What can evolutionary psychology tell us about cognitive architecture?

Certain evolutionary psychologists have argued that a massively modular cognitive architecture is the necessary outcome of natural selection. This argument appears to be predicated upon three core and questionable assumptions - 1) a Classicist Computational Theory of Mind; 2) Marr's (1982) tripartite explanatory project for the cognitive sciences; and 3) the view that adaptive behaviour must come under fairly direct cognitive control. This paper argues that, under these core assumptions, there is in fact no direct support for a modular architecture from an adaptationist analysis. This is because such analyses are always at the behavioural level and there is nothing in the behavioural data to help decide between possible architectures. Behavioural data can only be used to provide an extensional functional analysis and architectural claims are to some extent intensional. This leaves only an argument from theoretical cognitive science and this too is weak. The paper closes with a discussion about whether or not an evolutionary perspective can inform models of cognitive architecture at all and the conclusion is drawn that it can. Such an approach favours ground?up modelling of functions and thereby imposes a form of parsimony on algorithmic and architectural speculations