Recent years have witnessed considerable growth of research on the benefits of adult learning. Much of this is UK-based, and draws on evidence from large scale longitudinal data sets. Overwhelmingly, these studies have found clear evidence of economic, social and individual benefits as a result of participating in adult learning. While these claims are important and influential ones, there has to date been little discussion of the nature of the data and analytical techniques being used. Nor has there been sufficient attention to the possibility that learning may have negative outcomes. The paper identifies and explores some limitations of longitudinal research in the study of adult learning, but concludes that despite the problems, this body of work still represents an important departure in the field, with considerable international significance.