Rorty, Dewey and Semantic Minimalism

1. Two Teams

As all his readers know, Richard Rorty painted his picture of Western philosophy with a very broad brush. It resolved that history into a battle between two sides. In one corner we have rationalists and metaphysicians, including most analytic philosophers. These are followers of Plato, Descartes, and Kant, who conceive themselves as dealing with independent facts and structures and talk unblushingly of reasons, arguments, analysis, distinctions, objectivity and truth. These are wedded to the view that in many of their endeavours people manage to get things right, and that often enough later endeavours, by refining and building on earlier achievements, get them more nearly right, or sometimes more totally right, than their predecessors. In the other corner, Rorty places himself alongside pragmatists such as James or Dewey, not to mention Nietzsche, Heidegger, Wittgenstein, Sellars, Davidson, Quine, Foucault, and Derrida. These are said to think differently. They are ironically resigned to history, contingency, and forces of social and linguistic change whose effects cannot be anticipated. They have no interest in epistemology and reason; they see discourse in terms of politics not accuracy of representation. They may call themselves relativists or social constructivists or pragmatists or nihilists, but the labels are not important, and can safely be brushed aside. For in face of the crucial distance between the two orientations, fine details and potential differences between the members of these teams scarcely matter. All that matters is whether you share the delusive self-image of the first team, conceiving yourself as a servant of logos, or whether you are a clear-sighted player in the second team. As with all crusades, you have to choose sides. Rorty was nothing if not a great team builder.

I always recall at this point a remark the great literary critic F. R. Leavis said about a similar choice: ‘When people line up so promptly one suspects not only that the appeal of the chic has something to do with it, but that the differences are not of a kind that has much to do with thinking’. But do the teams in fact line up so quickly, or was the belief that they do so merely an illusion due to the broadness of Rorty’s brush? Others have effectively, and with some amazement, criticized a taxonomy that makes Descartes look just like Kant, and Kant look just like Plato.[1] I incline as well to worry about exactly what is supposed to unite the bad team, and exactly how their opposition to the good team works.

There are, of course, many pointers in Rorty’s work: the bad team might adhere to a correspondence theory of truth, to a myth of the given, to a realistic notion of “the world”, to an analytic-synthetic distinction, to truth in the surprising sense of “truth taken apart from any theory”, to a distinction between receptivity and spontaneity, or experience and theory, to a mysterious theory or irresoluble ‘hard problem’ of consciousness. The good team throws all these out.

But when we have no use for these things, what else goes missing?[2] Sometimes, it appears, quite a lot. For it becomes easy to move from abjuring high philosophical theory to appearing to forswear much of the everyday with what to my eye is often an alarming nonchalance. On his best behaviour Rorty can be clear-eyed about avoiding this slide. For example in one seminal paper, ‘The World Well Lost’, he contrasts the philosophers’ (alleged) notion of ‘the World’, the one that he wants us to lose, with the world of ‘the stars, the people, the tables and the grass’ which is supposed to be just fine—as presumably are events like the people observing the stars or the tables or the grass.[3] But in many other places the contrast seems to blur or disappear, such as when we are told that it is only the rag-taggle bunch of Kantians and that lot who share the ‘ingenuous image of themselves as accurately representing how things are’.[4] For when people minutely observe the grass and see, perhaps to their surprise, a bumblebee in it, what is so ingenuous or naive about them supposing that their report that there was such a thing ‘accurately represents how things are’? They put themselves, surely, in a position that, given their background abilities, makes it near enough certain that they believe that there is a bumblebee there if and only if there is, and telling this is exactly what is meant by accurately representing this bit of the world.

When Rorty asserts that ‘‘There is no way to get outside of our beliefs and our language so as to find some test other than coherence’ is he also forgetting that in a perfectly everyday sense you have to get outside and go where the grass grows in order to observe whether there is a bumblebee there? When he tells us that ‘Nothing counts as justification unless by reference to what we already accept and there is no way to get outside our beliefs and our language so as to find some test other than coherence’ is he denying that the activity of observation not only helps but often provides justification? [5] If asked whether there are bumblebees in the grass, am I supposed to answer simply ‘by reference to what I already accept’, excogitating an answer even if I have never been near the grass nor received intelligence about it?

Apart from the inverted commas, Rorty is on stronger ground when he tells us to ‘see justification as a social phenomenon rather than a transaction between “the knowing subject” and “reality”’.[6] He is right that justification is a social matter in the sense that it is only people who give verdicts allowing that people are justified in saying or believing what they do. Nobody ever has to defend themselves as being justified to a tribunal made up of the stars or the grass. We can say, with only a little imprecision, that justification is an element in solidarity, maintaining your status in the eyes of others, although of course you can also justify yourself to yourself. But very often the verdict of others hinges precisely on whether they suppose that the defendants had seen what they claim to have seen, as well as whether they interpreted it in the light of shared habits and principles. In a court of law that’s the difference between being a credible witness and being a fantasist, fraud, or impostor. When he lets his hair down Rorty is not enamoured of that distinction either: ‘To treat beliefs not as representations but as habits of action, and words not as representations but as tools, is to make it pointless to ask ‘Am I discovering or inventing, making or finding?’[7] As I write this in the United Kingdom a liar and fantasist, Carl Beech, has been awarded eighteen years in prison for neglecting exactly that distinction as he spun libellous stories incriminating many of the great and the good in horrible paedophile goings-on. I do not think this is proof that courts and jurors have been taken in by any of Plato, Descartes, or Kant.

To be fair Rorty might not have meant that sort of invention. Perhaps he meant the invention of new vocabularies, for example. But first of all, while lots of people make discoveries, such as the bumblebee in the grass, very few invent new vocabularies. And when they do it typically results from a discovery, when new terms (“New Zealand”, “muon”) are needed to refer to what has been discovered. [8]

In earlier work I puzzled over similar slides by one of Rorty’s heroes, Davidson, who says very similar things: ‘There is, then, very good reason to conclude that there is no clear meaning to the idea of comparing our beliefs with reality or confronting our hypotheses with observations’.[9] As I said, here too something that starts life supposedly as a deep philosophical objection to a high-flown correspondence theory of truth, instantly metamorphoses into what sounds like the rejection of a crucial everyday activity. Davidson goes on to explain that ‘No such confrontation makes sense, for of course we can't get outside our skins to find out what is causing the internal happening of which we are aware’.[10] But in any ordinary sense, confronting a hypothesis with observation is not a matter of starting with an awareness of internal happenings. It is a matter of, for instance, going and looking and possibly revising what we were inclined to believe in the light of what we find. I confront my happy belief that the grass is safe to sit upon with observation when I look and notice the bumblebee.

I postpone giving a diagnosis of these extraordinary doctrines for the moment.

Meanwhile similar worries beset some of the things that Rorty claims about truth and representation. He roundly tells us that ‘A pragmatist in the philosophy of science cannot use the truth of Galileo’s views as an explanation either of his success at prediction or of his gradually increasing fame.’[11] Perhaps ‘truth’ here is intended in the upper-case sense allegedly common to Plato’s team, and we know that Rorty would, rightly, not allow truth in that sense, completely beyond our ken, ever to explain anything. Nor would most other people. If we entertain Cartesian doubt, worried in case the stars and the grass and indeed the whole world of space and time is but a Matrix-like virtual reality, a veil obscuring the Truth that is beyond us, we would presumably admit that it cannot be that Truth, as opposed to truth about the stars and the rest, that explains Galileo’s success. Indeed it would be a standard part of the sceptical package that we couldn’t give any such explanation. But truth in the world of stars and people and tables and grass, Galileo’s world, the world we talk about, can certainly help to explain successes at prediction and gradually increasing fame. Perhaps the point is instead that the truth, for instance, that the acceleration of a falling body does not vary with its weight, does not by itself explain why Galileo got interested in the topic and found out about it. But then the claim is completely uninteresting, and certainly not the private insight of pragmatists, or any other philosophical school.[12] In any more interesting sense the claim is manifestly indefensible. The truth that there was a bumblebee in the grass clearly helps to explain why I believed it was there, even if the reason I was investigating the grass remains unclear to people. The astonishing accuracy of Captain Cook’s charts of different coastlines is exactly what explains his fame as a navigator, exactly what explained the habit of seamen seeking out and using those charts to navigate, and exactly what explained the fact that they coped so much better when they did so. And, fortunately, Cook did not excogitate the topography ‘by reference to what he already accepted’. He had to go and measure.

Of course it was already accepted by navigators long before Cook that you do indeed have to go and measure. But it wasn’t joining in that acceptance that explained Cook’s success. It was how he acted upon it, and because of the accuracy of the newly invented chronometer, how he was able to do so scrupulously and with unprecedented precision.

Rorty tells us that ‘Instead of seeking ‘vertical’ relationships between language, or ourselves as language users, and the world, we must concentrate upon ‘horizontal’ or inferential processes, whereby we advance and accept reasons from each other. Justification becomes a ‘social phenomenon’ rather than a transaction between a “knowing subject” and “reality”.’ Solidarity not copying. I would find it very difficult to see Cook’s procedures and success in terms of this opposition. As we have already seen justification is indeed a matter of justification to yourself and others. But this is not in contrast to putting yourself in touch with reality. It is largely determined by it. Not being, as sometimes mapmakers before him and since him have been, a fantasist or a fraud Cook carefully put himself into an interesting relationship with different coastlines of the world. It was only his having done so that justified him in publishing his charts and gave them their utility.

When he did so they were taken up and read. There wouldn’t have been much social to-and-fro about his justification for what they tell. He produced and others consumed—after all, they were in no position to conduct a trial of his sayings. They hadn’t been there. They trusted him, and their trust was well placed. Here semantic terminology swims into view. A good, precise way of putting it is that the people who could read knew what the symbols and signs on the chart represented. They had to learn to do this kind of thing, and it is not a trivial matter. The booklet that currently enables this in the United Kingdom is entitled Symbols and Abbreviations used on Admiralty Charts. My 1984 edition contains thirty-seven densely packed pages, interpreting twenty-two categories of symbols and abbreviations. Without knowing what those symbols represent, a chart is useless. Knowing what they represent is indeed a success in the social world, putting the reader in touch with the cartographer, and if you cannot navigate this bit of the social world, you cannot navigate at all. If you go to sea it would not be the world that is lost, but the sailor. [13] That is your motivation for putting yourself in touch with Cook or nowadays with the Admiralty, by learning their language.

One symbol on a chart represents a dangerous rock that covers and uncovers at different phases of the tide. If we find semantic terminology mysterious, or have some philosophical animus against it, we might look for analysis or paraphrase. Consider the non-accidental coordination between author and reader. The author has put into the public domain a sign and this enables the reader to make appropriate inferences and responses. We might wish to talk of intentions, as Grice did, or habits or conventions, as I would prefer. Neither approach would undermine the simple fact that the chart works by representing a rock that covers and uncovers at a certain place, and the reader has had to learn as much.

So why did representation, together with reference and truth, become such a target for Rorty’s second team? Part of the diagnosis must be that Rorty and Davidson have absorbed Sellars’s distinction between the ‘space of causes’ and ‘the space of reasons’, and taken it to imply that since observation is a causal process the processes of observation cannot have anything to do with reason and justification. It is as if there is no difference between looking and seeing the bumblebee and being caused by something else such as a blow on the head or a mad brainstorm to form a conviction that there is a bumblebee in the grass. The different causal etiology can make no difference to the justification of the belief. Cook would have been equally justified in publishing his charts if he had hallucinated the coastlines of New Zealand and New South Wales from his bed. Surely, however, this implication only has to be exposed to reveal its absurdity. Our senses are adaptations for increasing the probability that beliefs formed using them are true; brainstorms are not. We do not have to be bog-standard empiricists to be aware that it such an increase in probability has a great deal to do with justification. A glimpse, a bark, or a whiff might any of them justify me in suspecting the presence of Fido in my vicinity.

Even more discreditably perhaps the doctrine that observation is theory-laden confused some people into thinking that there is no significant difference between relying on observation and merely relying on the lore you bring with you, as if good observation is entirely a matter of shuffling words and thoughts.

Or is it that the bogey of the indeterminacy of translation led to despair whether the symbol really indicates the presence of a rock that covers and uncovers, or something else? It doesn’t do to despair over that when you are in a boat. And in spite of Nietzsche’s alignment of truth and metaphor there is nothing metaphorical about what the chart tells the sailor. If I said that a sometimes-submerged rock threatens to sink Rorty’s semantic boat, the metaphor would need to be unpacked, but there is no unpacking to be done when using the chart. Nor is there anything to deconstruct, only something to read. Irony is not in place.

Since these explanations only seem to give weak excuses it might take hard work from the admired Edinburgh school to understand why in the late twentieth-century people began to say these extraordinary things, and to be heard and applauded for doing so.

Perhaps, though, there is another issue in the background, which Rorty may have inherited from Dewey.

2. Misunderstanding Dewey?

John Dewey wrote that

The basic fallacy in representative realism is that while it actually depends upon the inferential phase of enquiry, it fails to interpret the immediate quality and the related idea in terms of their functions in inquiry. On the contrary it views representative power as an inherent property of sensations and ideas as such, treating them as “representations” in and of themselves. Dualism or bifurcation of mental and physical existence is a necessary result, presented, however, not as a result but as a given fact...psychological or mental existences which are then endowed with the miraculous power of standing for and pointing to existences of a different order. [14]

Dewey’s target here is the idea that representative power could be the intrinsic property of a particular thing. If we thought that, then we would soon notice that ordinary things around us, such as arrangements of marks on paper, or sounds in auditory space, or for that matter arrangements of furniture or flowers, have no such intrinsic powers. We might then be in danger of thinking that since such things are, as it were, disappointingly inert, the real power must come from some other kind of thing: mental existences of a different order, such as ideas or concepts. Wittgenstein had a similar target in the passages on rule-following. Our dispositions to apply or withhold terms in new contexts can only be contingently associated with any particular thing, whether on the page or in the head. It cannot be explained by the range or extent of a presence in the mind or anywhere else.

Dewey calls his target ‘representative realism’, but that was misleading. Firstly on the face of it the problem has nothing to do with realism: the difficulty over intrinsic or immediate representational power would be just as serious if it were applied to representations of Santa Claus or fairies instead of sometimes submerged rocks. But secondly, and I think more importantly for a discussion of Rorty’s philosophy, the target is not representation as such, but a particular account of what representation requires. The target is not our ability to represent coastlines with charts, foodstuffs with menus, or times of departure with timetables, but the idea that we do such things by making present to our minds one or another entity with an intrinsic, self-standing, and miraculous, power of doing it for us.[15]

I think there is no doubt that Rorty rightly absorbed Dewey’s criticism of this kind of theory of representational power. That much is visible in his dislike of semantic atomism and his rejection of semantic nominalism, that being any theory that construes all meaning in terms of a name-bearer relation, forgetting that it is only by having been given a use that a word has meaning. But there is a question whether Rorty himself misconstrued Dewey, by imagining that his target included the very idea of representation itself, as if, having decided that maps, menus, and timetables don’t work one way, we conclude that they don’t work, full stop. That would unfortunately reveal enslavement by the theory Dewey is attacking, supposing that if it does not tell us the way representation works, it cannot work any way at all.

Twentieth century pragmatism itself barred any such inference in the case of truth. Ramsey, Wittgenstein, Quine, and numerous followers offer deflationist or minimalist conceptions of this. They confirm the innocence of the notion by ensuring that it smuggles no metaphysical luggage, and implies no doubtful philosophy of its own. So it cannot be a sensible target of philosophical critique. Not too far from these deflationists stand other philosophers such as Davidson, who without subscribing to full-scale minimalism or deflationism, nevertheless admit that truth and representation are so tied in with concepts such as belief and assertion that they could not be jettisoned without at the same time abandoning all the rest of our intentional vocabulary, denying ourselves the title of thinking and believing things at all.

Nevertheless many pragmatists of a generation following Rorty have inherited some of his nervousness about “representationalism”. One manifestation of this, perhaps encouraged by deflationism about truth, is what Huw Price calls semantic minimalism, and in the rest of this paper I shall offer some remarks about any such program.[16]

In the hands of Paul Horwich, perhaps its most influential defender, deflationism about truth is applied to propositions, not sentences or inscriptions or words on a page. As such it comes at the same time as interpretation, not before it. That is, at exactly the same time and by exactly the same process as we learn to interpret a sentence on a page, we learn what does or would make the thought it expresses true. There is no grasp of the one without the other. But deflationism about truth is utterly silent about what might be needed for this process of interpretation to take place. In other words, it does not concern itself with what it is to learn to hear or read what is being said to you. If we are allowed to use the terminology, we could say that it is a priori that the proposition that there is a sometimes submerged rock at some place is jut the same as the proposition that it is true that there is a sometimes submerged rock at that place. But it is not a priori that any particular sign or symbol or sentence means that. It is entirely contingent whether anyone grasps that.

There are familiar kinds of thing that might be said by someone giving the semantics of some term, or terms, or sentences, or other fragments, of some language. Let us say that doing this is giving a semantic description of these terms or fragments or sentences. Such semantic descriptions offer interpretations. Familiar examples might be:

“snee” in German refers to snow.

“snow” in English refers to snow.

“London” in English refers to London

“Santa Claus” in English refers to Santa Claus

“good” in English is true of good things

“London is a good city” is true in English if and only if London is a good city.

“London is a good city” means in English that London is a good city.

Although familiar, such sentences can spawn confusion. One such might arise if we do not notice an ambiguity in what is being talked of as “English” or “German”. Are these identified syntactically, in the way that a logician would identify a formal language—that is, by their lexicons and by whatever tells competent speakers that such-and-such is a grammatical construction with in the language, corresponding to a well-formed formula in a logical calculus? Having specified a formal language in this way, logicians can choose whatever interpretation they wish for symbols of the language, and it is contingent which ones they do choose. By parity, if this is the model our specimen sentences are plainly contingent: had history been slightly different “snow” in English might have referred to hail. There are quite nearby possible worlds in which it does so. The speakers of English are, together, sovereign in just the same way that the logician is, and their collective habits could have been different.

A different policy would identify languages by some larger set of capacities exercised by their users—some combination of syntax and interpretation, so that strictly speaking it would not be English if “snow” did not refer to snow. In this case interpretation-giving statements, such as those just listed, are necessarily true, or made true by the definition of what counts as English. On this way of thinking, just as it wouldn’t be chess if the King could move two squares at a time (it would not really be a King, either) it wouldn’t be English unless the whole panoply of words and sentences used by English speakers had the meanings they do.

Presumably we are free to define English or German whichever way we like, just as we could decide to let in variant versions of chess as chess, not as some other game (we do this with bridge or poker, which have different versions with different rules). And it is familiar that the contingency that is subverted if we go in for a partially semantic identification of a language reappears at a different place. If, instead of saying that it is contingent that “snow” in English refers to snow, we say that it is necessary because it is an essential property of English that it does so, we have made English into an abstract structure implicitly defined by its totality of lexicon and semantics. But in that case, of course, it is contingent that English is the actual language implemented in the speech of any particular population— just as, if we define chess by its totality of rules, it become contingent that anyone has ever played it. So there is still a contingency that needs identifying and perhaps explaining and justifying, namely the contingency that people like me speak and read and understand English.[17] It cannot be something minimal, but something substantial in my dispositions and abilities that makes this true, just as it is a substantial and unfortunate gap in my dispositions and abilities that makes it true that I do not speak Mandarin, for example, and it must be something about a player’s dispositions and their guidance that makes it true that he is playing chess and not some little-known or private variant.

It might seem, then, that there is no space for real semantic minimalism. Nevertheless, one way of sympathizing with it (perhaps not the only one) takes off from a basically Davidsonian approach to thinking about meaning. In that we imagine a Tarskian truth theory for a language that simply gives interpretations for some basic expressions and then “chases truth up the tree of grammar” as Quine nicely put it. When our project is to describe the actual language of some population we proceed empirically and holistically adopting a methodological principle of charity. Tarski’s T-sentences are of course extensional, but the fact that a T-sentence is a suitable entry in an interpretative theory of English is not. That is, the sentence ‘ “snow is white” is true in English iff snow is white’ is a suitable entry, whereas the extensionally equivalent sentence ‘“snow is white” is true in English iff grass is green’ is not. The first is suitable because we could embed it in a systematic system for interpreting many, many, English sentences, delivering results that make English speakers appear both rational and truthful.

Reference comes into this story because as well as clauses for whole sentences we need clauses determining their make-up from their contributing constituents. Only if we do this can we track the way in which the contributing ingredients can be reshuffled in order to generate new sentences that are also part of English, and that can be more or less automatically understood by competent users of the language on the basis of what they have already absorbed. Then the idea is that these atomic entries are only implicitly defined by their role in the whole overall picture of the language. That is, there is no need to ask, and no point in asking, whether “reference” is the same whether we talk about an abstract object, a a physical place, a real person or a fictional person. Reference becomes a book-keeping notion, necessary to help us keep accounts rather than itself a source of semantic capital. It is as general as the capacity of humans to pay attention to things, or have their attention drawn to them.

It is perhaps easier to grasp this idea if we proceed in terms of Sellars’s dot notation. With this we say that “x” is a •X• where x is an expression of the object language, X is an expression in the language we are using to describe the object language (these may be the same), and the result of putting X between the dots is to denote the conceptual role that, contingently, is the one played by X. We might put this by saying that •X• is a rigid designator denoting the role of the term inside the dots, that works by giving us an instance or a sample of the role in question.

The dot notation gives us a rigid designator in the metalanguage, fixed to the role some specific word or sentence of the object language has. Using it gives us what Peter Strawson called degenerate contingent meta-statements: contingent (given that we think that English might have evolved with words having different roles than those they actually have), meta, since they are statements about object-language terms and sentences, and degenerate in the sense that unless you already know the role played by the terms or sentences inside the dots, they would not be intelligible to you. As information-giving sentences about those roles they are therefore useless, but they are true, and, of course, fitted to play a role in a systematic semantics, satisfying Quine's ambition.

These semantic clauses are themselves minimal—we might say, disappointing, degenerate or in one terminology, ‘modest’, in the precise sense that they take the roles of the expressions in the metalanguage for granted. They are simply silent about what those roles are, and they are similarly silent about the nature of the rapport between utterer and reader or hearer.

From this beginning minimalism might now transmute into the doctrine that I called ‘quietism’, denying that philosophers can find genuinely informative things to say about the terms central to our schemes of thought. It may not be possible to say anything interesting about •chair• or •cause• or •good•. We may be able to use these terms, like others, without being able to say what their use is. Any philosophical commentary could only be making explicit something that is normally implicit, and there is room for uncertainty whether this can be done.[18]

This is certainly not so across the board. If we want to know what these roles are, sometimes there is plenty of space for commentary to come flooding back.

So consider examples like "Fore!" in English is a •Fore!• or "Abracadabra" in English is an •Abracadabra•. These are both true, and both degenerate in the sense that each takes reference to its respective role for granted. It is up to the reader to supply knowledge of that. But of course we can go on to make explicit the information implicitly supplied by the knowing reader of such clauses. . The •Fore!• role is that of a warning most frequently used on a golf course to warn people of an oncoming ball. The •abracadabra• role is that of a term mostly used by stage magicians to herald the unveiling of a surprise. We might say that this commentary is itself part of semantics, or we might choose to call this an exercise of meta-semantics, a commentary on the minimal truths contained in semantics proper. I cannot see that anything hangs on this choice: either way, the explication of the roles obviously give us something more than the minimal clauses, and something that we clearly need in order to be able to say informatively how these terms function in English.

The role *refers* seems to be one which is quite amenable to a clarificatory story. Michael Williams has helpfully suggested that pragmatists should distinguish three elements in an explanation of meaning in terms of use (an EMU). There is to be an epistemological element describing how claims using the concept can be justified or established. There is to be an inferential element describing the conditionals that link them to other claims or other mental states, and there is a functional element, describing what job is done by having such terms in our repertoire. [19] In the case of reference the three components would firstly describe how you come to know that a term “a” refers to some object of thought a. The second would describe the inferences in which this piece of knowledge is embedded (this is what the predicate calculus provides for us). And the functional clause would explain the utility of knowing this kind of thing, which is that it enables you to understand what others are talking about, and in many cases enables you to anticipate what you will find as you move around the world. Reference, representation and their kin need not surface as primitives, but as useful concepts telling of the role that particular terms play in the communicative habits and communicative successes of the users of a language. In this way pragmatism should not be hostile to representation, but should enfold it.

We might now face a choice. We might want our concepts of reference and representation to be entirely catholic, covering anything that might be a topic or intended focus of shared attention—anything at all in fact, echoing the thought that if a reader or hearer is to understand the producer of a term then they must literally know what the producer is talking about. Or, we might wish to be discriminating about where such terms are appropriate. At the beginning of the Philosophical Investigations Wittgenstein makes it plain that he rejects Augustine's name-bearer picture of language in general. But he does not say that it never applies: indeed the "primitive" language of his builders one of whom calls for and the other of whom delivers deliver blocks, pillars, slabs and beams is presented precisely as wholly Augustinian. But:

Augustine, we might say, does describe a system of communication; only not everything that we call language is this system. And one has to say this in many cases where the question arises "Is this an appropriate description or not?" The answer is: "Yes, it is appropriate, but only for this narrowly circumscribed region, not for the whole of what you were claiming to describe." (PI §3)

Unfortunately it is not clear what circumscribes the region. Is it that the builders are dealing with visible objects, present in their environment and within a close causal range of them? It seems unlikely that Augustine did not realize that we refer to a huge variety of things outside our present ken. Wittgenstein goes on to say 'When we say: "every word in language signifies something" we have so far said nothing whatever; unless we have explained exactly what distinction we wish to make" (PI 13). In other words, taken by itself Augustine's picture is utterly bland rather than pernicious. Similar points can be made about all semantic vocabulary. There are truths to be had about every subject matter under the sun. Equally we can be said to express ourselves on any subject matter whatsoever, which raises the worry that “global expressivism” will no more identify a definite doctrine than “global representationalism”. After all, we are always trying to express something when we communicate. The detail has to come in the contrasts we draw and the distinctions we make. But which contrasts and distinctions is a poor philosopher equipped to make? Does a kind of minimalism resurface here?

We say that a symbol refers to or represents something when amongst some set of people, in virtue of conventions or regularities in their behaviour, it has the power reliably to coordinate attention on that thing. Hence, reference and representation are as wide-ranging as our abilities to fix attention on the same thing. Those abilities include having attention to things in an environment, like rocks and coastlines, but extend across the whole field of thought, including attention to abstract objects, fictional objects, properties, functions and so on. There will also be cases where we think we are focusing attention on one object, but there is no object there. This is when reference fails, and some version of the theory of descriptions is at hand to tell us what we are doing instead. There is therefore scope here for the semantics or metasemantics to lead us to exorcising unnecessary metaphysics: values, necessities, numbers, selves and other suspects can be revealed not, as appears at first sight, objects of attention, but as reifications deriving from our inferential or other practices.

It would be legitimate to talk of semantic minimalism if, but only if, the phenomena of shared attention were themselves wholly intralinguistic. In other words, the only criterion of success would be that one-time hearers or readers go on to imitate the linguistic behaviour of a producer, becoming themselves producers in turn. But as Sellars or Wittgenstein should have taught us, this is by no means so. To be sure, a symbol will be embedded in inferential practices, mastery of which is signalled by equivalence of linguistic behaviour. But it is hostage to shared entry rules (response to observation) and exit rules (coordination of non-linguistic activity as a result of sayings). Learners demonstrates that they can now read a piece of music not by saying the right things but by playing the right notes often enough, and mariners demonstrates they can read charts when they plot safe courses by using them. The prime criterion of understanding the notion of causation is that you use causal judgements to determine how you expect events to unfold or how you manipulate them to unfold in a desired way.

I see this story as pragmatist in spirit. It vindicates the idea that words are tools—tools for coordinating attention and action. It also vindicates the view that anything more substantial than needs saying will be at the local level where Wittgenstein hoped to uncover it. It redeems semantic vocabulary by placing it, where it should be placed, in the mundane communicative abilities of people. If it suggests that further understanding of those abilities is not to be found only in linguistic behaviour but will require an injection from the philosophy of mind, telling us more about our intensional powers, that may surely be no bad thing.

Wittgenstein also said that in some cases but not all, the meaning of a word is its use. I think he should have done without the qualification.

[1] Jay Rosenberg, ‘Raiders of the Lost Distinction’ Philosophy and Phenomenological Research, 53, 1993, pp. 195—214.

[2] As Rosenberg says ‘When such a torrent of bathwater hits the pavement, it's easy enough to overlook the occasional discarded baby that goes floating by.’

[4] ‘Philosophy as a Kind of Writing’ in CoP, p. 92–3.

[5] Philosophy and the Mirror of Nature, 178

[6] Philosophy and the Mirror of Nature, p. 9.

[7] Richard Rorty, Philosophy and Social Hope, London: Penguin Books, 1999, Introduction, p. xxv.

[8] It would be mischievous to dwell on the thought that Rorty’s paradigm intellectual activity, literary criticism, perhaps along with other hermeneutic studies, is one place where vocabulary inventions seems sometimes to be substantially independent of discovery.

[9] Davidson, D. ‘Empirical Content’ in Truth and Interpretation: Perspectives on the Philosophy of Davidson, E. LePore ed., Oxford: Blackwell, 1986, p. 324. My italics. I discussed Davidson’s pronouncements in Truth: A Guide. Oxford: Oxford University Press, 2005, pp. 148–62.

[10] Davidson, D. ‘A Coherence theory of Truth and Knowledge’ reprinted in Subjective, Intersubjective, Objective, Oxford: Clarendon Press, 2001, p. 144.

[11] Rorty, ‘Feminism and Pragmatism’, in Truth and Progress, p. 226

[12] In footnote 43 on the same page Rorty commends philosophers of the Edinburgh school of sociology of science for bringing in the background history of Galileo and his readers’ interest in such a matter. But the no doubt fascinating history of why I should be interested in the denizens of the grass does not topple the presence of the bumblebee from its role in explaining why I thought it was there.

[13] Rorty sometimes describes human beings as no more than ‘incarnated vocabularies’. But you can’t imagine a vocabulary of any kind hitting the rocks and drowning. See for instance ‘Private Irony and Liberal Hope’ in CIS, p. 88.

[14] John Dewey, Logic: the Theory of Inquiry, in Last Works, ed. Jo Ann Boydston, Carbondale: Southern Illinois University Press, 1968 v. 12, pp. 514–5.

[15] We might remember Frank Jackson's famous jibe that that "he has been at conferences in which people attacking representation nevertheless have in their pockets pieces of paper with writing on them that tell them where the conference dinner is and when the taxis leave for the airport". (Jackson 1997 p. 270).

[16] These remarks overlap to some extent with a discussion with a slightly different focus in ‘Wittgenstein and Brandom: Similarities and Divergences’, which appeared in Disputatio. Philosophical Research Bulletin Vol. 8, No. 9, Jun. 2019.

[17] Carnap was nicely clear about the choice between these two ways of thinking of a language, distinguishing between a “pure” semantics in which we have made the logician’s choice, identifying a language purely by its lexicon and formation rules, retaining the authority to interpret the symbols however we wish, and an “applied” semantics in which interpretation is beholden to the actual behaviour of some identified population (Carnap 1942).

[18] Price, for one, is not a semantic minimalist in this sense. He applauds the Wittgensteinian emphasis on differences of linguistic function, quoting the famous image of the levers in the locomotive to suggest the way forward.

[19] Michael Williams, ‘How Pragmatists can be Local Expressivists’ in Huw Price, ed. Expressivism, Pragmatism and Representationalism, Cambridge: Cambridge University Press, 2013, p. ?.