In this study, we examine when venture capital firms (VCFs) learn from their portfolio companies (PFCs). Relying primarily on learning and behavioral theories, we develop hypotheses regarding the effects of prior experience, knowledge overlap, trust, and PFC performance on learning by VCFs. We use a combination of primary and secondary data from 298 U.S.-based VCFs to test the hypotheses. Interview data are used to illuminate the results and to guide our discussion of implications. Many of our results were surprising. For example, we found that the VCF's overall experience is negatively related to VCF learning, and we found that trust in VCF-PFC dyads is also negatively associated with VCF learning. Whereas we expected to observe a curvilinear relationship between knowledge overlap and learning, we found that lower levels of knowledge overlap were associated with greater learning in a linear fashion. Finally, we found that VCFs perceive greater learning to occur in higher-performing PFCs. We discuss the limitations and implications of our findings and also suggest avenues for future research.