Like coastal cities in the third millennium, important areas of human discourse seem threatened by the rise of modern science. The problem isn’t new, of course, or wholly unwelcome. The tide of naturalism has been rising since the seventeenth century, and the rise owes more to clarity than to pollution in the intellectual atmosphere. All the same, the regions under threat are some of the most central in human life—the four Ms, for example: Morality, Modality, Meaning and the Mental. Some of the key issues in contemporary metaphysics concern the place and fate of such concepts in a naturalistic world view.1
True, some philosophers hold that at least some of these topics are not worth saving, and that the tide of science does us a favour by sweeping them away. Others hold that they do not need saving, being already out of reach of the waters of science—no part of scientific landscape, in effect, but no less respectable for that. To many contemporary philosophers, however, neither view seems appealing. The first—‘eliminativism’—seems to underestimate the value of what would be lost. The second—‘nonnaturalism’—sometimes seems to rely on a Canute-like faith in the limits of science, and to offer no satisfactory account of how there could be a region of the world both out of reach of science, and yet of relevance in human life.2 While affirming the vulnerability of the M-concepts, then, most philosophical naturalists look for some sort of rescue strategy—some legitimate place for the M-concepts, within a naturalistic framework.
In my view, however, contemporary naturalists have overlooked the most promising rescue strategy. The two main strategies currently on offer are noncognitivism and reductionism. I want to argue that the contemporary debate is incomplete, at the very least, in failing to recognise a third approach. As I’ll explain, this third strategy rests on two main premises, each of which, while controversial, is of some plausibility in its own terms. I won’t try to defend these premises in any detail here: I simply want to show that if both are granted, there is an attractive alternative to noncognitivism and reductionism.
The new approach has much in common with noncognitivism, however, so I’ll begin below with noncognitivism, and explain how the new strategy differs. I’ll then contrast it briefly to Frank Jackson’s reductionist program. I’ll also show that although it is naturalist in spirit, the new strategy offers an olive branch to nonnaturalists. In effect, it explains in the naturalists’ own terms how topics such as morality and meaning might remain high and dry, untouched and unthreatened by the rise of the scientific tide—it offers the benefits of nonnaturalism, without the metaphysical down side.
1. Noncognitivism and the Carnap thesis
Noncognitivism is often a response to the kind of philosophical concern just described. Some topic of human discourse—morality, say—seems difficult to accommodate in a naturalistic world view. The choices seem to be to accept the existence of moral aspects of reality distinct from the aspects of reality described by science, or to conclude that science has shown that moral talk is in error, in failing to connect with anything in the external world. Noncognitivism offers an escape from this dilemma. If moral talk isn’t in the business of describing reality—if its linguistic function is quite different—then we can leave it in place, without conflict with the ontological lessons of the naturalistic view.
This solution is pleasing to a naturalist in two respects. First, as noted, it entails that there are no moral facts or properties to be accommodated within the natural world. Second, the price of this escape—the notion of a linguistic function—is itself acceptable, in naturalistic terms. A naturalist need find nothing objectionable in the hypothesis that human languages serve a number of different functions.3
Noncognitivism appeals most in the moral case, but even here some feel that it devalues the talk it claims to save. If this is rescue, moral realists think, we do better to take our chances against the tide of naturalism on our own terms. This response is more common with respect to topics metaphysicians are inclined to take more seriously, such as modality, meaning and mind. In these cases, few people think that noncognitivism provides any sort of rescue worth having.4
Recently, noncognitivism has been under attack from a different direction. Noncognitivists need an account of the nature of the distinction they claim to find in language, between cognitive and noncognitive discourse. In practice this distinction gets cashed in a number of ways, some semantic and some psychological. On the semantic side, noncognitive discourse is said to be non-assertoric, non-descriptive, non-truthconditional, non-truth-apt or non-fact-stating, among other things. On the psychological side, the distinction is normally drawn in terms of a belief/non-belief distinction: cognitive utterances are said to be those which express genuine beliefs, rather than other psychological attitudes, such as desires.
However, it not clear that these distinctions can be drawn in the way the noncognitivist requires. One source of doubt has been minimalism about truth: roughly, the view that linguistic function of the truth predicate is not to refer to a substantial property, but to do something else—perhaps to provide a mere grammatical convenience, for example. This view shares with noncognitivism a concern with the functions of particular parts of language, but seems to leave little room for the claim that moral discourse, say, is not genuinely truth conditional (or truth-apt). If truth is merely a grammatical device of ‘disquotation’, for example, isn’t it immediate that any indicative utterance is truth-apt?5 What is the noncognitivist denying, when she denies that there are moral truths?6
This point is familiar with respect to noncognitivism and truth. There is an analogous but less familiar point about noncognitivism and ontology, stemming from the work of Quine and especially Carnap in mid-century. Carnap (1950) argues that there is no absolute, theory-independent ontological viewpoint available to metaphysics. Ontological questions about the entities mentioned in a particular theory or linguistic framework can properly be raised as what Carnap calls ‘internal questions’—questions posed within the framework or theory in question—but not as ‘external questions’, posed from a stance outside that framework. (Carnap follows Quine in taking it that the kinds of entities to which a theory is ontologically committed are those over which the theory quantifies, when cast in canonical form.)
Carnap himself applies this conclusion to argue that the traditional debate between platonists and nominalists about the existence of abstract entities is misconceived, except to the extent that it can be construed in one of following two ways. First, Carnap allows that there are legitimate (internal) issues as to what abstract entities, if any, are quantified over in the most satisfactory representation of the framework concerned—some nominalists might be construed as arguing that such quantification is always eliminable—and as to what particular entities there are of the recognised kinds.7 Second, Carnap also recognises a legitimate pragmatic issue which may be raised from the external standpoint: roughly, an issue concerning the utility of the framework concerned. What is disallowed, according to what I shall call the Carnap thesis, is the external ontological question as to whether there are ‘really’ sets, numbers, or whatever.
This thesis seems to make things difficult for a noncognitivist. If there is nothing more to the issue of the existence of moral properties, or possible worlds, or meanings, than is settled by the fact that these things need to be quantified over in the relevant parts of language—and the noncognitivist grants this, and does not want to say that these parts of language are in error—then how can it be denied that such things exist (and hence that the relevant parts of language do describe them)?8 Again, reflection on the function of existence claims leads to trouble for the very view that stakes its living on the thesis that linguistic functions may not be what they seem.
In my view, the difficulty stems from the fact that noncognitivism took a wrong turn early in its history. Noncognitivists were right in thinking that the notion of linguistic function provides a naturalistic solution to metaphysical concerns, but wrong to try to characterise the functions concerned in the terms normally used. To get things right, we need to retain the insight that different bits of language may serve different functions in a way which isn’t obvious at first sight, but set aside the usual attempts to characterise the functions concerned in terms of truth, factuality, belief, and the like.
I want to show that the resulting position actually derives great strength from minimalism about semantics and ontology. In particular, the Carnap thesis plays a crucial role in what follows. At various points the thesis blocks natural objections to my proposal. Often the block seems counterintuitive, but in a sense this was Carnap’s point: he thought that various ‘natural’ moves in philosophy turn out to be inadmissible, in presupposing an illegitimate externalist viewpoint. This suggestion has considerable plausibility, in my view, but I shall not attempt to defend Carnap’s thesis in this paper. I simply want to show that if Carnap is right, the issue of the fate of the M-worlds takes on a surprising new form.
2. Functional pluralism
I’ll call the new approach functional pluralism.9 A functional pluralist accepts that moral, modal and meaning utterances are descriptive, fact-stating, truth-apt, cognitive, belief-expressing or whatever—and full-bloodedly so, not merely in some ersatz or ‘quasi’ sense. Nevertheless, the pluralist insists that these descriptive utterances are functionally distinct from scientific descriptions of the natural world: they do a different job in language. They are descriptive, but their job is not to describe what science describes.
Of course, as my phrasing is meant to emphasise, this makes a functional pluralist sound like a Moorean nonnaturalist, who asserts that the world contains moral properties, in addition to the things talked about by science. The difference is that pluralism rejects the idea of a single world, containing both moral and natural entities. I don’t mean that it replaces this single world with a bare multiplicity of worlds, which would be equally unappealing. The point is that the judgement of unity or plurality could only be made from the framework-independent external stance, which the Carnap thesis disallows. Without that stance, functional pluralism is neither monist nor pluralist, in a primarily ontological sense—for there is no such sense.
The multiplicity the pluralist does recognise is one of linguistic function, and therefore naturalistically respectable.10 The multiplicity of worlds, such as it is, is merely what follows from the fact that more than one linguistic function may be exercised at the same time. From the evaluative stance it is correct (let us suppose) to say that there are values and evaluative properties. From the scientific stance it is correct to say that there are electrons. If I say that electrons are terrific little particles I occupy both stances—employ both frameworks, in Carnap’s terminology—but no metaphysical spooks are thereby required.11
The possibility of a functional pluralism has been obscured by a tendency to read into the notion of description an ontological monism for which the Carnap thesis implies there is no justification. When philosophers think of description, or representation, they often have in mind a familiar picture: on one side, the World, on the other side Mind, or Language (the medium of representation). Once this picture is in place, the path I want to take is practically invisible. Either the World in the picture is the natural world, in which case moral discourse must either be understood in naturalistic terms, or be treated as somehow less than full-blooded, successful, description; or it is the nonnaturalist’s world which contains moral facts, say, in addition to the kind of facts described by science.
Innocent as this picture may seem, I think it embodies two mistakes. The first is the mistake the Carnap thesis corrects, of imagining that there is an extra-linguistic stance available to metaphysics—a stance external to every linguistic framework or viewpoint, from which we may survey the World. The second is the mistake of failing to notice the plurality of linguistic functions, or frameworks, within descriptive discourse as a whole. Once we recognise that there is more than one framework in use in ordinary language, and that there is no framework-independent stance for metaphysics, it follows immediately that the naive picture is misleading, unless explicitly confined to one framework (in which case it can be thought of harmlessly, as a metaphorical picture of what can be said in internal terms).
In a limited way this conclusion is already drawn by Carnap himself. Carnap’s treatment of a fifth M-world, Mathematics, is the treatment a functional pluralist seeks to extend to other four. So why has this avenue not been explored? Mainly, I think, due to the grip of the noncognitivists’ idea that their functional distinction needed to be drawn elsewhere, in terms of a divide between descriptive and non-descriptive uses of language.12
I’ll come back to the Carnap thesis, for the picture I’m using it to oppose is very tenacious, and will try to re-assert itself at several points. For the moment I want to turn to the aspect of functional pluralism that may seem especially puzzling: the suggestion that in some philosophically interesting sense, description might serve more than one function in language.
3. Description as a multi-purpose tool
In one sense, the claim that descriptive discourse serves more than one function in language is entirely trivial. Even if we set aside the fact that we use descriptions for a wide variety of secondary purposes—to persuade, to shock, to entertain, for example—there is still a huge range of functional differences associated simply with differences of content. Descriptive utterances about aardvarks serve different purposes from descriptive utterances about zygotes, simply in virtue of the differences between (our relationship to) aardvarks and (our relationship to) zygotes. However, this kind of functional variety is no threat to the assumption that description itself constitutes the basic functional category—in effect, the assumption that a functional taxonomy of these uses of language would have description itself (or some cognate notion, such as assertion) as the core linguistic notion. Whatever else the various uses might be, they are all instances of this core descriptive use of language.
Not all human instruments wear their essential properties on the surface, however, Think of string, or cord, which also has many different uses: tethering animals, hanging washing, securing parcels, binding splints, producing musical notes, and so on. If we were interested in a functional taxonomy of the uses of human artefacts, how would we represent what these uses have in common? A proper theory would have to recognise that string has certain core properties—length, thinness, strength and flexibility, for example—all or most of which are exploited in each of these uses. A theory which didn’t identify these core properties would be inadequate in at least two ways. First, it would be unable properly to characterise the difference between a multi-purpose artefact, such as string, and a single-purpose artefact, such as a Phillips-head screwdriver. A screwdriver may be used on many different screws, but this isn’t the same as the functional diversity we find in the case of string.13 Second, such a theory would fail to exclude various uses of string which don’t rely on any of the core properties: the use of a ball of string as a paperweight, for example.14
Hence if we were interested in a functional taxonomy of human artefacts, the naive category ‘uses of string’ would not play a significant role in our final theory. On the contrary, the theory would appeal to the core properties themselves, and classify applications accordingly, as different ways of making use of these core properties.
Analogously, my functional pluralist suggests that notions such as ‘description’ and ‘assertion’ may be nothing more than relatively superficial labels for a linguistic category whose core properties remain to be discovered—a category which may turn out to have a multi-functional role in language, in the sense that its core properties serve a range of very different functions. One of these functions, perhaps, is that served by assertion in scientific discourse. (Actually, I think it is much more plausible that scientific discourse is already multi-functional. More on this below.) But if we fail to notice that there are other uses of the same linguistic instrument, we are likely to make the kind of mistakes that would be made by someone whose acquaintance with string was biased toward just one of its many uses. Imagine someone whose main acquaintance with strings has been in an orchestra pit, for example. He encounters a tethered goat, recognises that what connects the goat to the post is a string, and wonders how it is possible to make music on such an erratic instrument. He is right about the string, but wrong about what its use entails.
Thus the functional pluralist needs to answer two questions: What are the core properties of the descriptive or assertoric use of language? And what are (some of) the various uses to which a linguistic form with these core properties may be put? For the purposes this paper I shall simply suggest one possible answer to the former question, which I have defended in detail elsewhere.15 The suggestion is that the core function of assertoric language is to give voice to speakers’ mental states and behavioural dispositions, in a way which invites criticism by speakers who hold conflicting mental states. The key thing about assertoric discourse is thus that it embodies a normative idea of answerability to an external standard, the effect of which is to place an onus on speakers to be prepared to defend their views in the case of disagreements. The notions of truth and falsity play a key role, being the most pure expression of the relevant norms.
Again, I emphasise that what matters for present purposes is not whether this is the right characterisation of the core function of assertoric discourse, but simply that it provides an hypothesis about what the core function might be, in terms of which we can see that the same core function might have many different applications. However, I call attention to one feature of this proposal which would need to be shared by any satisfactory alternative: it avoids the kind of circularity that would be involved if we tried to characterise the core functions of assertoric discourse in terms of the very group of concepts whose functional significance is at issue.16 It isn’t helpful to be told that the key function of descriptive discourse is to ‘make statements’, or ‘express beliefs’, if these notions are simply part of the package whose deep significance we are trying to understand. It would be like being told that a string is a light rope, or a heavy thread. We would be no closer to an understanding of underlying nature of the tool in question.
Let’s turn to the second question: For what range of purposes might a linguistic device with this core property be employed? What different functions might description serve, in different cases? The striking thing is that these functions may be just those to which noncognitivists also appeal. They might include the function of giving voice to a range of psychological states with distinctive functional roles of their own, or distinctive dependencies on contingent human capacities and responses: motivational states (‘desires’), say, or states which play a distinctive roles in reasoning (‘conditional beliefs’), or perceptual states of various kinds. Of course, noncognitivists appeal to these psychological distinctions in order to argue that the attitudes concerned are not beliefs, and a functional pluralist thinks that this step is misguided. Concerning the psychological distinctions themselves, however, the two views might well agree.
4. Belief and ontology
Why do noncognitivists feel bound to say that psychological states distinguished in these ways are not beliefs? An important factor, I think, is the thought that because genuine beliefs have genuine truth conditions, someone who allows that moral beliefs (say) are genuine beliefs is committed to genuine moral facts or states of affairs (or to an error theory, of course). If reductionism is disallowed, so that these moral states of affairs are not simply natural states of affairs in disguise, then we seem to be left with the metaphysical mystery noncognitivism seeks to avoid: nonnatural facts, floating free of the physical world.
Again, however, the Carnap thesis tells us that this relies on an illegitimate conception of belief: roughly, the idea that beliefs are the mind’s attempt to stand in correspondence with a pre-existing World—‘pre-existing’ in the sense that it is thought of from an external stance which supposedly we occupy as semanticists and ontologists, asking to what our beliefs refer, and what makes them true. But there is no such stance, according to the Carnap thesis. We construct our semantics after the fact, relying on the theories to which we already subscribe.17 As users of moral language, holders of moral beliefs, we can describe their semantics, and may discover that when our beliefs are expressed in canonical form, they refer to moral properties. We thus find ourselves committed to the existence of moral properties, in the only sense of ontological commitment recognised by the Carnap view.
The key insight, shared by noncognitivists and my functional pluralists, is that at this point the remaining puzzle is merely sociolinguistic, and not metaphysical. It is the issue as to why creatures like us should find ourselves engaged in this particular linguistic practice, not the issue as to what these moral properties are. But whereas the noncognitivist thinks that the metaphysical issue is only deflected by a concession about language—a concession which entails that moral language does not involve genuine ontological claims—the functional pluralist sees that this concession may be unnecessary. If Carnap and Quine are correct about the nature of ontological claims, and the minimalists correct about truth, then the metaphysical issue is already empty, and the linguistic concession is unnecessary.
Doesn’t the metaphysical issue emerge again, from within the sociolinguistic perspective? This perspective is naturalistic, at least in initial orientation: it simply enquires into an aspect of the behaviour of natural creatures (ourselves). But when we ask for the function of moral discourse, aren’t there two possibilities? The first is that we need to appeal to moral properties or states of affairs, to which the speakers of moral discourse can be construed as responding. In this case we are in the same position as before—i.e., still faced with the problem of the nature of such things. Alternatively, if we don’t need to appeal to moral properties or states of affairs, then the sociolinguistic perspective seems to have shown us that moral talk is in error—unless, as the noncognitivist claims, it is not intended to be descriptive in the first place.
But the second outcome is misrepresented, and the source of the mistake is the same as before. The fact that there are no moral properties in the natural world does not entail that moral talk is in error, if its function is not to describe such properties. Enquiring into its function, we find (let’s suppose) that it is to give voice to certain mental states, distinguished by their motivational role in human psychology. Suppose—this is where the pluralist parts company with noncognitivism—that this function is one of those appropriately served by the core properties of the descriptive language. This would explain why moral discourse takes descriptive form, and hence why it involves truth claims, and existence claims. We would thus have a naturalistic understanding of what speakers are doing when they engage in moral discourse, and hence a reason to deny that such speakers are guilty of some global error, without any concession about the descriptive function of moral discourse. On the contrary, the account explains why such discourse does take (genuinely) descriptive form, in terms of an account of what core properties are genuinely distinctive of descriptive language.
5. The primacy of science
There is another factor which may have helped to favour noncognitivism at the expense of functional pluralism. Won’t any view which focuses on the role of (apparently) descriptive language in different areas of discourse want to say that science is special, or primary, in some way? Indeed, what would commitment to naturalism amount to, within this picture, if not something like this? And it is easy to take this as the thought that science (or perhaps some subset of it, such as its observational part) involves genuine descriptions, in contrast to the quasi-description of the other discourses. If we say this, we are committed to a form of noncognitivism, at the expense of functional pluralism. So a functional pluralist needs to resist this move, and yet to deal with the intuition that there is something primary about science.
In my view, the explanation lies in the fact that as functional pluralists, we speak from within the scientific framework, but about other frameworks. This gives the scientific framework a kind of perspectival primacy. Our viewpoint is internal to science, but external to morality, for example. It is a viewpoint which allows us to refer directly to the objects and properties countenanced by science, but not—given the Carnap thesis—to the objects countenanced by the moral stance. It is tempting to think that it thus makes the moral reality somehow ersatz, fictional, or second-rate, but I think this is a mistake. The apparent difference is explained as one of perspective, an artefact of the viewpoint. To mistake it for an absolute ontological difference is again to lose sight of the Carnap thesis, and to fall for the mirage of a framework-neutral stance: to think that science gives us that stance, in effect.
For a functional pluralist, then, the question ‘What is the distinguishing function of scientific discourse?’ is as important and non-trivial as the question ‘What is the distinctive function of moral discourse?’ The circularity involved in the fact that the former question is itself posed from a scientific standpoint is not vicious, of course—it is no more problematic than the use of human language in the study of the deep structure of human language, for example. However, scientific discourse may appear distinctive in that when we address this sort of question with respect to science, our answer will refer to the entities countenanced by the discourse in question—science itself—in a way which isn’t true of other discourses.
Even if we discount the perspectival factor just mentioned, it might seem that an important insight remains. In the case of science, or at least folk science, it seems we need to appeal to the existence of ordinary-sized physical objects to explain folk talk about such things, in a way in which we don’t need to appeal to moral properties to explain folk talk of moral properties. This might encourage the thought that genuinely descriptive discourse is whatever discourse needs to be explained as a response to a pre-existing world—a world which itself needs to be described in terms of the discourse in question.
This is an appealing idea, closely related to a familiar suggestion as to how we should settle claims between realists and antirealists in metaphysics: roughly, the suggestion that realism is appropriate when the putative entities do causal-explanatory work. However, if this were our criterion for genuinely descriptive language, it is far from clear that science, or even folk science, would count as genuinely descriptive. For example, science seems irreducibly modal at various points, and perhaps irreducibly committed to the notion of causation. Yet there might well be a naturalistic explanation of our use of our causal and modal discourse which didn’t presuppose the existence of modal facts—an explanation in which the existence of such facts played no causal role. (This isn’t as controversial as it might sound. A modal realist such as David Lewis is committed to the view that the origin of human modal talk is a causal product of nothing but the actual world; being causally isolated from the actual world, other possible worlds play no causal role. More generally, it is far from clear that causal relations themselves are among the causes of our talk of causal relations.) Even the use of kind terms and general concepts would be suspect, given that our causal contact is always with tokens. So the category of genuine description might be considerably narrower than that of (even) folk science. Indeed, it might turn out that there is no language explicable purely in these terms—no useful residue from this attempt to distill pure representation from functional reflection on human thought and language.18
To put the point slightly differently, we cannot take it for granted that all the conceptual tools used by science are themselves to be explained as bare responses to proper objects of scientific study. It might be a mistake to think that science could or should study causation in the same way that it studies gravitation, even if talk of causation turns out to be ineliminable from science. The appropriate scientific course might be to study talk of causation—to ask why we have it, what it does for us. As I’ve emphasised, this reflection on the functions of language need in no way devalue its original uses. Science has no more reason to be wary of a functional account of the concept of causation than it has to be wary of the deflationists’ functional account of truth, for example.
Some philosophical positions may have reason to be wary of a functional account of causation, of course. For example, some reductionist programs rely on the principle that all causation is at base physical causation. In the classic Armstrong-Lewis arguments for materialism about the mental, for example, this principle supports the ontological claim that mental states are identical with certain physical states. However, a study of the origins and functions of causal concepts might reveal that there is no more justification for this kind of bare physicalist monism about causation than—according to the Carnap thesis—there is for a bare physicalist monism about ontological claims themselves.
6. Jackson’s program
Mind-body reductionism of the kind just mentioned illustrates the second major contemporary strategy for rescuing the M-worlds from the rising waters of science. Contemporary reductionism is perhaps best exemplified in Frank Jackson’s elegant program, which relies on generalising Lewis’s treatment of the mental case, using the Ramsey-Carnap-Lewis approach to theoretical terms.19
In order to contrast functional pluralism with Jackson’s approach, it will be helpful to have a toy example. I’ll use the concept cool, as employed by speakers of English dialect (foreign, I suspect, to most of my readers) called ‘Street Cred’. When the Street community say that something is cool, they do not mean that its constituent molecules have low mean kinetic energy. They mean that it is good, in a particular way. In what particular way? The available answers depend in part on where we stand. Among themselves, Street speakers might explain it in terms of near synonyms: ‘Cool things are, like, wicked,’ they might say, using another positive evaluative term. But I’ll assume that this kind of answer either isn’t accessible to us, or isn’t what we were after, because the deeper theoretical issue we had in mind arises again with respect to the synonyms. If we want a naturalistic understanding of this aspect of Street usage it may be some help to be told that ‘cool’ and ‘wicked’ are near synonyms, but it doesn’t get us very far.
Let’s assume that the Street use of ‘That’s cool’ is descriptive. With this assumption in place, Jackson’s strategy is to assemble the various statements Street folk take to be true of the property of being cool, to form these by conjunction into a single statement of Street ‘Cool Theory’, and then to apply Ramsey’s technique, replacing all occurrences of the term ‘cool’ with a variable bound by an existential quantifier. The result is something of this form:
There is a property F such that CT(F).
Cool is whatever property makes this true—in other words, whatever property best realises the role of cool in the theory CT.
Up to this point, Jackson’s program involves nothing with which a functional pluralist need disagree. From the pluralist’s standpoint, there is no objection to marshalling the commitments of those who subscribe to the CT framework in this way. In effect, the marshalling shows us that those who employ this framework are committed to the existence of a property with certain properties, specified within the theory.
The difference comes at the next step. For the functional pluralist, the appropriate questions concern the function of the CT framework in the lives of Street speakers. There need be no further significant issue about what property cool is, and no mystery in the lack of such an issue. From inside the framework the issue does arise, of course, but has a trivial answer (once the marshalling has been done). What property is cool? Why, simply that property F such that CT(F). From outside the framework, the issue does not arise, as the Carnap thesis tells us. What does arise from outside is the issue as to why such a framework is in use in Street, and here the functionalist standpoint is the appropriate one to take up. (It may not provide a complete explanation, of course. To some extent, the framework may be simply an historical accident, maintained by linguistic convention.)
Jackson’s view is that there is a further ontological issue, however: the issue as to what physical (or natural) property F is such that CT(F).20 The view that this issue is legitimate—not simply a category mistake—seems to rest on two lines of argument.21 The first is one we have already encountered. If we fail to recognise that description may be a multi-purpose linguistic device, and reject the nonnaturalists’ ontological pluralism, then if ascriptions of cool are agreed to be descriptive, it seems that there is nothing for them to describe but the natural world. (We are also rejecting an error theory of cool, of course, but the ‘whatever fits’ character of Jackson’s program makes an error theory more than usually unappealing. ) What else could cool be but a natural property?
A functional pluralist meets this argument by denying that description is univocal in the right sort of way. There is a core function that descriptive uses of language share, according to the functional pluralist, but it may be one which permits a range of quite different applications. And as I’ve explained, the pluralist appeals to the Carnap thesis to prevent this view from degenerating into nonnaturalism of the metaphysical kind.
However, this leaves untouched Jackson’s second line of argument, which relies on the fact (agreed on all sides, let’s assume), that coolness supervenes on natural properties. If two things are the same in all physical respects, then either both are cool or neither is. In effect, then, coolness covaries with physical properties in some regular way. There is some (perhaps highly disjunctive) physical property which is simply what it takes for an object to be cool—a property which objects have if and only if they are cool. Supervenience requires that there be such a property. Given that there is, the fact that it co-varies with coolness ensures that it occupies the role described by CT: the various things that CT takes to be true of cool are true of this physical property.
The easiest way to see where the functional pluralist parts company with this argument is to see where a noncognitivist would part company, and then transpose into pluralist terms. Noncognitivists can allow that the use of evaluative terms such as ‘cool’ is supervenient on natural properties in at least two different senses. First, we may be able to say if two objects are physically identical, then normally functioning Street speakers will judge both to be cool or neither to be cool. In other words, our study of the practices of Street speakers may show that their use of the concept cool is sufficiently regular and rule-governed to ensure that this is the case. This shows that in principle there’s a naturalistic specification of the class of things normal Street speakers take to be cool. It doesn’t show that to be cool is to have the natural property that marks this class. These are quite different propositions, as shown by the fact that the latter one uses the term ‘cool’, whereas the former only mentions it—and just as well, for if the noncognitivist were to accept the latter proposition, it would undermine the claim that ‘That’s cool’ is non-descriptive, by providing this utterance with a descriptive content.
The second sort of supervenience that a noncognitivist might recognise is stronger. It might turn out that Street speakers themselves acknowledge that coolness supervenes on the physical, in the sense that they themselves will assent to the suggestion that if two things are physically indistinguishable, then either both are cool or neither is. Does this mean that in ascribing cool Street speakers are ascribing a physical property? Of course not, says the noncognitivist, for consider this analogy: We can easily imagine a group of rather insightful spectators (at a tennis match, say), who recognise that if two moves in the game are physically indistinguishable, then either it is appropriate to applaud both, or appropriate to applaud neither. This provides a sense in which applause supervenes on the physical, but obviously it doesn’t entail that to applaud is to ascribe a physical property. Similarly with ‘That’s cool’, according to the noncognitivist. Its function is more like that of applause than that of a property ascription, and the fact that it supervenes on the physical in the sense described does not show otherwise. It does not show that ascriptions of cool are ascriptions of a physical property.
The key idea here is that the attempt to extract reduction from supervenience is blocked by functional difference: the reduction does not go through, despite supervenience in the acknowledged sense, because the function of the supervening discourse is different from that of science (the subvening discourse). And this implies that the point works equally well for a functional pluralist, even if he affirms that the supervening discourse is descriptive, so long as there is a deeper sense in which the supervening and subvening discourses differ in function.
So I think a pluralist can account for the supervenience intuitions, without allowing the identity claim. It might be said that this doesn’t show that the identity claim is false, merely that it is unmotivated. However, is falsity what the pluralist requires? A better approach seems to be to argue that identity claims are simply not well-formed, unless they respect the lessons of the Carnap thesis. To ask whether an entity referred to in one framework is identical to an entity referred to in another framework seems to presuppose a framework-independent stance, from which the question can be raised. If we reject the idea of such a stance, the upshot seems to be that identity claims such as ‘Karl Marx is the father of Groucho Marx’ and ‘Karl Marx is the number 3’ go wrong in quite different ways. The first is false, from within the framework in which we talk about persons. The second is ‘not even false’: it simply involves a category mistake.
But where does the boundary between these two kinds of cases lie? What constitutes a framework? A functional pluralist needs an answer to these questions which admits as meaningful not only such identity claims as ‘Karl Marx is the father of Groucho Marx’ and ‘The morning star is the evening star’, but also, apparently, certain acceptable reductionist claims in science: aqua fortis is HNO3, for example. There are various important issues here. How are frameworks individuated? How do they combine in language, so that we may employ more than one in a single judgement? Can there be nested structures of frameworks, so that natural science might be unified by some overarching framework (which licenses reduction within science)?
I don’t have developed answers to these questions, or space here to try to present some. My aim has been much more modest: to point out that the reductionist approach to the problem of the M-worlds assumes that language is homogeneous in a way in which it may well not be, and that noncognitivism (as usually conceived) is not the only way to challenge this assumption.
The new challenge, that of functional pluralism, depends on the Carnap thesis. Without the Carnap thesis, naturalism seems to require ontological monism. This imposes homogeneity on descriptive language by default, as it were, by providing it with a single target—a single World to be described. In blocking this move, the Carnap thesis has radical implications for the project of saving the M-worlds from the tide of naturalism. It seems to permit a naturalistic alternative to noncognitivism and reductionism—a path with the attractions of nonnaturalism, in allowing talk of the M-worlds to be taken at face value, but without the metaphysical spooks.
The view that naturalism requires ontological monism is deeply entrenched, of course, and the suggestion that there might be a middle path of this kind is one that contemporary philosophical naturalists find hard to take seriously. However, I have argued that taking it seriously turns out to demand very little: simply the Carnap thesis, and a willingness to entertain the possibility that descriptive judgement might be a multi-purpose tool. I conclude that advocates of any of the more familiar positions—eliminativists and nonnaturalists, as well as noncognitivists and reductionists—should think about which of these moves they reject, and why. If both are admitted then contemporary metaphysical debates seem to be seriously incomplete, at the very least, in failing to engage with the possibility of a functional pluralism.22
Brandom, R. 1993: ‘Asserting’, Nožs, Vol. 17, pp. 637–650.
———1994: Making it Explicit (Cambridge, Mass.: Harvard University Press).
Boghossian, P. 1990: ‘The Status of Content’, Philosophical Review, Vol. XCIX, pp 157–184.
Carnap, R. 1950: ‘Empiricism, Semantics and Ontology’, Revue Internationale de Philosophie, Vol. 4, pp. 20–40.
Jackson, F. 1982: ‘Epiphenomenal Qualia’, Philosophical Quarterly, Vol. XXXII, pp. 127–136.
———1994: ‘Armchair Metaphysics’, in The Place of Philosophy in the Study of the Mind, ed. M. Michael and J. O’Leary-Hawthorne (Dordrecht: Kluwer), pp. 23–42.
Jackson, F., Oppy, G. and Smith, M. 1994: ‘Minimalism and Truth Aptness’, Mind, Vol. 103, pp. 287–302.
O’Leary-Hawthorne, J. and Price, H. 1996: ‘How to Stand Up for Non-cognitivists’, Australasian Journal of Philosophy, Vol. 74, No. 2, pp. 275–292.
Price, H. 1988: Facts and the Function of Truth (Oxford: Basil Blackwell).
———1992: ‘Metaphysical pluralism’, Journal of Philosophy Vol. 89, pp. 387–409.
Rosen, G. 1994: ‘Objectivity and Modern Idealism: What is the Question?’, in The Place of Philosophy in the Study of the Mind, ed. M. Michael and J. O’Leary-Hawthorne (Dordrecht: Kluwer), pp. 277–319.
Wright, C. 1992: Truth and Objectivity (Camb., Mass: Harvard University Press).
1What is naturalism? For the moment I take it to be the view that the project of metaphysics can properly be conducted from the standpoint of natural science. One reason for not being more specific is that the issue depends on the main project of this paper, which is to show that there is recognisably naturalistic program in metaphysics which is largely ignored by contemporary ‘naturalists’.
2There are important arguments for nonnaturalist views, of course, among them those of Frank Jackson (1982) concerning the status of qualia. As I think Jackson would be the first to acknowledge, however, these arguments do not remove the element of mystery in nonnaturalism.
3So long as the functions concerned can be characterised in naturalistically acceptable terms, of course. It isn’t obvious that orthodox noncognitivism passes this test. If the functions concerned are characterised in intentional terms, and intentionality itself is naturalistically suspect, noncognitivists may well have a problem.
4In the case of meaning it is not clear that noncognitivism is even coherent, for reason mentioned in the previous footnote; cf. Boghossian (1990).
5Jackson, Oppy and Smith (1994) answer ‘no’ to this question, arguing that one might be a minimalist about truth but not about truth-aptness, or the belief/non-belief distinction. This is quite true, but no help in defending noncognitivism against some of the more thoroughgoing forms of contemporary minimalism, which do extend to these notions; see O’Leary-Hawthorne and Price (1996).
6There is a clear statement of this argument in Wright (1992), ch. 1.
7In extreme cases the recognised categories might even be held to be empty, as in the view of the atheist, who accepts the conceptual framework of theism, but argues from within it that, in fact, there are no deities. Contrast this with the view of the person who regards the entire framework as meaningless or pragmatically irrelevant. Both views seem legitimate in Carnap’s terms, but they shouldn’t be confused.
8True, the noncognitivist might say that the Carnap thesis applies only to cognitive uses of language, and that the cognitive character of moral discourse is precisely what noncognitivism denies. But the point is that this denial seems implausible, if a moral ontology is so close at hand.
9In O’Leary-Hawthorne and Price (1996), John O’Leary-Hawthorne and I use this term, but refer to the view as a version of noncognitivism. There is no significant inconsistency with my present terminology, however. O’Leary-Hawthorne and I note that while the view in question is not standard noncognitivism, it is a product of the same intuition about the functions of language, and hence that a terminological decision must be made as to whether to call it a noncognitivist view. Here, in order to highlight the contrast with orthodox noncognitivism, I am using a different label. In Price (1992) I called the same view ‘vertical pluralism’, to distinguish it from the (‘horizontal’) kind of pluralism which merely allows a plurality of theories within a given area of discourse—a plurality of theories of electromagnetism, for example.
10Subject to the earlier caution—see fn. 3.
11Obviously a functional pluralist will have to allow frameworks, or linguistic functions, to ‘mix’ in this way. The task of making sense of this seems considerably less severe than that facing noncognitivists, however, who are required to account for the same mixed examples in the standard semantic terms.
12This idea was already firmly established when Carnap was writing, of course. Carnap himself refers to the ‘non-cognitive character’ of the external issue as to the acceptability of frameworks.
13The difference may be a matter of degree, of course.
14The test here is replaceability: the function served by the ball of string could be served by many objects with none of the core properties of string. The use of a screwdriver as a lever provides another example.
15 Price (1988). For a related account of assertion which might also be used for these purposes, see Brandom (1983, 1994).
16At least, it avoids this kind of circularity if notions such as conflict between mental states can be explicated in sufficiently basic terms, so that they don’t themselves depend on notions such as truth and falsity. I address this kind of concern in Price (1988), ch. 7.
17I think it is important not to confuse the pseudo-externality provided by what Quine calls ‘semantic ascent’ with the genuine externality with which Carnap is concerned. Semantic ascent is available to us from within a linguistic framework: it allows us to pose what are really internal questions by talking about the framework itself. Instead of asking ‘Is there a prime number greater than 100’ we can ask ‘Is the sentence “There is a prime number greater than 100” true?’ Questions of this kind are only sensible if the sentences in question retain their interpretation: otherwise, it would be like asking ‘Is the sentence “!@#$%^&*” true?’ Carnap’s external standpoint must be more remote: the issue as to whether to adopt a framework is like the issue as to whether to interpret in the first place.
18For an insightful discussion of related objections to the causal/explanatory criterion of reality, see Rosen (1994), pp. 309–313.
19There is a clear exposition of the program in Jackson (1994). It is also a major theme of Jackson’s 1995 John Locke Lectures.
20 Here, as elsewhere in this paper, nothing is meant to rest on the physical/natural distinction.
21 Perhaps there are three lines of argument, the third being the kind of appeal to a physicalist view of causation mentioned at the end of the previous section. If this appeal were invoked in the present case, on the basis that Street speakers take coolness to be a causally efficacious property, I think the pluralist’s response should be to ask whether the function of causal talk in Street usage provides support for the thesis that all causation is physical causation.
22I am grateful to Chris Daly, Max Kšlbel, Michaelis Michael, John O’Leary-Hawthorne, Graham Oppy, Nick Smith, Daniel Stoljar and Ed Zalta for many helpful comments on earlier versions.