This paper is a reflection on a three-year longitudinal study that explored the development of children's concept of a substance, for which detailed results concerning children's understandings have been reported elsewhere. The attention in this paper is on the methodological features related to the longitudinal nature of the study and the insights into pupils' learning that were afforded by its design. There is, of course, an extensive literature on children's understandings of scientific ideas, and the nature of conceptual change involved in learning is a contested area. I argue in the paper that longitudinal studies are uniquely placed to inform the debate, and by focusing on this particular study the aim is to demonstrate the kind of contribution that can be made.