The paper will deal in turn with three sets of modelling issues: the question of 'data'; the 'micro' problem of specifying market behaviour, and the. 'macro' issue of 'closing' the models in aggregate. I will conclude with some suggestions for future research. The basic theme of the paper is this: CGE modelling is essentially a conservative or 'neoclassical' scientific endeavour, and exhibits the strengths and weaknesses of neoclassicism. And as for the recent injection of apparently nonneoclassical imperfect competition or industrial organization (IO) concepts into CGE, though, as an 10 specialist myself I certainly welcome this in principle, I have doubts about the usefulness of the practice.