This study examined the effects of assigning high school students to a computer-delivered geometry program that either matched or did not match their preferred amount of instruction as measured by a preprogram questionnaire. Students could adjust their program length by adding screens in a lean version of the program or by bypassing them in a full version. Matching students with their preferred program length did not produce improved posttest achievement and was particularly ineffective with students who preferred a low amount of instruction. The full version of the program was somewhat more effective than the lean version, primarily because of the better performance of low-preference students in the full version (mismatched) than in the lean one (matched).