1 One of the few disappointments in an otherwise fine book is the little BonJour says about what makes one system of belief more coherent than another and the complete absence of any discussion of what makes an alternative system of belief available to one. BonJour cites as factors making for coherence: logical consistency, degree of probabilistic consistency, number and strength of inferential connections, degree to which the belief system contains subsystems of belief which are inferentially unconnected with each other, and degree of unexplained anomalies in the believed contents of the system. Are the members of any inconsistent system of actually held beliefs unjustified? If so, few, if any, of us would be justified in believing anything. Are only the members of the most coherent consistent subsystem of my actual system of beliefs justified? Yet, if my actual system contains inconsistencies, how can it be said that I have, even tacitly, a reflective grasp of the coherence of my most coherent subsystem and how can it be said that this grasp is my reason for accepting the members of the system? It would seem that considerations of coherence don't play all that much of a role in the generation of my beliefs.

Clearly the work of scholars in one's community - historians, social scientists, art critics, and natural scientists - provide alternatives to one's system of belief. If all such scholarly products - or the most coherent overall assemblage of them - are available to one, then no one's system of belief contains justified beliefs since any number of constructions from such scholarly products will constitute more coherent alternatives to what one actually believes. (Nor can it be said that all those beliefs which would be preserved in more coherent alternatives are justified. It can hardly be a grasp of the coherence of the most coherent alternative which constitutes one's reason for having the beliefs one does.) Even if one thinks alternatives must be understandable to one to be available, there will be few restrictions on the number of available alternatives. (Physics might not be understandable to many but history surely is.) Perhaps available alternatives are only those alternatives which are considered. Why, however, isn't it epistemically irresponsible not to consider certain alternatives so that a person's beliefs may be unjustified even if they are no less coherent than alternatives he considers? (BonJour raises a similar objection to Goldman [48).) Should we say that available alternatives are those it is reasonable for us to consider given what we already believe, especially given what we already believe about what sorts of factors are relevant to what and thus what possibilities are to be considered in answering questions? If so, it would seem to be impossible for there to be any radical alternatives available to one, alternatives containing different beliefs about relevance, and thus it would seem impossible for one to be justified in adopting such an alternative in the face of what one already believes. Yet, paradoxically, if one were to do so (inspired by some charismatic guru or visions of angels or just a blow on the head) one might have a more coherent and thus more justified belief system than if one had considered only alternatives it was rational for one to do so, given what one already believed. Isn't it self-delusory to regard one's present as the Triumph of Reason (by virtue of having a highly coherent belief system) if one's history is hardly the March of Reason? BonJour begs off questions about the elements of coherence on the grounds that most, if not all, foundationalist theories also must address them since most allow that coherence plays at least some role in justification. This is fair enough. However, it seems that foundationalists may have an easier time answering them since ways of answering them will be constrained by a base line of foundational beliefs.

2 The conclusion ‘Belief B is highly likely to be true’ will have premises of the form (1) ‘B has feature ϕ’ and (2) ‘Beliefs with ϕ are highly likely to be true.’ The need for justifiably believing (1) underlies BonJour's claim that justified believers must grasp the coherence of their beliefs; the need for justifiably believing (2) underlies BonJour's claim that justified believers must grasp the truth-conduciveness of coherence. It might be noted that in order to undermine foundationalism one need require only that justified believers justifiably believe (1), but not (2). However why should one require a grasp of (1) on the part of justified believers but not of (2) unless (2) is not essential to the epistemic relevance of ϕ On the other hand, why should one think (2) is essential but that a cognitive grasp of (2) is not, unless epistemic responsibility isn't the core of the notion of epistemic justification at all. (In that case a grasp of [1] is not required for justification either.) BonJour's requirement that justified believers justifiably believe both (1) and (2) seems well-motivated by his conception of epistemic justification.

3 It is important to realize that for BonJour externalism is simply the denial of premise (3) in the argument and foundationalism simply the affirmation of empirically basic beliefs as defined in premise (1). The former thus includes views according to which the empirical justification of beliefs about the world depends solely on logical or causal relations among internal states of believers, e.g. experiences or other beliefs, but (3) is still false. The latter thus includes views according to which beliefs of the same semantic type or content (e.g. first person pain report) may be empirically basic or not, depending on the circumstances in which they are held. (See Pollock, John Contemporary Theories of Knowledge (Totawa, NJ: Rowan & Littlefield 1987Google Scholar) for an alternative classification). Throughout this review I shall follow BonJour's usage.

Whatever the labels, premise (3) has important methodological significance. BonJour rejects Chisholm’s critical cognitivism whereby the task of the epistemologist is to discover the sources of the justification we have for our beliefs by reflecting on justified and unjustified beliefs and laying out an account of the conditions under which beliefs are justified and unjustified. He argues that unless the standards met by commonsense beliefs are given the sort of meta-justificatory argument he demands the results will be epistemologically arbitrary and represent a ‘question-begging and dogmatic’(13) dismissal of the possibility of scepticism. However critical cognitivists will argue that there are non-arbitrary starting points for reflection on the sources of justification and discovery of the criteria of justification. They will insist that justified belief requires that standards are met, but deny premise (3) as indicated above. (See Chisholm, R.M. ‘Knowing that one Knows’ in Foundations of Knowing [Minneapolis: University of Minnesota Press 1982]Google Scholar, Cleve, James van ‘Foundationalism, Epistemic Principles, and the Cartesian Circle’ Philosophical Review [1979]Google Scholar, and also see Alston, W. ‘Epistemic Circularity’ in Philosophy and Phenomenological Research [1986]Google Scholar.)

4 BonJour's account of a priori intuition and its relation to the understanding of a proposition is puzzling on many counts. He says, ‘to understand a proposition is to grasp the web of necessary connections with which it is essentially bound up’(207). Presumably understanding the necessary connections of a necessarily true proposition with other propositions requires grasping the necessary truth of the proposition in question: otherwise one would fail to grasp some of its necessary connections with other propositions by failing to grasp how its truth is connected with the truth or falsity of other propositions. Yet understanding a proposition cannot be a matter of grasping all of its necessary connections with other propositions, e.g. all of its logical consequences or all the contingent propositions which follow from it given the truth of other contingent propositions. For these two reasons, some independent account of understanding a proposition must be given to make clear the role of understanding in a priori justification.

5 BonJour says nothing in explication of the crucial notions of complete understanding and optimum case of understanding.

6 Alternatively, an epistemically ideal believer may be one who has good reasons for all of his beliefs, so that a less than ideal believer is one who approximates the ideal by having good reasons for some of his beliefs. However, simply by closely approximating this ideal, beliefs for which one nonetheless lacks good reason do not acquire ‘some positive epistemic status.’ The issue BonJour is faced with is how we have any good reason for any of our beliefs to acquire any ‘positive epistemic status.’

7 Cf. 154, quoted above in the opening section. Also see p. 152 where BonJour suggests that in many cases the elements of coherentist justification ‘are tacitly or implicitly involved in the actual cognitive state of a person who has empirical knowledge, even though he does not bring them to mind and indeed would normally be unable to do so even if explicitly challenged.’ In footnote 1 above, I raise some questions about the extent to which condition (i) is ever met and thus about the extent to which the ‘elements of coherentist justification’ are ever involved in ‘the actual cognitive state’ of persons.

8 Not all approximations to an epistemic ideal confer positive epistemic status on the approximations. True beliefs, even firmly held, constitute some sort of approximation to knowledge but don't thereby have any positive epistemic status unless they meet the requirements for some degree of justification, specifically according to BonJour's premise (2), that one have some reason to believe them true. Thus even if meeting condition (A), and thus conditions (i) and (ii), is part of an epistemic ideal of having beliefs on the basis of a fully explicit grasp of (i) and (ii), it won't thereby be an approximation which confers some positive epistemic status on the resultant beliefs unless one also has some positive reason for thinking (i) and (ii) obtain.

9 In a fifth case an historian draws conclusions on the basis of massive and apparently conclusive research but, nonetheless possesses a crystal ball which unbeknownst to him gives highly reliable answers to questions on the subject matter in question and would, in the case at hand, give an answer contrary to the his conclusion. The historian, BonJour suggests, is justified in his conclusion despite the ‘availability’ of a reliable procedure - consulting the crystal ball - which would have produced the opposite result. It seems open to BonJour's extemalist opponent to argue that in each of the cases the belief in question is produced by a process which (we know) is unreliable. This explains our intuitive responses apparently at odds with externalism. Believing something as a result of a process when one has massive evidence against the belief, no reason to believe that such processes are reliable either in one's own case or in general, or just no evidence for or against the reliability of the process, are themselves unreliable processes. (Perhaps even perception in such cases in not reliable, but in that case surely our intuitions are that perceptual beliefs are not justified in such cases.) The problem with these maneuvers is that there are any number of ways of describing the process which produces a belief on any given occasion and that relative to some descriptions the process is reliable but relative to others it is not. Yet there seems to be no principled way given by the externalist to determine which of the descriptions is the correct one relative to which the reliability of the process, indeed the nature of the process itself, is to be determined.) See R.M. Chisholm, ‘A Version of Foundationalism’ in Foundations of Knowing (Minnesota: University of Minnesota Press 1982) and John Pollock, ‘Reliability and Justified Belief’ Canadian Journal of Philosophy (1984).

10 Logically weaker because implied by but not implying the ‘Kantian’ model. Some philosophers would urge, that the epistemic justification of a belief does not require being guided by considerations of epistemic rationality in any sense. See, for example, Feldman, Richard and Conee, Earl ‘Evidentialism’ in Philosophical Studies (1985)CrossRefGoogle Scholar. My concern here is to determine whether plausible explications of BonJour's intuitions about epistemic justification and its requirements lead to his views about the structure of epistemic justification. (I think Feldman‘s and Conee’ s account of ‘well-founded belief’ - the notion they think authors like BonJour are really concerned with - should be amended in the light of what I say.)

11 One might say that one's beliefs are not being produced in an ‘epistemically arbitrary’ manner but rather than one is being guided by ‘epistemic rationality’ in one's believings so long as one's beliefs manifest habits of mind and reasoning whose function is to produce epistemically correct or virtuous belief. Larry Wright suggests an account of the relevant notion of function: ‘when we say that Z is the function of X, we are not only saying that X is there because it does Z, we are also saying that Z is (or happens as) a result or consequence of X's being there’ (‘Functions,’ Philosophical Review 1973, 160). (The potential relevance of Wright's work on functions for issues in epistemology was brought to my attention by my colleague Mohan Matthen.)

In any case if the account of epistemic correctness is ultimately deontological, correct standards will simply be standards conformity to which makes a proposition warranted or reasonable for one to believe. (See Firth, Roderick ‘Are Epistemic Concepts Reducible to Ethical Concepts’ in Values and Morals, Kim, and Goldman, eds. (Dordrecht: Reidel 1978)Google Scholar.) Deontologists must swallow, in an account either of epistemic responsibility or rationality guidedness, non-natural explanations of habits of mind - facts of reason serving, as Kant seems to think intelligible, as peculiar, sui generis causes. (In Philosophy and the Mirror of Nature [Princeton: Princeton University Press 1979] Richard Rorty has explored some puzzles concerning the Platonic version of the model according to which a rational person is guided in his actions and judgements not just by a representation of Forms but by the Forms themselves via a direct grasp of them.) If epistemic virtue is not irreducibly normative but, as teleologists like BonJour think, a matter of conduciveness to truth and the avoidance of error, then the explanation of the possession of cognitive habits in terms of their epistemic virtue may be thoroughly naturalistic.

12 See Russell's suggestions in Problems of Philosophy (Oxford: Oxford University Press 1967), Chapter 12, 77-8. Versions of the above possibilities are discussed in Alston, W. ‘Internalism and Externalism in Epistemology,’ Philosophical Topics (1986)CrossRefGoogle Scholar.

13 ‘The normal justificatory issue on this view is whether the believer is justified in holding a certain belief which he does in fact hold… since the basic unit of justification for a coherence theory is an entire system of belief, the analogous claim within the context of such a position is that the raising of an issue of empirical justification presupposes the existence of some specifiable system of belief’ (103). These claims are certainly peculiar as they stand. To be sure, one can't raise the issue ‘I believe that p. Am I justified in so doing?’ unless one thinks that one believes that p. But one need not actually believe that p to raise and discuss the issue. In any case, surely the theist is no more in the ‘normal position’ for raising the question of the justifiability of belief in God than is the agnostic or atheist.

14 Thus it is quite consistent with BonJour's Observation Requirement that no cognitively spontaneous beliefs be warranted because no class of such beliefs turn out to be in internal agreement to any significant degree. In this way BonJour's view is subtly different from a modest foundationalist theory according to which certain forms of observation are prima facie justified - epistemically acceptable so long as other acceptable observation judgments do not undermine them.

15 Many contextualists will say that it is a basic feature of ordinary epistemic practice that in raising and answering questions of justification we appeal to what we believe to be true including alleged facts about the world which are not in question. Thus sceptical questions about the adequacy of one's practice as a whole cannot coherently be raised and need not be taken seriously. BonJour forswears such contextualism on the grounds that it allows one to provide no non-questionbegging reason for thinking beliefs sanctioned by one's other beliefs (and by what one's peers believe and don't take issue with) are true and thus ‘destroys the claim of the theory to be an account of epistemic justification or acceptability in the first place’ (14). The force of this reply is obviously blunted by BonJour's later adoption of the Doxastic Presumption. I suspect BonJour's contextualist opponents - e.g. Rorty (see above), and Michael Williams, ‘Coherence, Justification and Truth,’ Review of Metaphysics (1980) - would wonder what could sanction strictures on the requirements of epistemic justification other than the practice of asking questions concerning what we call ‘justification’

Why then isn't BonJour's rejection of Rorty's contextualism and adoption of his own ad hoc? BonJour claims that his Presumption describes ‘from the outside something I unavoidably do’ (105) - assume my overall grasp of my system of beliefs correct - and ‘reflects’ an ‘aspect of cognitive practice.’ Does this mean BonJour construes the issue between himself and Rorty as partly an empirical psychological one about the nature and possibilities of epistemic practice? Certain empirical issues should give BonJour pause. Do persons ever have a ‘representation’ (104) of their system of beliefs as a whole at any given moment of epistemic reflection rather than a representation of certain specific beliefs - a very small portion of their total belief system - which they take to be relevant to the question of justification at hand? Yet, according to a coherence theory, it is always the former - the total belief system- which is relevant. Moreover, in information processing is there inevitably as much ultimate processing of (represented) non-doxastic information as of doxastic information? The issue between Rorty's contextualism and BonJour's are further explored in footnote 17.

16 BonJour claims that epistemic reflection begins with a ‘tacit representation’ (104) of what beliefs we have. If this is so, it also begins with a tacit representation of what apparent recollections we have of what we believed in the past. By what right can BonJour claim that until a specific question about our cognitive history is raised, there is a presumption in favour of the accuracy of our representation of our cognitive history without also conceding to Rorty and Williams that until a specific question is raised about our environment in general, there is a presumption in favour of the accuracy of our representation of our environment?

17 BonJour is appealing to the principle

In comparing rival hypotheses on the same evidence when the evidence is equally likely on either hypothesis, the issue turns on the prior probabilities of the hypotheses.

18 One might wonder how the system of belief could fail to represent its environment accurately if it developed in response to it. Why should one characterize the contents of the beliefs in the system so that they are inaccurate, rather than more broadly so that they are accurate? Possibly then the conceptual possibilities should be accounts of how systems of belief develop in response to an environment it accurately represents but which changes so that it no longer accurately represents it. So long as sufficient accuracy to secure survival is maintained e.g., enough food is detected to live whether what is thought to be food is food more often than not - then it may be easier not to develop ways of recognizing error - the fact that much of what one thinks to be food does not satisfy is either not recognized or is not pressing enough - than to develop more accurate systems of thought. Clearly the coherence of such possibilities can be challenged from the standpoint of strong causal theories of the content of beliefs or conceptual role theories of the content of belief. Why don't the contents of the beliefs alter as the environment to which they respond alters or as the way they affect and are affected by sensory stimuli and other beliefs alters? However, I doubt that BonJour can help himself to such accounts of content since they make his whole meta-justificatory project unnecessary. First, verificationists don't need metajustificatory arguments for criteria of justification. Second, if the content of beliefs is determined by the situations which typically give rise to them then knowing what one believes gives one a reason for thinking one's belief likely to be true. That one's beliefs were likely true would be guaranteed for BonJour by the Doxastic Presumption.