IX* THEORIES OF KNOWLEDGE: AN ANALYTIC FRAMEWORK by Susan Haack Contemporary epistemology must choose between the solid security of the ancient foundationalist pyramid and the risky adventure of the new coherentist raft. (Sosa 1 ) One seems forced to choose between the picture of an elephant which rests on a tortoise (What supports the tortoise?) and the picture of a great Hegelian serpent of knowledge with its tail in its mouth (Where does it begin?). Neither will do. (Sellars 2 ) It is commonly supposed that one is obliged to choose, in the theory of knowledge, between two, fundamentally opposed models: the foundationalist, according to which the rest of our knowledge is supported by the foundation of a relatively small class of epistemically privileged beliefs, and the coherentist, according to which the various parts of our knowledge are mutually supportive. But it is rarely spelled out 3 what the characteristic features of foundationalism and coherentism are, or why they should be thought to be the only feasible options. My first object, in this paper, is to construct a scheme of classification within which one can locate two extreme kinds of theory, which I will call 'Pure Foundationalism' and 'Pure Coherentism', and a whole range of intermediate kinds of theory in between. My second object is to examine, in the context of this framework, the arguments which have dominated the debate between foundationalism and coherentism. This discussion will, I believe, indicate that there are two intermediate types of theory capable of withstanding the arguments of both sides. I shall make it clear which of these I favour; but space will permit me only to hint at the reasons for my preference. Meeting of the Aristotelian Society held at 5/7 Tavistock Place, London WC1, on Monday, 14 March, 1983 at 6.00 p.m.
144 SUSAN HAACK I Let me acknowledge, at the outset, that there will of course be an element of arbitrariness in the characterisations I shall give. But I hope that the characterisations given here will at any rate serve adequately to bring into focus the issues that have been debated between those who call themselves 'foundationalists' and those who call themselves 'coherentists', without distorting the history of this debate any more than is unavoidable. The use of the terms 'foundationalism' and 'coherentism' in the literature has, however, been ill-defined and shifting to such an extent that I doubt that any characterisation could square with all of it. All the theories to be characterised in this section Pure Foundationalism, Pure Coherentism and the intermediate theories share, the assumptions that we do have knowledge, and that our having knowledge requires (at least) our having beliefs which are justified as well as true; where they differ is on how our justified beliefs get their justification. By 'Pure Foundationalism' I shall mean any theory according to which: and: (PF1) Some beliefs are, epistemically, absolutely secure (PF2) All beliefs which are justified, but not absolutely secure (in the sense of (PF1)) are justified wholly by the support, direct or indirect, of these absolutely secure beliefs. I shall call (PF1) without pejorative intent 'Dogmatism'. 4 I have stated Dogmatism in a very general way, because I intend it to cover not just one, but a range of views. All Dogmatists hold that there are some privileged beliefs which enjoy immunity to certain kinds of epistemic defect; they differ, however, with regard to the kind of epistemic defect they have in mind. For example, some have in mind immunity to error (the privileged beliefs are 'certain', that is, roughly, it is impossible that they should be mistakenly held); some, immunity to unwarrantedness (the privileged beliefs are 'self-warranting', that is, roughly, it is impossible that they should be unjustifiedly held); some,
THEORIES OF KNOWLEDGE: AN ANALYTIC FRAMEWORK 145 immunity to ignorance (the privileged beliefs are 'selfmanifesting', that is, roughly, it is impossible that they should fail to be believed when they are true); others, again, hold that there are privileged beliefs which enjoy some two, or even all three, kinds of immunity. (Here, even in the remarks in which I elaborate on my original, vague statement, I am, of course, greatly simplifying an extremely complex matter; a full account would call for careful textual analysis of what has been meant by saying that some beliefs are 'indubitable', 'incorrigible', 'directly evident', 'selfevident', etc., etc.) 5 According to (PF2) the justification of each justified but unprivileged belief depends on the support of some privileged belief or beliefs: either directly (privileged beliefs justify the unprivileged belief in question) or indirectly (privileged beliefs justify unprivileged beliefs,.. which justify the unprivileged belief in question). The crucial point is that all the justification of any justified but unprivileged belief derives, strictly linearly, from the support of privileged beliefs. So I shall call (PF2) the 'Strong Thesis of One-Directional Justification'. (PF1) does not entail (PF2); it could be that, though there are absolutely secure beliefs, they do not serve to support the rest of our knowledge. Neither does (PF2), by itself, entail (PF1); 6 together with the assumption that there are some justified beliefs, however, it does. The negation of (PF1) is: (NPF1) No beliefs are, epistemically, absolutely secure. I shall call (NPF1) 'Fallibilism'. Like'Dogmatism', 'Fallibilism' refers to a family of views, rather than to a single thesis. 7 While each form of Fallibilism is incompatible with the corresponding form of Dogmatism, Fallibilism with regard to one concept of epistemic privilege is compatible with Dogmatism with regard to another. The negation of (PF2) is: (NPF2) Not all beliefs which are justified, but not absolutely secure, are justified wholly by the support of absolutely secure beliefs.
146 SUSAN HAACK Any theory which accepts (or entails) either of (NPF1) or (NPF2) is incompatible with Pure Foundationalism (a qualification with respect to the kind of epistemic privilege concerned is understood here). By 'Pure Coherentism' I shall mean any theory according to which: and: (NPF1) (Fallibilism) (PCI) No belief is, in and of itself, epistemically more secure than any other (PC2) Any belief that is justified is justified wholly in virtue of its belonging to some set of beliefs which has certain properties (consistency, comprehensiveness). I will call (PCI) 'Epistemic Egalitarianism'. Like'Dogmatism' and 'Fallibilism', 'Epistemic Egalitarianism' refers to a class of views, differing with respect to the particular interpretation given to 'epistemically secure'. I will call (PC2) the 'Thesis of Mutual Justification'. It is necessary to stipulate Fallibilism, separately, as a characteristic thesis of Pure Coherentism, because Epistemic Egalitarianism is compatible, strictly speaking, with Dogmatism as well as Fallibilism (but on the Dogmatist assumption that some beliefs are absolutely secure, Epistemic Egalitarianism would have the very peculiar consequence that all beliefs are absolutely secure). However, Fallibilism with respect to warrantoriented privilege is actually a consequence of the Thesis of Mutual Justification, since if all justified beliefs get their justification in the same way, purely by means of mutual support, none are privileged with regard to justification. It is obvious that Pure Coherentism, as I have characterised it, is incompatible with Pure Foundationalism. For, first, it is committed to Fallibilism. And, second, on the assumption that there are justified beliefs, (PC2) entails (NPF2); indeed, that all justified beliefs are justified wholly by mutual support entails not just that not all, but that no beliefs are justified in the way claimed by (PF2).
THEORIES OF KNOWLEDGE: AN ANALYTIC FRAMEWORK 147 In fact Pure Coherentism, as this suggests, is a particularly strong rival of Pure Foundationalism. For, first, it rejects both the characteristic theses of Pure Foundationalism. And, second, its denial of each characteristic thesis of Pure Foundationalism is as strong a denial as it could be: not only are there no beliefs which are epistemically privileged, there aren't even any beliefs which are at least epistemically better off than others; and not just not all, but no justified beliefs get their justification in the way the Pure Foundationalist claims. So we can expect there to be other, intermediate kinds of theory which are rivals both of Pure Foundationalism and of Pure Coherentism. 8 Among the intermediate theories, I shall distinguish two categories: those which retain (PF1) but reject (PF2), and those which (like Pure Coherentism) reject both. (Since, on the assumption that there are justified beliefs, (PF2) entails (PF1), there are no (non-sceptical) theories which retain (PF2) but reject (PF1).) In the first category falls the type of theory I shall call 'Impure Foundationalism', and characterise as holding both: (PF1) (Dogmatism) and: (IPF2) All beliefs which are justified but not absolutely secure are justified at least in part by the support of absolutely secure beliefs; some are also justified in part by mutual relations of support among themselves. Like the Pure Foundationalist, the Impure Foundationalist requires that every unprivileged but justified belief derives from some privileged belief(s); unlike the Pure Foundationalist, however, he does not require that the justification for any justified but unprivileged belief derive exclusively from privileged beliefs. For this reason, I shall call (IPF2) the 'Impure Thesis of One-Directional Justification' (or perhaps 'Thesis of One- Directional Justification with Twiddle' would have been more perspicuous, if less elegant). Impure Foundationalism still allows that epistemically
148 SUSAN HAACK privileged beliefs play a specially important role in the justification of other beliefs. But there is another possible type of position according to which, while there are epistemically privileged beliefs, their support is not essential to the justification of other beliefs: (PF1) (Dogmatism) (NPF2*) No beliefs which arejustified but not absolutely secure are justified even in part by the support of absolutely secure beliefs. ((NPF2*) is, of course, incompatible with (IPF2) as well as with the stronger (PF2).) I shall refer to this as the 'Idle Epistemic Wheels' view. (I say 'view' because, since it is characterised only negatively, it seems inappropriate to call it a theory.) Theories in the second category (rejecting (PF1) as well as (PF2)) deny that there are any absolutely secure beliefs. But it is possible to do this, to be a Fallibilist, without accepting Epistemic Egalitarianism. Fallibilism is quite compatible, also, with the thesis that some beliefs are more secure than others, 'Epistemic Inegalitarianism'; this is: (NPC1) Some beliefs are, in and of themselves, epistemically more secure than others. I shall distinguish three types of Inegalitarian intermediate theory. One type sees justification as essentially one-directional, from the relatively secure to the relatively insecure. I will refer to this as 'Weak Foundationalism', and characterise it as the conjunction of: (NPF1) (NPC1) (Fallibilism) (Epistemic Inegalitarianism) and: (WF2) The justification of less secure beliefs always depends upon the support of more secure beliefs, and not vice-versa
THEORIES OF KNOWLEDGE: AN ANALYTIC FRAMEWORK 149 which I shall call (noting that (PF2) implies (WF2), but not viceversa) the 'Weak Thesis of One-Directional Justification'. The two other types of Inegalitarian theory reject this linear conception of justification, seeing it as, rather, a matter of mutual support. The difference between them is that one, like Pure Coherentism, requires that justification be wholly a matter of relations among beliefs, while the other, like Foundationalist theories, does not. The first type I shall call 'Weak Coherentism', and characterise as the conjunction of: (NPF1) (NPC1) (Fallibilism) (Epistemic Inegalitarianism) and: (WC2) Any belief that is justified is justified wholly in virtue of its relations to other beliefs, the justification of more secure beliefs possibly depending upon the support of less secure beliefs, as well as vice-versa which I shall call the 'Thesis of Weighted Mutual Justification'. The last type of theory to be characterised is, as it were, intermediate between Weak Foundationalism and Weak Coherentism, so I shall call it 'Foundherentism'; (NPF1) (NPC1) (Fallibilism) (Epistemic Inegalitarianism) and: (FH2) The justification of more secure beliefs may depend upon the support of less secure beliefs, as well as vice-versa. I will refer to (FH2) as 'Up-and-Back-ism'. (I note that (WC2) implies (FH2), but not the other way around.) My classification of types of epistemic theory is not exhaustive, and it is exclusive only in a rather restricted sense. It is not exhaustive, because the types of theory sketched all share two assumptions: that we do have knowledge, and that
150 SUSAN HAACK knowledge is (at least) justified, true belief. So sceptical theories, and theories which reject this kind of definition of knowledge, fall outside its scope. 9 (Nor do I claim to have characterised every possible type of theory between my two extremes, though of course I hope I have characterised the most serious options.) The classification is exclusive in this sense: each of the types of theory I have sketched is incompatible with each of the others with respect to any one conception of epistemic privilege, and with respect to the same class of beliefs. The former qualification is needed because, as I have emphasised, there is more than one conception of epistemic privilege (certainty, self-warrant, etc., etc.), and it is consistent to subscribe to different types of theory with respect to different conceptions of epistemic privilege. The latter qualification is needed because, on the assumption of, for example, a radical distinction between logical and empirical knowledge, it would be consistent to subscribe to one type of theory with regard to the one, and another with regard to the other. It might be useful, in view of this, to distinguish between local and global epistemic theories the latter but not the former aspiring to apply to all human knowledge. And with this distinction in hand, it would be possible to characterise 'Contextualism' 10 as a form of local foundationalism: within a context ('form of life', society or whatever) there are privileged beliefs and justification is one-directional. II Perhaps Pure Foundationalism and Pure Coherentism, as I have characterised them, are so extreme that no philosopher has ever seriously held either position. But the extreme theories are nonetheless important to an understanding of the arguments traditionally deployed in the debate between 'foundationalism' and 'coherentism'. (This debate has, of course, been conducted without benefit of the scheme of classification I have just offered; so, in what follows, my use of the terms 'foundationalism' (with a small T) and 'coherentism' (with a small 'c') should not be understood as presupposing that scheme.) Unless foundationalism is true, its sympathisers sometimes
IDLE WHEELS VIEW (PFI) Dogmatism (NPF2 # ) No beliefs are justified asin(pf2) IMPURE FOUNDATIONALISM (PFI) Dogmatism (IPF2) Impure Thesis of One- Directional Justification All beliefs which are justified but not absolutely secure are justified at least in part by the support of absolutely secure beliefs; some also in part by mutual support WEAK FOUNDATIONALISM (NPF1) Fallibilism (NPC1) Epistemic Egalitarianism (WF2) Weak Thesis of One- Directional Justification Justification of less secure beliefs always depends on more secure, not v-v. SCEPTICISM 1, I do not belong to Pure NON-JUSTIFICATIONALl Foundationalism/Pure DEFINITIONS OF ( Coherentism dimension KNOWLEDGE / FOUNDHERENTISM WEAK COHERENTISM PURE COHERENTISM (NPF1) Fallibilism (NPC1) Epistemic Incgalitarianism (FH2) Up-and-Backism Justification of more secure beliefs may depend on less secure beliefs, as well as v-v. (NPF1) Fallibilism (NPC1) Epistemic Incgalitarianism Some beliefs are more secure than others (WC2) Thesis of Weighted Mutual Justification Any justified beliefis justified wholly by mutual support; more secure may depend on less secure beliefs (NPF1) Fallibilism No beliefs are absolutely secure PURE FOUNDATIONALISM (PCI) Epistemic Egalitarianism No beliefis more secure than any other (PC2) Any justified belief is justified wholly in virtue of belonging to a coherent set (PF1) Dogmatism Some beliefs are absolutely secure (PF2) Strong Thesis of One- Directional Justification All beliefs which are justified but not absolutely secure are justified wholly by support of absolutely secure beliefs
152 SUSAN HAACK argue, there is no way in which our beliefs could be justified, and we should have no knowledge. Foundationalism must be accepted, because the only alternative is a surrender to scepticism. I will call this the 'No Tolerable Alternatives Argument'. Of course, the argument depends upon the assumption that scepticism is intolerable. But I shall leave aside the issues raised by this assumption, to concentrate on the claim that foundationalism is the only alternative to scepticism. The claim is that foundationalism offers the only possible account of how a person's beliefs could be justified. Suppose, the argument goes, someone believes something on the basis of certain other things he believes, and that he holds these beliefs on the basis of certain further beliefs,... and so on. If this process eventually ends with beliefs which do not themselves depend on further beliefs, and yet which are themselves secure, the original belief can be justified. But not otherwise; not if the process comes to an end with 'terminal' beliefs which are insecure, or if it does not come to an end at all, but goes round in a circle, with one belief supported by others to which, in turn, it lends support. Now, first, the sort of epistemic privilege in the terminal beliefs which is called for by this argument must, surely, be warrant-oriented. The plausible assumption implicit in the argument is that a person cannot be justified in believing something if he believes it on the basis of something else he is not justified in believing. More particularly, second, what the argument requires is that there be terminal beliefs which are justified, but not justified by the support of other beliefs. For unless the process of justification eventually reaches beliefs which are not dependent on further beliefs, it has not come to an end. So what the No Tolerable Alternatives Argument requires is that there be some beliefs which are, as I shall say, 'independently warranted'. Some foundationalists have, I think, had in mind that experiential beliefs have this status: they are justified, not by the support of other beliefs, but by experience. Other foundationalists, of a less empiricist 11 stripe, have offered other candidates. 'Justified, but not by the support of other beliefs' is how I would make sense of the prima facie rather puzzling claim that
THEORIES OF KNOWLEDGE: AN ANALYTIC FRAMEWORK 153 some beliefs are 'self-justifying'. It is also a possible interpretation of the suggestion that some beliefs 'do not stand in need of justification': they do not need justification by other beliefs, since they get their justification elsewhere. (I suspect, however, that if, at times, the claim that purely experiential beliefs do not stand in need of justification has been meant more literally than this, it has been because of a confusion between an experience itself, and a belief about the experience.) Beliefs which have some other kind of epistemic privilege, but are not independently warranted, may be epistemically Idle Wheels. (I do not want to enter, here, into the very complicated considerations involved in the question of the logical relations between independent warrant and other forms of epistemic privilege. It will suffice to point out that if, if a belief is independently warranted, it follows that it has some other type of epistemic privilege, the weaker type of privilege is, so to speak, implicitly required by the No Tolerable Alternatives Argument; but if, if a belief has some other type of privilege, it follows that it is independently warranted, the stronger type of epistemic privilege is not required, even implicitly, by the No Tolerable Alternatives Argument.) The No Tolerable Alternatives Argument simply dismisses out of hand the idea that the process of justification might 'go round in a circle'. Sympathisers with coherentism might protest that they propose nothing so crass as one beliefs being justified by the support of another, which depends on a further belief,... which depends on the first; the'mutual support'they envisage, to which the foundationalist disingenuously refers as 'going round in a circle', is, rather, a matter of beliefs deriving their justification from the fact that they belong to a suitably cohesive and comprehensive set of beliefs. The second argument I want to consider is an explicit critique of this proposal. The coherentist, the argument goes, tries to give an account of justification purely in terms of the relations among a person's beliefs, with no reference to anything external, over and above the beliefs themselves. But this simply cannot be done; for no such internal relations within a set of beliefs could be sufficient to show them to be justified. There are too many 'coherent' belief sets, and the coherentist is quite unable to discriminate among them. (In what I regard as its most plausible form, this argument
154 SUSAN HAACK points out, in particular, that the coherentist has to allow that a person might be justified in beliefs which are wholly disconnected from his sensory experience, provided only that they cohere internally in an appropriate way.) I will call this the 'Drunken Sailors Argument' (justification, in the coherentist account, is like two drunken sailors back to back what are they standing on?). This makes explicit what, as I suggested, was implicit in the No Tolerable Alternatives Argument: that an acceptable theory of justification must make reference to something beyond a person's beliefs. Neither a Pure nor a Weak Coherentist theory can allow such a reference. Foundationalist and Foundherentist theories, however, can allow it; and those which spell out the notion of (absolute or relative) epistemic privilege in terms of independent warrant do, implicitly, appeal to something outside one's beliefs. Neither of the arguments considered here, however, shows that an acceptable theory must admit absolutely privileged beliefs. One may, I take it, be more or less justified in believing something. The No Tolerable Alternatives Argument does not establish that there must be terminal beliefs which are absolutely secure, in the sense of completely justified, independently of the support of further beliefs. Generally, I take it, justification is, so to speak, 'diluted' as one passes from supporting beliefs to the beliefs they support; I mean, that even if one is completely justified in some initial beliefs, one will usually only be justified to some extent in the further beliefs one holds on their basis. Now, if it were required that in order for a belief to count as knowledge one should be completely justified in holding it, then only epistemically privileged beliefs would qualify as knowledge anyway, and the foundationalist structure of justification would be simply irrelevant. If, on the other hand, it is required only (as seems more plausible) that, in order for a belief to count as knowledge, one must have sufficient justification for holding it, then it is not necessary for the terminal beliefs to be completely justified independently of the support of other beliefs; it is necessary only that they be justified enough to pass on sufficient justification to the rest.
THEORIES OF KNOWLEDGE: AN ANALYTIC FRAMEWORK 155 And it would be sufficient, to meet the Drunken Sailors Argument, to allow that there are some beliefs which are at least partially justified otherwise than by the support of further beliefs; it is not necessary that there be beliefs which are completely justified in this way. So if these two arguments are accepted, they leave some Inegalitarian theories (Weak Foundationalism and Foundherentism) as well as the Dogmatist theories (Pure and Impure Foundationalism) in the field. Let me next consider an argument used by sympathisers with coherentism. Foundationalism, it is sometimes argued, requires there to be beliefs which are both secure, and at the same time rich enough in content to be capable of supporting the rest of one's knowledge. But no beliefs could be both secure enough and rich enough; if any are absolutely secure, it can only be because they have been so stripped of content as to take no epistemic risk, and such beliefs would be quite incapable of serving as the foundations for the rest of one's knowledge. I shall call this the 'Swings and Roundabouts Argument' (what you gain on the swings of security you lose on the roundabouts of content). Though it is vaguer than one would like as to exactly what the relation is between security and content, still, I think, this argument has some plausibility (especially, perhaps, in view of the tendency for foundationalists to restrict the content of their basic beliefs when their claim to security is under threat). The chief point I want to make about the argument here is a simple one: that it is the more plausible the more secure the privileged beliefs are supposed to be, and the greater the burden of the support of unprivileged beliefs they are supposed to bear^this means that it is most persuasive against Pure Foundationalism, which requires its privileged beliefs to be both absolutely secure and wholly responsible for the justification of other beliefs. It is not quite so persuasive against Impure Foundationalism, which, though it requires its privileged beliefs to be absolutely secure, does not require them to be quite wholly responsible for the support of other beliefs. And it need not be a serious threat to Weak Foundationalism, which, not requiring absolute security, can start with more content in its basic beliefs than either the Pure or the Impure forms. So, of the types of theory I have characterised, it looks as if
156 SUSAN HAACK. two Weak Foundationalism and Foundherentism (both: with respect to independent warrant) may be able to survive the arguments of both sides in the traditional debate between foundationalism and coherentism. Of these two types of theory, I regard Foundherentism as the more plausible. Central among my reasons for this preference is that this kind of theory seems to me able to offer a better account of the role of experience in justification. Like Quine (and before him, Dewey and Peirce), I conceive of the theory of knowledge naturalistically, as centrally concerned with the ways in which human beings learn about the world by means of their interactions with it. And, like them, I think of experience as playing an important role in the justification of a person's beliefs. The human subject has experiences, and in the light of those experiences he modifies his beliefs. Some of his beliefs depend largely upon experience, some largely upon other beliefs of his. But none depend solely on experience; for the beliefs he already has affect what beliefs he acquires as a result of experience. Weak Foundationalism, like Foundherentism, is Inegalitarian, and so allows that some beliefs get more of their justification than others from experience. Weak Foundationalism, again like Foundherentism, is Fallibilist, and so denies that any beliefs are completely justified by experience alone. But while Foundherentism can explain that this is because even beliefs very close to experience depend in some small measure upon the support of more 'theoretical' beliefs (e.g., about the conditions of perception), Weak Foundationalism cannot; which is a reason to prefer Foundherentism. This, of course, cuts a very long story very short indeed; but a comprehensive defense of Up-and-Back-ism is beyond the scope of this paper. 12 * * The final version of this paper was written up while I was a Visiting Fellow at the Australian National University. I am very grateful to Howard Burdick for detailed comments on a draft.
THEORIES OF KNOWLEDGE: AN ANALYTIC FRAMEWORK 157 NOTES ' Sosa, E, 'The Raft and the Pyramid', in Midwest Studies in Philosophy, V, ed. French, P. et at., Minnesota UP, p. 3. 2 Sellars, W. 'Empiricism and the Philosophy of Mind', in Science, Perception and Reality, RKP, 1963, p. 170. 3 But cf. Sosa, op. cit., and Pollock, J., 'A Plethora of Epistemological Theories', in Justification and Knowledge, ed. Pappas, G., Reidel, 1979, 93-113. 4 A usage I derive from Peirce. 5 Cf. Alston, W. P., 'Varieties of Privileged Access', American Philosophical Quarterly, 8, 1971,223-41. 'Taking the 'all' as in the predicate calculus. 'In 'Fallibilism and Necessity', Synthese, 41, 1979, 37-63, without benefit of this framework, I used the term 'fallibilism' to refer to one member of this family. 8 Any Egalitarian theory which held that all beliefs got their justification in some other way than that specified by (PC2) would be a rival of Pure Foundationalism as strong as Pure Coherentism; but I know of no serious candidate. 9 See e.g., Goldman, A., 'A Causal Theory of Knowing', Journal of Philosophy, 64, 355-372. Popper would be another example of a writer who rejects the justification requirement; he is a sceptic with respect to justification, but not with respect to knowledge. '"See e.g., Wittgenstein, L., On Certainty, Blackwell, 1969; Annis, D., 'AContextualist Theory of Epistemic Justification', American Philosophical Quarterly, 15, 1978, 213-9. " I would not want this to suggest that I regard the distinction between 'empiricist' and 'rationalist' approaches as unproblematic; it might possibly be amenable to a treatment roughly analogous to the one I have employed in part I of this paper, with approaches according to which all knowledge derives from experience at one extreme and approaches according to which all knowledge derives from reason at the other. 12 1 have discussed this matter in much more detail in two unpublished papers, 'Taking Away the "given"' and 'An Empire State Building Made Out of Toothpicks' (a critique of C. I. Lewis's Impure Foundationalism).