John O'Leary-Hawthorne and Huw Price

[This is a draft of a paper which appeared in The Australasian Journal of Philosophy, June, 1996.]

Is non-cognitivism compatible with minimalism about truth? A contemporary argument claims not, and therefore that moral realists, for example, should take heart from the popularity of semantic minimalism. The same is said to apply to cognitivism about other topics--conditionals, for example--for the argument depends only on the fact that ordinary usage applies the notions of truth and falsity to utterances of the kind in question. Given this much, minimalism about truth is said to leave no room for the view that the utterances concerned are non-cognitive in nature.[1]

In this paper we want to derail this fast-track route to cognitivism. We want to show that with a proper understanding of what is essential to non-cognitivism, the position turns out to be largely untouched by the adoption of any of a range of minimalist views about truth. The issue as to the nature of non-cognitivism is crucial, however, and we begin in Sect. I below by defending a broader characterisation of the position than is common in contemporary literature. The nature of minimalism also calls for clarification, and in Sect. II we distinguish two importantly different strands which are both prominent in contemporary debates. Against this background, we go on to explore two possible strategies for standing up for non-cognitivism in the face of minimalism.

One of these strategies has been propounded in a recent paper by Frank Jackson, Graham Oppy and Michael Smith.[2] It turns on the idea that minimalism about truth is quite compatible with a non-minimalism about truth-aptness, and that the latter can be used to ground non-cognitivism. In Sect. III we discuss this strategy in some detail. We argue that it is less general and more vulnerable to a minimalist counterattack than Jackson, Oppy and Smith suppose; and, worse, that one of its central planks seems constitutionally unsuited to the weight it is required to bear.

Our own strategy, which we introduce in Sect. I and go on to elaborate in Sect. IV, depends on our broader than usual conception of the essential character of non-cognitivism. In Sect. IV we illustrate this broader conception with a number of examples from the contemporary literature. We show that it makes non-cognitivism--or a doctrine recognisably descended from non-cognitivism--a very tenacious position, though one which does the same sort of philosophical work as its more traditional ancestors.

In the last two sections of the paper we turn to some important ramifications of this broader conception of non-cognitivism. Sect. V is concerned with some important questions about language--about truth, in particular--that seem especially salient from the theoretical perspective associated with the second strategy. We argue that some minimalists and many non-cognitivists are at fault in ignoring these questions. And in Sect. VI, finally, we turn to an issue that provides a broader motivation for the present discussion. We think that our second strategy for defending non-cognitivism points to a serious difficulty for an increasingly popular reductionist strategy in metaphysics--an approach based on the Ramsey-Lewis approach to theoretical terms--of which Frank Jackson is one the leading contemporary proponents. Briefly, the preconditions for an application of the Ramsey-Lewis reductionist program are much harder to establish with non-cognitivism understood from our theoretical perspective than under more conventional characterisations. The linguistic issues raised in this paper thus turn out to be of considerable importance to contemporary metaphysics.


I. What is Non-Cognitivism?

Non-cognitivism about a particular family of terms presupposes a more general distinction: that between cognitive and non-cognitive discourse in general. In effect, the standard literature offers two possible answers to the question as to how this distinction is to be characterised. One view is that it is primarily a semantic distinction, to be defined in terms of possession of truth conditions, or some such. The other view is that it is primarily a psychological distinction, to be defined in terms of the nature of associated psychological states--whether they are genuine beliefs, for example. Either view might well hold that the characterisation offered by the other is correct, although derivative or secondary. Indeed, a possible third view is that the relevant semantic and psychological notions are tied together in such a way that neither can really claim precedence. In practice, writers tend to slide from semantic to psychological characterisations and back again, without saying where, if anywhere, they think the priority lies.

Let us call these two varieties of non-cognitivism the semantically defined form and the psychologically defined form, respectively. What do they have in common? Indeed, by what right do we treat these two positions as different forms of the same basic view? Most basically, we suggest, what these positions have in common is that they characterise a linguistic function, or category, in terms of which the non-cognitivist may claim that the disputed sentences serve a different function from, or belong to a different category to, other parts of language (and in particular, to paradigmatic causal-explanatory parts of language). In one case the category is characterised in terms of truth, in the other case in terms of belief, but the common factor is the functional or categorial distinction itself. In support of this diagnosis, note that it is a categorial distinction of this sort that underpins the philosophical impact of non-cognitivism, particularly in opposition to reductionist or eliminativist moves. Non-cognitivists argue that these programs, and the philosophical concerns from which they arise, rest on a distinctive kind of mistake about language--on a misidentification of the linguistic category within which particular families of concepts have their home. (More on this in Sect. VI below.)

Once we see that in principle there might be a number of different ways to characterise the relevant distinctions in language--we cannot be sure in advance that the only possibilities are those provided by the familiar semantic and psychological notions--it is easy to see that non-cognitivism is really a collection or genus of possible positions, sharing the thesis that philosophy is prone to category mistakes of these kinds, but differing in how the categories should be characterised. The familiar semantic and psychological forms are two of the species of this common genus, but there might well be others.

Thus we want to suggest that non-cognitivism is best construed in more general terms than is usual in the literature--as the genus, rather than as either or both of its common sub-species. Nothing hangs on the terminology, of course. We could quite well use another name for the genus, if anyone preferred to reserve `non-cognitivism' for the semantically or psychologically defined versions. What matters is that it is the functional characterisation of the genus as a whole that captures what is philosophically most interesting here, and that to fail to notice the genus is to fail to see the potential strength of the general philosophical strategy. Once we notice the genus, moreover, it is easy to appreciate the possibility of a powerful strategy for defending non-cognitivism against minimalism: simply find some third way of characterising the cognitive/non-cognitive distinction, which will be capable of bearing the weight the non-cognitivist wants to place on it, even if notions such as truth and belief are conceded to the minimalist. In the ethical case, for example, the non-cognitivist will be immune, prima facie, from the objections raised by minimalists about truth or belief, provided that she characterises the distinctive function of ethical judgements in other terms.

We might call this way of defending non-cognitivism the `third leg' strategy, for it rests on the idea that non-cognitivism itself might be defined in neither of the two ways common in the contemporary literature. The third leg strategy is easy to appreciate, at least in abstract, once the general functional characterisation of non-cognitivism is in view, but is almost invisible otherwise. All the same, there are many proposals in the contemporary literature which illustrate how the strategy might operate in practice, and we shall describe some of these in Sect. 4. The strategy itself has been clearly identified by Paul Horwich. Responding to the charge that semantic minimalism would trivialise `a substantive issue in meta-ethics', Horwich says


the moral here is not that minimalism and emotivism are incompatible, but that emotivism should be re-formulated. For a minimalist could quite easily accept the correctness and philosophical importance of the emotivists' central insights: namely, that the function and assertibility conditions of certain ethical claims are fundamentally different from those of empirical, explanatory descriptions, and that an appreciation of the difference will help to resolve philosophical problems surrounding the notion of an ethical fact. My point is that this position need not, and should not, be formulated in such a way as to preclude the minimalist conception of truth.[3]

We shall come back to the third leg strategy in Sect. IV. Before that, however, in order to display the comparative strength of this approach, we want to consider the nature of minimalism, as it figures in these debates; and we want to criticise a rival approach suggested by Jackson, Oppy and Smith.


II. What is Minimalism?

Semantic minimalists reject traditional philosophical concerns with the nature of truth, arguing that the traditional project rests on a misconception. This rejection comes in two importantly different forms, however. On one view, the traditional mistake is to assume that there is a `thick' or `substantial' property of truth, the understanding of the nature of which is the main goal of a philosophical theory of truth. This view does not deny that there is a place for philosophy in understanding the function of the truth predicate in socio-linguistic practice--it merely denies that this function is a matter of making reference to a substantial property, whose nature is therefore a matter for philosophical concern.

What is it for a property to be `substantial', or `thick'? The following gloss fits a number of minimalist sources: a substantial property is one that enjoys a causal-explanatory role in mature scientific theory. Thus the minimalist's thought is something like this: serious science puts the category tree in explanations of sundry phenomena, including indeed our use of the term `tree'; whereas serious science doesn't need to refer to any real property of truth, either to explain our ordinary use of the term `true', or for any other causal-explanatory purpose. Instead, for example, it might be suggested that the primary function of the truth predicate is as a grammatical device that facilitates a kind of generality in language which would otherwise be difficult to express.

Views of this kind might be developed in a number of ways. For present purposes, however, we want to draw attention to a distinction between this general strand of minimalism, and a second strand, which is in some ways more radical. Often called `quietism', this second strand rests on the view that philosophy really has nothing to say about certain of its traditional concerns. In particular, it asserts that philosophy cannot be revisionary of folk usage about the matters in question, taken at face value.

In contemporary philosophy this quietist view is often associated (perhaps wrongly[4]) with the work of the later Wittgenstein. It is easy to see how it is thought to challenge non-cognitivism. After all, one of the standard characteristics of traditional non-cognitivism is the claim that ordinary usage is not to be taken at face value, on pain of category mistakes. So if we are to defend the possibility of a non-trivial non-cognitivism against contemporary challenges, quietism should be one of the views we consider.

These two forms of minimalism really are distinct. Quietists might well reject the idea that there is anything insubstantial about truth, finding no basis for this view in ordinary usage itself. For their part, minimalists of the first kind--`insubstantialists', we might call them-- need hardly be quietist, since the view that truth is not a deep explanatory property entails nothing about the priority of ordinary practice. Indeed, some theories of this kind are radically anti-quietist. Hartry Field has recently suggested that in opting for a insubstantial conception of truth, we may have to overturn our ordinary conceptions of meaning and synonymy, for example.[5]

In effect, then, we have characterised three views of the project of a philosophical theory of truth--quietism, insubstantialism, and the non-minimalist position we may call substantialism. Moreover, it is easy to see how there can be an analogous range of positions about belief. An insubstantialist about belief will deny that beliefs need play a causal-explanatory role--whether of folk belief-talk or of anything else--in mature scientific theory.[6] And a quietist about belief will refuse to countenance any philosophical perspective on belief that might overturn folk practice, taken at face value. We thus have six possible positions before us:


              Substantialism  Insubstantialism  Quietism        
About truth        A                 B              C               
About belief       D                 E              F               

In light of this table, we can make some preliminary points about the prospects for non-cognitivism, given minimalism. We have already indicated how quietism makes trouble for non-cognitivism, since one of the key tenets of traditional non-cognitivism is that folk usage should not be taken at face value. Hence semantically defined non-cognitivism seems incompatible with position C above, while psychologically defined appears to be incompatible with position F.

In the face of minimalism at positions B or E, however, the problems for non-cognitivism are more subtle. The basic tension seems to be something like this: if there is an important distinction in language whose defining characteristic concerns the possession of truth conditions (or something similar), then the notion of a truth condition thereby plays an important role in linguistic theory--which seems incompatible with the insubstantialist's claim that truth itself plays no significant causal-explanatory role. In other words, a semantically defined non-cognitivism seems bound to reject the central claim of insubstantialism about truth; and similarly, mutatis mutandis, for psychologically defined non-cognitivism and insubstantialism about belief.

Quietism and insubstantialism thus lead to rather different kinds of problems for non-cognitivism. However, it is easy to see that the third leg strategy avoids both kinds of problem. If the cognitive-non-cognitive distinction is drawn in terms of linguistic functions which are not themselves couched in terms of truth or belief, then the non-cognitivist will not be relying on substantial notions of truth or belief, and will have no immediate need to challenge the quietist's claim that philosophy has nothing interesting to say about the folk use of these notions.

We shall illustrate these virtues of the third leg strategy in Sect. IV. We turn first to a rival strategy for defending non-cognitivism in the face of semantic minimalism, recently suggested by Jackson, Oppy and Smith (hereafter, `JOS').


III. Aptitude and Platitude: the JOS Strategy

JOS take non-cognitivism to be the view that the sentences of a disputed class are `not truth apt'--`not in the business of being either true or false'. (p. 287) Thus they operate, at least initially, with what we have called a semantic definition of non-cognitivism. With this characterisation of non-cognitivism in place, JOS argue for four main points:

(i) To claim (as semantic minimalists do) that truth ascriptions are not ascriptions of a substantial property, does not commit one to the truth-aptness of any disputed class of sentences;

(ii) Only an appropriate minimalism about truth-aptness would provide a blanket resolution of these disputes in favour of cognitivism;

(iii) Considerations about `folk platitudes' serve to defeat minimalism about truth-aptness, and thus show that there is no cheap victory for cognitivists on the score of the truth-aptness of any disputed class;

and hence

(iv) Non-cognitivist proposals do indeed need to be settled case by case--minimalism provides no blanket victory.

JOS's argument for point (iii) depends on the claim that at least in central cases, a sentence is truth-apt just in case its normal use would be to express a belief.[7] Once this is granted, it follows that so long as there is no cheap victory for the cognitivist concerning which sentences express belief, there can be no cheap victory concerning which discourses are truth-apt--`in the business of being either true or false.'

In other words, JOS's point is that a non-cognitivist may accept that truth is an insubstantial property, but define non-cognitivism in terms of truth-aptness, and ground a substantial notion of truth-aptness on a substantial notion of belief. Though defined semantically, the resulting non-cognitivism will actually be defended on psychological territory: in the ethical case, for example, the non-cognitivist will need to show that moral claims or sentences do not normally express beliefs. Thus the non-cognitivism whose possibility JOS defend is close to the variety which simply defines non-cognitivism in psychological terms to begin with. In terms of our earlier table, JOS's strategy amounts to pointing out that it is possible to be a non-cognitivist while endorsing position B (insubstantialism about truth) so long as one also endorses position D (substantialism about belief).

It is possible to meet the challenge to non-cognitivism from quietism about truth in a similar way. That is, one might endorse position C and yet base a non-trivial non-cognitivism on position D. The trick is simply to operate with psychologically defined version of non-cognitivism, immune from the enervating effects of quietism about semantic notions. It is doubtful whether this option is available to JOS themselves, however. Given that they define non-cognitivism in terms of truth-aptness, rather than explicitly in terms of belief, a quietist about the notion of truth is likely to baulk at the sort of refusal to take things at face value that comes with such claims as that ethics is not truth-apt. As it stands, then, JOS's argument does not show that non-cognitivism is compatible with the quietist form of minimalism about truth, but at best that it is compatible with the insubstantialist version.[8]

Note also that JOS do not argue that non-cognitivism is compatible with the conjunction of positions B and E, let alone with that of C and F. Their defence of the compatibility of non-cognitivism and B depends on position D, since it turns on considerations about the sort of causal-explanatory role proper to the notion of belief. So their conclusion that non-cognitivism is not excluded by minimalism is a rather limited one. It does not apply to the more radical of the two main forms of minimalism about truth in the current literature, and does not apply to either of the two corresponding forms of minimalism about belief. At the psychological level they rely more on rebutting minimalism than on trying to show that it is compatible with non-cognitivism.

These points might be of little significance, of course. Quietism about any subject matter might be held to be an independently implausible philosophical position,[9] in which case if JOS's case for position D is successful, it does all we need ask of a defence of the possibility of non-cognitivism. All the same, we think there is more to be said. For one thing, with a clearer view of the conceptual landscape now before us, it is at least worth asking whether a more general response to contemporary minimalism is possible--whether there is a version of non-cognitivism that is compatible with positions C, E and F, as well as B. As we noted at the end of Sect. II, our broader functionalist conception of non-cognitivism has this desirable generality. So long as the non-cognitivist relies on a functional distinction which is not itself cashed in terms of truth or belief, non-cognitivism seems compatible with either form of minimalism about these notions.

The issue of generality aside, however, we think that JOS's strategy is unsatisfactory even in its own terms. In the remainder of this section, we want to show that there is an internal tension in the strategy, to which JOS fail to attend sufficiently.[10] The tension arises from the fact that JOS's own methodology is in some respects very close to quietism. JOS share with quietists a great respect for ordinary usage, taking it that the prime task of a philosophical account of truth is to collate and present what ordinary folk take to be platitudinous about truth. This provides a very questionable basis for a defence of non-cognitivism in the face of semantic minimalism, however. As we shall show, it is doubtful whether the platitudes-based approach can sustain the distinction between belief-expressing discourse and other discourse in such a way as to give non-cognitivism a fighting chance.

The platitude-based approach to philosophy has an evident appeal. As JOS put it,


The beauty of this approach to analysis is that it is hard to see how anyone could object to it. An objection to an analysis consists in pointing to something intuitively evident that conflicts with it. But if we have included everything platitudinous, there is no such intuition to be found. (p. 295)


All the same, this very quality of the approach seems likely to alarm any thoughtful non-cognitivist. By its very nature, non-cognitivism is a view that challenges first appearances and prima facie intuitions. Non-cognitivists hold that things are not as they seem, and that philosophers, let alone ordinary folk, are easily misled by linguistic appearances.

In JOS's terms, for example, the non-cognitivist needs to argue that moral and conditional claims are not truth-apt--not `in the business of being either true or false'. At first blush, however, it doesn't seem questionable that the folk take them as being so. Isn't it a platitude if anything is that some ethical claims are true and some false? Or that it is appropriate to wonder whether it is true that if the prime minister loses the next election then his party will dump him? And similarly with belief, so that there is no advantage to the non-cognitivist in moving to the psychological level. Commonplace references to the moral beliefs of others are hardly marginal features of folk practice. If we allow ourselves to be guided by what ordinary folk take to be uncontroversial, then, it seems that non-cognitivism of either the semantic or psychological kind will simply be ruled out of court.

JOS recognise this danger this, of course, but their attempt to meet it seems to us to be unsatisfactory. It depends on a crucial qualification to the platitude-based program of analysis, which they introduce in the following passage:


The right view is that there is no natural stopping point in analysing a concept short of finding an analysis that captures the whole network of the central, equally appealing platitudes surrounding the concept we are trying to analyse--provided, of course, that they really are central and that they can all be satisfied. (p. 297)


This sentence embodies what we see as the central tension of JOS's platitude-based approach. On the one hand the sentence emphasises the all-inclusive nature of the approach. (It occurs in a paragraph in which JOS criticise Crispin Wright for failing to take sufficient account of platitudes linking truth and belief.) But it also introduces the qualifications whose function is to prevent the approach being so all-inclusive that it necessarily embraces cognitivism. JOS seek to exploit these qualifications in the following passage:


The issue is not one about common parlance, but rather about the centrality of platitudes and their robustness under scrutiny. In common parlance we ... apply the words `true' and `false' to ethical sentences and to conditionals. This does not mean that we should conclude straight away that given the platitude preserving approach to analysis, ethical and conditional sentences are truth apt. On that approach the issue does not turn on the prevalence of the words `belief' and `true' and `false' in everyday talk about ethics and conditional. The issue turns rather on the appeal of the claim that ethical sentences and conditionals are truth apt and give the content of beliefs after reflection and explanation of the issues in the light of all the relevant platitudes. After all, the word `valid' is often enough used to describe statements rather than arguments in everyday talk, and the words `thinking' and `memory' are used to describe what is going inside an electronic calculator. But that does not mean that statements are valid and that calculators think and remember, and what reveals this is what we say after considered reflection and explanation of the issues. (p. 297, italics in the original)


The analogies at the end of this passage are surely red herrings, however. `Valid' is a term of art in philosophical discourse, and we don't pretend to be using a folk concept when we talk about logical validity. Logicians would hardly be embarrassed to discover that in ordinary discourse `valid' was interchangeable with, say, `reasonable'. But non-cognitivists of the kind JOS are discussing would certainly be embarrassed to discover that the literal meaning of `true' in ordinary discourse was such that it applied to moral judgments.

The case of thinking calculators is different but no more decisive. It doesn't take `considered reflection and explanation of the issues' to see that calculators don't literally think--we know that it is intended as a metaphor all along. (Perhaps some people don't know this, but they are people who are mistaken about whether calculators literally think. This is not analogous to non-cognitivism, however. The non-cognitivist is not someone who thinks that folk practice is mistaken, but someone who says that folk practice isn't doing what it seems to be doing.) Ordinary folk find it odd if we suggest that there is something non-standard about the use of `believe' in a sentence such as `Most of us believe that torture is wrong.' They don't find it odd if we suggest that calculators don't literally think.

Turning to the more general issues raised by the quoted passage, JOS's suggestion is that the platitudes that seem problematic for the ethical non-cognitivist may turn out to conflict with other central platitudes. As philosophers we might notice this conflict, and see that some central folk platitudes need to be sacrificed in the interests of consistency. Hence we might vindicate non-cognitivism as the theory that does best after `due reflection and explanation of the issues in the light of all the relevant platitudes.' We might discover, say, that on pain of contradiction, one cannot hold that what the folk call `moral beliefs' really are beliefs.

Such a discovery of contradiction among the folk platitudes seems to be the only way that a platitude-based approach could possibly lend any weight to a denial of truth-aptness in any of the disputed domains. But is it a real threat to the cognitivist? In JOS's terms, the problem is to show that after `due reflection', we might have good reason to reject the following platitude (and thereby to avoid cognitivism `on the cheap'):

(1) There are moral beliefs.

JOS suggest (p. 298) that a non-cognitivist might argue as follows:

(2) The states we call `moral beliefs' are essentially motivating.

(3) Real beliefs are not essentially motivating, and can only produce action by combining with desires.


(4) The states we call `moral beliefs' are not real beliefs (so that (1) is not a `central platitude', or at least is not to be taken literally).


But where are premises (2) and (3) supposed to come from, according to the platitude-based analysis? Ordinary folk do not explicitly endorse these principles, and so presumably the idea must be that they are implicit in folk practice. But given that (2) and (3) together contradict a literal reading of (1), then the fact that the folk regard (1) as a platitude means that they are implicitly committed to the denial of the conjunction of (2) and (3). After all, implicit commitment is simply what consistency demands. As Moore pointed out long ago, a claim to have discovered a contradiction in common sense provides good evidence that one has mis-characterised common sense. The upshot seems to be that the technique of extracting implicit platitudes can never yield the kind of conflicts that JOS's proposed defence of non-cognitivism requires. Given consistency at the level of explicit platitudes, the quietist always has the upper hand, for what lies on the surface does not need to be unearthed as an implicit commitment.

Hence it seems to us that there is a fatal flaw in the particular strategy that JOS suggest for defending non-cognitivism. We emphasise that the problem does not lie in the argument for the claim that non-cognitivism is compatible with semantic minimalism--that follows immediately, once we appreciate that non-cognitivism can be framed in psychological terms, and that minimalism about truth need not imply minimalism about belief. The problem concerns JOS's own argument for a non-minimalist position about belief. The methodology of a platitude-based approach seems to require that explicit platitudes always take precedence over claimed implicit platitudes, where the two threaten to conflict.[11] If so, then the cognitivist is always on secure grounds in appealing to the fact that the folk say there are moral beliefs, as well as moral truths.


IV. Non-Cognitivism as Functional Pluralism

Despite our reservations about JOS's argument, we agree with their main conclusion, at least in spirit: we too endorse a conception of non-cognitivism according to which non-cognitivist proposals do indeed need to be settled case by case, and according to which minimalism provides no blanket victory. We have argued that so long as non-cognitivism is correctly construed in the first place, this conclusion falls out in a simple and general form. The third leg strategy shows that non-cognitivism is compatible with a broad range of minimalist positions: in the notation of our table, with positions C, E and F, as well as position B.

The key to the third leg strategy is the observation that most generally construed, non-cognitivism is a doctrine about the functions of parts of language. In ethics, for example, the non-cognitivist's essential claim is that the function of ethical discourse is different from that of, say, scientific discourse, in some philosophically significant respect--in such a way as to make attempts to reduce ethical talk to scientific talk inappropriate, for example. Once our attention is drawn to the possibility, it is not difficult to see that the functional distinctions that do this sort of philosophical work might be neither the semantic nor the psychological distinctions of folk usage. After all, this is little more than the possibility that not everything of interest about the functions of language lies in view on the surface, where the ordinary folk can see it.

In practice, however, we find that this is a sticking point. Our opponents claim to have difficulty in seeing what kinds of things these linguistic functions could be, and hence how there could be a non-cognitivist position which was not cast in the usual semantic or psychological term. Part of the difficulty may lie in the terminological problem we mentioned earlier: because the term `non-cognitivism' has direct links to the semantic and psychological vocabulary, it is in one sense an inappropriate label for a position which renounces these links. We have chosen to let this be outweighed by the fact that the new position represents a genus of which traditional forms of non-cognitivism were the early species, but nothing really hangs on this choice. The genus is the philosophically more interesting position, and it is this position whose clear compatibility with minimalism we want defend, whatever one wants to call it.

As an aid to the visibility of this position, we want to show that it is already well exemplified in the philosophical literature. In this section we describe three well-known views which are naturally read in these terms. Each case involves a claim about the distinctive function of a particular part of language--a claim which the author of the view takes to do the philosophical work of traditional non-cognitivism, such as blocking reductionist moves. The functions concerned are not described in the non-cognitivist's usual semantic or psychological terms, however, and hence the resulting views are thoroughly compatible with minimalism about both truth and belief. (Our use of these examples does not require that we endorse the particular functional claims concerned, of course. We present them simply as an aid to those who have trouble visualising the possibility of our functional brand of non-cognitivism, unless it is couched in the usual semantic or psychological terms.)

We turn first to Robert Brandom's seminal paper, `Reference Explained Away'.[12] Brandom argues that `refers' is primarily a device for constructing complex anaphoric pronouns. The notion of anaphora is familiar: when we say `Jones went to the shop and he bought a leek', the pronoun `he' depends anaphorically on the term `Jones'. Brandom contends that the primary use of `refers' is to be found in such linguistic episodes as this:


Here `The person to whom Evans referred' is anaphorically dependent upon `Jones' in a manner analogous to that of `he' on `Jones' in the first example.

This is an account of the linguistic function of the term `refers', then, and Brandom is explicit about its relevance for certain more traditional approaches:

The anaphoric approach will not tell us how to understand sentences such as

Reference is a physical, causal relation.

The reason is clear. On the anaphoric account although `... refers to ...' plays a syntactically relational role, its semantic role is anaphoric and pronominal rather than relational. Philosophers have misconstrued the plain man's use of `refers' ... Such a mistake is of a piece with the search for the objects corresponding to each expression that syntactically plays the role of a term. (pp. 487-8)


We emphasise that Brandom doesn't deny that we have beliefs about what refers to what, or that claims about what refers to what are sometimes true. All the same, however, he claims to identify a crucial functional difference between `reference talk' and ordinary descriptive discourse--a difference in virtue of which there would be a sort of category mistake involved in seeking to reduce one to the other. Since the functional difference that he is interested in is not couched in terms of the distinction between having and lacking truth conditions, or between expressing and failing to express a belief, he is certainly not relying in a covert way on a substantialist conception of belief or truth. Nor is he offering the sort of challenge to folk practice that the quietist is likely to baulk at. His imputation of a category mistake in attempting to understand `refers' on the model of causal-relational predicates thus appears perfectly compatible with the sundry varieties of minimalism we have looked at.

Now to our second example. In Wise Choices, Apt Feelings, Allan Gibbard takes over from emotivism the view that the basic function of normative claims is that of expressing attitudes of approval and disapproval on the part of the speaker.[13] He argues that naturalistic analyses of normative claims miss `a general element of endorsement--an element an expressivist analysis can capture.'(p. 10) But Gibbard is quite happy to say that we have normative beliefs, so long as we recognise that normative beliefs and empirical beliefs are not the same sort of thing. And he takes it for granted that distinctive function of normative discourse is to be characterised from the standpoint of theoretical psychology (in alliance with human evolutionary biology), not in terms of the folk notion of belief.

Gibbard himself goes on to defend a semantically characterised version of his version of non-cognitivism about normative discourse: he argues that `normative judgements are not pure judgements of fact.' (p. 105) But he is more sensitive than many non-cognitivists to the question as to what this claim amounts to: `On the account as I am now giving it, after all, normative terms act much like other terms. What are we denying if we claim that normative judgements are not strictly factual?' (p. 105) The answer he goes on to suggest is reminiscent of the terms in which we have characterised the issue between substantialists and insubstantialists: he sees the crucial point as the claim that `our normative capacities can be explained without supposing that there is a special kind of normative fact to which they typically respond.' (p. 107)

In effect, then, Gibbard addresses the issue as to whether normative judgements are factual in terms of a substantialist account of factuality. The point we which to emphasise is that this move is not a necessary condition of the claim that normative judgements have a distinctive functional role. Gibbard could quite well have taken the same deflationary attitude to factuality as he takes to belief, and accepted that there are normative facts--and yet continued to insist that in virtue of the fact that normative judgements express motivational states, there can be no reduction of normative claims to naturalistic claims.[14] (As our page references indicate, indeed, Gibbard presents his anti-reductionist view of normative discourse long before he addresses the issue as to whether normative claims are factual. The functional point stands on its own, and does not depend on the later gloss.) Once again, the general point is that the philosophically interesting work of non-cognitivism--the work of blocking reductionist moves, in particular--is done by the functional characterisation. With this in place, a non-cognitivist has no need to insist on a distinction in terms of the folk semantic and psychological notions.

For our third example we turn to the insubstantialist theory of truth itself. An insubstantialist theory of truth is typically grounded in a story about the function of the truth predicate. Let us focus briefly on Paul Horwich's version of the view, for example.[15] According to Horwich,

the truth predicate exists solely for the sake of a certain logical need. On occasion we wish to adopt some attitude towards a proposition--for example, believing it, assuming it for the sake of argument, or desiring that it be the case--but find ourselves thwarted by ignorance of what exactly the proposition is. ... In such situations the concept of truth is invaluable. For it enables the construction of another proposition, intimately related to the one we can't identity, which is perfectly appropriate as the alternative object of our attitude. (pp. 2-3)


Horwich does not conclude that we do not have beliefs about truth, or that claims about what is true cannot be true. But he does deploy his account of the distinctive functional role of the truth predicate to the same sorts of philosophical ends as classical non-cognitivism: on the one hand, we get a story about the continuing need for the truth predicate in our practice; on the other, an insistence that any attempt to reduce truth to some naturalistic property involves a kind of category mistake. The latter moral is particularly clear in the following passage:

What the minimalist wishes to emphasize. . . is that truth is not a complex or naturalistic property but a property of some other kind. (Hartry Field suggests the term `logical property'.) The point behind this jargon is that different kinds of property correspond to different roles that predicates play in our language, and that unless these differences are appreciated we will be tempted to raise questions regarding one sort that can legitimately arise only in connection with another sort. (pp. 38-9)


Hence Horwich himself is a non-cognitivist, in our generalised sense of the term, and a live exemplar of the fact that non-cognitivism of this kind is compatible with minimalism about truth.[16] Indeed, insubstantialism about truth is almost automatically a non-cognitivist position, in claiming that the function of the truth predicate is not--as it seems--to refer to some substantial property. To understand the general character of non-cognitivism is to see that at least in this special case, it cannot be incompatible with this common form of minimalism about truth.

The special case aside, the three examples above provide ample illustration of the main points we want to make. It is quite possible to be a non-cognitivist, in a philosophically well-motivated sense of that term, without characterising one's position in terms of either truth or belief. As a result, non-cognitivism is compatible not only with insubstantialism about truth or belief, but also, at least in general,[17] with quietism.


V. The Functions of Truth

Insubstantialist theories of truth raise a range of questions concerning the functions of the folk concept of truth, questions not often given their due prominence. What is the truth predicate for? Why do its limits fall where they do in language; that is, why are there large classes of sentences with respect to which the construction `P is true' is disallowed? If the truth predicate is applied to discourses which themselves perform different linguistic tasks, does it have a common function in relation to the various tasks concerned?

Questions of this kind ought to have been addressed by traditional non-cognitivists, who often characterised their position in terms of truth, and yet wanted to draw a deep (`cognitive-non-cognitive') distinction within the class of sentences to which ordinary usage applies the truth predicate. In some respects, however, we think that the lack of the broader functional perspective we have advocated has prevented these issues from being given their due prominence. In this section we want to identify some of the proper concerns of philosophy in this area, bringing out along the way the merits of the functional perspective, and some of the errors of traditional non-cognitivism. JOS's discussion again provides a useful foil. Given that folk practice is the explanandum here, we might expect their platitude-based methodology to be sensitive to questions of this kind. In some respects, however, we think that they perpetuate some of the errors and oversights of traditional non-cognitivists, whom they themselves criticise on similar grounds. By calling attention to what is missing in JOS's account, we hope to highlight the general desiderata for work in this area.

Non-cognitivists who characterise their position in terms of truth often seem curiously immune to the thought that there must be some foundation for the fact that ordinary usage applies `true' and `false' to the sentences for which they advocate a non-cognitivist interpretation. After all, unless non-cognitivism is willing to ascribe some deep mistake to the folk--a move out of keeping with the view's typically conservative tendencies--it will have to allow that there is some legitimate basis for folk usage in this respect. A natural suggestion is that the use of truth in these cases is somehow idiomatic or non-standard, but any serious attempt to make good this assumption would lead to the sort of reflection on the role of truth envisaged above. After all, it is implausible to tell us that truth does something different in say ethical discourse, without telling us just what it does there, and what it does elsewhere, in supposedly cognitive discourses.

Prima facie, these questions are difficult for an insubstantialist about truth: in so far as the concept of truth is thin, there seems little room for a distinction between the two kinds of cases. As long as non-cognitivism is characterised in terms of truth, insubstantialism about truth leaves little room for the non-cognitivist to manoeuvre. Indeed, this is the basis of the claim that insubstantialism about truth excludes non-cognitivism.

Our third leg strategy avoids the problem by characterising non-cognitivism in completely different terms. In a more limited way, the same is true of an orthodox psychologically defined non-cognitivism. But does JOS's brand of non-cognitivism escape the problem? If so, only because it too is psychologically grounded, though semantically defined. Unless JOS are prepared to be more explicit about the psychological foundation than they actually are in their paper, however, the problem seems bound to re-emerge. The non-cognitivist they envisage would be claiming that ethical utterances are not truth-apt, while conceding that ordinary usage applies truth and falsity to ethical claims (and not seeking to revise this aspect of folk usage). But what is the concept of truth supposed to be doing in these cases, if the claims concerned are not truth-apt? And whatever the answer, how does it relate to what the concept of truth does with respect to claims which are truth-apt?

Presumably, JOS envisage that the non-cognitivist will claim that the ordinary application of `true' to ethical claims is somehow idiomatic, or less than full-blooded. But if truth is insubstantial to start with--a logical device for expressing generality, for example, as many insubstantialists have suggested--then it is easy for its use to be full-blooded, and this response seems implausible. As a device for generality, truth seems just at home in ethical discourse as it is elsewhere: after all, what difference does it make to the logical function of `true' in `What the Pope said is true' whether the Pope was giving voice to a moral opinion? The same goes for any brand of insubstantialism: if truth is insubstantial, there is very little room to deny that the use of the truth predicate is full blooded, in any discourse in which it normally appears--in other words, very little room to deny that the sentences of the discourse in question really are truth-apt.[18] JOS's strategy for combining semantically defined non-cognitivism with insubstantialism about truth thus seems unstable--in contrast, note, to its close relative, the position which defines non-cognitivism psychologically from the start.[19]

If non-cognitivism is characterised in the functional terms we recommend, the questions about the functions of truth with which we began this section still need to be addressed, of course. The non-cognitivist still needs to confront the fact that truth applies in a uniform way to parts of language which are being claimed to have different functions. Here the non-cognitivist's interests coincide with those of insubstantialists about truth, however. Both sides are interested in the functions of parts of language, and they have a common interest in producing a coherent account--an account in which the function of truth talk can be seen to mesh with that of other parts of language, in such a way as to explain the data of ordinary language. We take it to be an advantage of the more general functional perspective we advocate, and in terms of which we have suggested that non-cognitivism is best formulated, that it does give these questions about truth their due prominence.[20] Finally, we emphasise that there is no reason to think that the issues raised from this perspective can be addressed by a platitude-based strategy. The linguistic functions to which such accounts appeal might be no part of the folk repertoire, and hence be unmentioned in any platitude, explicit or otherwise.[21]


VI. Non-Cognitivism and Reductive Metaphysics

We close with a note about the wider significance of the issue as to how non-cognitivist positions should be formulated. There has been considerable recent interest in a reductionist program in metaphysics which takes its lead from an argument for the mind-brain identity theory advanced by David Lewis in the 1960s. In outline, the general proposal is that we treat folk platitudes about particular topics in the way that Ramsey suggested we treat scientific theories: we replace problematic terms with existentially quantified bound variables, and take the folk to be talking about whatever makes the resulting `Ramsey sentence' true. As their interest in folk platitudes might suggest, the JOS team includes some of the proponents of this program.[22] (Some of the program's advocates have taken up an ironic suggestion we made in an earlier version of the present paper, and now call it `the Canberra Plan'.[23])

The program depends in any particular case on a crucial linguistic assumption--roughly, the assumption that the reduced theory is doing the same linguistic job as the reducing theory. Unless this assumption is valid in the case in question, the proposed reduction involves a kind of category mistake. (This is just the point that non-cognitivism is antithetical to reductionism.) Jackson himself recognises the need for this assumption, of course. In a recent paper in which he and Philip Pettit seek to apply the strategy to the case of value, for example, they explicitly presuppose cognitivism.[24] However, they take for granted that this is an assumption to be cashed in the standard terms. (As it turns out, they employ the psychological version of the standard story, saying that `cognitivists ... maintain that practical evaluations ... are expressions of judgement and belief.') The point we have emphasised above is that this may well address the issue at the wrong level. A sophisticated non-cognitivist may simply deny that the relevant distinctions in language and linguistic psychology lie at this accessible level.

Our examples in Sect. IV illustrate this point. It is quite clear that Brandom, Gibbard and Horwich would deny that reference, rationality and truth are fit topics for the Ramsey-Lewis-Jackson program. In each case, they would say that to apply the program would be to make a mistake about the linguistic category of the target discourse, the kind of mistake traditional non-cognitivists urged philosophy to avoid. The linguistic categories are not now drawn in terms of truth, but the continuity with traditional truth-grounded non-cognitivism is quite clear. What survives unchanged is the idea that a reductionist program is liable to category mistakes, to avoid which we need to reflect on the linguistic function of the parts of language concerned. We conclude that to miss the functional standpoint is to fail to notice the true character of the assumption on which the Canberra Plan depends. Indeed, the functional standpoint threatens to undercut the motivation for reductionism of this sort. Once we have an adequate explanation for the fact that the folk talk of Xs and Ys and Zs, an explanation which distinguishes these activities from what the folk are doing when they do physics, why should try to reduce the Xs and Ys and Zs to what is talked about in physics?[25]

Finally, it is important to appreciate that the functional perspective is itself scientific, and naturalistic. Hence it can't be dismissed a priori--particularly by physicalists--and the reductionist program turns out to depend on an a posteriori theoretical thesis about language. As noted above, moreover, folk intuitions may give us very little access to the matters addressed from this theoretical perspective. Such intuitions provide the explanandum--what we want is an account of the origins and functions of our folk intuitions, in effect--but not the explanans. In particular, therefore, these are not issues that can be addressed by the platitude-based methodology of the Canberra Plan itself--even setting aside the obvious concern about circularity--for almost certainly they concern things that the folk have simply not thought about. So there really is no route into these issues other than the hard one of standing back from the concepts in question and asking how they arise, what functions they serve in the lives of the creatures who employ them, and so on. Despite their importance, however, these are issues whose visibility in contemporary philosophy is tenuous at best, a fact well illustrated by the terms in which JOS's discussion of the possibility of non-cognitivism is couched. In drawing attention to a more general perspective on the same problem--in showing in particular that a functionally-grounded non-cognitivist is a much more sure-footed opponent--we hope to have bestowed on these issues some of the visibility they deserve.[26]