All contributions in this section deal with the question of conceptual structure. Religious ideas are organised by concepts which carry certain tacit assumptions and constrain the range of inferences and conjectures subjects can make about religious entities and processes. The problem is to describe the mental structures which make it possible for subjects to use religious categories. The diversity of models presented here is a reflection of the variety of approaches in current linguistic and psychological theories of conceptual representation. Keller and Lehman argue for a ‘theory-based’ understanding of religious categories. Their argument starts with the observation that such concepts are intuitively more ‘complex’ than most everyday categories. This complexity can be explained by the fact that religious categories are constrained by underlying theories, rather than by external similarities. They then examine the relevance of this hypothesis on Polynesian categories of ‘magical’ connections. Keesing argues for the relevance of ‘conventional metaphors’ to the description of symbolism. He argues that cultural symbolism is typically organised around focal metaphors, and draw both their ‘evocative’ power and their vagueness from metaphorical structures found in everyday language. In both Bloch's and Boyer's contributions, a third approach is presented, which examines the connections between intuitive, universal knowledge of natural entities and processes, on the one hand, and cultural elaborations, on the other. On the basis of Malagasy data, Bloch's chapter shows how intuitive notions of natural kinds are both and partially suspended in religious symbolism.