was whether the behaviors prompting referral differ by domain (e.g., academic vs. behavior) or by degree for children differing by ethnicity or gender. The total sample of 150 children (55 white, 43 black, and 52 Hispanic) was selected using a stratified random sampling design. While the design sought to compare these groups, it should be noted, children from all three ethnic groups who had been referred exhibited intellectual, achievement, and behavioral measures indicative of students with real problems. For example, the mean score in reading and spelling for all three ethnic groups was approximately 2 standard deviations (SD) below the general population mean. On the Critical Events Index (Walker and Severson, 1990), tapping “behavioral earthquakes,” the authors reported, “relative to the CEI standardization sample mean and SD (M = 0.12, SD = 0.46), the effect size for males was 5.20 and for females 2.30, indicating a level of differentiation of greater than 99 percent” (p. 147). Clearly teachers referred a group of children who presented significant achievement and behavior problems. The number of children presenting problems of similar severity who were not referred cannot be ascertained from this study.
Returning to comparisons of referred students on the basis of ethnicity and gender, there were differences on the basis of ethnicity. White referred students differed significantly from black referred students on the Wechsler verbal IQ, performance IQ, and reading achievement and differed from Hispanic referred students on verbal IQ, reading, and spelling. In all instances, white students scored significantly higher. Comparison of black and Hispanic referred students failed to detect any significant differences on intellectual or achievement measures. On problem behaviors and social skills, racial/ethnic comparisons revealed only one difference—black referred students exhibited more problem behaviors than did Hispanic referred students.
Gender comparisons failed to reveal any significant differences on either IQ or measures of achievement. However, in the behavioral domain, referred males were rated significantly higher on problem behaviors (i.e., conduct problems), critical events (low frequency, high salience—e.g., fire setting), and hyperactivity, while they were rated significantly lower on social skills. One intriguing finding is the failure to find male-female differences in academic achievement using standardized measures. However, teacher ratings of overall academic competence yielded significantly higher ratings for females than for males, despite the finding of no differences on the standardized measures.
MacMillan et al. (1996b) concluded that the students referred by teachers exhibited severe academic and behavioral deficiencies, validating “teachers as tests” in the referral process. This held across gender and racial/ethnic groups. Contrary to the fear that teachers indiscriminately and unfairly refer minority group children who are actually doing well, the findings