This paper explores the effect of reviewing items and altering responses on the efficiency of computer adaptive tests (CATs) and the resultant ability measures of examinees. Subjects included 712 medical students: 220 subjects were randomly assigned to the review condition; 492 were randomly assigned to a review control condition (the usual CAT administration method). The test, designed to establish student ability above or below a pass/fail point, included items from a pool of 726 Rasch calibrated items. Subjects in the review condition were instructed to answer each item when presented, but were allowed to review and alter responses at the end of the test. Subjects in the control condition were not permitted to review and alter responses. Student ability measures before and after review were correlated at 0.98. The average efficiency of the test was decreased by 1% after review. About 32% of the students improved slightly after review, but did not change their pass/fail status. These data do not support prohibition of review on CATs administered under these rules. Results indicate that the students allowed to review performed significantly, if slightly, better than did control students. Three tables are included. (TJH)