This paper describes an experiment attempting to replicate and broaden the findings of three previous studies in which two measures of compulsivity were taken in order to determine their effect as a "moderator variable" in the prediction of course grades by an interest scale. The hypothesis underlying these studies was that the more compulsive students would put a great deal of effort into their course work, regardless of their interest in the courses, but the effort of the less compulsive students would depend upon their interest. Hence, a higher correlation between interest measures and grades was expected for the less compulsive students. Two indirect measures of compulsivity were employed. One was the Accountant scale of the Strong Vocational Interest Blank (SVIB)—those with high scores, resembling accountants in their interest, were presumed to be compulsive. The second was a residual score based on the Speed of Comprehension and Vocabulary scores of the Cooperative Reading Comprehension Test—those reading more slowly than predicted from their vocabulary scores were presumed to be compulsive. The most important findings were that 1) the two compulsivity variables did not act as moderators for liberal arts students of either sex, although they had for men in engineering; 2) the joint use of the two variables did not enhance the moderator effect; and 3) the two measures moderated different scales; and 4) the moderated scales reflected different interests. It is concluded the moderator variables are still a valuable concept, but not as stable or generalizable as originally hoped. Paper presented at APA meetings, Chicago, September 1965.