A semantic differential (SD) instrument, modified by replacing words or noun phrases with phrases describing a behavior, was administered to male freshmen students. Six items discriminated between two groups, 97 science majors and 161 non-science majors, on three axes, labelled as evaluation, potency, and activity. To test whether the instrument could be used to group individuals 471 entering male freshmen were tested and grouped by a method profile analysis. The 14 groups with fiv e or more members were examined in detail, and four response categories (general science, medical science, physical science, and non-science) identified. For 15 of 18 students interviewed, their interview data and the models from the SD were complementary. The projected college major of the other three cases was an imitation of the father's actual or projected career. The lack of relationship between SD measured interests and science achievement, measured by a 60-item test administered just after classes commenced, is interpreted as indicating that the SD instrument is mainly measuring the affective domain. (Author/AL)