Purpose: The purpose of this paper is to explore the extent to which project preferences and social capital constrain mobility in project-based careers. Design/methodology/approach: The paper analyzes the careers of 352 individuals who entered the motion picture industry between 1988 and 1990. It uses motion picture credit histories to generate role sequence paths. The paper quantifies differences between paths using optimal matching techniques and cluster analysis to classify paths into clusters. It validates the classification by testing hypotheses about differences between path clusters. Findings: In addition to a large group of individuals who exit the industry after the initial credit, the paper identifies three distinct clusters of career paths that exhibit differences in the sex of individuals on them, in the persistence of relationships with employers, in employer characteristics, and in the nature of subsequent projects. Research limitations/implications: Because the paper is exploratory, general hypotheses are tested. Motion picture production may be an extreme example of project-based production, which would limit generalizability. Practical implications: Managers, individuals and career experts should recognize that mobility can be constrained and channeled by preferences in project type and by social capital. Employer celebrity appears to play no role in the careers of assistants, but control over many projects plays a significant role. Originality/value: The paper demonstrates non-organizational constraints on mobility in project-based, apparently boundaryless, self-managed careers. (Contains 6 tables.)