In the collection under review, Boghossian (2008) assembles 14 of his papers from the last 20 years.1 They are presented in four groups. The first three groups are focused on, respectively, the nature of mental content, the links of content with self-knowledge and the links of content with a priori knowledge. The two papers of the last group, written with David Velleman, deal with colour and colour concepts. Each group of papers is followed by a bibliography, where responses and possible further reading are listed. Since Boghossian has been a notable contributor on all the topics mentioned, at the very least this collection provides a useful entry point to worthwhile literature on good problems. But one can say much more. Boghossian writes with acuity and ingenuity and his style is clear and direct. He is prepared to challenge and rethink venerable doctrines and is also prepared to revise his own views when good arguments are brought against them. He shows admirable persistence, in returning to topics from different angles, bringing out more implications of the positions explored and the complexity of the issues. He does not at any point gloss over difficulties or pretends to more completeness or decisiveness than the material warrants. So we have here much rigorous and judicious discussion of a thought-provoking kind.
The fourth group of papers, on colour, is not closely linked in theme to the rest of the collection. And although the issues merit further consideration we will not pursue them here. Let me just note that in these papers Boghossian and Velleman lay out two of the more attractive realist accounts of colour, dispositional and physicalist, and argue that no current version of such theories is acceptable.
In the sections which follow, I shall first consider briefly the papers on content and self-knowledge, and after that will turn to the nine papers of the first and third sections on the interlinked topics of content, rule-following and the a priori.
In his Introduction Boghossian writes, with reference to the papers on the nature of content, ‘For all the attention they have received, in my view both Wittgenstein's discussion and Kripke's exposition of it remain underappreciated, to the detriment of current work in philosophy of mind. That is not to say, by any means, that a proper engagement with these arguments would result in a moral that is recognisably ‘Wittgensteinian.’ On the contrary, on my view what emerges from such an engagement is a conception of meaning that is anything but: realist, anti-reductivist and in no obvious way hostile to the idea of a private language. Nevertheless, it seems to me that both Kripke's and Wittgenstein's texts contain considerable insights and it is a pity that they have come to be relatively neglected.’
My own view is that Boghossian is right that there is much of value in Wittgenstein's discussion of meaning and rule-following. But I would not characterize Wittgenstein's ideas (as Boghossian does implicitly in the quoted passage) as anti-realist or reductive. Just possibly Wittgenstein's thought might offer help in seeing ways through some of the perplexities, particularly the metaphysical difficulties, which Boghossian brings to our attention? We shall return to this thought in the final section. But first some more on the themes of the collection.
2. Content and self-knowledge
Boghossian was among the earliest to note the implications of combining externalist accounts of the content of concepts with traditional views about our possession of privileged self-knowledge. Suppose that just sitting in my armchair here and now, I can rightly be wholly certain that some thought of mine is about water. Suppose also (as concept externalism suggests) my thought can be about water only if water exists – and moreover I can from my armchair also rightly be wholly certain about that. Then it seems that, sitting in my armchair here and now, I can refute any external world scepticism, e.g of the malicious demon or brain in a vat kind, which has the non-existence of water as a consequence. Surely something has gone wrong here. Using concept externalism to leverage from epistemic privilege vis-à-vis the inner to epistemic privilege vis-à-vis substantive outer world matters seems a deeply dubious manoeuvre. But what should we give up?
Suppose that what we choose to surrender is traditionally conceived total and privileged knowledge of the nature and content of our own thoughts. The consequence, it seems, is that other parts of our accepted ways of thinking of ourselves come under pressure. With some externalistically individuated concepts (e.g. those of named individuals or natural kinds), if the world changes in ways unnoticed by me, then my concepts change also, in ways unnoticed by me. If I grant this then I must also grant that I cannot, with certainty, tell whether a concept I exercise at one point in thought is the same as or different from a concept I exercise at another point. And if that is so, then what happens to the idea that I can think rationally about individuals or natural kinds? My ability to do so depends on recognizing contradiction and validity in argument, which in turn depends on ability to discriminate identity and difference in concepts used.
In the papers collected here (‘Content and Self-Knowledge’ 1989, ‘The Transparency of Mental Content’ 1994 and ‘What the Externalist Can Know A Priori’ 1997), Boghossian expounds and explores these difficulties. He does not, however, say what way out of the tangle he favours, only that he thinks there are tensions between concept externalism on the one hand and familiar ideas of privileged self-knowledge and rational thought on the other. My own suspicion, however, is that going, broadly speaking, the way externalism indicates would not be as uncomfortable as seems to be imagined and allows us to construct a more realistic view of ourselves than trying to rehabilitate a strong version of privileged self-knowledge or a strong conception of our rational capacities.
To accept concept externalism, in whatever exact form, is to accept that our thinking is more at the mercy of factors outwith our reflective control than, in moments of grandiosity, we might like to suppose. A thoroughgoing concept externalist will take it that there are few, perhaps no, risk-free concepts, since most, perhaps all, concepts build in substantive presuppositions. Moreover, there is no hope of identifying and discharging all of these presuppositions and thereby arriving at a set of concepts which pick out the logico-metaphysical building blocks of all possible worlds and are therefore always guaranteed to be safe to use. A corollary of acknowledging this is acknowledging also that if and when we come to doubt things hitherto taken for granted, then any concepts which presuppose those things should equally become for us suspect cognitive tools.
But granting all this will have unlivably disruptive consequences for our practices of authoritative self-report, and for our sense that attempting to reason is worthwhile, only to the extent that we suppose that the presuppositions underpinning our current concept use actually are widely and undetectably mistaken. And do we suppose that? If we do not, then carrying on as we do now, but with a decent dose of humility in awareness of the fact that our minds and conceptual repertoire are limited and faulty, is both the only thing we can do and something which, on reflection, seems wholly defensible.
Boghossian is surely right to say that there is much more to be thought and said about these topics. But defending the total transparency of the mind to itself, on the grounds that only so can we make sense of any form of first person authority, or only so can we make sense of any rational responsibility for thought, does not seem a promising avenue. The better way forward, I suggest, is to work out what sort, or sorts, of externalism we need to acknowledge, how they affect different types of concepts, and exactly how the traditional picture of the transparent mind needs to be altered.
3. The nature of content
We turn now to the papers in the first and third groups. The collection starts with ‘The Rule-Following Considerations’, a justly classic discussion of the still influential and provocative ideas launched by Kripke (1982). As Boghossian notes, despite the title of Kripke's book, the central question is not most helpfully approached as about rule-following. The best way to get hold of the force of Kripke's considerations is to think of him as trying to elucidate mental content. What is it for a person to have a thought with a determinate content at some time? Amid the deft exposition of the views of Kripke, the heading off of misreadings, the pithy summaries and placement of the ideas of major commentators, two important claims are advanced in this piece, and are carried forward in further discussion.
The first is that Kripke is right in claiming that a reductive account of content in naturalistic terms is not to be had. This theme is pursued in the later ‘Naturalizing Content’, where Boghossian urges that notions like causal co-variation, however fancied up and conditionalized, cannot provide a naturalistic explanation of content. The second is that Kripke's conclusion of meaning scepticism is not acceptable. That theme too is pursued further in a later piece ‘The Status of Content’, in which the incoherence of irrealism about meaning is urged. We will return to the arguments of this paper in Section 6. For the moment let us just note the conclusion.
Putting these two claims together, the upshot is that we are to be non-reductive realists about content. And at this point new challenges appear. Boghossian notes at the end of ‘The Rule-Following Considerations’ (2008: 49–50): ‘But, now, how is an anti-reductionist about content properties to accord them a genuine causal role without committing himself, implausibly, to the essential incompleteness of physics? This is, I believe, the single greatest difficulty for an anti-reductionist conception of content.’ This metaphysical unease about making sense of non-reductive realism about meaning surfaces at other points, and we shall return to it in Section 6.
4. A priori knowledge
The bold and attractive suggestion pursued in the third group of papers (led off by ‘Analyticity Reconsidered’) is that we can rehabilitate from Quinean attack the idea of analytic truth, that is, the idea of a claim of which one can have a priori knowledge in virtue of understanding its meaning.
An initial move is to note that if indeterminacy of meaning is not acceptable (as argued against Kripke) then the same thoughts will show that we cannot accept Quine's arguments for indeterminacy either. Many philosophers seem to think that they can combine the view that meaning is determinate with acceptance of Quine's hostility to the analytic/synthetic distinction. But, says Boghossian (with some plausibility) this is incoherent. Given determinate meaning then there will be determinate sameness of meaning. And, on the basis of that, Frege-analyticity, namely being transformable into a logical truth by the substitution of synonyms for synonyms, can be introduced.
But what about logical truths themselves? Are they too knowable a priori on the basis of meaning? And if so, what meaning is at issue and how does it underpin the epistemology? Boghossian proposes that accepting conceptual role semantics offers the right way to approach these questions. ‘An inference form’ he says ‘can be constitutive of the meaning of one of its ingredient constants C’ (2008: 222). As I read him, what this means is that concepts build in a dynamic element, and the exercise of a concept is to be thought of as the imposition of a distinctive shape on forward-moving development. To make a judgement is to become poised to take up certain further cognitive positions, given the appropriate circumstances. For example, one who believes, ‘If he drops it, it will break’ is poised to judge ‘it will break’ on coming to judge ‘he drops it!’ And semantic theory should include characterizations like ‘A logical constant C expresses that logical object, if any, that makes valid its meaning-constituting inferences’ (2008: 222). (The attractions of conceptual role semantics are easiest to appreciate in the case of logical concepts. But in ‘Blind Reasoning’ Boghossian seems sympathetic to it in connection with substantive descriptive predicates also.)
But how exactly do we parlay conceptual role semantics into an account of how a person is justified in making an inference? And how do we parlay it into an account of how someone might have a priori knowledge that something is a good inference or that something is a logical truth? The first challenge, one which faces us in finding a convincing answer to either question, comes from the fact that we might build mistakes or muddles into the constitution of our concepts. (This is concept externalism cropping up again.) For any concept where this is the case, inferences employing it may be ones we unhesitatingly buy into, through our mistaken confidence in the concept, but they are not ones we are justified in making. And if somehow we came to the conscious reflective view that those inferences were valid that would represent not a priori knowledge but a serious misapprehension. Supposing this first challenge to be dealt with, there is still the second challenge of explaining how mastery of a good concept, conceived as a skill of performing certain cognitive moves, can be transformed into explicit reflective knowledge about the status of moves of that form.
Boghossian begins to confront these challenges, particularly the first of them, with much ingenuity in ‘How are Objective Epistemic Reasons Possible?’ and ‘Blind Reasoning’. Thus, he considers how attempts to discriminate soundly constituted concepts from rogue concepts might seem to be vitiated by circularity, and argues that the difficulty is illusory. He also urges that some concepts are so fundamental that we cannot dispense with them in our thinking and that therefore we must be blameless in using them. But, as he recognizes, there is more to be done. He writes, in the Introduction to the volume, ‘I consider this neo-rationalist program to be in its very earliest stages. We are very far from knowing whether there are any satisfactory meaning-based explanations of the a priori and, if there are, exactly how they are to be formulated and how far they can be made to extend’ (2008: 5). This remark is typical of the avoidance of bombast or overblown claims, which is evident throughout the collection. But Boghossian has done enough to put some interesting ideas on the table and to suggest that the general form of approach here, accepting determinacy of meaning and conceptual role semantics and then developing these materials into an account of a priori knowledge, is promising.
5. Epistemic Rules
‘Epistemic Rules’ is the most recent of the papers in the collection and represents Boghossian's latest thoughts on rule-following. Its main concern is to argue that it is much more difficult than one might suppose to make sense of the idea that when we make inferences we follow epistemic rules. The difficulty is that a natural way of unpacking what it is to follow a rule turns out to be applicable to the case of making an inference only at the cost of provoking an infinite regress. So if this unpacking is the only option, we may have to abandon the idea that making an inference is following a rule.
To see how the problem arises, let us start from a case where no one would doubt that it was intelligible to talk of following a rule. To take Boghossian's example, I might make a rule ‘Answer any email that calls for it immediately upon receipt.’ In this kind of case, my doing the right thing is (plausibly) mediated by my recalling the rule, and inferring from it, together with awareness of the current situation, what I ought now to do. It is the fact that an inferential step takes place in applying the rule which produces a problem. When we apply the same account to inference itself, on the supposition that all inference is rule-following on the just-sketched model of what that involves, we get an infinite regress. Making the inference requires me to consult the rule as to how to do so, which in turn requires me to infer from the rule what I am to do, which requires me to consult another rule. And so on.
Considering possible responses, Boghossian writes on the closing page of the paper (2008:134):
Is Boghossian right to suggest here that primitivism about rule-following goes well beyond taking an anti-reductionist view of mental content? As noted earlier, it is easiest to get hold of the question Kripke confronts us with by talking of content. But if conceptual role semantics is the order of the day then it may turn out that ‘rule-following’ was after all not such a misleading label for Kripke's topic, and an anti-reductionist view of content may turn out to be at the same time a ‘primitivist’ view of rule-following. Boghossian's thoughts about meaning and the a priori cry out for linking with the thoughts about rule-following. If we better understood conceptual role semantics and how it could be parlayed into knowledge of a priori truths, then we might ipso facto better understand rule-following. Both puzzles are arguably, at root, about the kind of multiplicity in unity (aka holism), which has often been noted as distinctive of the mind and about the potential for articulation of that multiplicity which comes with cognitively sophisticated consciousness. Perhaps there are solutions to Boghossian's puzzles to be got without taking on these topics. But I doubt it.
Perhaps we should embrace rule-skepticism, denying that our reasoning is under the influence of general rules?
The trouble is that this seems not only false about reasoning in general, but also unintelligible in connection with deductive inference. It is of the essence of deductive inference that the reasons I have for moving from certain premises to certain conclusions are general ones.
So what we are contemplating, when we contemplate giving up on the rule-following picture of deductive inference, is not so much giving up on a rule-following construal of deductive inference as giving up on deductive inference itself. But that is surely not a stable resting point – didn’t we arrive at the present conclusion through the application of several instances of deductive inference?
The only other option … is to try taking the notion of following – or applying – a rule as primitive …. Notice that this goes well beyond the sort of anti-reductionist response to Kripke's arguments that I was already inclined to favor – an anti-reductionism about mental content.
It would involve a primitivism about rule-following or rule-application itself: we would have to take as primitive a general (often conditional) content serving as the reason for which one believes something, without this being mediated by inference of any kind. It is not obvious that we can make sense of this, but the matter clearly deserves greater consideration.
6. The metaphysics of meaning
The kinds of argument Boghossian pursues lead to conclusions about meaning and rule-following – the untenability of naturalistic reductionism together with the untenability of irrealism – which may be uncongenial given a certain philosophical outlook which favours scientific naturalism and correspondence-theory style realism. As we have already noted, Boghossian voices uneasiness at various points as to whether the conclusions he feels drawn to are really defensible or intelligible. Does some ontologically extravagant and unacceptable form of dualism lurk here?
One way of attempting to deal with the uneasiness would be to look again into the toolbox available to the analytic philosopher of scientific naturalist bent. This toolbox is extremely well equipped. It contains a large set of notions drawn from philosophical logic, such as modality, quantification, sense, reference, etc., in all their varieties, also notions from a familiar metaphysical picture, such as property, individual, law, causation, etc., and many other kinds of device as well. Because it is so well equipped, the extent of the logical space of options it makes available for discussion is vast. So perhaps, somewhere in that logical space, there is a way of dealing with the uneasiness induced by having to postulate mental content which is real but not naturalistically reducible. Perhaps such content can, after all be reduced. Or perhaps the idea of its being irreal can be made more palatable?
It is, however, worth noting that countenancing things which are real but not naturalistically reducible produces pressure for unpalatable dualism, of a kind which postulates extra stuff or forces to compete with those recognized in physics and chemistry, only against the background of correspondence-style assumptions about how to understand talk of reality and truth. Other options become available if we are prepared to move into the territory of deflationary accounts of truth and metaphysical outlooks which reject the idea that the world comes pre-sliced.
Boghossian skirts this territory in ‘The Status of Content’, where he discusses the contrast between robust and deflationary conceptions of truth. He argues that the conditions for a linguistic move being up for assessment as true or false on a deflationary conception of truth are very weak – it is enough if the move looks like a statement. Hence, he urges, given some linguistic move which satisfies this weak condition, one can make a contrast between a factual and a non-factual reading of it only if one is prepared to agree that a deflationary (e.g. a merely disquotational) reading of ‘true’ can be contrasted with a more robust notion of truth, to be explained in terms of whether the predicates in the sentence do or do not denote real properties. But the upshot of accepting this is that non-factualism (irrealism) about meaning trips itself up. Its overt content is that there are no facts about meaning, and hence ‘true’ does not refer to a real property. But in being a non-factualism about a subject matter, that of meaning, it presupposes that ‘true’ does refer to a property.
In the final sections of the paper, Boghossian asks whether the non-factualist about meaning can arrive at a consistent position by becoming a global deflationist about truth – one who denies that there are facts about anything, including meaning. He notes that this goes against the usual motivations for non-factualism about meaning, which bases it in some supposed contrast between the ontologically solid (e.g. the physical) and the ontologically suspect (e.g. the meaningful). But his final move is to argue that deflationism about truth falls victim to the same problem as non-factualism about meaning, namely being self-undermining. In setting itself up as a non-factualism it too presupposes the very notion it purports to deny.
The conclusion to which this piece drives is that we need to be realists about meaning, that no form of non-factualism about it makes sense. But, as we have seen elsewhere, the non-reductive realism which other considerations indicate is something Boghossian is uneasy about. So we are back at the impasse. Where to go from here?
As we have just noted above, the possibility of being deflationist about all kinds of truth is a position discussed briefly in ‘The Status of Content’. Let us consider its corollaries further. As we saw, deflationism does not allow us to contrast the physical and the meaningful in ontological status – one ‘real’ the other ‘not real’. But the general thrust of Boghossian's discussion (as of other discussions of Kripke's ideas) is that the idea of such a contrast of status is incoherent. So although this implication of deflationism is unwelcome to the non-factualists about meaning, to the rest of us it would appear an advantage. Moreover, Boghossian's larger scale position is that of wanting to be a non-reductive realist about meaning. And from this point of view the fact that the position does not provide for a contrast in ontological status between meaning and the subject matter of natural science is also welcome. So for Boghossian's larger project it looks as if deflationism about truth, and the philosophical outlook of which that would be a part, is a promising move.
Why does Boghossian not advance into this territory? The reason may be that there is a price for doing so. That price is making the move which will disable the argument against deflationism advanced in the final pages of ‘The Status of Content’, namely the argument that the deflationary conception of truth undermines itself, since it is a form of non-factualism about meaning. The move which will disable this argument is taking deflationism's criticisms of the opposing view, ‘correspondence theory’ or ‘robust realism about truth’ or whatever we call it, as an unmasking of nonsense rather than as denial of something intelligible which turns out to be false. This is the line of thought which Wittgenstein (at least in many philosophers’ readings of his later writings) invites us to contemplate. This Wittgenstein is not an anti-realist or a reductivist, but an uncoverer of hidden nonsense. Perhaps he can lead to the sunny uplands of an intelligible non-reductivist realism about meaning? But the price is, as indicated, that we must honestly engage in the therapy which enables the unmasking and repudiation of the nonsense. And that therapy, with all its talk of ‘forms of life’ and ‘grammar’ and what not, can seem to be sloppy or murky stuff which tough-minded people should keep away from. On the other hand, if we do keep away from it, the other options seem to be delving in the analytic toolbox again (just more buzzing of the fly in the ever-bigger and more richly furnished fly bottle?) or biting on the bullet of substantive dualism. It is difficult to know which way to go. As Boghossian remarks at several points, there is plenty more work to be done. In these clear-headed and undogmatic papers he sets a fine example of the right style in doing it.