In order to screen out items which may be biased against some ethnic group prior to the final selection of items in test construction, a statistical technique for assessing item bias was developed. Based on a theoretical formulation of R. B. Darlington, the method compares the performance of individuals who belong to different ethnic groups, but have equal scores on a subtest containing the item. A chi square technique is used to evaluate differences in performance and is demonstrated on data collected during the item analysis phase of the current revision of the Metropolitan Readiness Tests. (Author)