The percentage of school students specializing in math and science is particularly low. The current research suggests that this is due to prototypes about math and science being highly dissimilar from self-prototypes students have or want to have of themselves. Going beyond previous studies on self-to-prototype matching, we assumed that students compare their self-views to both a prototypical student liking a certain subject ("favourite-subject-prototype") and a prototypical student disapproving of it ("least liked subject-prototype"). Results show that for humanities (German and English language), favourite-subject-prototypes were judged more positively than least-liked-subject-prototypes, whereas for science (math and physics), least-liked-subject-prototypes were perceived as more positive than favourite-subject-prototypes. As expected, only if a student's self-prototype was quite clear (high self-clarity) was it used as a standard against which school-subject prototypes were compared with respect to their degree of overlap. Our results showed that the better the match between self and favourite-subject-prototype, the stronger were the subject preferences.