BIAS IN IQ TEST PREDICTIONS OF SUBTRACTION SKILLS LEARNING AND ACHIEVEMENT.

Abstract

The present study was conducted to address the need to compare the validity of intelligence tests in predicting learning and achievement for Mexican Americans and Anglo children. In addition, the study examined the effects of variations in the language of test administration on Mexican American children with different linguistic competences (predominantly Spanish, bilingual, predominantly English). A widely used individual intelligence test, the Wechsler Intelligence Scale for Children Revised (WISC-R), the Spanish English Screening Instrument, a subtraction skills pretest, five subtraction learning trials and five subtraction posttests were administered to a random sample of 150 Mexican American children and 50 Anglo children. The WISC-R and subtraction pretests and posttests were used as the predictor and criterion, respectively. The Cleary definition of bias in test use provided the basis for all analyses. Regression analysis were also performed in order to examine the effects of achievement and immediate prior learning on learning scores. In addition the Tukey Honestly Significant Difference test (HSD) was used following a one way analysis of variance to determine differences in IQ among the Anglo and Mexican American children tested under varying language conditions (English, Spanish, Bilingual). The results indicated that IQ test scores are not suitable predictors of learning for the Mexican American group and suggested that IQ scores can be used as predictors of learning for Anglo children. It was also found that IQ scores have utility in predicting achievement for both, Mexican American and Anglo children. In addition, analysis of variance data obtained from this study revealed significant differences in IQ associated with language of test administration. It was found that Mexican Americans tested in both Spanish and English obtained significantly higher scores than Mexican American children tested in English or Spanish alone. Findings from this study suggested that regardless of linguistic competence Mexican American children appear to benefit from bilingual approach to test administration. In short findings from this study revealed that the predictive validity of the WISC-R for Mexican American children is seriously impaired when a learning criterion is used. Implications of these findings were discussed and suggestions were made for the development of an assessment model based on learning as the criterion.