A study involving 50 experimental and 99 control subjects (graduate education majors) was undertaken to assess the interchangeability of knowledge of correct response feedback (KRC) and answer until correct feedback (AUC) in computer-assisted instruction. P. L. Smith's model (1988) suggests that AUC in better for high-ability students. W. Dick and R. Latta (1970) found AUC to be better for high-ability students and KCR to be better for low-ability students. The study was designed to determine whether high-ability students benefit most from the deeper processing required by AUC, while low-ability students perform best with KCR, which requires less processing. Experimental subjects were randomly assigned to a AUC or KCR treatments, resulting in a 2 (treatment: AUC or KCR) by 2 (ability: low or high) factorial design. One week later, the subjects took an identical posttest that provided the final exam score for the course. Analysis of covariance revealed that low-ability students performed best with KCR, while high-ability students performed best with AUC. The interaction between ability and feedback form was significant. These findings suggest that low-ability students should be provided with KCR and high-ability students should receive AUC. The study also provides support for Smith's model of feedback by learner ability. Three data tables and three figures are included. (TJH)