To help teachers who must produce test items to measure instructional objectives, 54 teacher education candidates participated in an experiment where easily understood constraints on item production were manipulated. Four forms of a test item writing exercise sheet were randomly distributed, each asking for the production of eight sample test items, two for each specified topic. The subjects produced 16 items, to be used for seventh grade students. Two 16 item tests were constituted, one on subtraction and one on current events. The tests were administered to 51 junior high school students. Means and standard deviations of the items were computed, and analysis of variance for the subtest means was conducted for each replication. Significant differences (F=8.3, df=3, 12) were observed for subtraction. For the current events data differences were not significant. Findings are limited by the number of items on each subtest. Further staff studied are investigating how to produce items truly congruent with objectives and how best to translate these findings into practical procedures for teacher. (Author/CJ)