We explore and critically reflect on the process of science and engineering research assistant skill development both within laboratory-based research teams and, when no team is present, whithin the faculty supervisor-research assistant interactions. Using a performance-based measure of research skill development, we identify research assistants who, over the course of an academic year of service as a researcher, markedly developed, modestly developed, or failed to develop their research skills. Interviews with these research assistants and their faculty supervisors, seen through the lens of cognitive apprenticeship, provide insight into this variation. We found that within the contours of supervisory relationships and research teams, research skill development is indelibly shaped, for better or worse, by supervisor influence and abundant trial-and-error.