This paper examines the relationship between methodological naturalism and the standard practice within philosophy of constructing theories on the basis of our intuitions about imaginary cases, especially in the work of Alvin Goldman. It is argued that current work in cognitive science presents serious problems for Goldman's approach.