The relative effectiveness of multiple-choice (MC) and constructed-response (CR) test formats in computer-managed instruction (CMI) were compared using four test groups of 30 trainees each who were assigned nonsystematically from the basics course at the Propulsion Engineering School, Great Lakes Naval Training Center. Group A took module tests in the standard CR format with answer cues and then converted their answers to MC answer sheets for CMI scoring. Group B took CR tests with answer cues, but the research staff converted the answers. Group C took CR tests but without answer cues, and the staff converted the answers, while Group D took tests in the MC format. Before and after the tests, skills and knowledge were measured to compare factors such as learning, retention, time to complete the course, and attitudes. There were no differences in learning among the groups, but Group C, with the CR questions without cues, had the best retention, but took longer to complete the course and rated their tests as being more difficult than did students in the other groups. The attitude questionnaire and the results of ANOVAs comparing the groups on measures of learning are appended, and five references are provided. (Author/BK)