In order to measure the value of student feedback in contributing to the refinement of instructional frames in computer-assisted instruction (CAI), a stand-alone CAI frame was written around a question designed to elicit a wide variety of replies. Half of a real sample of fifty handwritten responses were submitted to the frame, with a match score of 96% and a sense score of 48%. After revision, the frame attained a match score of 100% and a sense score of 88%, on the other half of the test set. Conclusions drawn from the study were that (1) for an open question with a large number of legal answers and an adult, heterogeneous student set, a striking improvement in meaningful intercept can be obtained through close analysis of a modest number of live responses; and (2) as a corollary, the rate of improvement to be derived from close analysis beyond this modest number must decline quite rapidly. In the report of the study, selected-string-match CAI strategies and the mechanics of the CAI "macroframe" are discussed. Appendices contain supporting materials. (Author/MF)