Saturday 30 August 2008

Travel narrative

I’ve just got back from a cycling holiday: cycling the Sustrans Rote 7 from Carlisle to Inverness. There’s something very powerful about the ‘narrative structure’ of such a holiday. If it’s Monday it must be Pitlochry. Further, there’s a simple injunction every day to get on to the next stop. Given the practicalities of coping with the terrain, mechanical failures, the British weather and, in my case, dodgy tendons, such holidays also fit Charlie Brown’s thought: ‘That’s the secret to life... replace one worry with another’.
But, obviously, to think of one’s thoughts on, and about, a cycling holiday as having a ‘narrative structure’ is simply to borrow the structure of travel fiction. Though I seldom seek it out, a handful of travel books, in the broadest of senses, read in my youth, have in turn long suggested a characteristic shape for my attitude to such holidays.

Starting from my earliest memory:

Tove Jansen’s Comet in Moominland. I can’t now recall why, when faced with an impending comet collision, our heroes feel the need to travel to see an astronomer since they only arrive there as the comet swoops down. I guess spurious travel in the face of potential disaster is the shape also of Wells’ War of the Worlds.

Tolkien’s The Lord of the Rings. The early stages of this, before it gets rather more exciting, is a great description of a journey out of the ordinary. The awkwardness of arriving at the Prancing Pony where things are just a little differently done is delightfully described. Although it was more explicitly merely a journey, as a child, The Hobbit never had the same resonance for me.

There must have been some travel in all the Herman Hesse I read in my teens but I suspect it’s best forgotten.

Pirsig’s Zen and the Art of Motorcycle Maintenance. Although memorably described by my sister in law as ‘that dreadful hippie book’, this was the book that first got me interested in philosophy. But it’s better as a great, if slightly nerdy, book about how to travel by motorcycle.

Kerouac’s On the Road. The book, a conversation about which, got me through my university undergraduate interview without therefore having to reveal my ignorance of my chosen subject.

Hunter S Thompson’s Fear and Loathing in Las Vegas. The flipside of On the Road.

But then, since my twenties, I seem to have stopped reading such things. Perhaps Sebald's The Rings of Saturn counts. Instead of fiction, travel guides have become much more important. I can constantly return to reading A.W. Wainwright’s Pictorial Guides to the Lake District. They stand in for the real thing when I'm not there. (After only a moment’s deliberation it was obvious I’d buy all of the new revisions by Chris Jesty, a fellow Kendalian who is walking every path, as they come out. I’m awaiting the eight volume Fellranger books (pictured) by Mark Richards in much the way I’d await a new Radiohead album or McDowell paper.)

On my recent cycle holiday, Sustrans’ wonderfully efficient pair of maps to cover the whole of the 430 miles took the role of a graphic representation of what we had covered as well as what was to come. They were not just a tool to navigate by but rather seemed to carry within them the past and future of the trip itself in the way that Wittgenstein suggests that, for those who are not aspect-blind, the picture of a loved one stands in for that person, goes proxy, whilst a car blueprint does not stand in for the car.

But the most extreme case of this was when, years ago, I walked Wainwright’s Coast to Coast with my partner, Lois. The guidebook itself took on a central role in conversations in the pub at the end of each day. It became an example of the idea of the extended mind as our conversations increasingly took the form of simply pointing to the pages to come, or pages past. (The book thus took on an essential role in constituting our thoughts.) And thus I became increasingly anxious that if we lost the book, it would not only be a practical problem navigating but also would disrupt the conversational structure of the holiday.

Tuesday 19 August 2008

SEESHOP 2 postscript

As I was leaving the workshop today, I had a quick further conversation with Harry about what he thought of Theresa Shilhab’s presentation. His main thought was that there might be things which, for merely contingent reasons, contributory experts (mothers, in the example) might not have commented upon and were thus not available to interactional experts (non-mother midwives). Thus, I think, he aimed to defuse the threat of there being something in principle only available to the former by suggesting that the evidence presented might of something merely in practice unavailable.

He didn’t, however, like my yesterday’s generalisation of that thought: that experience equips subjects with knowledge in specific ways and those ways may be detectable in the tests. He suggested that the guidelines for running imitation-game tests included not asking about specific information to sidestep that sort of worry. Rather, the claim was that interactional experts could acquire the ‘conceptual structure’ of relevant contributory expertise. This seems to me to be a tricky point, however.

Of course, by connecting the theoretical claim to a grasp of conceptual structure, one could now invoke a fairly standard reading of the later Wittgenstein to argue that experience alone is not sufficient for grasp of grammar. (That’s the point of the critique of a philosophical use of ostensive definition to explain reference in the early stages of the Investigations. It is also in play – albeit with some other assumptions about what is available – in the private language argument.) So, if experience alone is what distinguishes mothers from non-mothers, this fact would seem to be insufficient to equip only the former with a grasp of the conceptual structure, or grammar, of the experience of motherhood (whatever that would be). But whilst that helps defend Collins’ point with some a priori philosophy, it makes me suspect that that this isn’t really what he wants. I understand his claims about interactional expertise to be contingent. Hence all the experiments.

Surely, therefore, the claim isn’t just about conceptual structure or grammar but rather about the ability to make some moves in accord with that grammar. Interactional experts can make the very same moves as contributory experts. So I wonder whether there is a distorting effect here of the empirical test used: the imitation game. Testing the claim that interactional experts can in principle acquire the full linguistically codified knowledge of contributory experts (not their non-linguistic tacit knowledge, of course) by seeing whether in practice they always, or even usually, pass the imitation game seems wrong.

One problem is, to repeat, that experience equips subjects with a particular route to knowledge and this might affect any actual imitation game. Good interactional expertise might be aspirational, it might take lots of ongoing work. If so, the claim would be this: that there are no knowledge claims in principle unavailable to that stance. In other words, there is no propositional knowledge that cannot be acquired by testimony. For a few seconds that seemed to me, if true, to rule out specific embodied but still propositional knowledge of what motherhood is like and thus to be a substantial claim. But on reflection it seems a piffling point. Testimony enables one to know any propositional knowledge, given the right interlocutor. So what can the claim really be?

For a change to the official picture, see this update.

Monday 18 August 2008

Studies of Expertise and Experience

I’m very grateful to have been invited by Harry Collins (pictured)and Rob Evans to the second international workshop on studies on studies of expertise and experience (SEESHOP2) in (a very wet) Cardiff. For me the specific interest is in the connection between expertise and tacit knowledge. I’m interested simply to learn what is going on in this area of sociological theory especially because of Collins’ Wittgensteinian leanings.

The role of expertise is central within the so called Third Wave of sociological analysis of science. As I discovered yesterday, the Third Wave came about in part as a reaction to a conversation Harry had on a bus in the US. He was challenged by some cultural anthropologists to say what help sociology of scientific knowledge could be in the case of AIDS policy in Africa, given that, eg., some men believed that sex with a virgin would cure them. The Second Wave adopts a methodological relativism: it is confidant that any historical belief could be shown to be rational by suitable contextualisation. Scientific beliefs are not treated as in any sense a special kind of belief. But that seems no help in guiding ongoing public policy decisions.

The Third Wave aims to take account of the special value generally ascribed to science and to suggest some policy tools for making judgements in the face of scientific debate and disagreement. Central to that is the issue of working out who has the role of an expert and how experts, of different sorts, can be differentiated from others. So expertise isn’t merely a matter of attribution. It is a genuine feature or ability of some people in the debate. (Collins suggested with some sense of irony that in setting out the factors that should guide policy decisions, he found himself approaching Karl Popper’s Open Society and its Enemies, Popper’s methodological philosophy of science having been left behind by sociology of science thirty years ago.)

So described, I wondered why the Cardiff movement didn’t simply give up Wave Two’s relativism and say that these were tools for getting, albeit fallibly, to the truth, since truth is surely the aim of science policy. (Over a beer, Rob Evans suggested that that might be acceptable providing that truth were pretty minimally construed. As long as it wasn’t a correspondence notion of truth with a capital T. But that’s fine, since that’s the right thing to think about truth.)

Equally, I wonder why a commitment to a non-relational, non-attributional theory of expertise – realism about expertise! – doesn’t leech into realism about ground level science claims; whether, in other words, Wave Three is inconsistent with Wave Two. Perhaps that will become clearer to me. In the meantime, however, I’ve heard an interesting mix or methodological papers and papers examining the transfer and management of tacit knowledge in real settings, from nickel factories to physics laboratories.

As set out in Collins and Evans 2007, there are a variety of kinds of expertise. Especially relevant to science policy and management is ‘interactional expertise’: non-coal face, non-contributory expertise but an ability to speak the relevant language and thus, if necessary, to bluff. Empirically testing this, Collins and Evans spend quite a bit of time running Turing-style imitation games to see, eg, whether blind people can bluff sighted people and vice versa (theory predicts the blind can fool the sighted, having been linguistically socialised, but the sighted cannot fool the blinds). Collins expressed some surprise that the journal that published their much cited first paper on the Third Wave, refused to publish this work, but I can imagine that it doesn’t seem much like sociology of knowledge.

In an unintended reflective application of interactional expertise, Collins presented the slides of a psychologist who could not be present and whose specific training he lacked. Ironically, no one seemed to think that this presentation went very well.

One presentation that took a line against Collins and Evans was by Theresa Schilhab. She reported an imitation game study on mother and non-mother midwives and non-midwives. Apparently, although non-mother midwives could fool non-midwives into thinking they were mothers, non-mother midwives could not fool midwives. This, she suggested, was probably because mothers had a kind of embodied cognition of what giving birth was actually like which they could reveal and which could not be imitated. Such embodied cognition could not in general be captured in merely interactional ways.

I’m not sure what to make of this experiment but two things strike me. Firstly, it might not be a matter of mothers having an embodied cognition. It might be merely that giving birth has non-cognitive, perhaps affective, effects on a subject that can later be detected. Perhaps being a mother has emotional effects – such as calmness under pressure – that are hard to fake. If so, the result would not have interesting conclusions for the study of expertise. ((Theresa said she called any effect ‘cognition’ but that has the advantages merely of theft over honest toil, in Russell’s phrase.))

Secondly, and thinking more generally, having particular experiences need not give rise to a distinct kind of (embodied) knowledge but can clearly give rise to a distinct source of knowledge. So a visual memory of a seating plan may enable a ‘contributory expert’ to say that Evan was to the right of Kyle and that Barry was opposite both. But an interactional expert who hears this can know – via standard testimonial transmission – just the same facts. They need not know them with a different kind of – as it were embodied – content, especially if they have vivid visual imagination (although their justifications will be different). But, and this was always implicit in the contrast between interactional and contributory experts, the former have always to await the claims of the latter. If so, the idea of embodied knowledge may not be as interesting as it sounds and may not run counter to Collins and Evans’ claims.

See also the postscript.

Monday 11 August 2008

Reply to Neil Levy

(Christian Perring has kindly offered me a right of reply to a very negative review which he published. Here's my draft.)

I am grateful to Neil Levy for engaging with the underlying philosophical themes of my book even if not with the details of its arguments.

To take one example of this lack, Levy ascribes to me the argument that since ‘we live in a space of reasons, a non-normative account of mental illness cannot be given’. This is merely a caricature of the discussion of chapter 1 which turns on the question of whether biological functions can be used to reduce the normative notions particular to illness, especially mental illness. Having, it seems, more faith than I have in evolutionary theory to defend reductionism, Levy must disagree with my conclusions but it is a pity he has not engaged more closely with the arguments.

In order to set up one of his two challenges, he ascribes to me the same kind of brisk dismissal of reductionist accounts of meaning: ‘since we live in the space of reasons we cannot give accounts of significance according to which they are built up from non-significant elements’. But again this ignores my actual arguments, [eg the 5 page discussion ibid: 129-133] (against what I concede in the text is still a live research project [ibid: 129]) which look, inter alia, to the problems reductionists have answering the ‘disjunction problem’ or showing how causal theories of reference are of the right form to explain intentionality.

Levy likewise merely parodies my argument for moral particularism. I cannot claim originality for this line of thought (I cite John McDowell) but the argument [ibid: 68-71] starts from the prima facie appearance that such value judgements cannot be codified and then considers why this appearance is often disregarded in philosophy. The discussion of codified judgements is meant to challenge a deep seated prejudice about rationality which can blind us also to the nature of uncodified judgements. Thus when Levy ascribes to me the view that arithmetic also is uncodifiable he simply ignores the text itself which says:

Thus even in the case of a judgement that can be codified – by the axioms of Peano arithmetic, for example – the principle itself has to be applied through a kind of practical judgement. Whatever principles can be used to encode such practice their application relies on practical judgements which are not themselves codified. Wittgenstein’s discussion should undermine the prejudice that wherever there is a rational judgement it must be encoded in a principle, since principles themselves cannot govern judgement unaided. Without the prejudice, however, there is no reason to doubt the appearance that value judgements are made without a close framework of principles. [ibid: 71]

Levy suggests that the main target of my book is ‘an entire metaphysics and philosophy of mind’ rather than an engagement with particular views within philosophy of psychiatry. Whilst he is right that I oppose reductionism, his description is again an unhelpful way of presenting the way the general and particular inform each other. To take just one area, my discussions of the different accounts of delusion offered by Jaspers, Maher, Frith, Sass, Campbell, Davies and Bolton and Hill turn on the particular details of their theories and models. There can be no one size fits all response. The overall argument for a relaxed (non-reductive) naturalism emerges out of piecemeal discussion of key authors and key issues within the philosophy of psychiatry.

This is not the place to attempt to resolve the longstanding debate about the prospects of reductionist naturalism but let me suggest a way to respond to Levy’s remaining challenge. He asks how we could have evolved to grasp real values, worrying that ‘If there are values independent of us, we have no reason to think that what we take to be values actually are’. If, however, one takes seriously the idea of values as genuine features of the world then there might well turn out to be an evolutionary account of grasp of them. (Since we have no wish to explain values in other, eg evolutionary, terms, this issue is not pressing for non-reductionist, relaxed naturalists. Reductionists face a bigger challenge.) But if so it had better not eliminate the potential gap between being right and merely seeming right that characterises any judgement that aims to track real features of the world. Levy’s sceptical worry is both under motivated and misplaced.

In addition to misrepresenting my book, I fear Levy also misrepresents the abilities of those with an interest in the philosophy of mental health and psychiatry. My ten years’ experience of teaching psychiatrists, mental health nurses and service users, with a passionate interest in the fundamental issues surrounding mental healthcare but without a philosophy background, has in turn taught me that such students thrive by getting stuck into the complexities of the real debates and real figures, from Jaspers to EBM. Understanding my little book should be no problem at all, to them at least.

Wednesday 6 August 2008

Charles Travis ‘The twilight of empiricism’

Following on from Adrian Moore’s Points of View, I’ve been looking at Charles Travis’ (pictured) (2003) ‘The twilight of empiricismProceedings of the Aristotelian Society 104: 247-272. Although I find Travis a struggle to read (working out whether, eg., ‘that’ is functioning to introduce a clause or as a demonstrative adjective seems oddly difficult in his writing; or the oddness, to my mind at least, of his verb, object, infinitive construction (“hear him to say” rather than “hearing that he says”)), there is always an argument worth investigating. In this paper there are a number of main themes (bigger summary here):

1: There’s an articulation and diagnosis of Quine’s argument for his pragmatist empiricism, suggesting it stems from an application of the argument from illusion. The argument from illusion constrains what Quine thinks can properly be experienced and hence calls for a revisionary treatment of how experience properly constrains judgement.

2: Later there’s an argument to the effect that Quine’s attempt to moderate the effects of the argument from illusion through the notion of pragmatic and holistic adjustments to beliefs in response to recalcitrant experience fails because he has no right to the notion of recalcitrance. (This is akin to a discussion of Quine by Crispin Wright.) The argument from illusion needs, instead, to be blocked.

3: Sandwiched between these is a description of two arguments from Frege against the kind of private experience that the empiricist application of the argument from illusion gives rise to. One concerns the failure of private ideas to have the right kind of open ended systematic import to model thoughts and meanings. The other concerns the fact that private ideas cannot be used to generate the model of maximally general and universal structure that Frege thinks is required by logic.

4: Finally Travis examines the idea of parochial rather than private thoughts: thought that can only be shared by some subjects (humans rather than Martians, for example). Whilst Frege’s argument from (his account of) the nature of logic would rule these out (they would be immune to Frege’s other argument against privacy), Travis argues that Austin provides good grounds for thinking that the truth of descriptions is occasion-sensitive and thus parochial. Thus we need a different model of logic that can accommodate occasion-sensitivity. Meanwhile occasion-sensitivity can be used to frame a response (not a million miles from a disjunctivist response) to the argument from illusion.

Putting to one side the discussion of Quine, I’d like to understand whether this defence of parochialism against what Travis calls the Martian principle runs counter to Adrian Moore’s defence of absolute representations. The Martian Principle is this:

For any answerable stance, any thinker must be able, in principle, to grasp when it will have answered; to see something in the way things are or might be, on which its having answered or not turns—thus to see it to be answerable. [265]

Travis argues that occasion-sensitivity (as defended by him and by Austin) conflicts with this principle.

[O]ne can be all a thinker must be, and all that logic captures of that, and, further, one may grasp all that one must grasp to count as having the relevant concepts—one may know all there is to know as to what being coloured red would be as such—and for all that, one may not yet be in a position to see whether a given judging, or stating, say, that the drapes are red is to count as having answered to the way things are. [262]

I take it that what is going on here is that a subject cannot tell whether the answerable stance of another subject, expressed in a verbal judgement, say, has been answered by a state of affairs.

So now suppose a given candidate thinker simply could not grasp certain standards for a given stance’s having answered—certain things it might be, say, for drapes to count as red. Suppose he was intractably blind to what those standards demanded for truth. He could not, say, catch on to the way stains would matter to the truth of the particular judgement in question. Would that disqualify him as a thinker, or even as a grasper of the concept red, etc.? Austin’s point entails that it need not. [262]

My hunch, however, is that this point is orthogonal to Moore’s defence of absolute representations, representations that can be added without danger of conflicting points of view. Moore distinguishes between the the conditions of the production of a representation ‘and the role that the representation can play in such processes as… integration’ [Moore 1997: 89] He argues that perspectivalness of the former has no effect on the latter. So we can imagine responding to Travis by agreeing that the words alone, expressive of an answerable state / a representation of the world, do not pin down what it means context-free. But given the context and thus given the content, which might be expressed in those or other words in a different context, the representation might or might not be absolute in Moore’s sense.

More generally, in trying to get the relation of contingency – as said to be found at the heart of the later Wittgenstein’s discussion of the conceptual order – and an absolute conception of the world clear, my hunch is that there is a much less direct connection than I used to think.

Travis’ version of such contingency – occasion-sensitivity – seems perfectly describable and sayable without putting at risk ground level normativity (centrally, in this case, of how an ‘answerable stance’ answers to the world). But it does not conflict with absolute representations. Darker, transcendent, claims about contingency may conflict with absoluteness – they imply an ineliminable point of view – but at the same time they have problems of their own as Williams pointed out some years ago.

[I]f our talk of numbers has been determined by our decisions, then one result of our decisions is that it must be nonsense to say that anything about a number has been determined by our decisions. [Williams, B. (1982) ‘Wittgenstein and idealism’ in Moral luck, Cambridge: Cambridge University Press: 95]

I still don't believe in absolute representations, but not because of contingency.

See this entry on ‘A sense of occasion’, this on ‘Reason’s reach’, this on ‘The twilight of empiricism’, and this on the discussion of rule following in Thought’s Footing.

Tuesday 5 August 2008

Contingency and perspectivism in Wittgenstein

I’ve been reading Adrian Moore’s Points of View (Moore, A.W. (1997) Points of View, Oxford: Oxford University Press) as part of a general attempt to get clear on Wittgenstein and transcendental idealism. In the first instance, I wanted to see what he takes the connection to be between that and Williams’ Absolute Conception. My assumption was that his defence of the possibility of absolute representations would simply conflict with any kind of role for idealism in Wittgenstein but things are not so simple.

The first main thread is the argument for absolute representations. Defined in part in opposition to (essentially and inherently) perspectival representations and partly through the argument for their existence, they are representations that can be added together without problems stemming from presupposing distinct and incompatible points of view. There is a very brisk argument in chapter 4 for their existence [73-4]. What seems to drive the argument as well as motivate it is Williams’ original idea that a commitment to an absolute conception grows out of a self-conscious understanding of how knowledge answers to something that is ‘there anyway’.

If reality is something substantial that representations answer to, the same reality in every case, then not only must it be possible to provide an account of the … for any possible true representation, but the part of this account that is used for the indirect endorsement of the representation must be combinable with every other part into a single conception of reality – call it C… The requirement of combinability here is the requirement that the representations constituting C should all be from the same points of view, if any… Now consider any possible true representation ρ from any point of view π. One of the members of C must be derived from the account of how ρ is made true by reality. This account, since it serves for pitting ρ against any possible true representation from a point of view incompatible with π, cannot itself be from π. So, given that all the members of C are from the same points of view, none of them can be from π. But π was chosen arbitrarily. So none of the members of C can be from any point of view. Absolute representations are possible. [74]

Underpinning this argument is the questionable idea that distinct perspectival representations can be endorsed from a kind of highest common factor conception of reality (contrast McDowell’s discussion of Williams). In addition, two other thoughts strike me: that there is something odd about how true perspectival representations relate to the non-perspectival world (since Moore, contra Nagel, denies that perspectivism or perspectivalness attaches to reality itself); and that the independence of reality from representations has to inflate in this way. Why should a conception of an objective world have to say more than a compendious statement of instances of the T-schema?

But as well as arguing for the possibility of absolute representations, Moore also wants to explain why a contrary view is attractive. He does this in chapters 5 and 6 in which he examines arguments for the perspectivalness of our representational practices and our language. One interesting comment is that our meta-level concepts of truth and representation may be perspectival but that that need not affect the status of our representations.

One key move is to distinguish between the conditions of the production of a representation ‘and the role that the representation can play in such processes as… integration’ [89] and to argue that the perspectivalness of the former has no effect on the latter.

A similar approach is sued to head off the worry that the meanings of sentences used to frame representations of the world are essentially context sensitive and thus expressive of a point of view. ‘There is no inconsistency in the idea that an absolute representation should be produced by using a sentence which can also be used, in another context, metaphorically perhaps, to produce a representation with some quite different content – and thus of some quite different type. All that matters, as ever, is the role that the original representation plays, or might play.’ [99]

But the real surprise begins in chapter 6. The main line of argument here is that transcendental idealism (in Kant) or something like it (which can perhaps be found in Wittgenstein) is self stultifying and thus simply nonsensical. But it gives rise in the next chapter to consideration that it might be true but unsayable. Again, Moore rejects this but suggests that a description of transcendental idealism would be an apt thing to try to say to put into words something ineffably known. He is explicit that he is not attempting to rehabilitate the truth of transcendental idealism:

The claim is not that transcendental idealism, though true, cannot be stated. The claim is rather that transcendental idealism, though incoherent, is the result of an unsuccessful attempt to state what cannot be stated. [158]

But there is something positive to be said about its role in what is a genuine insight:

Now for Wittgenstein… clarity of understanding was the main goal of philosophy. Clarity of understanding was to be sought in those cases where, for whatever reason, we misconstrue the logic of our own language and become bemused and confused by nonsense.. To achieve such clarity of understanding we must focus self-consciously on our understanding, that is on our understanding of our language. We must self-consciously re-activate it. But then… we not only have ineffable knowledge, we are shown something. What we are shown depends on what we are led to when we try to put what we clearly understand into words. And that, here as in Kant, is transcendental idealism. [161]

Why is there anything akin to transcendental idealism in play in the characterisation of the ineffable knowledge (in the failure to put it into words, in the form of words that is, in the end, strictly nonsense in an everyday, resolute, sense)? Moore thinks that one positive effect of reading Wittgenstein is to see that our concepts depend on contingent practices. That is where Wittgenstein’s opposition to Platonism takes us.

Focusing self-consciously on our understanding, we recognize the deep contingencies that sustain it… Our understanding has nothing to answer to. It is part of how we receive the world… If we do achieve such clarity, then what we actually get into focus is an arrangement of interlocking, mutually supporting practices that are grounded in one another’s contingency, a complex knotted structure that might easily have been different. [162]

The book then goes on to characterise a little further how a ‘saying versus showing’ distinction can help characterise the ineffable knowledge; the way such knowledge does not answer to something independent of a person’s having it; and the way ordinary nonsense has to be mentioned in the forlorn attempt to say what is, properly, only shown to a person who has the knowledge, in a forlorn, again, effort to put the knowledge into words.

But the key thing seems to be that there is some genuine, even if ineffable, knowledge, somewhere in the region of transcendental idealism (insofar as that that nonsense is what we reach for to try to put the ineffable knowledge into words) and which stems from the Wittgensteinian contingency about our representational practices, even whilst this does not rule out there being absolute representations.

For reasons I must try to get clear, I begin to doubt that it is possible to say anything positive about a supposed lesson about contingency in the later Wittgenstein. As soon as one invokes Cavell's whirl of organism, or Lear's mindedness, to explain why we go on in just one of the myriad ways that make up, eg., Kripkean sceptical possibilities and thus make the conceptual order depend on contingencies (that we have our whirl of organism, our mindedness), one lapses into something false.

Even for empirical concepts, theorising that the correct application of ‘red’ depend on such contingencies, rather than simply tracking its extension, looks wrong. Firstly, invoking such contengincies threatens the normativity of the correct use of ‘red’ to speak of red things (well, roughly). Secondly, it looks to be an instance of scheme-content dualism. The account of contingency explains the application of ‘red’ in purely scheme-side terms, not in terms of what is red.

So whilst talk of contingency might be a way to counter a content-side (of scheme-content dualism) invocation of rampant platonism, it surely cannot be a positive move in its own right. But if not, how precisely, should we say that there is any contingency in a properly Wittgensteinian description of language use?