, distribution, and memory, prompt persons to think about individuals. Other tasks (and measures), such as stereotyping and similarity, prompt persons to think abstractly about the group as a whole. Tasks (and measures) differ further in whether they require the possession and utilization of accurate information. The memory tasks in particular define variability in terms of a participant's ability to accurately encode and retrieve information—i.e. low variability is indexed in regard to high confusion errors. Of course, the tendency for the tasks (and measures) to detect similar patterns of perception across the minimal vs. non-minimal variable and dissimilar patterns across levels of group status cannot be explained simply in regard to the different mental activities required by the multiple tasks and measures. It is plausible, however, that such differential activities function in conjunction with social processes that are activated by socio-structural variables. Stated otherwise, due to their requisite mental activities, tasks (and measures) potentially vary in sensitivity to different social processes activated by socio-structural moderators.

Relative group status, for example, plausibly activates motivational (Tajfel & Turner, 1979) and cognitive (Fiske, 1993) processes that regulate intergroup perception. Social identity theory (Tajfel & Turner, 1979), for example, suggests that membership in a low status group provides unfavorable intergroup comparisons that threaten the positive distinctiveness of an internalized social identity. In the absence of a mutable status structure, the identity threat motivates efforts toward individual mobility (i.e. seeking membership in higher status group; e.g. Ellemers, van Kippenberg, de Vries, &

Wilke, 1988) or a biased social perception that enhances and redefines the ingroup's positive distinctiveness (e.g. Ellemers et al., 1997). As others have argued, perceptions of group variability can be altered in the service of a motivated identity enhancement strategy. For example, exaggerating the perceived variability of the ingroup and outgroup on comparison dimensions that threaten the ingroup might blur intergroup distinctions and soften the threat of the unfavorable comparison (Doojse, Spears, & Koomen, 1995). Likewise, perceiving outgroup homogeneity (i.e. greater variability among ingroup than outgroup members) on dimensions that threaten the ingroup might protect a positive self-identity by allowing greater discrepancy between self and ingroup members (Rubin, Hewstone, & Voci, 2001).

Cognitive processes activated by group status are suggested by perspectives that interpret status effects as reflecting underlying power differences between groups (i.e. higher status groups tend to be higher power groups; e.g. Guinote,Judd, & Brauer, 2002; Sidanius & Pratto, 1999). Cognitive accounts suggest that power affects relative attention to the idiosyncratic characteristics and attributes of ingroup vs. outgroup members. Fiske and colleagues (Fiske, 1993; Fiske & Depret, 1996; Neuberg & Fiske, 1987), suggest that members of the higher power group—whose outcomes are affected slightly, if at all, by the lower power group—have little reason to attend to the lower power outgroup. However, members of the lower power group, whose outcomes are greatly affected by the actions of the higher power group, are apt to attend to the traits and behaviors of outgroup members. Such differential attention provides members of the lower power group with a more diverse perception of the outgroup.5

So, how might these motivational and cognitive processes interact with the mental activities prompted by the variability tasks (and measures) to produce discrepant patterns of perception across the measures? Measures that require accurate encoding and retrieval of information (such as those of the memory task) might be particularly sensitive to cognitive processes that affect variability judgments. If, for example,

members of low power groups (i.e. which typically are low status) are more attentive to information about outgroup (and ingroup) members than are members of high power groups, they would arguably have more accurate intergroup memories and, consequently, evidence fewer ingroup and outgroup errors on memory tasks. The pattern of perception evidenced by the memory measures for members of low and high status groups is certainly consistent with such an explanation (see Table 2). The stereotyping and similarity tasks ostensibly do not require the utilization of accurate information, which may account for the discrepancy between the patterns evidenced by those tasks and the patterns of the memory tasks. Perhaps aspects of the stereotyping and similarity tasks (e.g., prompting group-leveljudgments) enable those tasks to tap into motivated processes that bias variability judgments in an effort to maintain a positive social identity. Unfortunately, the data of the current meta-analysis do not enable us to test conclusively these speculations.

In conclusion, this meta-analytic research indicates that the multiple measures of perceived variability systematically yield different patterns of perception across changing socio-structural factors. These shifting patterns pose a challenge to theory development and stand as a warning that choice of measure contributes to resulting perceptions of intergroup variability. We suspect that characteristics of the measures (or tasks from which the measures are derived) render the measures differentially sensitive to social processes triggered by social contexts. Such a suspicion awaits empirical scrutiny and we suggest that a fruitful avenue for future research is to examine systematically the mental activities initiated by the various measures.

1. Relative group size (i.e. majority vs. minority) has been identified as a moderator of perceived variability (Mullen & Hu, 1989). Unfortunately, studies that assessed group size included

an insufficient array of perceived variability measures for meta-analytic purposes.

2. Two issues preclude inclusion of between-subject ratings of ingroup vs. outgroup. Within-person

ratings control for idiosyncratic perceptions across persons (which between-person ratings do not), and necessarily provide the relative perceptions of the ingroup vs. the outgroup. Ignoring the latter issue, it is possible to estimate the relative ingroup vs. outgroup perception from between-person ratings, however, effect sizes from such between-person ratings cannot be combined with effect sizes from within-person ratings unless the two ratings are transformed into a common metric (Morris & DeShon, 2002). Transforming the between-person ratings requires an estimate of the standard deviation of: (a) ingroup ratings, (b) outgroup ratings, and (c) the difference between ingroup and outgroup ratings, and all three estimates must be at the same level (i.e. within cells of the primary study or aggregated across cells of the primary study). Those three estimates were not concurrently available from the primary studies.

3 Four studies required participants to rate multiple groupings (e.g. male vs. female and Blacks vs. Whites; Brigham & Barkowitz, 1978; Cross, Cross, & Daly, 1971; Jalbert & Getting, 1992; McKelvie, 1981). Consequently, each of those studies contributed two effect sizes and introduced minimal non-independence.

4. A detailed description of the distribution of each characteristic is available from the authors upon request.

5. We discuss this as a cognitive process because of our emphasis on the processing of intergroup information (i.e. attention to and retrieval

of information). We concur with Fiske and colleagues who suggest that differential encoding of information is motivated by concerns of outcome dependence.

Acknowledgments

This research comprised a portion of the first

author's doctoral dissertation, which was directed

by the second author. We thank Lynda Nguyen and

Amanda Amidon for assistance with data coding.

References

Bardach, L. & Park, B. (1996). The effect of ingroup and outgroup status on memory for consistent and inconsistent behavior of an individual. Personality and Social Psychology Bulletin, 22, 169-178.

*Boldry, J. G., & Gaertner, L. (2006). Separating status from power as an antecedent of

intergroup perception. Group Processes and Intergroup Relations, 9, 377—400.

*Boldry, J. G., & Kashy, D. A. (1999). Intergroup perception in naturally occurring groups of differential status. A social relations perspective. Journal of Personality and Social Psychology, 77, 1200—1212.

*Brauer, M. (2001). Intergroup perception in the social context: The effects of social status and group membership on perceived outgroup homogeneity and ethnocentrism. Journal of Experimental Social Psychology, 37, 15—31.

*Brewer, M. B., Dull, V., & Lui, L. (1981). Perceptions of the elderly: Stereotypes as prototypes. Journal of Personality and Social Psychology, 41, 656—670.

*Brewer, M. ., Weber, J. G., & Carini, B. (1995). Person memory in intergroup contexts: Categorization versus individuation. Journal of Personality and Social Psychology, 69, 29—40.

*Brigham, J. C. & Barkowitz, P. (1978). Do 'they all look alike?' The effect of race, sex, experience and attitudes on the ability to recognize faces. Journal of Applied Social Psychology, 8, 306—318.

*Brigham, J. C., Maas, A., Snyder, L. D., & Spaulding, K. (1982). Accuracy of eyewitness identifications in a field setting. Journal of Personality and Social Psychology, 42, 673—681.

*Brigham, J. C., & Ready, D. J. (1985). Own-race bias in lineup construction. Law and Human Behavior, 9, 415—424.

*Brigham, J. C., & Williamson, N. L. (1979). Cross-racial recognition and age: When you're over 60 do they still 'all look alike?' Personality and Social Psychology Bulletin, 5, 218—222.

*Brown, R., & Smith A. (1989). Perceptions of and by minority groups: The case of women in academia. European Journal of Social Psychology, 19, 61—75.

*Brown, R., & Wootten-Millward, L. (1993).

Perceptions of group homogeneity during group formation and change. Social Cognition, 11, 126—149.

*Buckhout, R., & Regan, S. (1988). Explorations in research on the other-race effect in face recognition. In M. M. Grunberg & P. E. Morris (Eds.), Practical aspects of memory: Current research and issues: Vol. 1. Memory in everyday life (pp. 40—46). New York: Wiley.

*Byatt, G., & Rhodes, G. (1998). Recognition of own-race and other-race caricatures: Implications for models of face recognition. Vision Research, 38, 2455—2468.

*Cabecinhas, R., & Amancio, L. (1999).

Asymmetries in the perception of other as a function of social position and context. Swiss Journal of Psychology, 58, 40—50.

Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the Multitrait— multimethod matrix. Psychological Bulletin, 56, 81—105.

Carli, L. L, & Eagly, A. H. (1999). Gender effects on social influence and emergent leadership. In G. N. Powell (Ed.), Handbook of gender and work (pp. 203—222). Thousand Oaks, CA: Sage.

*Carroo, A. W. (1987). Recognition of faces as a function of race, attitudes, and reported cross-racial friendships. Perceptual and Motor Skills, 64, 319—325.

*Castano, E., & Yzerbyt, V. Y. (1998). The highs and lows of group homogeneity. Behavioural Processes, 42, 219—238.

*Chance, J. E., Goldstein, A., & Anderson, B. (1986). Recognition memory for infant faces: An analog of the other-race effect. Bulletin of the Psychonomic Society, 24, 257—260.

*Chance, J. E., Goldstein, A., & McBride, L. (1975). Differential experience and recognition memory for faces. Journal of Social Psychology, 97, 243—253.

*Chance, J. E., Turner, A. L., & Goldstein, A. G. (1982). Development of differential recognition for own- and other-race faces. Journal of Psychology, 112, 29—37.

*Chiroro, P., & Valentine, T. (1995). An

investigation of the contact hypothesis of the own-race bias in face recognition. Quarterly Journal of Experimental Psychology, 48A, 879—894.

*Cross, J. F., Cross, J., & Daly J. (1971). Sex, race, age, and beauty in the recognition of faces. Perception and Psychophysics, 10, 393—396.

*De la Haye, A. (2002). Unpublished data, University of Poitiers, Poitiers, France.

Doojse, B., Spears, R., & Koomen, W. (1995). When bad isn't all bad: Strategic use of sample information in generalization and stereotyping. Journal of Personality and Social Psychology, 69, 642—655.

Eagly, A. H., Wood, W., & Diekman, A. B. (2000). Social role theory of sex differences and similarities: A current appraisal. In T. Eckes & H. M. Trautner (Eds.), The developmental social psychology of gender (pp. 123—174). Mahweh, NJ: Erlbaum.

Ellemers, N., van Knippenberg, A., & de Vries, N., & Wilke, H. (1988). Social identification and permeability of group boundaries. European Journal of Social Psychology, 18, 497—413.

Ellemers, N., van Rijswijk, W., Roefs, M., & Simons, C. (1997). Bias in intergroup perceptions: Balancing group identity with social reality.

Personality and Social Psychology Bulletin, 23, 186-198.

*Eller, A. (2001). Putting Pettigrew's reformulated model to the test: The intergroup contact theory in transition. Unpublished doctoral dissertation. University of Kent at Canterbury, UK.

*Ellis, H. D., Davies, G. M., & McMurran, M.

(1979). Recall of white and black faces by white and black witnesses using the photofit system. Human Factors, 21, 55-59.

*Ellis, H. D., & Derogowski, J. B. (1981).

Within-race and between-race recognition of transformed and untransformed faces. American Journal of Psychohgy, 94, 27-35.

*Ellis, H., Shepard, J., & Bruce, A. (1973). The effects of age and sex upon adolescents' recognition of faces. Journal of Genetic Psychohgy, 123, 173-174.

*Feinman, S., & Entwisle, D. R. (1976). Children's ability to recognize other children's faces. Child Development, 47, 506-510.

Fiedler, K., Kemmelmeier, M., & Freytag, P. (1999). Explaining asymmetric intergroup judgments through differential aggregation: Computer simulations and some new evidence. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology (Vol. 10, pp. 1-10). New York: Wiley.

Field, A. P. (2001). Meta-analysis of correlation coefficients: A Monte Carlo comparison of fixed- and random-effects methods. Psychological Methods, 6, 161-180.

Field, A. P. (2003). The problems using fixed-effects models of meta-analysis on real-world data. Understanding Statistics, 2, 77-96.

Fiske, S. T. (1993). Controlling other people: The impact of power on stereotyping. American Psychologist, 48, 621-628.

Fiske, S. T., & Depret, E. (1996). Control,

interdependence, and power: Understanding social cognition in its social context. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology (Vol. 7, pp. 31-61). Chichester, UK: Wiley.

*Frable. D. E. S., & Bem, S. L. (1985). If you are gender schematic, all member of the opposite sex look alike. Journal of Personality and Social Psychology, 49, 459-468.

*Galper, R. E. (1973). Functional race membership and recognition of faces. Perceptual and Motor Skills, 37, 455-462.

Garcia-Marques, L., & Hamilton, D. L. (1996). Resolving the apparent discrepancy between the incongruency effect and the expectancy-based illusory correlation effect: The TRAP model. Journal of Personality and Social Psychology, 71, 845-860.

*Going, M., & Read J. D. (1974). Effects of uniqueness, sex of subject, and sex of photograph on facial recognition. Perceptual and Motor Skills, 39, 109-110.

»Goldstein, A. G., & Chance, J. E. (1979). Do foreign faces really look alike? Bulletin of the Psychonomic Society, 13, 111-113.

»Goldstein, A. G., & Chance, J. E. (1980). Memory for faces and schema theory. Journal of Psychology, 105, 47-59.

»Gramzow, R.H., Gaertner, L., & Sedikides, C. (2001). Memory for ingroup and outgroup information in a minimal group context: The self and an informational base. Journal of Personality and Social Psychology, 80, 188-205.

»Guinote, A. (2001). The perception of group variability in a non-minority and a minority context: When adaptation leads to outgroup differentiation. British Journal of Social Psychology,

40, 117-132.

»Guinote, A., Judd, C. M., & Brauer, M. (2002). Effects of power on perceived and objective group variability: Evidence that more powerful groups are more variable. Journal of Personality and Social Psychology, 82, 708-721.

*Haslam, S. A., Oakes, P. J., McGarty, C., Turner, J. C., Reynolds, K J., & Eggins, R. A. (1996). Stereotyping and social influence: The mediation of stereotype applicability and sharedness by the views of ingroup and outgroup members. British Journal of Social Psychology, 35, 369-397.

*Haslam, S. A., Oakes, P. J., Turner, J. C., & McGarty, C. (1995). Social categorization and group homogeneity: Changes in the perceived applicability of stereotype content as a function of comparative context and trait favourableness. British Journal of Social Psychology, 34, 139-160.

Hedges, L. V. (1981). Distribution theory for Glass's estimator of effect size and related estimators. Journal of Educational Statistics, 6, 107-128.

Hedges, L. V., & Vevea, J. L. (1998). Fixed-and random-effects models in meta-analysis. Psychological Methods, 3, 486-504.

»Hewstone, M., Hantzi, A., & Johnston, L. (1991). Social categorization and person memory: The pervasiveness of race as an organizing principle. European Journal of Social Psychology, 21, 517-528.

Hewstone, M., Islam, M. R., & Judd, C. M. (1993). Models of crossed categorization and intergroup relations. Journal of Personality and Social Psychology, 64, 779-793.

»Howard, J. W., & Rothbart, M. (1980). Social categorization and memory for ingroup and outgroup behavior. Journal of Personality and Social Psychology, 38, 301-310.

Hunter, J. E., & Schmidt, F. L. (2000). Fixed

effects vs. random effects meta-analysis models: Implications for cumulative knowledge in psychology. International Journal of Selection and Assessment, 8, 275-292.

»Jackson, L. A., & Hymes, R. W. (1985). Gender and social categorization: Familiarity and ingroup polarization in recall evaluation. Journal of Social Psychology, 125, 81-88

»Jackson, J. W., & Smith E. R. (1999). Conceptualizing social identity: A new framework and evidence for the impact of different dimensions. Personality and Social Psychology Bulletin, 25, 120-135.

»Jalbert, N. L., & Getting, J. (1992). Racial and gender issues in facial recognition. In F. Loesel & D. Bender (Eds.), Psychology and law: International perspectives (pp. 309-316). Berlin: Walter De Gruyter.

Jones, E. E., Wood, G. C., & Quattrone, G. A. (1981). Perceived variability of personal characteristics of ingroups and outgroups: The role of knowledge and evaluation. Personality and Social Psychology Bulletin, 7, 523-528.

»Jost, J. (1995). Psychology and the social order revisited. Unpublished doctoral dissertation.

»Judd, C. M., & Park, B. (1988). Outgroup homogeneity: Judgments of variability at the individual and group levels. Journal of Personality and Social Psychology, 54, 778-788.

»Judd, C. M., Park, B., Ryan, C. S., Brauer, M., & Kraus, S. (1995). Stereotypes and ethnocentrism: Diverging interethnic perceptions of African American and white American youth. Journal of Personality and Social Psychology, 69, 460-481.

Judd, C. M., Ryan, C. S., & Park, B. (1991). Accuracy in the judgment of ingroup and outgroup variability. Journal of Personality and Social Psychology, 61, 366-379.

Kashima, E. S., & Kashima, Y. (1993). Perceptions of general variability of social groups. Social Cognition, 11, 1-21.

»Kelly, C. (1989). Political identity and perceived intragroup homogeneity. British Journal of Social Psychology, 28, 239-250.

*Lee, Y. T. (1992). Ingroup preference and homogeneity among African-American and Chinese-American students. Journal of Social Psychology, 133, 225—235.

Lee, Y. T. (1993). Perceived homogeneity and

familial loyalty between Chinese and Americans. Current Psychology: Developmental, Learning, Personality, Social, 12, 260—267.

*Lindsay, D. S., Jack, P. C., & Christian, M. A. (1991). Other-race face perception. Journal of Applied Psychology, 76, 587—589.

Linville, P. W., Fischer, G. W., & Salovey, P. (1989). Perceived distributions of the characteristics of ingroup and outgroup members: Empirical evidence and a computer simulation. Journal of Personality and Social Psychology, 57, 165—188.

Linville, P.W., Fischer, G. W., & Yoon, C. (1996). Perceived covariation among the features of ingroup and outgroup members: The outgroup covariation effect. Journal of Personality and Social Psychology, 70, 421—436.

Linville, P. W., Salovey, P., & Fischer, G. W. (1986). Stereotyping and perceived distributions of social characteristics: An application to ingroup-outgroup perception. In J. F. Dovidio & S. L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 165—208), New York: Academic Press.

*Lorenzi-Cioldi, F. (1993). They all look alike and so do we...sometimes: Perceptions of ingroup and outgroup homogeneity as a function of sex and context. British Journal of Social Psychology, 32, 111—124.

*Lorenzi-Cioldi, F. (1998). Group status and perceptions of homogeneity. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology (Vol. 9, pp. 31—75). New York: Wiley.

*Lorenzi-Cioldi, F., Deaux, K., & Dafflon, A. (1998). Group homogeneity as a function of relative social status. Swiss Journal of Psychology, 57, 255—273.

*Lorenzi-Cioldi, F., Eagly, A. H., & Stewart,

T. L. (1995). Homogeneity of gender groups in memory. Journal of Experimental Social Psychology, 31, 193—217.

*Lui, L., & Brewer, M. B. (1983). Recognition accuracy as evidence of category-consistency effects in person memory. Social Cognition, 2, 89—107.

*Malpass, R. S., & Kravitz, J. (1969). Recognition of faces of own and other race. Journal of Personality and Social Psychology, 13, 330—334.

»Malpass, R. S., Lavigeur, H., & Weldon, D. E. (1973). Verbal and visual training in face recognition. Perception and Psychophysics, 14, 285-292.

»Mason, S. E. (1986). Age and gender as factors in facial recognition and identification. Experimental Aging Research, 12, 151-154.

»McKelvie, S. J. (1981). Sex differences in memory for faces. Journal of Psychology, 107, 109-125.

Morris, S. B., & DeShon, R. P. (2002). Combining effect size estimates in meta-analysis with repeated measures and independent-groups designs. Psychological Methods, 7, 105-125.

Mullen, B. & Hu, L. (1989). Perceptions of ingroup and outgroup variability: A meta-analytic integration. Basic and Applied Social Psychology, 10, 233-252.

Neuberg, S. L., & Fiske, S. T. (1987). Motivational influences on impression formation: Outcome dependency, accuracy driven attention, and individuating processes. Journal of Personality and Social Psychology, 53, 431-444.

»Ostrom, T. M., Carpenter, S. L., Sedikides, C., & Li, F. (1993). Differential processing of ingroup and outgroup information. Journal of Personality and Social Psychology, 64, 21-34.

Ostrom, T. M., & Sedikides, C. (1992). Outgroup homogeneity effects in natural and minimal groups. Psychological Bulletin, 112, 536-552.

»O'Toole, A. J., Deffenbacher, K. A., Valentine, D., & Abdi, H. (1994). Structural aspects of face recognition and the other-race effect. Memory and Cognition, 2, 208-224.

Park, B., & Hastie, R. (1987). Perception of variability in category development: Instance versus abstraction based stereotypes. Journal of Personality and Social Psychology, 53, 621-635.

Park. B., & Judd, C. M. (1990). Measures and models of perceived group variability. Journal of Personality and Social Psychology, 59, 173-191.

Park, B., Judd, C. M., & Ryan, C. S. (1991). Social categorization and the representation of variability information. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology (Vol. 2). Chichester, UK: Wiley.

Park, B., & Rothbart, M. (1982). Perception of outgroup homogeneity and levels of social categorization: Memory for the subordinate attributes of ingroup and outgroup members. Journal of Personality and Social Psychology, 42, 1051-1068.

*Park, B., Ryan, C. S., & Judd, C. M. (1992). Role of meaningful subgroups in explaining differences in perceived variability for ingroups and outgroups. Journal of Personality and Social Psychology, 63, 553—567.

*Pickett, C. L., & Brewer, M. (2001). Assimilation and differentiation needs as motivational determinants of perceived ingroup and outgroup homogeneity. Journal of Experimental Social Psychology, 37, 341-348.

*Platz, S. J., & Hosch, H. M. (1988). Cross-racial/ ethnic eyewitness identification: A field study. Journal of Applied Social Psychology, 18, 972-984.

Quattrone, G. A., & Jones, E. E. (1980). The

perception of group variability within ingroups and outgroups: Implications for the law of small numbers. Journal of Personality and Social Psychology, 38, 141-152.

Rosenthal, R. (1994). Parametric measures of effect size. In H. Cooper & L. V. Hedges (Eds.), The handbook of research synthesis (pp. 231-244). New York: Russell Sage Foundation.

Rothgerber, H. (1997). External intergroup threat as an antecedent to perceptions of ingroup and outgroup homogeneity. Journal of Personality and Social Psychology,

73, 1206-1212.

Rubin, M., Hewstone, M., Crisp, R. J., Voci, A., & Richards, Z. (2002). Gender outgroup homogeneity: The roles of differential familiarity, gender differences, and group size. Unpublished manuscript, University of Newcastle, Callaghan, Australia.

*Rubin, M., Hewstone, M., & Voci, A. (2001).

Stretching the boundaries: strategic perceptions of intragroup variability. European Journal of Social Psychology, 31, 413-429.

*Rubin, M., Richards, Z., & Hewstone, M. (2000). Unpublished data, Cardiff University, Cardiff, Wales.

*Ryan, C. S., & Bogart, L. M. (1997). Development of new group members' ingroup and outgroup stereotypes: Changes in perceived group variability and ethnocentrism. Journal of Personality and Social Psychology, 73, 719-712.

*Ryan, C. S., & Judd, C. M. (1992). False consensus and outgroup homogeneity: A methodological note on their relationship. British Journal of Social Psychology, 31, 269-283.

*Ryan, C. S., Judd, C. M., & Park, B. (1996). Effects of racial stereotypes on judgments of individuals: The moderating role of perceived group variability. Journal of Experimental Social Psychology, 32, 71-103.

»Sedikides, C., (1997). Differential processing of ingroup and outgroup information: The role of relative group status in permeable boundary groups. European Journal of Social Psychology, 27, 121-144.

Sedikides, C., & Ostrom, T. S. (1993). Perceptions of group variability: Moving from an uncertain crawl to a purposeful stride. Social Cognition, 11, 165-174.

»Shepherd, J. W., Deregowski, J. B., & Ellis, H. D. (1974). A cross-cultural study of recognition memory for faces. International Journal of Psychology, 9, 205-211.

Sidanius, J., Levin, S., & Pratto, F. (1996).

Consensual social dominance orientation and its correlates within the hierarchical structure of American society [special issue]. International Journal of Intercultural Relations: Prejudice, Discrimination, and Conflict, 20, 385-408.

Sidanius, J., Liu, J. H., Shaw, J. S., & Pratto, F.

(1994). Social dominance orientation, hierarchy attenuators and hierarchy enhancers: Social dominance theory and the criminal justice system. Journal of Applied Social Psychology, 24, 338-336.

Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory of social hierarchy and oppression. Cambridge, UK: Cambridge University Press.

Sidanius, J., Pratto, F., & Rabinowitz, J. L.

(1994). Gender, ethnic status, and ideological asymmetry: A social dominance interpretation. Journal of Cross-Cultural Psychology, 25, 194-216.

Simon, B. (1992). The perception of ingroup and outgroup homogeneity: Reintroducing the intergroup context. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology (Vol. 3). Chichester, UK: Wiley.

»Simon, B., & Brown. R. (1987). Perceived

intragroup homogeneity in minority-majority contexts. Journal of Personality and Social Psychology, 53, 703-711.

»Simon, B., & Hamilton, D. L. (1994). Self-

stereotyping and social context: The effects of relative ingroup size and ingroup status. Journal of Personality and Social Psychology, 66, 699-711.

»Simon, B., & Hastedt, C. (1999). Self-aspects as social categories: The role of personal importance and valence. European Journal of Social Psychology, 29, 479-487.

»Simon, B., & Mummendey, A. (1990). Perceptions of relative group size and group homogeneity: We are the majority and they are all the same. European Journal of Social Psychology, 20, 351-356.

»Simon, B., & Pettigrew, T. F. (1990). Social identity and perceived group homogeneity: Evidence for the ingroup homogeneity effect. European Journal of Social Psychology, 20, 269-286.

»Stephen, W. G. (1977). Cognitive differentiation in intergroup perception. Sociometry, 40, 50-58.

Stewart, T. L., & Vassar, P. M. (2000). The effect of occupational status cues on memory for male and female targets. Psychology of Women Quarterly, 24, 161-169.

»Stewart, T. L., Vassar, P. M., Sanchez, D. T., & David, S. E. (2000). Attitude toward women's societal roles moderates the effect of gender cues on target individuation. Journal of Personality and Social Psychology, 79, 143-157.

Tajfel H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In S. Worchel & W. G. Austin (Eds.), The social psychology of intergroup relations (pp. 33-47). Monterey, CA: Brooks/Cole.

»Thompson, S. C., Kohles, J. C., Otsuki, T. A., & Kent, D. A. (1997). Perceptions of attitudinal similarity in ethnic groups in the US: Ingroup and outgroup homogeneity effects. European Journal of Social Psychology, 27, 209-220.

»Valentine, T., & Bruce, V. (1986). The effect of race, inversion, and encoding activity upon face recognition. Acta Psychologia, 61, 259-273.

»Valentine, T., & Endo, M. (1992). Towards an exemplar model of face processing: The effects of race and distinctiveness. Quarterly Journal of Experimental Psychology, 44A, 671-703.

»Verkuyten, M., & Kwa, G. A. (1996). Ethnic self-identification, ethnic involvement, and group differentiation among Chinese youth in the Netherlands. Journal of Social Psychology, 136, 35-48.

Voci, A. (2000). Perceived group variability and the salience of personal and social identity. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology (pp. 177-221). New York: Wiley.

»Vokey, J. R., & Read, D. (1988) Typicality, familiarity, and the recognition of male and female faces. Canadian Journal of Psychology, 42, 489-495.

»Weimann, G., Fishman, G., & Rattner, A. (1988). Social distance and misidentification. Research Communications, 27, 217-225.

»Wilder, D. A. (1984). Predictions of belief homogeneity and similarity following social categorization. British Journal of Social Psychology, 23, 323-333.

* Indicates articles included in the meta-analysis.

Paper received 1 April 2004; revised version accepted 17 June 2004.

Biographical notes

Jennifer g. boldry received her PhD in social psychology at Texas A&M University in 2001. Her main areas of research include intergroup and interpersonal perception, particularly with respect to the effects of status and power.

Lowell gaertner is an assistant professor of psychology at the University of Tennessee. In addition to his interests in intergroup and intragroup relations, he studies self processes and mass violence.

Jeffrey m. quinn received his PhD in social psychology from Duke University in 2005. His primary research interests include attitudes and habitual behavior.