Pursuing earlier results which indicated that response-set scores might be important predictors of grades in law schools, data from four more schools were obtained and analyzed. These data suggest that while response sets may be useful predictors, they are subject to certain important restrictions. First, response sets do not seem to be effective when ability variables are already doing the job well. Second, a set cannot operate as a predictor if it scarcely exists in the inventory-taking behavior of the students. The latter restriction might be overcome by allowing or encouraging students to display the set more extensively. What particular kind of set will be helpful in predicting grades in a particular school may sometimes be forecast by someone who knows the school well.