This study determined whether item forms ( rules for constructing items related to a domain or set of tasks) would enable naive item writers to generate multiple-choice items at three taxonomic levels--knowledge, comprehension, and application. Students wrote 120 multiple-choice items from 20 item forms, corresponding to educational objectives related to diabetes mellitus and renal disease. The items were then administered to 20 graduate-level dietetic interns. Results were analyzed by a two-way mixed effects factorial design with one observation per cell. No significant differences between test items were reported for nine of the 20 forms. A comparison of a chi-square index of item homogeneity with the analysis of variance indicated agreement between the procedures in 18 of the 20 item forms. Homogeneous items shared the following characteristics: short item stems, distractors, and correct responses; an average of 10 replacement set structures within item forms; and high average item difficulty indices. Results did not support the use of item forms in generating homogeneous items. (A history of norm-referenced, criterion-referenced, and domain referenced testing; the item forms; and the diabetes and renal disease tests are included.) (Author/CP)