Two issues have become increasingly apparent in attempts to apply the prevailing notions of' economic rationality' to the theory of the organisation of business firms. The first goes under the rubric of bounded rationality. This is hardly a new idea, and has been forcefully brought to our attention in the work of Simon and others, although awareness of the problem (in economics) goes back at least to Clark (I9I8). I shall try to provide a more detailed taxonomy of bounded rationality than is usually done. In particular, it is important to distinguish between (i) costly rationality, like the costs of observation, communication, and even computation, that require only an extension of the standard 'Savage Paradigm', and (2) truly bounded rationality, like not knowing the implications of everything that one knows, which as far as I know goes far beyond the Savage paradigm. The second issue, which I shall call indeterminacy, arises in attempts to apply the theory of strategic games to models of organisations, namely, one often faces a very large multiplicity of solutions, which significantly weakens or even destroys the predictive power of the theory. By 'solution' I mean here the socalled non-cooperative equilibrium, usually associated with the names of Auguste Cournot and John Nash (and extended and refined byJohn Harsanyi, Reinhard Selten and others to cover games in which the players have incomplete information). Although indeterminacy can arise even in static games with complete information, it seems to be especially prevalent in dynamic games and/or games with incomplete information. Both of these issues have profound implications for the organisation of the firm. In discussions of this topic, it has become commonplace to set up the straw figure of the firm as a black box, mysteriously choosing production and investment plans to maximise profits. It is true that there are still undergraduate price theory textbooks that present this picture, but economists have long been familiar with the phenomenon of the separation of ownership and management, at least since the publication of the book by Berle and Means (I932), if not before; in fact, Adam Smith had some trenchant comments on it (I776, pp. 264-5). This phenomenon is also referred to as the separation of ownership and control (see Fama and Jensen, I983). Today, one is likely to see this analysed in terms of the 'principal-agent' model, with the owner(s) cast as the principal, and the manager(s) as the agent.