Scepticism and Justification - page 2

Written by A.C. Grayling.

There is a platitudinous but rather vague answer, already suggested. It is that our sharing a conceptual scheme which none of us individually could have generated, is a function of our belonging to a community whose chief instrument of community is language. Communal possession of language plays the major role in enabling community members to apportion epistemic tasks, to process and record the results, and to put them to use. There is no suggestion here that there cannot be any sort of shared conceptual scheme in the absence of language, for manifestly there are good naturalistic accounts to be given of languageless creatures displaying concept-applying behaviour in common with others of their kind, frequently in ways indispensable to their co-operative interactions. But we have no reason to suppose that a non-language-mediated conceptual scheme has complexity above a certain level, and it is barely controversial that the one we (humans) possess would be impossible without language.

It would not strictly be incorrect to describe the finitary predicament in terms of the deficiency or incompleteness of our information relative to our practical epistemic needs, in particular our having to choose courses of action. This is a formulation from which discussions of the ampliative character of induction often begin, and it serves as a statement of the dilemma posed by the joint fact of our pressing need for techniques of ampliative inference and their imperfection relative to deductive standards. But it would or at least could be misleading to begin this way, because the problem which confronts us is not so much the deficiency of our state of information about the world as, at a quite different level, its completeness. At the level of detail–of particular matters–our knowledge is radically deficient. But at a highly general level there is a background of assumptions, some of a structural nature, against which our ordinary thought about the world proceeds and which makes it possible; and these jointly take the character of an overarching picture of the world which we hold steady as the interpretative frame for the first, deficient, state of knowledge. It is the character of the overarching picture–the framework–and its relation to the latter, which invites attention here.


So: we are taking seriously the fact that the cognitive capacities of individuals are finite, and asking what follows from this for an understanding of the global theories we apply to the domains over which we exercise those capacities. This is not a question about how, in the light of epistemic finitude, individuals come to have and use a putatively inclusive explanatory theory of the world; rather, it is a question of what work that theory – that conceptual scheme – does. Privileging the second question over the first is something we have been taught by Kant to do: he pointed out that the crucial question concerns not how we get our concepts, but what role they play.

The question can be formulated in alternative ways to bring out others of the concerns which the fact of epistemic finitude prompts. For one important example, we can approach the task by asking what status we can suppose our conceptual scheme to have, given its radical underdetermination by the evidence which the subscribers to that scheme can acquire in the course of the activities which bear upon the verification or falsification of the commitments (the beliefs and theories) in which the scheme consists.

It is evident from the finitary character of individual cognitive powers that the conceptual scheme we employ would be at least extremely difficult to acquire–arguably: impossible to acquire–if it were left to individuals to construct it for themselves from their own resources–if the notion of such a proceeding were, in the light of the Private Language Argument, intelligible. The range of an individual's powers is restricted to his current perceptual environment and whatever of past experience and future expectation his limited powers of recall and inference can provide. Without supplement, these powers would (whether or not equally) at best very weakly support a large number of widely divergent interpretations of what they give their possessor access (or supposed access) to.

This shows the special interest of the discrepancy between the finitary predicament of any of us taken individually, and the richness of the conceptual scheme we each in fact employ. What makes possession of such a scheme possible? One ready and persuasive answer comes from reminding ourselves that we are not isolated individuals, but members of an epistemic community whose chief instrument of community is language. Language enables members of the community to share epistemic tasks and to process, record and utilise the results. There are certain obvious ways in which that process can be portrayed: Popper, for example, with a certain literal-mindedness, thinks of libraries (or, more generally, data banks of various kinds) as embodying the outcomes of the community's joint epistemic activities over time. This must be partly right. But what is more to the present point is that the central role of linguistic competence, in making possible the difference between a rich conceptual scheme and individual finitude, suggests that the key lies in what goes into possession of that competence. And the thought must be that linguistic competence essentially involves or instantiates a theory of the world which enables any speaker of the language to interpret, or indeed to have, experience of the world in that way. Another way to put this is to view linguistic competence as a sixth epistemic modality, where by 'epistemic modality' I mean a means of acquiring, interpreting, processing, storing and transmitting the information yielded by experience and reflection on experience. So considered, linguistic competence is a vastly more powerful epistemic modality than the five other, strictly sensory, modalities; it is what provides them with their framework of operation. What is needed is an account of the scheme instantiated by application of linguistic competence.


One thing we immediately recognise about the scheme is that, as already noted, it is realist in character. Despite appearances and much misleading debate, realism is an epistemological thesis asserting the independence of the objects of discourse from discourse itself. More precisely, it asserts that relations between thought and its objects, perception and its targets, knowledge and what is known, acts of referring and referents–call them 'mind-world' relations, although in fact they are all different if intimately connected–are external or contingent ones.

We can recognise, as fundamental to understanding the way in which our discourse works, assumptions of the kind listed earlier about the world being a law-governed realm of causally interacting spatio-temporal particulars, and events or other entities typically individuated by reference to these. Moreover, commitments in these respects reveal why ordinary discourse invites accounts of reference and truth which are distinctively realist, for a view of the foregoing kind about the domain over which our discourse ranges is very naturally interpretable in terms of those familiar views about the links between referring devices and things, and between sentences and objectively obtaining truth-conditions, which are presupposed to much recent discussion in this area: namely, that reference works on a naming-paradigm in causally direct ways, and that truth in some sense consists in fit between what we say or think about the domain over which discourse ranges, and the domain itself.

Whatever particular difficulties affect giving an exact account of the ontology and the semantics invited by this realist picture, it is at least clear that it constitutes a simple and powerful view which on the whole successfully sustains the demands made on it by experience–a strong pragmatic justification for it. That is a fact which is independent of debates about whether the realist commitments of the scheme are literally true or not, a question upon which much turns; but for present purposes they can be left aside, because we need only note that we are construing the commitments weakly as assumptions of the scheme. Whether they are taken as literally true or merely as assumptions, the scheme's justificatory character remains. It becomes a matter for the second task, identified earlier, to settle this question of 'literal truth', that is, how these first-order facts about the scheme are to be interpreted in the light of the sceptical problem which gives that second task its content.


Collecting the suggestions already made, we can venture the following as a first approximation of what the framework of ordinary epistemic practice looks like, treated as a justificatory scheme. Such a scheme is in effect an inferential scheme, representable as providing security for familiar practices of basing judgements on evidence. In standard thinking about these matters, empirical judgements are thought of as inductively based on the evidence for them, and as being defeasible to the degree that the evidence is partial. But we have just noted that the scheme consists of a set of assumptions about the nature of the world over which our experience ranges, and we have further noted that these assumptions include some to the effect that the world is lawlike and independent. Add these assumptions to statements of evidence as supporting premises, and the logical picture changes: we see that the form of reasoning being employed is enthymematic deduction on the covering-law model.

At its roughest, the picture is something like this. A judgement about some particular matter of fact is inferred from the evidence for that judgement (reported by evidential premises) in the presence of more general premises about the kinds of things in question, and even more general standing premises about the world (background premises). As such the form of reasoning is representable as deductive: the conjunction of evidential and background premises entails the judgement. But, of course, empirical judgements are defeasible, which appears to conflict with the idea that inferences to them take deductive form. The answer lies in noting, firstly, that background premises have to carry ceteris paribus clauses, or clauses about normal conditions; and secondly, that evidential premises are only as good as the evidence they report, and here the usual finitary constraints apply. Accordingly we can be, and often enough we are, wrong in our judgements. But we can often measure the degree of confidence that we repose in our judgements, by taking into account the relevant defeating possibilities inherent in either or both the evidence that evidential premises report, and the stability of the normal conditions assumed in background premises. This is where this picture saves what is persuasive about conceptions of probability, and in effect 'solves the problem of induction' by suggesting that all reasoning is always deductive in form. (For example, inferences by analogy assume uniformity of nature grounds–and so for other non-enumerative inductive forms.)

It is illuminating to think of Aristotelian classification as, obliquely, among the forerunners of this idea of an inferential scheme. Two reasons why there is only heuristic value in remarking the connection are these: the Aristotelian system of classification by genus and species is too neat (too simplistic) even for the domain where it has most plausibility, namely, the biological domain; and secondly, discussion of it came to be distracted, perhaps not unnaturally, into discussion of definition, and in that guise the objections to it are many and obvious. Some are tellingly summarised by Locke in the Essay (not every term can be precisely explained by two other terms giving genus and differentia; and some words cannot be lexically defined, on pain among other things of regress and failure to constrain their meanings by reference to extralinguistic considerations ). But there is much that is suggestive there (and in later logic: see the Kneales on medieval theories of assumption). Note for example the striking resemblance between the Tree of Porphyry and the structure of reasoning employed in the game of Twenty Questions, in which moderately skilled questioners can identify any individual spatio-temporal object usually in fewer than twenty steps, exploiting classificatory conventions governing our picture of the world together with appropriate cognitive strategies. What both seem to capture is the sense that inferences about matters of fact proceed to their conclusions by way of the deductive inferential structure outlined: premises of relatively great generality are conjoined with premises of relatively lesser generality, including particular ones, in any number of steps within practical constraints, to yield, in a highly reliable way even in the face of defeasibility considerations, judgements about matters of fact.


These remarks are merely schematic, and only gesture in the direction of a research programme. But the implications of such an approach are clear. The quest for an account of justification is satisfied by this picture, as explaining how it consists in inherence in a scheme or framework. Full statements of justification proceed, via a report of the relevant evidence in the case, to appeal to the scheme as a whole: 'this', we might in the end say in such a transaction, 'is how (we think) the world is', and that has to satisfy the sceptic at that level of enquiry. He might then–he should–raise his sights to the question of the justification of the scheme as a whole, but that, as noted is a different and further matter. (Some, like Carnap and Wittgenstein in their different ways, would take it that there can be no such higher task.) But then the problem of scepticism comes to be seen as arguably it ought to be seen: as the problem of relativism.

The standard difficulties concerning justification can be taken implicitly to specify desiderata that have to be satisfied by any adequate theory of justification. They are not best satisfied by attempting to defeat the defeaters proposed by sceptical arguments of the familiar sort, as earlier epistemology often tried to do, not untypically on a blow-by-blow basis. Rather, they are satisfied by a positive theory which shows how justification is secured. A theory like the present one, which postulates ultimate justification by reference to the assumptions of the scheme, serving as foundational premises from which, together with other premises, judgements of lesser generality are deduced, accordingly satisfies these desiderata. It does so conditionally upon resolution of the higher order question about the overall justification of the scheme itself, of course, which is where the determining connection of the first to the second order enterprises becomes manifest: it shows that something like a transcendental deduction of the assumptions of the scheme, together with an argument that any alternative scheme can only be a variant of this one (the modality is seriously intended), is required to deal with scepticism fully. Once again, this higher-order task is sometimes claimed to have been effected already in the lower-order one, on the grounds that only the lower-order one is possible anyway; but that is precisely one of the central controversies of recent epistemology.

Among much else that might be said about the strategy outlined here, I shall mention in conclusion just one. In earlier debates about empirical knowledge it was pointed out that reasoning about matters of fact does not take the form of conjunctions of evidential and background premises entailing judgements. Judgements, especially perceptual ones, are typically immediate, appearing to consist in the exercise of well-rehearsed, experience-based recognitional capacities. As a description of how empirical judgement feels (so to speak), this is surely correct; but it involves a confusion of psychological and logical facts about its structure. True, we do not as a matter of psychological fact usually to go through processes of inference in such cases; but we can see that if challenged to justify a judgement we would have to state our grounds and, if pressed, the background assumptions that constitute them as grounds. In a full story of this kind we would have the conceptual scheme qua inferential framework fully present, and would be able to trace inferential routes to the judgement itself.


The nub of the claim here is that we can redescribe the problem of justification as the problem of epistemic finitude, and, by seeing how such finitude is overcome–viz. by our possession and employment of a realistic conceptual scheme which serves as a geographical-historical explanatory framework designed to make experience coherent, serving as a framework of inference, specifically deductive in form, in which the general assumptions of the scheme play an undischargeable role–we thereby see how we come by justification for our workaday epistemic judgements. And we thereby also identify where the major philosophical task in this region lies: namely, in justifying the scheme itself, which is the same thing as refuting scepticism in its most interesting and substantial guise, namely, as relativism. This is a different problem of justification, but it only comes to the fore when epistemology's traditional problem of justification has been dealt with in the way suggested here.