The objectives of this study were to first determine whether or not the empirical item analysis of domain referenced tests (DR) was justified; and second, in the event that it was, which of a set of recommended procedures was most effective for determining item quality. The analysis that followed led to the conclusion that empirical procedures were highly desirable. When these empirical procedures were applied to test data, the results indicated that four different techniques provided almost identical information: Rasch statistics, instructional sensitivity indexes, traditional statistics, and Baysian indexes. Based on these results, it would seem that any one of these four would serve adequately. (Author/RC)