Previous studies have found that students' concept-building approaches, identified a priori with a cognitive psychology laboratory task, are associated with student exam performances in chemistry classes. Abstraction learners (those who extract the principles underlying related examples) performed better than exemplar learners (those who focus on memorizing the training exemplars and responses) on transfer exam questions but not retention questions, after accounting for general ability. We extended these findings to introductory biology courses in which active-learning techniques were used to try to foster deep conceptual learning. Exams were constructed to contain both transfer and retention questions. Abstraction learners demonstrated better performance than exemplar learners on the transfer questions but not on the retention questions. These results were not moderated by indices of crystallized or fluid intelligence. Our central interpretation is that students identified as abstraction learners appear to construct a deep understanding of the concepts (presumably based on abstract underpinnings), thereby enabling them to apply and generalize the concepts to scenarios and instantiations not seen during instruction (transfer questions). By contrast, other students appear to base their representations on memorized instructed examples, leading to good performance on retention questions but not transfer questions.