Several alternative indexes of item usefulness were empirically compared. They were: (1) item discrimination based on high and low groups on a postinstructional measure, (2) shift in item difficulty from a pre- to a postinstruction measure, and (3) item discrimination based on pre- and posttest performances. A typical classroom final examination was administered to 28 students on their first day in a multivariate statistics course and again at the end of the term. The results, using the various indices, indicated that shift in item difficulty from pre- to postinstruction yielded data significantly more similar to the pre-post discrimination index than did the high-low group posttest discrimination index. It is recommended that the conceptually more ideal pre- to postinstruction discrimination index be used even when instruction is not performance oriented. (Author/JS)