An empirical review of test items is described as an essential step in criterion-referenced test development. The concept of test items' instructional sensitivity is introduced, and research is briefly reviewed which describes four theoretical contexts in which instructional sensitivity indexes have been observed: criterion-referenced; classical test theory; Bayesian statistics; and latent trait theory. The study described in this report uses a wide range of item characteristics that transcend the four previously-mentioned approaches; encompasses instructional settings that vary in effectiveness; and has sufficient numbers of items and examinees for each test. The relationship of putative instructional sensitivity indexes to the traditional concepts of item difficulty and item discrimination is explored, using 17 item characteristic indexes. Results strongly suggest that instructional sensitivity is a unique and useful concept in criterion-referenced test development. (Author/GDC)