I've noticed a common tendency among certain sectors of the blogosphere (in the direction of the Less Wrong crowd) to assume as exhaustive a division of rationality into epistemic and instrumental rationality. The idea is that there is a rationality concerned with how one's beliefs fit the world (epistemic) and there is a rationality concerned with achieving one's goals (instrumental); and the exhaustiveness, which is often assumed with astounding dogmatism (why would one think it obvious that the only rational activities are checking your beliefs and getting what you want?), seems to be based on some notion that it is established by cognitive science somehow. This extends at times to virtual plagiarism of cog. sci. reference books: one gets repeated assurances that "cognitive scientists recognize two kinds of rationality".
Of course, a look at the literature shows that what this actually means is not that cognitive scientists have established that there are no other kinds of rationality, but that they have found it handy for sorting out different kinds of research; that is, it's a practical rather than established distinction. There is no great concern for precise definition of what counts as either -- which is why there are a number of different non-equivalent definitions of both -- and no work to establish exhaustiveness. Nor need there be; it's really there just be useful for sorting out different kinds of research and results into rough groups without worrying too much about precise boundaries.
There is, in short, no good reason to accept the division as exhaustive. More than that, however, trying to treat the division as exhaustive runs into a serious problem that would need to be directly addressed. The question comes down to this: What is the rationality that is involved in coordinating the two rationalities? There are three possibilities.
(a) It could be epistemic.
(b) It could be instrumental.
(c) It could be some third kind that is neither properly epistemic nor properly instrumental.
(a) and (b) both have the result of collapsing the distinction. If the coordinating rationality is epistemic, for instance, than the standards of assessment for how to be instrumentally rational (e.g., given certain beliefs) are epistemic, which means that we are actually assessing the instrumentally rational by epistemic standards. Likewise, if the coordinating rationality is instrumental, the standards of assessment for how to be epistemically rational (e.g., given certain goals) are instrumental, and we are actually assessing the epistemically rational by instrumental standards. There are, in fact, positions on the subject in both direction: i.e., people who think all rationality is really epistemic and people who think all rationality is really instrumental. On both of these accounts the epistemic-instrumental distinction can be nothing but a practical convenience in certain kinds of cases -- an economic classification for certain epistemic situations, or a useful heuristic for certain practical purposes.
If, however, we preserve the distinction as something rooted in reality, the only way to do this is to say that there is another kind of rationality with standards that are not properly either epistemic or instrumental. We could hold that the distinction is brute and primitive, of course, but this would require us to say that there is no way to assess how belief should relate to action, which is absurd in about a thousand different ways, and completely contrary to anyone's actual experience of planning and inquiry, to name just two areas of life where both have to come together. So if the distinction is to be preserved, there has to be at least one other kind of rationality coordinating the distinct provinces in a way suitable for things like inquiry and planning, or, for that matter, just living in general.
This is all old hat, actually. The scholastics recognized that there need to be a sapientia that cut across practical/speculative lines. Kant recognized that in addition to a Critique of Pure Reason and a Critique of Practical Reason, he also needed a Critique of Judgment (which, one notices, quite clearly includes most of the Kantian theory of inquiry). And it is not surprising that this is so; to make the distinction is to say that there is some set of standards defining the universe of discourse, making both epistemic and instrumental rationality kinds of rationality and not just things that happen for reasons of purely historical accident to be called 'rationality'. Likewise, in actual life, the two have to interrelate somehow, and the serious possibilities are just that this mutual coordination is either in one province (in which case that province dominates entirely) or that (neither of the two dominating) it is in some third province. Thus our theory of rationality should either recognize, at root, only one kind of rationality, or it should recognize at least three kinds.
Or rather, to be more exact, one should say that rationality is not exhausted by the epistemic/instrumental distinction: however one glosses the two, one of the two just becomes equivalent to rationality itself or rationality involves more than these two facets, epistemic and instrumental.