After graduation from secondary school, academic tasks (i.e., learning contents) are no longer structured in terms of school subjects (i.e., English, mathematics). Therefore, learners lack past performance and mastery experience to inform their expectancy of success (i.e., ability beliefs) regarding novel tasks. In this paper, we investigate how individuals establish expectancy of success regarding novel academic tasks. We hypothesize that individuals draw on ability beliefs regarding known tasks that are deemed similar to novel tasks to estimate expectancy of success (generalization hypothesis). Participants were first-year students (n = 354) in the field of business administration (Study 1), and (Study 2) psychology students predominantly (n = 174). In Study 1, we analysed relations between ability beliefs (i.e., academic self-concepts of ability) regarding four school subjects and four fields of study varying in similarity. In Study 2, we assessed mastery experience regarding two school subjects and expectancy of success (i.e., self-efficacy) regarding a fictitious course manipulating participants' similarity judgement. We analysed the data using mainly structural equation modelling. Results support the generalization hypothesis regarding both indicators of expectancy of success (i.e., self-concept and self-efficacy). Subject-specific self-concepts of ability predict study-related self-concepts of ability according to individuals' similarity judgements. Subject-specific mastery experience predicts expectancy of success only if the respective school subject is emphasized in the course description. Individuals apparently draw on established ability beliefs regarding known tasks to inform their expectancy of success regarding novel tasks. Findings further our understanding of the development of motivation to learn in adulthood.