Detection of Item Bias Using Analyses of Response Patterns

Item bias, when present in a multiple-choice test, can be detected by appropriate analyses of the persons × items scoring matrix. Five related schemes for the statistical analysis of bias were applied to a widely used, primary skills multiple choice test which was administered in either its English- or Spanish-language version at each of the two levels, to 1259 students in bilingual education programs. The results indicate that from one-fifth to one-third of the items in the tests show strong evidence of bias, corroborated by a separate analysis of linguistic and cultural sources of bias for both the biased items and those items with no statistical findings of bias.