Stereotypes

Stereotypes occupy a venerable position in research on prejudice. Their invidious character has been so well established that the term immediately conjures up trait attributions to groups that are categorical, rigid, and invective. Stereotypes have been inextricably bound to the concept of prejudice—they are "the language of prejudice" (Ehrlich 1973, 21).

In identifying those aspects of stereotypes that are especially damaging to intergroup relations, analysts have focused primarily on the following: the categorical attribution of personality traits to a group, the sharp setting-apart of a subordinate group from the dominant group, the derogation of subordinates, and the reliance on erroneous information (folk knowledge) rather than rational, scientific information about a group. These properties have been regarded as so endemic to stereotyping that researchers have felt no compunction to assess their presence empirically in standard measures of stereotypes. Instead, these properties were simply loaded into the two prevailing measures of stereotypes (stereotype checklists and agree-disagree categorical statements) rather than providing subjects with the opportunity to express more modified intergroup beliefs. I briefly discuss analysts' concern with each of the key properties of stereotypes, in turn, and then outline how the standard measures of stereotypes have served to reinforce, rather than to test, common assumptions about the content and form of damaging intergroup beliefs.

Categorical Attributions

For many scholars, the description of groups in categorical terms is perhaps the most fundamentally damaging feature of stereotypes. Most definitions of stereotypes emphasize categorical attributions as a critical element (for example, Allport 1954; Fishman 1956; Richter 1956; Vinacke 1957; Secord 1959; Koenig and King 1964; Williams, Jr. 1964; Sherif 1966, 38–41; Campbell 1967; Harding et al. 1969; Tajfel 1969, 1982; Cauthen, Robinson, and Krauss 1971; Brigham 1971; Mackie 1973; Hamilton 1981; Hamilton and Trolier 1986). The tendency to simplify and classify the world is often regarded as a basic perceptual operandum as well as a functional coping mechanism for bringing some order to the confusion of reality (Lippmann 1922; Allport 1954; Tajfel 1969; Ehrlich 1973, 38; Hamilton and Trolier 1986). In

the context of intergroup beliefs, however, this process is believed to become emotionally invested, exaggerated, and rigidified.

Use of the term stereotype to denote trait attributions to groups is borrowed from the printing industry where it means "a one-piece printing plate cast in type metal from a mold taken of a printing surface, as a page of set type" (Webster's New World Dictionary 1968). Stereotypes thus preempt raw perception with a rigidly fixed stamp. The categorical perception of groups "provides the mold which gives shape to intergroup attitudes" (Tajfel 1969, 91).

A frequent line of reasoning has been that any perception that unequivocally lumps all the members of a group into one unvaried category is intrinsically damaging because it credits group membership with an overwhelming role in determining individual personality characteristics, and because it provides the perfect grounding for discriminatory policies and practices. Most important, the categorical description of a group represents a closed and insensitive perception that routinely pigeonholes group members and leaves the perceiver unreceptive to contrary evidence, which is either overlooked or dismissed as a mere "exception." Categorical descriptions of groups are thus seen as inherently inaccurate and irrational (Lippmann 1922; Allport 1954; Richter 1956; Williams, Jr. 1964, 36; Ehrlich 1973; Mackie 1973; Tajfel 1969, 1978; Prager 1982). Thus, although there has been more interest in categorical attributions that are explicitly derogatory, any kind of categorical attribution is regarded as damaging, whether the specific traits are positive, negative, or even vague. An extreme example of a belief item that reflects this perspective is found in Wilner et al.'s study (1955, 62): "There may be a few exceptions, but in general, Negroes are pretty much alike." A similar item is found in the ethnocentrism scale of Adorno et al. (1950).

It is frequently assumed that such categorical attributions are more probable when people are describing groups other than their own, especially groups with which the individual has little or no personal familiarity (for example, Allport 1954; Pettigrew 1982; Hamilton and Trolier 1986; Linville, Salovy, and Fischer 1986; Miller and Brewer 1986; Hewstone 1989). This assumption ties in with the recurrent theme that prejudice is a product of ignorance. It is assumed that familiarity would yield the inescapable observation of variation within the target group, and categorical attributions would thus crumble.

Setting Groups Apart and Derogating Subordinates

A second feature of stereotypes that has been regarded as damaging is the attribution of positive traits to one's own group and negative qualities to other groups. Such attributions set groups apart and assert the superiority of one group over another in a single maneuver (Vinacke 1957; Allport 1954; Williams

1964, 40; Sherif 1966, 37–38; Campbell 1967; Harding 1968; Tajfel 1969, 1978; Eldridge 1979, 23–27). In Tajfel and Turner's theory of social identity and social comparison, it is postulated that individuals seek to enhance their self-esteem by comparing themselves favorably to other individuals and their group favorably to pertinent out-groups: this poses a problem for subordinated groups, whose self-esteem is damaged by their position as the object of consensual dominant derogation (Tajfel and Turner 1979).

The extent to which people distinguish between the basic personality attributes of their own and other groups provides the meaning for group labels. In addition, such distinctions carry an implicit evaluation of each group's worth and serve as part of an attributional process that accounts for the varying fates of different groups. A positive self-image is believed to come easily to dominant groups as a product of their privileged position in society; besides, assertions of their superior endowment of personality traits constitute a convenient rationale for their privileged position. Thus, the common assumption has been that dominant groups will assert large and invidious distinctions between their own and subordinate groups as part of a logical attempt to vindicate the status quo unequal relations between them.

The extent to which subordinate groups accede to or deny dominant group assertions of superiority has been regarded as more problematic. This is a critical question, since the former implies a consensus about fundamental perceptions that bolster the status quo, whereas the latter implies a rift between groups that would provide a force for social change. It is generally assumed that the dominant view, because it emanates from the group with more power and prestige, has the advantage, and that it is difficult for subordinate groups to counter with claims of their own superior or even equal endowment of personal attributes. The extent to which a subordinate group does challenge the dominant group's definition of the distinctions between them gives a fundamental indication of their development of positive group consciousness as well as of the dominant group's ideological control over the relations between groups.

Erroneous Beliefs

A common but often unspoken assumption is that stereotypes are erroneous, and that a good part of the damage they wreak comes from their unscientific, irrational genesis. Students of prejudice generally place implicit faith in rational, scientific knowledge as the enlightened replacement for parochial folk knowledge. In Allport's initial delineation of prejudice, he illustrates the problem by drawing a pivotal distinction between two different types of avoidance behaviors: his first case is the anthropologist who does not allow his children to have

contact with the American Indian tribe he is studying, and his second case is the hotel proprietors who refuse accommodation to "Mr. Greenberg" because of his Jewish name. Allport reasons that the anthropologist's behavior is untainted by prejudice because it is based on a realistic assessment of the risk of his children contracting tuberculosis from the Indian children in the village, whereas the hotel proprietors are acting out of prejudice because their behavior is based on an irrational, categorical rejection.

If the innkeepers were basing their rejection on facts (more accurately, on a high probability that a given Jew will have undesirable traits), their action would be as rational and defensible as the anthropologist's. . . . Certainly they had not consulted scientific studies concerning the relative frequency of desirable and undesirable traits in Jews and non-Jews." (Allport 1954, 5–6)

Allport's exposition is consistent with the reigning assumption that prejudice is born out of ignorance. Although there is some ambiguity as to the source of such ignorance (that is, whether it is imposed on the individual by circumstances beyond her individual control, or whether the individual fails to seek out available information about the target group), ignorance itself is regarded as a key to the problem of prejudice. In 1944, Myrdal wrote:

The race prejudice of the typical Northerner . . . is based mainly on ignorance, both simple and opportune. . . . The Northerner seldom gets a chance to see the Negro's good points, and he does not understand the social background of the Negro's bad points. The Southerner's prejudice also has much ignorance in it, but the Southerner's ignorance is more opportune because it is tied to fundamental motives. (Myrdal 1944, 1142)

The theme has persisted: stereotypes contradict knowledge and reason (see, for example, Harding et al. 1969; Selznick and Steinberg 1969; Simpson and Yinger 1972; Pettigrew 1982; Stephan and Stephan 1984).

Despite this emphasis, the verity of stereotypes has been a troublesome issue that has thrust an uncomfortable wedge into scholarly inquiry from time to time. Some scholars have wrestled with the "kernel of truth" hypothesis that stereotypes are exaggerated attributions based on a germ of truth (Allport 1954, 189–196; Mackie 1973). Others have pondered the distinction between incorrect beliefs and negative beliefs (Schuman and Harding 1964). Discussions of Jensen's argument (1969) that ethnic groups vary in IQ also have hinged on whether his argument is true or false. And moral judgments about Jensen's line of research have hinged on whether it is scientifically motivated or racially motivated, as though science automatically cleanses human observation. This issue has been

troublesome because stereotypes have been regarded as the antithesis of science and "rational" knowledge.

Measurement of Stereotypes

Two measures of beliefs about group traits became the standard, the stereotype checklist and agree-disagree categorical statements about a group's attributes. Both measures reflected analysts' assumptions about the hallmark characteristics of stereotypes. Rather than testing the properties of people's intergroup beliefs, analysts instead relied on measures that had the expected properties loaded into them. Reliance on these measures thus served to perpetuate analysts' preconceptions about the specific properties that go into the manufacture of discriminatory intergroup beliefs.

The most frequently replicated research on trait attribution to groups has used the stereotype checklist, usually administered to small samples of select subgroups (generally college undergraduates). This measure presents respondents with a long list of traits and instructs them to assign the traits to specific groups; the positive or negative values attached to the traits are often obtained from another group of respondents (see, for example, Katz and Braly 1947; Meenes 1943; Gilbert 1951; Sherriffs and McKee 1957; Karlins, Coffman, and Walters 1969; Broverman et al. 1972; Smedley and Bayton 1978).

In cross-sectional, public-opinion surveys, the measure that has been used most frequently presents respondents with categorical statements about the trait attributes of a group (such as "poor people are lazy," "blacks are musical"), and respondents can express agreement or disagreement with those statements (see, for example, Adorno et al. 1950; Deutsch and Collins 1951; Wilner, Walkley, and Cook 1955; Marx 1967; Selznick and Steinberg 1969).

These two standard measures share one important feature—they both force respondents to express themselves in categorical, all-or-none terms, thereby precluding the expression of more qualified beliefs.^[1] Because of this constraint, these measures cannot reflect the degree to which people do in fact think about groups in a categorical or qualified way. The concern of analysts with this specific issue has exacerbated the problem by encouraging interpretations that far outstrip the information yielded by these measures. Thus, it is commonly assumed that respondents who agree with the statement "blacks are lazy" think all blacks are lazy. Similarly, the very name of the "stereotype checklist" carries the clear assumption that, in requiring respondents to label groups with

[1] Procedures used by Broverman et al. (1972) did allow respondents to make finer distinctions between men and women, but these distinctions were ignored in the variables that were constructed from the data.

traits, the analyst has captured stereotypical thinking. Indeed, one analyst has made this assumption explicit: Ehrlich argues that "the statement 'Atheists are cynical' may be correctly transliterated 'All Atheists are cynical'" (Ehrlich 1973, 21).

The categorical phrasing of the measures not only restricts the information they can yield, but it may also make them more susceptible to response-bias effects. The blatant character of the task may cue some respondents (especially the well-educated) to social-desirability pressures, and make them unwilling to present their beliefs in an unfavorable light.^[2] Meanwhile, other respondents with simpler cognitive styles (especially the poorly educated) may be trapped by the loaded and unbalanced construction of the measure into saying "yes" or into labeling groups with traits because these response options are "near enough." Thus, these measures may be seriously confounded with response sets caused by varying cognitive styles or susceptibility to social-desirability pressures (Campbell et al. 1960; Ehrlich 1964; Jackman 1973; Jackman and Muha 1984).

Several analysts have registered dissatisfaction with these measures of stereotypes. As early as 1951, LaViolette and Silvert noted that "the attributes of stereotypes have not been examined critically by social psychologists" (1951, 260). In 1954, Harding et al. pointed out that standard stereotype measures do not differentiate between respondents who are indeed ignoring individual differences and those who are merely making a statistical generalization. These complaints have been reiterated over the years (Hyman 1969, 10; Brigham 1971; Jones 1972, 70; Jackman and Senter 1980, 1983; Ashmore and Del Boca 1981). Ehrlich (1964) experimented with the introduction of response options to the standard public-opinion measure that allowed for the expression of strong or mild agreement or disagreement, indecision, and "don't know": he reported that the standard format overstates the degree of acceptance of prejudiced statements. Ehrlich and Rinehart (1965) experimented with a variant of the stereotype checklist that used an open-ended format: they reported that the standard format elicited the assignment of more traits to groups and more consensus among respondents, and that the two formats produced different listings of traits. Brigham (1971) suggested that respondents should be asked to assess what percentage of a group has a particular trait, in order to get a more valid measure of the form of people's attributions to groups. McCauley and Stitt (1978) and McCauley, Stitt, and Segal (1980) have suggested that the standard stereotype check-

[2] Some evidence of this is provided by the replications of the original Katz and Braly design among Princeton undergraduates. Gilbert (1951) and Karlins, Coffman, and Walters (1969) report that many subjects protested the unreasonableness of making such simplistic generalizations.

list oversimplifies human perception and that stereotypes can be understood better as probabilistic predictions that distinguish one group from another. More recently, the General Social Survey (Davis and Smith 1990; Bobo and Kluegel 1991) measured stereotypes of whites, blacks, Hispanics, and Asians with 7-point bipolar scales in which respondents indicated whether they thought each group was characterized more by one trait (for example, hardworking ) or by its opposite (for example, lazy ). But despite the criticisms and suggested revisions over the years, the standard measures of stereotypes have not been seriously discredited. Studies using the traditional measures continue to be cited as classic evidence of the pervasiveness of stereotyping, and, as Ashmore and Del Boca (1981) report, the Katz and Braly checklist has continued to be the reigning measure in the prolific literature on stereotypes.

As a consequence, the complaint made by LaViolette and Silvert in 1951 that "we may call their [stereotypes'] characteristics 'claims' rather than established attributes" (1951, 260) remains as true today. First, as already noted, the standard measures do not yield direct evidence on the extent to which people's images of social groups are indeed categorical or qualified. Second, we lack critical information on the pattern of categorical thinking. Is categorical thinking more evident when people are describing groups other than their own, or do people tend to describe their own group in positive categorical terms and other groups in negative categorical terms? In other words, are categorical descriptions used to differentiate the homogeneity of another group from the individual variance within one's own group ("they are all alike"), or to maximize the difference between the bad qualities of the other group and the good qualities of one's own group ("they are all bad, we are all good")? Third, by the same token, the standard measures cannot give information on the degree to which people draw distinctions between groups. Do people tend to draw large or small distinctions between their own and other groups? Do people tend to minimize within-group variance among outgroups and exaggerate between-group differences, as Tajfel and others have claimed? Fourth, the stereotype checklist is often accompanied by an evaluative scale for the traits on the list, but the standard measure used in public-opinion surveys makes no direct assessment of the derogatory or flattering intent of a person's beliefs.