A LOOK AT CONTENT BIAS IN IQ TESTS

Abstract

Sexual and racial biases persist even in the face of legal sanctions. While the past century has seen females and nonwhites enfranchised, their equality has yet to be realized. Pejorative stereotypes abound in present-day society, fostering and maintaining discriminatory attitudes. While the more obvious forms of discrimination have been outlawed by the Civil Rights Act of 1964 and by Title IX of the Educational Amendment Act of 1972, more subtle forms of discrimination have yet to be eradicated. In the arena of educational assessment, one potential form of subtle bias may exist in the content of IQ test items. Although research has substantiated the existence of bias in achievement test item content (Donlon, Ekstrom, Lockheed & Harris, 1977; Tittle, McCarthy & Steckler, 1974) and performance differences (Donlon et al., 1977; Strassberg-Rosenberg & Donlon, 1975), research on IQ test bias has been limited to performance (Jencks, Smith, Acland, Bare, Cohen, Gintis, Heyns & Michelson, 1972; Jensen, 1971; Laosa, 1977; Samuda, 1975). Perhaps a reason that so much more research has been done on bias in test performance than in test item content relates to the relative ease with which performance bias can be detected. Disparities in mean IQ scores are straightforward. However, differences in the ways which stereotypes are depicted in test items require clear definitions of a biased representation. In addition to assessing the extent to which individual items are stereotyped, an analysis of content bias in IQ tests requires measurement of the frequency with which members of different races and sexes appear. Although stereotyping can be readily identified on an item by item basis, the cumulative effects of disproportionate sexual and racial representations are not apparent until the test is considered in its entirety. Regardless of whether item content bias affects subsequent test performance, the issues of sexual and racial stereotyping and disproportionate representations warrant scrutiny in and of themselves. Just as blatant forms of bias have no place in education, more subtle forms are equally intolerable even if they do not in turn effect performance. To the extent that test items are biased, they foster inappropriate and inaccurate perceptions. Considering the widespread use and respectability of IQ tests, it becomes important to ask: To what extent do test items accurately represent the distribution of sexes and races, and to what extent are members of these groups stereotyped?