Two primary dimensions of stereotype content

Upon encountering out-group members, we posit that people ask two questions: Do they intend to harm me; and are they capable of harming me? The two core dimensions of general stereotype content that we propose, warmth (e.g. friendly, good-natured, sincere, and warm) and competence (e.g. capable, competent, confident, and skillful), respectively, answer these questions (for reviews, see Cuddy, et al., 2008; Fiske, et al., 2007). These dimensions emerge in classic person perception (Asch, 1946; Rosenberg, Nelson, & Vivekananthan, 1968), social-value orientations (e.g. self- and other- profitability; Peeters, 2002), construals of others’ behaviours (Wojciszke, 1994), and voters’ ratings of political candidates (Kinder & Sears, 1981). In fact, they account for 82% of the variance in global impressions of other individuals (Wojciszke, Bazinska, & Jaworski, 1998). Related dimensions also describe national stereotypes and prejudices towards specific groups (both reviewed below). Data from US surveys illustrate how stereotypes array on the two dimensions (Figure 1) and support SCM hypotheses (Cuddy, et al., 2007; Fiske, et al., 2002).

Open in a separate window

Figure 1

Group competence–warmth stereotypes (Fiske et al., 2002, Study 3). Key: Stars indicate cluster centres. H, M, and L, respectively, indicate high, medium, and low; W, warmth; C, competence.

Ambivalent stereotypes

Support for SCM’s ambivalent stereotypes hypothesis – that many groups are tagged as proficient in one sphere (i.e. either warmth or competence) and inferior in the other – stems from stereotypes of specific social groups. Two types of stereotyped groups materialize in this literature – those viewed as kind but helpless, and those viewed as skillful but cunning. ‘Envious’ prejudice targets the latter, seen as threateningly competent and untrustworthy (Fiske, et al., 2002; Glick, 2002; Glick & Fiske, 2001a, 2001b).¹ For example, non-traditional women, such as career women and feminists, allegedly possess agentic but not communal traits, and are respected but disliked – embodying envious prejudice (Eagly & Kite, 1987; Glick & Fiske, 1996, 2001a; MacDonald & Zanna, 1998; Spence & Helmreich, 1979). Envious prejudice also targets Asian Americans and other “model minorities”, stereotyped as excessively competent (too ambitious, too hardworking) and lacking sociability (Ho & Jackson, 2001; Hurh & Kim, 1989; Kitano & Sue, 1973; Lin, Kwan, Cheung, & Fiske, 2005; Maddux, Galinsky, Cuddy, & Polifroni, 2008). Similarly, stereotypes of Jews combine business acumen with interpersonal self-interest (Allport, 1954; Glick, 2002).

Groups seen as benevolent but incapable of competing in mainstream society sit in the opposite corner of the map. This ‘pitying’ prejudice reflects liking but disrespect (Glick & Fiske, 2001a, 2001b; Jackman, 2001) and often targets traditional women (Cuddy, Fiske, & Glick, 2004; Glick & Fiske, 1996, 2001a) and older people (Cuddy & Fiske, 2002; Cuddy, Norton, & Fiske, 2005; Heckhausen, Dixon, & Baltes, 1989; Kite, Deaux, & Miele, 1991), both perceived as high on communal, but low on agentic traits.

Still, some groups receive evaluatively consistent stereotypes. Groups perceived as both hostile and indolent most elicit the antipathy traditionally associated with derogated groups. Conversely, in-groups and mainstream social groups are favoured as both warm and competent, eliciting pride and admiration (Fiske, et al., 2002).

Overall, these four clusters of stereotype content, defined in the competence £ warmth space, should be universal. Our first goal examines this universality of content across cultures and groups.

Social structural correlates of stereotypes

All complex societies organize hierarchically and compete for resources. These structural relations forecast groups’ location on the SCM’s competence × warmth space. Stereotypes are shaped by perceived and actual economic, geographic, normative, and power relationships (Eagly & Kite, 1987; LeVine & Campbell, 1972; Linssen & Hagendoorn, 1994; Poppe, 2001). In the SCM, judged warmth and competence stem from appraisals of, respectively, (a) the potential harm or benefit of the target’s intent and (b) whether the target can effectively enact that intent. In the stereotype space, competitors lack warmth, while non-competitors are warm; high status people are competent, while low status people are incompetent. Besides US surveys (Cuddy, et al., 2007; Fiske, et al., 2002), the structural predictors appear in experiments on intergroup perception (Caprariello, Cuddy, & Fiske, 2007; Oldmeadow & Fiske, 2007) and interpersonal perception (Russell & Fiske, 2007).

Social structure determines competence and warmth throughout the interpersonal and intergroup perception literatures. Impression-formation research (Wojciszke, 2005) demonstrates that perceived status predicts perceived competence, and self-interest (akin to competition) predicts perceived morality (akin to warmth). Similarly, European nations’ perceived economic power (i.e. status) predicts perceived competence, and conflict (i.e. competitiveness) predicts perceived warmth (Poppe, 2001).

People link perceived traits with social structure for several reasons. First, they legitimate unfair social structures over which they feel they have no control (for a review, see Glick & Fiske, 2001a). For example, just-world thinking leads people to view outcomes as deserved (Lerner & Miller, 1978): groups with high status, well-paying jobs must have earned them through talent and hard work. And indeed, the status–competence correlation varies by individual differences in just-world beliefs (Oldmeadow & Fiske, 2007). Similarly, system-justification legitimates group-level sociopolitical and socio-economic inequalities (Jost & Banaji, 1994). Superordinate groups justify their advantage by viewing the status quo as fair, and even subordinate groups may endorse this view because it explains their own outcomes. In short, status predicts competence stereotypes because it justifies the apparent meritocracy. Competition negatively predicts warmth stereotypes because it can exclude groups with goals that conflict with those of the in-group. If they are not nice, why include them?

Additionally, people may simply infer a group’s traits from their social position. In interpreting behaviour, westerners tend to over-use internal dispositions, ignoring the influence of the situation (Gilbert & Malone, 1995; Jones, 1979; Ross, 1977). Thus, when a group is supposedly overrepresented in high status positions, people may attribute this to the group’s perceived competence. Likewise, when a group is viewed as competing for resources, such as tax dollars, people may attribute this behaviour to the group’s alleged malice. Thus, our first goal also addresses potentially universal principles in predictors of stereotypes.

Out-group derogation and reference-group favouritism

Our second goal revisits a heretofore universal principle of prejudice, namely that in-group favouritism underlies out-group derogation (Allport, 1954; Brewer & Brown, 1998; Sumner, 1906). The SCM can re-examine this idea across societies, by comparing societal in-groups, relative to societal out-groups. If many out-groups are ambivalent, as the SCM predicts, then they may be positive on at least one dimension, calling into question the necessity of absolute in-group superiority. Because of potential cultural differences in views of the in-group (see below), a cross-cultural comparison allows us to explore this question.

Our societal level of analysis revives a construct with a rich history in social psychology: reference groups (e.g. Hyman, 1942; Merton, 1957). In US SCM maps, the only unambivalently positive stereotypes describe reference groups, including personal in-groups (e.g. students) and societal prototype groups (e.g. Christians, middle-class, Whites). Defined as ‘psychologically significant for one’s attitudes and behaviour’ (Turner, 1991, p. 5), reference groups are those with which an individual identifies, and are often, but not always, in-groups (Allport, 1954; Mackie & Wright, 2001; Turner, 1991). They also can be societal prototype groups – valued mainstream groups that are not necessarily in-groups (e.g. middle-class and Whites in the US). Reference groups have theoretically served as normative standards for social comparison and most often, social aspiration (Hyman, 1942; Merton, 1957; Sherif & Sherif, 1953). Because the current studies ask participants to report societal stereotypes, we consider reference groups rather than demographic in-groups. The term ‘reference group’ applies to both valued personal in-groups (e.g. students) and to societal prototype groups.

Reference-group favouritism – preferring personal in-groups (Allport, 1954; Brewer & Brown, 1998; Fiske, 1998; Sumner, 1906) or societal prototype groups (cf., Jost & Burgess, 2000) – has always seemed to underlie bias, more than out-group derogation. In the US, the Protestant work ethic exhorts people to work hard towards goals (Weber, 1905/1958); everyone can be optimistic about assimilating to the mainstream if some reference groups are apparently ideal (i.e. both warm and competent). The SCM predicts reference groups to lodge in the high–high cluster, separated from other groups, and the US data consistently support this (Cuddy, et al., 2007; Fiske, et al., 2002). Goal 2 explores the role of reference-group favouritism in out-group derogation.

How culture affects perceivers

Countless cultural differences in how people perceive the world could undermine potentially universal principles of stereotyping. Within this vast literature, a few findings could limit the SCM’s potentially universal principles: category use; dispositional bias; legitimating ideologies; valued traits; multiculturalism; and most central, individualism- collectivism (IC).

First, culture affects basic cognitive processes (Nisbett, Peng, Choi, & Norenzayan, 2001; Norenzayan, Smith, Kim, & Nisbett, 2002). For instance, compared with East Asians, North Americans tend to think more with the aid of categories (Chiu, 1972; Choi, Nisbett, & Smith, 1997; Norenzayan, et al., 2002), a social-cognitive process at the root of stereotyping (Fiske, Neuberg, Beattie, & Milberg, 1987; Taylor, Fiske, Etcoff, & Ruderman, 1978). If people in non-US cultures think less with categories, a skeptic might propose that they are less likely to fall prey to the social consequences of categorical thinking (i.e. stereotyping), thus rendering SCM dimensions inapplicable.

Second, North Americans more than members of other cultures attribute people’s behaviours and outcomes (competitiveness and status) to dispositions (warmth and competence) as opposed to situations (Choi, Nisbett, & Norenzayan, 1999; Miller, 1984; Morris, Nisbett, & Peng, 1995; Morris & Peng, 1994). As noted, the perceived relationship between social structure and stereotypes might rest on this attributional bias – people earn high status because they are competent, and people who are competitive must also be cold. According to the skeptic, cultures without a strong dispositional bias might not use structural outcomes to infer traits.

Third, culture shapes the ideologies that legitimate prejudice (Cohen & Nisbett, 1994; Crandall & Martinez, 1996; Crandall, et al., 2001; Glick, et al., 2000). Cultural ideologies govern what people see as good and bad, thereby stipulating which groups will become the targets of prejudice. For example, North Americans score particularly high in just-world thinking (Loo, 2002), which provides moral justifications for good and bad outcomes. Especially relevant to the perceived relationship between status and competence, cultures lower in just-world thinking might not perceive the same link.

Fourth, cultures come with different values that influence which traits reflect goodness, morality, and social acceptance (e.g. Markus, Mullally, & Kitayama, 1997; Morling, Kitayama, & Miyamoto, 2002; Sedikides, Gaertner, & Toguchi, 2003; Wheeler & Kim, 1997). For example, southern Mediterranean cultures (e.g. France, Greece, and Spain) especially value characteristics that reflect excitement-seeking and appreciation of novelty (Aaker, Benet-Martinez, & Garolera, 2001). Contrary to universality, some societies might not value highly enough SCM’s traits – competence and warmth – to use them in forming stereotypes.

Fifth, a culture’s political ideals influence how its citizens perceive out-groups. Political correctness, for instance, might lead US citizens to bestow on out-groups a partially positive stereotype, increasing the frequency of ambivalent stereotypes. Political correctness results in part from the US status as an immigrant society, with the world’s longest continuous history of receiving immigrants (Dovidio & Esses, 2001). Thus, multicultural ideals also infuse US political dialogue, making diversity an objective (Berry, 2001), which might lead US perceivers to differentiate groups on more than one evaluative dimension. A skeptic could argue that this uniquely US context underlies the SCM.

In short, ample evidence would suggest to a skeptic that cultural influences should preclude universal principles of stereotyping. The preceding five issues all predict cultural differences at least between US and East Asian perceivers, consistent with Goal 3.

Finally, cultural analyses raise an issue consistent with both Goals 2 and 3: the generality of reference-group favouritism for collectivist cultures. Interdependent (i.e. collectivist) selves do not self-promote to the same extent as independent (i.e. individualist) selves (Kitayama, Markus, Matsumoto, & Norasakkunkit, 1997), to ensure social harmony (Markus & Kitayama, 1991; Triandis, 1995). Collectivists self- enhance less than individualists (Heine, Lehman, Markus, & Kitayama, 1999; Kitayama, et al., 1997), rate failures as more self-relevant (Heine et al., 2001; Lee & Seligman, 1997), and self-criticize more (Kitayama, et al., 1997). This might generalize to the group level (Heine & Lehman, 1997; cf. Muramoto & Yamaguchi, 1997; but also see Hewstone, Bond, & Wan, 1983; Taylor & Jaggi, 1974). In-group favouritism may diminish in collectivist settings, befitting cultural ideals of maturity as modest, humble, and self-improving. Group ideals may depend on moderation: What is average is good. Collectivist cultures show less personal optimism (Heine, et al., 1999; Markus & Kitayama, 1991) and less interpersonal positivity bias (Yamagishi, Cook, & Watanabe, 1998), perhaps particularly towards reference groups. Considerable evidence thus suggests that more collectivist cultures may not uniformly favour reference groups. If so, out-group derogation might also weaken in these cultures.

A variety of perspectives suggest cultural differences. But much of this research compares North American (more individualist) with East Asian (more collectivist) perceivers. Most cultural psychology assumes that other western perceivers, namely Europeans, would resemble North Americans, in contrast to East Asia. These are empirical questions we explore here.

Method

Questionnaire and procedure

The questionnaire listed the then-current 15 EU nations (Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, The Netherlands, Portugal, Spain, Sweden, and UK). Translations for non-Anglophones were satisfactory in independent back-translations.

Participants read, ‘We are studying how different European countries are perceived by citizens of the European Community (EC). We are not interested in your personal beliefs, but in how you think citizens of the EC view these countries. Using the scale below, write the number that best represents how you think these countries are viewed by citizens of the EC’. The instructions aimed to reduce social desirability concerns and to draw on perceived cultural stereotypes, as in our earlier work (e.g. Cuddy, et al., in press; Fiske, et al., 2002). Participants rated the 15 groups on scales of competence, warmth, status, and competition (see Appendix A for items and reliabilities).

Appendix A

Scales, both studies

Competence

As viewed by society, how… are members of this group?

Competent, confident, capable, and skillful

Warmth

As viewed by society, how…are members of this group?

Friendly, warm, good-natured, and sincere

Status

How prestigious are the jobs typically achieved by members of this group?

How economically successful have members of this group been?

How well-educated are members of this group?

Competition*

How much does special treatment given to this country make things more difficult for others in the EC?

How much do market resources that go to citizens of this country take away from the market resources of others in the EC?

As this country gains power, to what extent do others in the EC lose power?

Open in a separate window

Note. All items used five-point scales (1 = not at all, 5 = extremely). Study 1 reliabilities were sufficient for all scales in all samples: competence α = .67–.85; warmth α = .67–.83; status α = .69–.84; competition α = .59–.75 (The Netherlands α = .40).

^*After Study 1, Study 2 restored the standard competition scale items: If members of this group get special breaks (such as preference in hiring decisions), this is likely to make things more difficult for people like me. Resources that go to members of this group are likely to take away from the resources of people like me. The more power members of this group have, the less power people like me are likely to have.

Results

Competence and warmth scores for each target group (omitting in-group ratings) were first averaged across participants within samples, then across samples.

Hypothesis 1: Competence–warmth

Two types of cluster analysis examined whether stereotypes of groups array along the competence and warmth dimensions. First, hierarchical cluster analysis (Ward’s method, minimizing within-cluster variance and maximizing between-cluster variance) determined an appropriate number of clusters. Second, K-means cluster analysis (parallel threshold method) assigned each group to a cluster (Hair, Anderson, Tatham, & Black, 1995).²

We validated the hierarchical cluster analyses’ most credible number of clusters by specifying a decision rule and checking that solution independently (Blashfield & Aldenderfer, 1988). First, t tests examined relationships among clusters on the two dimensions. Second, separate variables (status, competition) also differentiated the competence–warmth clusters.

For the combined data, the cluster analysis pointed to a three-cluster solution (see Figure 3, which also includes separate in-group ratings). As predicted, the clusters spread across two dimensions, using both equally. Respondents everywhere used the full range (1–5, standard deviations 0.60–0.84, median = 0.70). Warmth and competence had equivalent ranges and variances, as in the US. If either dimension had not sorted groups, the cluster analysis would yield a single-cluster solution, reflecting no differentiation on one dimension. Moreover, correlations will further support the dimensions’ validity by providing two separate variables (status, competition) that correlate differentially with competence and warmth. Hypothesis 1, the utility of the competence × warmth space to differentiate these groups, is supported.

Open in a separate window

Figure 3

Study 1, EU nations, three-cluster solution. Key: Stars indicate cluster centres. In-group ratings, which are not reflected in the aggregated ratings, are separate, italicized, and in larger font. HC–LW, high-competence/low-warmth; HHC–LLW, highest-competence/lowest-warmth; LC–HW, low-competence/high-warmth.

Hypothesis 2: Ambivalent stereotypes

Three analyses tested the ambivalent stereotypes hypothesis (i.e. that many groups are stereotyped as significantly higher on one dimension than on the other): (a) independent samples t tests comparing warmth and competence between clusters; (b) paired t tests comparing warmth and competence within clusters; and (c) paired t tests comparing warmth and competence within groups. To be identified as ambivalent (high-competence/low-warmth or low-competence/high-warmth), clusters had to meet two conditions: (1) warmth and competence means differed significantly and (2) the mean for their high dimension was higher than groups low on that dimension, and the mean for their low dimension was lower than groups high on that dimension.

Table 2 provides cluster warmth and competence means. Cluster HHC–LLW, highest on competence and lowest on warmth, comprised two groups: Germany and UK. This cluster was marginally more competent than warm, diff = 1.28, t(1) = 8.32, p = .07, (surely the marginal p value was due to the fact that the analysis had only one degree of freedom). It scored higher on competence than the other two clusters, both ps < .01, and lower on warmth than the other two clusters, both ps < .01. By our definition, this cluster embodied an ambivalent stereotype.

Table 2

Competence and warmth means for each cluster; both studies

Sample	Competence		Warmth
Study 1: EU nations combined (N = 755)
HC–LW	3.49b	>	3.10b
HHC–LLW	4.01a	>*	2.73c
LC–HW	3.13c	<	3.53a
Study 2: Hong Kong (N = 60)
HC–LW	4.28a	>	2.65c
LC–HW	2.52c	<	3.65a
MC–MW	3.35b	=	3.10b
LC–LW	2.34c	=	2.55c
Study 2: Japan (N = 83)
HHC–LLW	4.55a	>	2.68b,c
HC–LW	3.41b	>	2.94b
LC–HW1	3.04c	<	3.51a
LC–LW	2.34d	<	3.03b
LC–LW	1.60e	<	2.31c
Study 2: South Korea (N = 91)
HC–LW1	3.75a	>	2.82b
HC–LW2	3.50a	>	2.21c
LC–HW	2.98b	=	3.20a
LC–LW	2.14c	=	2.56b,c

Open in a separate window

Note. Within each row, within each sample, means differ (p < .05)if>or < is indicated; for *p = .07 (but note that this analysis had only 1 df). Within each column, means not sharing a superscript differ (p <. 05). Key: HC–LW, high-competence/low-warmth; HHC–LLW, highest-competence/lowest- warmth; LC–HW, low-competence/high-warmth; MC–MW, middle-competence/middle-warmth; LC–LW, lower-competence/lower-warmth; LLC–LLW, lowest-competence/lowest-warmth.

Cluster HC–LW included Austria, Belgium, Denmark, Finland, France, Luxembourg, The Netherlands, and Sweden. This cluster scored higher on competence than on warmth, diff = .39, t(7) = 7.48, p < .001, lower on competence than cluster HHC–LLW, t(8) = 4.31, p < .01, but higher on competence than cluster LC–HW, t(11) = 3.44, p < .01. Its warmth score was higher than cluster HHC–LLW, t(8) = 4.24, p < .01, and lower than cluster LC–HW, t(11) = 5.53, p < .001. Thus, this cluster embodies an ambivalent stereotype.

Cluster LC–HW included Greece, Ireland, Italy, Portugal, and Spain. It scored significantly higher on warmth than on competence, diff = .40, t(4) = 3.95, p < .05, outscored both other clusters on warmth (ps < .01), and scored lower than both other clusters on competence (ps < .01). This cluster also embodies an ambivalent stereotype.

Also supporting the ambivalent stereotypes hypothesis, paired t tests revealed that 13 of 15 (87%) groups differed significantly on competence and warmth, all ps < .05. Nine groups were rated more competent than warm (highest to lowest difference): Germany; UK; France; Austria; The Netherlands; Luxembourg; Sweden; Denmark; and Finland. Four groups were rated as more warm than competent (highest to lowest difference): Portugal; Greece; Spain; and Italy. Belgium and Ireland did not differ on warmth and competence.

Paired t tests compared competence and warmth means within perceiver nation. Differences were interpreted as significant for p values of < .01, to correct for possible family-wise error. The results identified 4 (France) to 11 (Portugal, Spain) high-competence/low-warmth groups. Low-competence/high-warmth groups ranged from 2 (Portugal) to 5 (Belgium, UK). Across samples, 53 (France) to 100% (Spain) of groups differed significantly on competence and warmth.

Method

Results

Hypotheses 1 and 2: Competence–warmth space and ambivalent stereotypes

Each group received competence and warmth scores averaged across scale items and participants. Because each sample rated its own groups, we were not able to collapse across samples, as in Study 1. Following the same cluster analysis procedure used in Study 1, cluster solutions ranged from 4 (Hong Kong, South Korea) to 5 clusters (Japan). All samples recovered a low–low cluster, unlike EU data, but like previous US data. All samples recovered the two ambivalent clusters, at least one HC–LW and one LC–HW cluster. Everywhere, at least half the groups received ambivalent stereotypes.

Hong Kong

In the Hong Kong sample, a four-cluster solution best described the data (Figure 4b). The solution included one HC–LW cluster, one LC–HW cluster, one middle- competence/middle-warmth (MC–MW) cluster, and one LC–LW cluster (see Table 2).

Open in a separate window

Figure 4

(a) Study 2, Hong Kong, four-cluster solution. Key: Stars indicate cluster centres. HC–LW, high-competence/low-warmth; LC–HW, low-competence/high-warmth; MC–MW, middle competence–middle warmth; LC–LW, low-competence/low-warmth. (b) Study 2, Japan, five-cluster solution. Key: Stars indicate cluster centres. HC–LW, high-competence/low-warmth; LC–HW, low-competence/high-warmth; HHC–LLW, highest-competence/lowest-warmth; LC–LW1, low- competence/low-warmth 1; LC–LW2, low-competence/low-warmth 2. (c) Study 2, South Korea, four-cluster solution. Key: Stars indicate cluster centres. HC–LW1, high-competence/low-warmth 1; HC–LW2, high-competence/low-warmth 2; LC–HW, low-competence/high-warmth; LC–LW, low-competence/low-warmth.

Fifty-nine percent of the groups (16 of 27) differed on warmth and competence ratings at p < .01. Eight were rated as significantly more competent than warm (in descending order by difference scores): rich; professionals; white-collar; Hong Kong locals; men; adults; married; and single. Eight were rated as significantly more warm than competent (in descending order by difference scores): children; mentally disabled; elderly; Christians; janitors; poor; students; and unemployed. Eleven groups – Asians, blue-collar, Chinese, Filipino maids, foreigners, immigrants, undergraduates, Mainlanders, Pakistanis, women, and youths – received equal ratings on warmth and competence.

Japan

In the Japanese sample, a five-cluster solution best fit the data, containing one HHC–LLW cluster, one HC–LW cluster, one LC–HW cluster, and two LC–LW clusters (see Table 2 and Figure 4c). Eighty-six percent of the groups (18 of 21) differed significantly on competence and warmth ratings and p < .01. Eight were rated as significantly more competent than warm (in descending order by difference scores): professionals; (my) university; civil servants; entertainers; full-fledged members of society; men; (my) company; and students. Ten were rated as significantly more warm than competent (in descending order): children; disabled; senior citizens; poor; homeless; (my) family; (my) friends; women; (my) clubs; and odd-jobbers. Only three groups – (my) hometown, Japanese, and office workers – did not differ.

South Korea

A four-cluster solution best fit the South Korean data, with two HC–LW clusters, one LC–HW cluster, and one LC–LW cluster (see Table 2 and Figure 4d). Seventy-eight percent of the groups differed significantly on warmth and competence at p < .01: Nine were rated as significantly more competent than warm (in descending order by difference scores): rich; professionals; intellectuals; employers; merchants; teachers; public functionaries; men; and employees. Nine were rated as significantly more warm than competent (in descending order by difference scores): children; elderly; poor; illegal immigrants; housewives; unemployed; women; Buddhists; and ministers. Five did not differ: middle-class; protestants; college students; blue-collar workers; and junior high students.

Summary

Both warmth and competence differentiated the groups and did so equally. All samples used the scales’ full 1–5 range, and although standard deviations for competence were slightly higher than for warmth (M diff = 0.07), the difference was less than comparable US data (M diff = 0.19). Thus the competence × warmth space describes collectivists as well as individualists, consistent with Hypothesis 1. Moreover, ambivalent groups dominate both, consistent with Hypothesis 2.

Hypothesis 4: Favouritism–derogation

One striking cultural difference emerged. Collectivists rated reference groups, including societal prototype groups (e.g. college graduates, full members of society, educated, middle-class) and clear-cut in-groups (e.g. students), less favourably than did US (Cuddy, Fiske, & Glick, in press; Fiske, et al., 2002), German (Eckes, 2002), and Belgian respondents (preliminary study). Reference groups, occupying individualists’ HC–HW quadrants, migrated towards the collectivists’ middle (Figure 4a–4c). Some reference groups occupied the HC–LW clusters (e.g. Japan: my university, students); some occupied the LC–HW clusters (e.g. Japan: family, friends, Japanese, my clubs; South Korea: college students, middle-class), and some occupied a MC–MW cluster (e.g. Hong Kong: Asians, Chinese, students). Not one collectivist sample included a HC–HW cluster, the spot typically reserved for reference groups in western samples.

A more fine-grained analysis isolated one group guaranteed to be in-group for all samples, namely students. Table 3 compares these three collectivist samples with three individualist samples that rated similar groups. The individualists rate students significantly higher than the average other group (ps < .01), whereas none of the collectivists do. Collectivists demonstrating less reference-group favouritism fits well-documented IC differences concerning self. Thus, midward migration of reference- group clusters fits collectivists being less likely to rate themselves, their in-groups, and others too positively.

Table 3

Reference-group favourtism ratings, all samples

Sample	Competence	Warmth	General positivity (warmth + competence mean)
Collective samples
Hong Kong	−.06	.39*	.17
Japan	.10	−.07	.02
Korea	.19	.17	.18
Individualist samples
Belgium	.36*	.51*	.44*
United States (Fiske, et al., 2002, Study 2)	.39*	.17*	.29*
United States (Cuddy, Fiske, & Glick, in press)	.72*	.46*	.49*

Open in a separate window

Note. Values represent difference between the mean for the in-group ‘students’ and the mean for all other groups in that sample. High numbers indicate higher scores for students (i.e. more reference-group favouritism).

^*p < .01

These samples show clear out-group derogation on one dimension (LC or LW) or both (LC–LW), even without reference-group favouritism. Out-group derogation thus might not require explicit reference-group favouritism, to our knowledge, the first demonstration of this.

Towards pancultural principles of stereotyping

Across culturally varied perceivers and targets, cluster analyses corroborated the two dimensions hypothesis that perceived competence and warmth differentiate stereotypes. Comparing warmth and competence – within groups, within clusters, and between clusters –supported the ambivalent stereotypes hypothesis that many out-groups receive ambivalent stereotypes as competent but not warm or as warm but not competent. Across studies and samples, most groups (59–100%) significantly differed on warmth and competence; both high-warmth/low-competence and low-warmth/high-competence combinations occurred everywhere.

Correlational analyses supported the social structural correlates hypotheses that high status groups are perceived as competent and non-competitive groups are perceived as warm. Everywhere, status strongly correlated with competence; mostly, competition significantly, negatively correlated with warmth.

Warmth and competence may not be the only dimensions of stereotypes; we argue they are two central, theoretically meaningful, predictable dimensions. What’s more, they appear to be fundamental dimensions of social perception in general (Cuddy, et al., 2008; Fiske, et al., 2007). This claim builds on their observed prevalence in person perception and stereotypes, theoretical functions related to social structure, and relationships to intergroup emotions and behaviours, as we elaborate below. The current data fortify this claim by demonstrating the dimensions across cultures, perceivers, and groups.

The result was not a priori obvious. We introduced the cultural comparison by describing previous evidence of western versus Asian differences in category use, dispositional bias, legitimating ideologies, valued traits, multiculturalism, and, most central here, IC. The exploration of individualist European settings was more exploratory, as large cultural differences have not been identified. We found no evidence that the Asian–western contrast indicated any of the following: less category usage (consensual images of groups prevailed); less dispositional bias (structure-trait correlations replicated); legitimating ideologies (ditto); differently valued traits (warmth and competence ranges were comparable); or multiculturalism (ambivalent stereotypes appeared everywhere). We did find evidence consistent with collectivist modesty, locating the societal reference groups in a more moderate, humble location instead of the in-group-enhancing high–high quadrant, as in the western within-country samples.

Despite the main similarities in use of warmth and competence, these dimensions might be construed idiosyncratically to complement cultures’ social values. For instance, warmth entails social harmony, central to collectivists, while competence entails individual-focused success, central to individualists. However, warmth also conveys social skill, necessary for individual-goal advancement. Likewise, competence conveys information about capability at helping in-groups achieve collective goals. These data do not preclude emic (culture-specific) construals of warmth and competence.

A critic might argue that our trait ratings are descriptive, but not evaluative. Our US research suggests otherwise. Warmth and competence ratings correlate strongly and reliably with distinct emotional prejudices (Cuddy, et al., 2007; Fiske, et al., 2002). Each quadrant carries a unique signature of prejudiced emotions: admiration; contempt; envy; and pity. To recap the introduction briefly, envy targets groups with supposedly controllable, coveted outcomes (e.g. economic success; Parrott & Smith, 1993); envy indeed hits high-competence/low-warmth groups. Pity targets groups with apparently uncontrollable stigmas (e.g. age, disability; Weiner, Perry, & Magnusson, 1988); pity indeed hits incompetent-warm groups. Those whose positive outcomes reflect on self receive admiration and pride (Alexander, et al., 1999; Smith, 2000), as do high-competence/high-warmth groups. Finally, contempt targets groups with allegedly avoidable negative outcomes (Barnett, Quackenbush, & Pierce, 1997; Weiner, et al., 1988); contempt indeed targets low-competence/low-warmth groups. Moreover, these intergroup emotions link to unique patterns of intergroup behavioural tendencies (Cuddy, et al., 2007). At least in the US, warmth and competence stereotypes relate to emotional and behavioural aspects of bias.

Cross-cultural stereotyping differences

Collectivist samples consistently lacked clear high-competence/high-warmth clusters. No personal in-groups or societal prototype groups were uniformly favoured over other societal groups. Nevertheless, these cultures demonstrated the same relative out-group derogation as individualists.

In the EU context, European samples neither nominated any exemplary shared reference group, nor consistently favoured personal in-groups. Nevertheless, they too derogated out-groups on either competence or warmth. Some have suggested that reference-group favouritism does not necessarily lead to out-group derogation (Brewer & Brown, 1998; Mummendey & Wenzel, 1999), so in-group favouritism is not a sufficient condition for prejudice. We show here that in-group favouritism is also not a necessary condition for prejudice. To our knowledge, these two studies are among the first to demonstrate out-group derogation without reference-group favouritism.

We explain the absence of the HC-HW societal-reference-group quadrant by two related reasons in these two samples. We suggest that collectivist cultures moderate the positive attributes of the societal in-group/reference groups, in keeping with East Asian well- documented modesty, humility, and harmony norms described earlier. Similarly, EU citizens deny any single country the status of EU in-group/reference group, to maintain the strict ideology of equality within the Union – so necessary to EU harmony. We do not suggest that EU nations, within their own borders are collectivist, because cultural psychology finds no evidence for this. But, in an apparent violation of parsimony, we note that they can follow collectivist norms in the EU setting as equal members, none being privileged over the others, at the same time as being western individualists towards groups within their own borders.

The SCM framework identifies basic principles that help explain cultural idiosyncrasies in stereotype contents within each country. For example, US respondents (Fiske, et al., 2002) stereotyped economically disadvantaged groups (i.e. welfare recipients) as lacking both competence and warmth, whereas Belgian respondents charitably viewed them as less competitive/exploitative, therefore as warm although still not very competent. Likewise, Blacks, blue-collar workers, and gay men gained both warmth and competence among Belgian respondents likely being perceived as less competitive and higher-status than in the US.

We thank Tristan Colangelo, Laura Heller, Lisa Leslie, Liza Moldonado, Lisa Pugh, Mark Rogers, and Lisa Taylor for help with data entry. We also thank Leticia Garcia and Diana Raysz for help collecting data.

We would especially like to thank our colleague, Mariko Yamamoto of the University of Tsukuba, who contributed to the manuscript but sadly passed away before it was published.