Recent decades have seen renewed attention to issues of causal inference in the social sciences, yet implications for life course research have not been spelled out nor is it clear what types of approaches are best suited for theoretical development on life course processes. We begin by evaluating a number of meta-theoretical perspectives, including critical realism, data mining and experimentation, and find them limited in their potential for causal claims in a life course context. From this, we initiate a discussion of the logic and practice of 'natural experiments' for life course research, highlighting issues of how to identify natural experiments, how to use cohort information and variation in the order and timing of life course transitions to isolate variation in exposure, how such events that alter social structures are the key to identification in causal processes of the life course and, finally, of analytic strategies for the extraction of causal conclusions from conventional statistical estimates. Through discussion of both positive and negative examples, we outline the key methodological issues in play and provide a road map of best practices. While we acknowledge that causal claims are not necessary for social explanation, our goal is to explain how causal inference can benefit life course scholarship and outline a set of practices that can complement conventional approaches in the pursuit of causal explanation in life course research.