How are one's empirical beliefs justified? Reflection on this question has led some theorists to what has come to be called the regress problem. Thus it has been held that because the justification of any belief requires appeal to additional beliefs and the justification of these beliefs appeal to still further beliefs and so on, attempts at justification can never be carried through. Others, less pessimistic, have pictured justiflcatory chains as looping back on themselves. Here a belief is thought to be justified not by being fastened to some secure mooring, but by its contribution to a selfcontained edifice comprised of other, related beliefs. Like a stone in an arch, it is supported by a collection of a certain sort of which it itself is a crucial element.
Such metaphors have an undoubted appeal. What is less obvious, however, is the extent to which they provide a satisfactory portrayal of the structure of epistemic justification. One is sometimes encouraged to accept the picture on the grounds that it is the only one seriously in the running. An endless chain of justification seems both unworkable and unhelpful: a belief moored on an infinite tether is not moored at all. On the other hand, the prospects of locating foundational beliefs, beliefs capable of generating warrant without themselves requiring the support of other beliefs, seem dim indeed.
Recently, Laurence Bonjour has argued that these prospects are not simply dim, but that they are altogether unimaginable (Bonjour, 1978(Bonjour, , 1980)). The possibility of epistemic justification depends on one's capacity to connect beliefs to other, justified beliefs. Where such linkage is absent, so is justification. The notion that some beliefs might initiate justificatoly chains, the notion, that is, that some beliefs might be epistemically basic or foundational is, according to Bonjour, 'extremely paradoxical' (Bonjour, 1978, p. 5).