Forty-five graduate students in two educational psychology classes were randomly assigned to three groups for a classroom test on the course content. One group was asked to go to another room and then was privately excused from the test. The other two groups took tests with the same 32 items, but one group used a multiple-choice format and the other group a short answer format. Four weeks later, all three groups were given both forms of the same test (the short answer form first) on a unannounced basis. In this test one half the items were measures of knowledge and one half were measures of concepts. Resulting scores were analyzed using a 3x2x2 factorial analysis of variance design with repeated measures in the last two factors. The treatment effect was not significant. Scores were higher on the multiple-choice test than on the short-answer test. There also was a significant interaction between test format and question complexity (knowledge questions versus concept questions); that is, concept questions were comparatively easier in multiple-choice form and harder in completion form. (Author/CTM)