Possible criterion-related bias of the WISC-III with a referral sample

This study examined whether ability scores from the Wechsler Intelligence S cale for Children-Third Edition (WISC-III) showed criterion-related bias. P articipants (N = 832) comprised a referral cohort between ages 8 and 16 yea rs that differed by race (Anglo, African American) and gender. Verbal, Perf ormance, and Full Scale IQs from the WISC-III were used to predict Reading, Mathematics, Language, and Writing Composites from the Wechsler Individual Achievement Test (WIAT). Unlike previous research with the WISC-III, 50% o f the analyses (12 out of 24) showed statistically significant effects. How ever, in all instances where bias was found, differences in regressions wer e statistically significant for intercepts, but not for slopes, and served to overpredict the performance of minority groups (African Americans, femal es) relative to majority groups (Anglos, males). Results are discussed in t he context of how instances of differences in intercepts could mislead psyc hologists in interpreting children's ability scores.