In spring 1986 Anglo students received higher scores than Blacks or Hispanics on the Iowa Tests of Basic Skills (ITBS), on all subtests and at all grade levels in the Denver (Colorado) Public Schools. This study, by attempting to locate patterns of item response, seeks to identify items in the ITBS that suggest differential pupil achievement by racial or ethnic group. Discriminant analyses of items missed on each subtest of the ITBS by members of each ethnic group were done for pupils in grades 1, 2, 5, 7, and 8. The following results were discovered: (1) there is a nonchance relationship between a pupil's racial or ethnic designation and his or her right-wrong response pattern; (2) knowing nothing more than the pupil's response pattern, one can predict with from 58 to 73 percent accuracy the pupil's racial or ethnic designation; (3) the accuracy of the prediction increases as the age of the pupil increases; (4) the Vocabulary subtest provides the best means to predict whether a pupil is Black, Hispanic, or Anglo; (5) the subtests in Spelling, Capitalization, Punctuation, and Mathematics Computation have the least predictive power; and (6) those subtests requiring knowledge of specialized vocabulary and reading skills were generally stronger predictors than those that require application of rules or the use of memory or recall. The single easiest and potentially most effective instructional activity that will improve minority test scores is vocabulary building, with emphasis on nonvernacular, specialized words. Appendices provide data on graphs and tables, and describe the method of analysis. (BJV)