For some time now, formal modelling has been touted by its supporters as a panacea for political science – or at least as a major step forward in the discipline's development. Certainly, it embodies a number of praiseworthy elements. Its insistence on starting with a parsimonious and precisely formulated set of assumptions cannot help but constrain slippery thinking, for example, and its rigorous working out of implications, while often demonstrating the obvious, occasionally leads to unanticipated and intriguing results. Moreover, the combination of precision and rigour holds forth the promise of generating relatively clear-cut tests of rival explanations, a major boon – if it proves true – in a discipline more inclined to abandon theories than to disconfirm them.

How much better the analytical or formal orientation is, then, than the ‘funnel of causality’ approach of empiricists whose quest for the highest explained variance seldom produces more than a miscellaneous grab-bag of influences on the dependent phenomenon. Empirical work of that sort may have some limited utility in identifying possible causes, to be sure, but at some point the scholarly enterprise must move to the higher level of elaborating a clear logical structure among causal factors. Here, empirical success in accounting for observed phenomena cannot be the sole guide: the best theory is the one that provides the most accurate idea of what is actually going on in the real world, not the one with the best correlations.