Performance of bilingual students on translated and non-translated versions of an ability test

The present study investigated cultural bias in the WISC-R and problems that arise from translating the WISC-R from one language to another. Four Verbal subtests--Information, Similarity, Vocabulary, and Comprehension--were split in half by their odd and even items. The even items were translated into Bengali, a language spoken in India. The subtests were then administered as a group, pencil and paper test to 80 East Indian children, whose age ranged from 13 to 16 years. The obtained data was compared to that of an American sample of 51 students. Split half Reliabilities, T tests, Mixed Design ANOVAs, P-values and Chi-Squares were used to analyze the data. Results suggest that both groups performed better on the odd items, overall. Translation does affect the difficulty level of items. Further, some items were identified as biased, positively and negatively, against each of the two sample groups