Continuing professional development (CPD) evaluation in education has been heavily influenced by "level models", deriving from the work of Kirkpatrick and Guskey in particular, which attempt to trace the processes through which CPD interventions achieve outcomes. This paper considers the strengths and limitations of such models, and in particular the degree to which they are able to do justice to the complexity of CPD and its effects. After placing level models within the broader context of debates about CPD evaluation, the paper reports our experience of developing such models heuristically for our own evaluation practice. It then draws on positivist, realist and constructivist traditions to consider some more fundamental ontological and epistemological questions to which they give rise. The paper concludes that level models can be used in a number of ways and with differing emphases, and that choices made about their use will need to reflect both theoretical choices and practical considerations. (Contains 2 notes and 1 figure.)