Using item responses from an in-training examination in diagnostic radiology, the application of a strength of association statistic to the general problem of item analysis is illustrated. Criteria for item selection, general issues of reliability, and error of measurement are discussed. (Author/LMO)