Philosophical Studies

, Volume 161, Issue 3, pp 349–366

Pragmatic antirealism: a new antirealist strategy


    • Philosophy, School of Social SciencesThe University of Manchester, Arthur Lewis Building
  • Philip Brown
    • Philosophy, School of Social SciencesThe University of Manchester, Arthur Lewis Building

DOI: 10.1007/s11098-011-9742-1

Cite this article as:
Scott, M. & Brown, P. Philos Stud (2012) 161: 349. doi:10.1007/s11098-011-9742-1


In everyday speech we seem to refer to such things as abstract objects, moral properties, or propositional attitudes that have been the target of metaphysical and/or epistemological objections. Many philosophers, while endorsing scepticism about some of these entities, have not wished to charge ordinary speakers with fundamental error, or recommend that the discourse be revised or eliminated. To this end a number of non-revisionary antirealist strategies have been employed, including expressivism, reductionism and hermeneutic fictionalism. But each of these theories faces forceful objections. In particular, we argue, proponents of these strategies face a dilemma: either concedes that their theory is revisionary, or adopt an implausible account of speaker-meaning whereby the content of certain types of utterance is opaque to their speakers. In this paper we introduce a new type of antirealist strategy, which is thoroughly non-revisionary, and leaves speaker-meaning transparent to speakers. We draw on work on pragmatics in the philosophy of language to develop a theory we call ‘pragmatic antirealism’. The pragmatic antirealist holds that while the sentences of the discourses in question have metaphysically contentious truth conditions, ordinary utterances of them are pragmatically modified in context in such a way that speakers do not incur commitment to those truth conditions. After setting out the theory, we show how it might be developed for both mathematical and ethical discourse, before responding to some likely objections.



Take some discourse D, the sentences of which appear to posit some metaphysically dubious class of entities. Sceptics raise problems about the nature and existence of these entities and about our epistemological access to them. If we take these problems seriously, we might be led to doubt whether the sentences of D are true, or that if they are true we are justified in believing them. But these problems, while of concern to philosophers, seem remote from the communicative intentions of most people who utter sentences of D. Nevertheless, those who use the discourse (at least the non-philosophers) seem to take what they say at face value and are not intentionally engaging in fiction.

Numerous areas of discourse have been thought to be like D: ethics, mathematics, folk psychology, talk about colour, time, truth, etc.1 There are three standard responses to D’s predicament:
  1. (i)

    Argue that, while D-sentences commit us to the existence of the allegedly dubious entities, the sceptical problems can be solved.2 We will call this position metaphysical realism.3

  2. (ii)

    Accept that the objections cannot be met and either purge D-talk,4 or revise our attitudes to D to accept rather than believe D-propositions—where accepting a proposition does not involve believing that it is true.5 The first of these options is eliminativism, the second revolutionary fictionalism. These are revisionary antirealist accounts.

  3. (iii)

    Provide a non-revisionary antirealist account of D. Such responses aim to preserve our everyday use of D-sentences, but without incurring commitment to metaphysically contentious entities or revising our attitudes to the discourse itself.

We believe that responses of type (iii) provide the most charitable account of our everyday use of D-sentences, in light of the issue raised in the first paragraph. D-sentences appear to be adequate for the communicative intentions of ordinary speakers, and this adequacy does not seem to be threatened by the outcome of philosophical controversies. However, extant responses of type (iii) face numerous challenges to their non-revisionary credentials. In general, these objections press proponents of (iii) in one of two directions: either to concede that their theory is revisionary after all, or to adopt an implausible account of what speakers mean when they engage in the target discourse. However, we believe that there is a way of developing (iii)—a novel variety of non-revisionary antirealism—that avoids these objections.

We will briefly review the main type (iii) theories and the problems with them in Sect. 1, set out a new theory in Sect. 2, give some examples in Sect. 3 and consider some objections in Sect. 4.

1 Section one

What in general can be said about the meaning of the sentences of D and of normal utterances of those sentences? The leading options are set out in the following table:

Do the sentences have truth-apt content?





1. Expressivism

Do the sentences have truth-conditions that depend on the existence or instantiation of metaphysically contentious properties or facts?





2a. Reductionism


2b. Subjectivism


2c. Minimalism

Do utterances of indicative sentences of D express belief in the content of those utterances?





3. Hermeneutic fictionalism

Are the sentences of D systematically untrue?





Metaphysical realism

Revolutionary fictionalism

We will be ignoring the last row of options as they are responses of type (i) and (ii). The rest are well-known varieties of response (iii): attempts at non-revisionary theories that do not interpret normal utterances of the target discourse as expressing belief in its apparent subject matter. In the remainder of this section we will provide sketches of these theories and some of the most forceful objections to them.6
  1. 1.

    Expressivist theses involve a negative and positive component. First, sentences of the discourse lack truth-apt content,7 despite their formal similarity with ordinary, descriptive sentences. Second, in uttering D-sentences, speakers express their non-cognitive attitudes.

  2. 2.

    Theories in the second row maintain that sentences/utterances of the discourse do not—despite appearances—have metaphysicaltruth-conditions (which we will use as shorthand for truth-conditions that posit metaphysically contentious properties or facts). (a) Reductionists propose that the truth-conditions of sentences and utterances in D can be given in a distinct, metaphysically respectable discourse. (b) Subjectivism is a special case of reductionism according to which the truth-conditions of D-sentences are determined by the mental states of the speaker (or of some class of people). (c) According to minimalists, if the sentences of the discourse seem to have metaphysical truth-conditions that we have reason to think systematically fail to obtain, that is a reason for thinking that we are attributing too robust an analysis of truth to that discourse. Crispin Wright argues that truth can be more minimally construed in terms of justification within a discourse and—at least in the cases of ethical and mathematical discourses—we can avoid the charge of massive error (Wright 1992).8

  3. 3.
    According to hermeneutic fictionalists, sentences in D have metaphysical truth-conditions; however, speakers do not use sentences of D to convey their semantic content but to perform some other speech act. The following two types of hermeneutic fictionalism have been defended:
    1. (a)

      In uttering sentences of D, speakers express noncognitive attitudes. Kalderon (2005) argues that while moral realists are correct about the meaning of moral sentences, expressivists are right about the way they are used. We will call this type of position expressivist fictionalism.

    2. (b)

      Utterances of indicative D sentences are truth-apt but quasi-assertoric. A quasi-assertion is a speech act with the outward appearance of assertion where the speaker does not endorse the uttered sentence but presents it as adhering to some norm other than truth. In Yablo’s (2000, p. 214) mathematical fictionalism, abstract objects function as props that facilitate description, but speakers do not, in availing themselves of these props, incur commitments to them. This is accounted for by construing abstract objects as playing a metaphorical role: “Rather as ‘smarts’ are conjured up as metaphorical carriers of intelligence, ‘numbers’ are conjured up as metaphorical measures of cardinality”. We will call this type of position figurative fictionalism.


Theories in the first and second tiers of our chart face well-canvassed difficulties that identify either some revisionary consequence of the antirealist theory in question, or some feature of D that the theory cannot explain. (1) Expressivism appears to have the revisionary result that a sentence changes its meaning between asserted and embedded contexts, and has problems with explaining the use of certain logical connectives such as negation. The former is brought out by the Frege-Geach argument: The utterance of the sentence ‘It is wrong to break promises’ might be taken by the expressivist to be the expression of an attitude of disapproval towards breaking promises, but it does not express that attitude when embedded in the conditional ‘If it is wrong to break promises then it is wrong to make promises that you do not intend to keep’. Consequently, an argument that used these sentences as the premises of a modus ponens with the conclusion ‘It is wrong to make promises you do not intend to keep’, will be logically invalid on an expressivist theory. Expressivists have devised a number of strategies for dealing with these problems, but it is controversial whether any of them can succeed.9 (2a) Reductionism has difficulty in providing a plausible reduction of the target discourse that does not carry the same metaphysical commitments. Reductionists can claim that the reducing statements capture the kosher content of D-statements, while discarding the commitment to any contentious entities; this, however, would be a revisionary theory.10 (2b) Subjectivist accounts have particular difficulty with explaining how disagreement is possible. If, for example, an ethical utterance is true just in case the speaker believes it then we cannot disagree with someone’s ethical judgement (assuming they have sincerely expressed it) on the grounds that their opinion is false.11 (2c) Minimalism has a prima facie problem with accounting for the validity of arguments composed of sentences from discourses that employ different truth predicates.12

Antirealist theories in the third tier of our chart, that accept a standard semantics for D, are not susceptible to objections that they are revisionary about D-language. However, the most serious objections to hermeneutic fictionalism present the second horn of the dilemma given in the introduction: they seem committed to an implausible account of speaker meaning. Expressivist fictionalism can account for the continuity in meaning between embedded and non-embedded occurrences of D-sentences. But, as Eklund (2009) has pointed out, someone uttering a moral modus ponens, for example, intends the thoughts that she expresses by her premises to give us a reason to accept her conclusion. If these utterances express non-cognitive attitudes, the embedding problem transfers from D-sentences to the thoughts expressed by D-utterances. Similarly figurative fictionalism, while it adopts a standard semantic theory of D-language, runs into difficulty in giving us a plausible account of what users of D-discourse are communicating. As Jason Stanley (2001) has pointed out, we would expect competent speakers of a D-discourse such as arithmetic to deny that they were using figurative language. If figurative fictionalism is true, however, then speakers are often seriously misguided about what speech acts they are performing when they make D-utterances (and about what mental states those utterances express).

Antirealists over the last 30 years have attempted to develop non-revisionary theories to explain the apparent similarities and logical relationships between realist and target antirealist regions of language. Hermeneutic fictionalism is perhaps the most successful recent development in this debate. But there are also similarities between the types of communicative intention and speech act engaged in by competent speakers when using target antirealist and realist areas of discourse. Speakers freely move between different discourses and combine sentences and characteristic expressions from different discourses, without any apparent change in either the type of mental states that are being conveyed or the force of what is said. Here, however, fictionalists are unsympathetic to appearances. We are—contrary to how it seems to us—wrong to think that we believe the sentences that we are uttering, unaware of what we are really communicating, and not fully appreciative of the speech acts in which we are engaging.

This is not, of course, a serious problem for antirealists if it can plausibly be argued that our meanings are, to a degree, inaccessible to us (or, at least, very difficult to access). However, it does commit the antirealist to special pleading on behalf of the target discourses. When using discourses the sentences of which have metaphysically pukka truth-conditions, we know when we are pretending, speaking in code, or saying something purely for affect; but in the target discourse(s), the antirealist contends, this ability goes awry. The problem for the antirealist is not that the sentences of D have truth-conditions of which speakers are unaware—the attribution of the correct truth-conditions to sentences often takes a considerable amount of theorizing—but that utterances should have or lack content in way that is opaque to speakers.

Is it possible, therefore, to develop a non-revisionary antirealist account of D language which escapes the implausible consequences that beset existing forms of antirealism? In the next section, we will argue that this can be done with a pragmatic theory of utterance meaning—a standard method of interpretation in philosophy of language that has yet to be applied to the realism debate.

2 Section two

On a standard semantic theory, the content of utterances—what issaid—is determined by the meaning of the uttered sentence (analysed in terms of its truth-conditions) after its component ambiguous expressions have been disambiguated and its indexical or context sensitive expressions have been assigned values. Take for example the utterance
  1. 1.

    He is in the grip of a vice.

What is said by this utterance is determined by the meaning of the sentence after the disambiguation of ‘vice’ and ‘grip’ and the saturation—the assigning of values—of the indexical ‘He’. In addition, there is an important but separate pragmatic story to be told about what the utterance suggests or implies; this most clearly comes into play with the interpretation of non-literal utterances.

All of the antirealist theories that we considered in Sect. 1 adopt the framework of a semantic theory. They seek to avoid commitment to contentious metaphysical entities by either introducing modifications to the semantics of D-sentences (expressivism, minimalism, reductionism); or by introducing a distinctive pragmatic account of D-sentences, without modifying the semantics (fictionalism).

There is, however, an alternative framework for analysing the meaning of utterances. On a pragmatic theory, or pragmatics, what is said by an utterance is determined not only by the semantic processes of disambiguation and saturation, but also by the context driven pragmatic processes of loosening, narrowing and transfer.13 Semantic processes are generally linguistically mandated, in that they are prompted by ambiguous or indexical expressions that require additional information to give the utterance a determinate propositional content. Moreover, the type of information required for a given context-sensitive expression is standardised: ‘I’, for example, takes the speaker of the utterance as its value in all contexts of utterance. Pragmatic processes, in contrast, are prompted by contextual factors and may result in non-standard, ad hoc modifications to the information content of the concepts encoded by the constituent expressions of the utterance.14 A concept is narrowed by introducing some contextually salient restrictions into its content and is loosened when a restriction is stripped out; with transfer, the modified concept refers to contextually salient but different properties to those picked out by its linguistically encoded counterpart. In pragmatics, these modifications occur on the fly in the process of communicating and interpreting utterances, they can apply to concepts that are not linguistically signposted as context sensitive, and they are pervasive.

Consider, for example, the sentence
  1. 2.

    The city is constructed on granite

uttered in the context of a discussion about Scotland, and the sentence
  1. 3.

    The audience fell silent

uttered in the context of a discussion about London music venues. On a pragmatic analysis, the concepts CITY and AUDIENCE can be seen as narrowed to give CITY* (meaning: ABERDEEN CITY) and AUDIENCE* (meaning, given the relevant context: AUDIENCE IN THE 12 BAR CLUB) while some of the conditions specified in the literal application of concepts GRANITE and SILENT are dropped to give GRANITE* (meaning: PREDOMINANTLY GRANITE) and SILENT* (meaning: QUIET ENOUGH NOT TO BE NOTICEABLE). With the sentence
  1. 4.

    The ham sandwich has left without paying

uttered by a waiter about a customer leaving the restaurant, the content of the concept HAM SANDWICH is pragmatically modified, by transfer, to HAM SANDWICH* (meaning: HAM SANDWICH ORDERER). Pragmatics, like semantics, distinguishes between what is said by an utterance and what an utterance implies, but employs an enriched account of what is said that is supposed to accord with normal speakers’ understanding of the truth-conditions of their utterances.

The pragmatic framework provides us with a new way of developing an antirealist account of D-type discourses that is fully non-revisionary. We will call this pragmatic antirealism. The position is as follows: Indicative sentences in D have metaphysical truth-conditions.15 However, D-talk is pragmatically modified in context by loosening, narrowing or transfer of the information content of the distinctive concepts of D, such that the truth-conditions of D-utterances (‘what is said’, pragmatically) are metaphysically harmless. In a simple case, this might involve stripping out or modifying a component in the concept’s logical information content (i.e. one of the rules for its correct use), such as it standardly referring to a certain (metaphysically suspicious) property; all the other logical and encyclopaedic information content in the concept remains unchanged.

The pragmatic antirealist need not dispute a realist analysis of the truth-conditions of sentences of ethics, mathematics, etc.; she does, however, propose that pragmatic effects intrude on the content of utterances to a degree and in ways that the semanticist rejects. Philosophers, on this view, have analysed ‘what is said’ in the semanticist’s sense, and then tended to treat utterances of those sentences as if they expressed similar content, modified by only linguistically mandated contextual information. Once pragmatic effects are factored in, the truth-conditional content of the utterance may be substantially modified. For some discourses the modification will not be realism-relevant: both sentences and utterances of them will have the same metaphysical truth-conditions. However, where it can be argued that D-utterances pragmatically screen out the metaphysical commitments of the D-sentences they express, then pragmatic antirealism becomes a live option.

Pragmatic antirealism follows the same route as hermeneutic fictionalism up to the third line of our chart: D-utterances do not have the metaphysical truth-conditions of the D-sentences they express. However, it diverges on the question of whether or not speakers assert or believe what they are saying. This is possible because pragmatics allows that differences between the truth-conditional content of an utterance and the sentence it expresses are a routine occurrence in normal literal, assertoric discourse. In contrast, by working with a semantic theory, hermeneutic fictionalism requires supplementary theories of fiction, metaphor, quasi-assertion, pretence, etc., to show that D-utterances do not have the same metaphysical commitments as D-sentences, and to distinguish it from the revisionary and realist options on the fourth tier of our chart. Since, as we argued in Sect. 1, these supplementary theories give rise to what we take to be the most serious weakness in hermeneutic fictionalism—namely, its implausible account of speaker meaning—we see this as an important advantage of pragmatic antirealism over fictionalism. Pragmatic antirealism, therefore, answers ‘Yes’ to the question ‘Do utterances of indicative sentences of D express belief in the content of those utterances?’ It opens up a new antirealist option between the third and fourth tiers of our chart by answering ‘No’ to the question ‘Do utterances of indicative sentences of D express beliefs in the content of those sentences?’ The last tier should properly be in response to the question ‘Are both or neither of the sentences and utterances of D systematically untrue?’16

3 Section three

Our main aim in this paper is to set out a framework for developing pragmatic antirealism rather than defend any particular variety of the theory. However, let us sketch out how pragmatic antirealism might be developed for two of the most frequently discussed discourses in the realism versus antirealism debate: mathematics and ethics. We will first consider Joseph Melia’s fictionalist account of mathematics and show how the evidence about speaker meaning that he uses can be incorporated into a pragmatic framework to support pragmatic antirealism about mathematical discourse. Then we will look at how recent attempts by some moral antirealists to deflate the metaphysical commitments of ethical language provide a route into pragmatic antirealism about moral discourse.

Mathematical realists like to argue that our best scientific theories are true if and only if mathematical objects exist; and that given that they are our best theories, we ought to believe that they exist.17 Melia (2000) agrees that scientists use theories that entail the existence of mathematical entities and that these theories may be essential in a true description of the physical world. However, he does not think that this establishes mathematical realism because the commitments that are apparently taken on when a scientific theory is endorsed can be disavowed—a manoeuvre he calls weaseling. For example, in asserting some scientific theory T*, that quantifies over sets, we can escape commitment to sets by saying:
  1. (i)

    T*-but there are no such things as sets.

Melia thinks that this is not an unusual part of our linguistic practice. For example we might find ourselves saying:
  1. (ii)

    Everyone who Fs also Gs,

but subsequently weasel our way out of the universal generalisation:
  1. (iii)

    Except Harry—he’s the one exception.


Melia claims that weasels are not hypocrites and that in taking back what they say they have clearly not ‘spouted contradictory nonsense’ (468), though he only hints at why this is so. He proposes that weasels say things—like (ii)—that they do not believe, but also that they are not dishonest (469); he suggests that weasels are using fictions and what they are doing is akin to telling a story (470). So he appears to be committed to figurative fictionalism about mathematics. However, there is a pragmatic antirealist interpretation of the phenomenon that Melia identifies in (i) that avoids the problems that we have identified with fictionalism.

For pragmatic antirealist purposes, (i) and similar utterances provide salient evidence about the pragmatic content of mathematical utterances in scientific and other practical contexts. In response to philosophical questioning, scientists, engineers, architects, etc., may deny the platonic implications of what they say; Melia plausibly suggests that accusations by philosophers of ‘double-think—denying by night what they believe by day’ (269) are misplaced. However, we propose that Melia has misidentified what scientists and others say as weaseling: (i) does not show that T* sentences are quasi-asserted, rather, it corrects the impression that the assertion of T* communicated anything with platonic implications. We take claims like (i) as evidence that the metaphysical truth-conditions of the sentences of T* have systematically lapsed for the utterances of T* sentences in practical contexts. This could result from filtering out references to mathematical entities in the constituent mathematical concepts employed in T* by the pragmatic process of loosening. How would this work? Consider the concept ONE. Plausibly, ONE includes in its logical information:
  • Log1: Single; not many

  • Log2: The first cardinal number

  • Log3: a component of arithmetic that is used like this …

  • etc.

In practical contexts where Log2 is not relevant to what is being communicated, ONE* could be constructed by loosening: dropping Log2. Since the other standard rules for the use of the concept (as well as other encyclopaedic information) are retained, the loosening is harmless with the respect to the concept’s practical application.

From the pragmatic antirealist’s perspective Melia’s comparison between (i) and the generalisation in (ii) and (iii) is inapt because we lack sufficient contextual information to interpret the latter two utterances. The speaker’s second utterance in (iii) could be a genuine retraction of a false implication conveyed, intentionally or otherwise, by (ii). Alternatively, it could be more like our reading of (i): not weaseling but a clarification of what was intended in response to the questioner’s misinterpretation.

To make this latter distinction clearer, consider the following utterances:
  1. 5.

    The house was silent.

  2. 6.

    The fridge is empty.

On the pragmatic approach, in most contexts (5) will be a standard case of loose talk: SILENT (meaning: SOUNDLESS) is loosened in context to SILENT* (meaning: VERY QUIET); (6) is a standard case of narrowing: EMPTY is modified to EMPTY* (meaning: DEVOID OF FOOD). But consider the following dialogues between A and B:
  • A: The house was silent.

  • B: But surely there was at least some sub-auditory atmospheric noise?

  • A: Ok, the house was silent—but there may have been some sub-auditory noise.


  • A: The fridge is empty.

  • B: Where have the shelves gone?

  • A: It’s empty of food.

In both dialogues B has failed, intentionally or unintentionally, to perform the pragmatic operations needed to understand what A has said. A’s responses are neither retractions nor weaseling: A is clarifying what she said in response to B’s tin ear (or facetious) questions. In both cases, A could have added ‘… and you know very well, that’s not what I meant’. This, the pragmatic antirealist can argue, is the way in which we should interpret (i) when given in response to a philosopher’s questions about mathematical realism.

If we are working with a semantic theory, the utterance (i) looks like a retraction: a correction of ‘what is said’ (semantically). The only way to avoid the charge of inconsistency within the terms of semantic theory is to propose that the speaker did not really assert the claim that she appears to qualify; instead, some different speech act was employed. This, in effect, is what Melia proposes about T* in suggesting that (i) and (iii) are weasel words rather than retractions. However, positing non-assertoric speech acts generates the further problems that we identified in Sect. 1. For the pragmatic antirealist, scientists can be understood as merely clarifying what they have said. That is, scientists can be absolved of platonic commitments without taking what they have said as quasi-assertions, fictions or double-speak.

There is, of course, much more to say and argue about mathematical pragmatic antirealism and we will provide some additional evidence and arguments in the following section. This is not, however, the place for a detailed defence of one variety of pragmatic antirealism; for the rest of this section, we will look at how the theory might be developed for ethics.

Pragmatic antirealism will be well positioned in respect to discourses where it can be argued that pragmatic modifications of its constituent concepts result in utterances lacking truth-conditional content that is necessary for realism about the discourse. Ethics is a case in point. We think it unlikely that an analysis of the pragmatic content of ethical utterances will satisfy the standard conditions for ethical realism and that there should be a defensible form of pragmatic ethical antirealism.

To see how pragmatic ethical antirealism might be developed, let us review one of the more effective non-revisionary antirealist strategies in recent work in ethics: giving deflationary, metaphysically innocuous interpretations of the central concepts that are taken to be characteristic of ethical realism. Most contemporary formulations of ethical realism have a semantic and a metaphysical component:

(S) Ethical utterances have truth-apt content; they describe ethical facts and/or attribute ethical properties and are true or false depending on whether they accurately represent those facts or properties.

(M) At least some ethical facts obtain.18

However, deflationary analyses are available for all of the key terms of this theory, including truth, fact, description, property, truth-apt, etc. Take the familiar example of deflationism about truth, according to which everything that can in general be said about truth is exhausted by the disquotational schema. With this analysis of truth, even an expressivist can express the judgement ‘Breaking promises is wrong’ as ‘It is true that breaking promises is wrong’ without taking on additional metaphysical commitments. This deflationary or ‘quasi-realist’ strategy can be extended as follows: An utterance describesx as A just in case it says that x is A; it is a fact that x is A iff it is true that x is A; x has the property of being A iff x is A.19

Non-revisionary anti-realists, and in particular expressivists, have exploited quasi-realism to present problems for ethical realists in two ways. First, since on a deflationary reading of (M) and (S) these conditions can be endorsed by antirealists, realists need either to provide a more sophisticated statement of those conditions or to identify supplementary conditions to distinguish realism from antirealism. Second, by providing an interpretation of our talk of moral truth, facts, etc., that is compatible with antirealism, it follows that we cannot ‘read off’ a realist theory from our use of realist-sounding expressions in moral discourse. Realists need to justify a realist interpretation of these expressions.

Moral realism is often supplemented with an objectivity or independence condition, such as

(O) Moral facts obtain independently of the attitudes and beliefs of moral appraisers.

Here too, a deflationary reading is available. As Simon Blackburn points out, someone who disapproves of cruelty to animals will presumably also disapprove of our moral sensibilities about animal cruelty hinging on facts about our beliefs and attitudes rather than, say, the pain cruelty causes animals. This second-order attitude can be expressed by saying that the wrongness of animal cruelty is mind-independent (Blackburn 1984, pp. 217–220); more generally, it seems that antirealists should be able to endorse (O). Various suggestions have been put forward for a more robust construal of the objectivity condition, i.e. one which could not be endorsed by antirealists using quasi-realist tactics. According to Jenkins (2005) independence should be understood ‘essentially’ rather than modally: the realist should argue that that it is no part of what it is for such and such to be wrong that we have a negative attitude towards it (as opposed to such and such being wrong even if our attitudes were different). Shafer-Landau (2003, p. 15) argues that a realist account of objectivity requires that ‘the moral standards that fix the moral facts are not made true by virtue of their ratification from within any given actual or hypothetical perspective’; FitzPatrick (2008, p. 166) proposes that the independence of ethical facts requires that ‘they are not constituted by the actual or hypothetical results of any ethically-neutrally specifiable set of conditions or procedures applied to our beliefs, desires, attitudes, etc.’ Miller (2009) contends that realist moral facts are independent from our intentional attitudes and conceptual schemes such that they obtain in nearby possible worlds where human beings have different moral attitudes and beliefs.

Now, pragmatic antirealists about ethics do not need to take a position in this debate, either on how best to define moral realism, or on how far or how effectively the deflationary strategy can be taken. The relevance of the debate for the pragmatic antirealist is that quasi-realism has imposed on realists the need for a more sophisticated statement of their theory. What is important here for the pragmatic antirealist is not whether realists are able to successfully distinguish their position from quasi-realism, but rather that they have been led to such sophisticated refinements of their theory at all. For the more elaborate and technically subtle ethical realism becomes, the less plausible it is as an account of the meaning of ethical utterances.

Take the discussion of condition (O). One can (sometimes) elicit from a moral appraiser the opinion that a given moral judgement expresses a fact independent from anyone’s beliefs or attitudes; but this, as we have seen, can easily be captured by the quasi-realist. What of the proposed robust modifications of (O)? When speakers aver their belief in some moral fact, they are not generally committed to a view about the relationship between those facts and our ethically-neutral decision procedures, or whether they obtain in nearby possible worlds where human beings have different conceptual schemes, etc. Why is this? First, it is doubtful that we can attribute to moral speakers even a grasp of the distinctions involved in the more sophisticated accounts of objectivity; such distinctions usually require a course in metaethics to appreciate. Second, it would be surprising if the sophisticated conditions needed for robust realism were part of the content of ethical utterances, since these conditions have no relevance in the practical decision making in which ethical utterances are usually employed. Even in contexts where more critical reflection is involved, such as when we make a moral judgement about the moral practices of people in other cultures or times, the kind of culture-independence of moral facts required for such judgements to be true is still no more robust than that given by the quasi-realist. Third, let us suppose that we could find robust realism about objectivity in the pragmatic content of moral utterances. Since moral realism is a package deal of necessary and jointly sufficient conditions, we would still have to contend with making a similar case for (S) and (M), i.e. that speakers are using robust conceptions of truth and fact, etc.

Note that the argument developed here for pragmatic antirealism does not rely on a positive story about the pragmatic content of moral utterances but only on the observation that this content—whatever it may be—does not satisfy the conditions for robust realism. In fact we suspect, consistently with the pragmatic model we proposed in Sect. 2, that the concept GOOD varies in its contribution to utterance content depending on the context of the utterance, ranging from a minimum (loosened) registering of approval to (in more critical contexts) the attribution of a mind-independent property. Moreover, there is empirical evidence for this context sensitivity. One set of experiments (Goodwin and Darley 2008) has measured subjects’ judgements about moral objectivity, in part by presenting to them a range of moral claims and assessing their views on whether those claims are determinately true or false or whether there can be differences of opinion where neither party is mistaken. A principal finding of the experiment was that the commitments of a subject to the objectivity of a given moral claim vary according to the moral issue being addressed as well as the subject’s degree of agreement or disagreement with the claim in question.

To summarise, the quasi-realist/realist debate in metaethics has been driven in part by the aim of providing a realist semantics for moral discourse that cannot be captured by the quasi-realist. As the semantic account has become more complicated, it has become less plausible as an account of what is being communicated when speakers use moral concepts. This presents an opening for the pragmatic antirealist to argue that the content of moral utterances falls short of what is required for moral realism.

4 Section four

Let’s conclude by considering some objections to pragmatic antirealism.

1. Pragmatic antirealism is claimed to lack the revisionary defects of other antirealist theories. But is this achieved at the expense of jettisoning semantic theory as an account of utterance meaning?

We do not see this as a theoretical cost. Pragmatics is a well-entrenched theory in both philosophy of language and linguistics, and we suggest that the use of semantic theory as the assumed framework of debates about realism and antirealism is a matter of philosophical tradition rather than conceptual necessity.

2. The pragmatic antirealist proposes that the predicates that are distinctive of a disputed discourse undergo pragmatic modification. But how does one know, in general, when a pragmatic modification has occurred and what form it takes? Unless we can tell when and whether a pragmatic modification has occurred, the pragmatic antirealist’s theory looks entirely speculative.

A pragmatic modification occurs when an encoded concept, employed in a certain context, has its information content changed by pragmatic processes. That a pragmatic modification has occurred should be discoverable with the combination of empirical evidence about speaker meaning and linguistic analysis. Analysis is needed to uncover the information content of an expression independently of any particular context of application; evidence about what speakers mean when they use the concepts and how hearers interpret their utterances, along with information about the contexts in which those concepts are used, should allow us to establish whether they have been modified.

In Sect. 3 we gave a partial outline of an analysis of a numerical concept and evidence—the willingness of speakers in scientific contexts to deny the seeming platonic implications of their utterances—of its pragmatic modification. Here are three other sources of evidence. First, in most contexts in which number concepts are used, the question of whether or not numbers exist is irrelevant to what is being communicated and understood. The utterance that, for example, ‘the number of goals scored in the 2010 world cup final is one’ conveys information about goals and a football match; its truth is neither intended nor understood to depend even in part on the outcome of the platonism/nominalism debate. In general, the truth of utterances employing number concepts in practical contexts appears independent from metaphysical truths about the status of mathematical objects. This lends support to the theory that the constituent mathematical concepts have been loosened to remove this extraneous information. Second, a piece of evidence that is noted by Yablo is that utterances that seem to imply the existence of numbers have a ‘translucent’ quality whereby they are understood to refer to concrete objects rather than numbers: ‘You “see through” my statement that the number of zebra mussels has doubled in a year to the fact that I was trying to get across: there are twice as many zebra mussels as a year ago’. (2001, p. 89) Yablo suggests that this can be explained if numbers are representational aids; the pragmatic antirealist contends that it can be explained by the loose use of number concepts, where reference to numbers has been filtered out.

Third, and following from the first point, consider the fact that speakers employing numerals are seemingly indifferent to the question of whether they exist. This point is brought out by Yablo’s variation on the oracle thought experiment. (Yablo 2000; Burgess and Rosen 1997, p. 3) The oracle, an unimpeachable source of information, tells you that there are no abstract objects and that everything is concrete. But it seems clear that your subsequent attempts to raise concerns about the truth and legitimacy of the claims and practices that rely on mathematical theory would be met with irritation and derision. Yablo thinks that this evidence lends support to figurative fictionalism, but it can equally be accommodated and explained by pragmatic antirealism. In the same vein, Melia suggests that our confidence with mathematics is not vulnerable to doubts about the existence of abstract objects: “Show me a metaphysician who has tried telling a large number of people ignorant of philosophy (such as a class of first year undergraduates) that there are such things as numbers …, which we cannot see and with which we cannot interact, and I will show you a person familiar with a wide variety of incredulous stares and disbelieving sneers” (Melia 1995, p. 223). That such metaphysical scepticism is easily elicited, while scepticism about the truth of basic mathematical claims is not, adds further support to the view that mathematical utterances in practical contexts do not posit abstract entities.

3. Pragmatic modifications are ad hoc and context specific conceptual changes. But the changes proposed by the pragmatic antirealist appear to be much more systematic, for example, applying across the standard use of mathematical concepts in practical contexts. How can there be a modification that systematically modifies the meaning of classes of concepts?

Pragmatic modifications are pervasive and contextually driven. They can also be systematic if the context that prompts the modification of a concept is in place in most circumstances in which the concept is used. Consider the examples of the narrowing of ‘silent’ or the loosening of ‘empty’ given in Sect. 2—these modifications occur not just in the specific context given, but in most contexts in which they are used. We can add: ‘Everywhere’, ‘full’, ‘vacuum’, ‘nothing’, etc. These concepts are rarely used in their strict, unmodified sense except in specialised scientific or philosophical discussion. A similar phenomenon can occur, we propose, in type-D discourses.

4. Does pragmatic antirealism give us license to continue to engage in any area of discourse unimpeded by metaphysical objections to the claims of the discourse?

As indicated in response to question (1), pragmatic antirealism will only be defensible for an area of discourse if a plausible case can be made that there is a realism-relevant pragmatic modification in operation. Pragmatic antirealism can meet objections to its non-revisionary status that present serious problems to other forms of antirealism; it is not a way to extend the reach of antirealism to fields of discourse where the theory lacks supporting evidence. Moreover, there are discourses that do have (contentious) metaphysical commitments in the pragmatic content of their utterances. It seems to us that religious discourse—or at least talk of God—is a case in point. The pragmatic content of talk about God seems to have the existence of a supernatural entity among its truth-conditions: speakers understand themselves as referring to and communicating information about a supernatural entity. Consequently, reasons for scepticism about the metaphysical commitments implicated in using religious discourse—i.e. that God does not exist—are also reasons for either eliminating the discourse or revising our attitude towards it.

Similarly, we are doubtful that the earlier proposed pragmatic modification of mathematical concepts occurs in some pure contexts. For statements of, say, theorems of combinatorics or number theory and mathematical statements contributing to their proofs, the evidence of pragmatic modification does not stack up so well. We expect that the speakers of these utterances—mostly pure mathematicians—would not endorse a denial of the existence of mathematical objects, that the existence of such entities is not irrelevant to what they are communicating, and that while they might not regard the pronouncements of the oracle as causing a serious crisis in mathematics (pure mathematicians not being inclined to overdramatise) they would regard it as a serious challenge and revise their opinion of the oracle’s authority. For mathematical discourse in such contexts, as with religion, we think that the antirealist will have to take a revisionary option.

5. It may seem that pragmatic antirealism runs into some of the same difficulties as figurative fictionalism. Take, for example, Yablo’s theory according to which mathematical utterances are metaphorical: the realist’s mistake is to interpret mathematical claims as asserting a literal (and anti-nominalistic) content that is not intended. As Burgess (2004, p. 26) points out, it is a problem with this theory that there is a defeasible but reasonable assumption that we should take speakers’ utterances at face value: ‘The burdenof proof is on those who would suggest that people intend what they say only as a good yarn, to produce some actual evidence that this is indeed their intention.’ As indicated in Sect. 2, we are sympathetic to this argument: the figurative fictionalist attributes to apparently competent users of mathematical discourse speech acts of which they are not aware, and this seems unwarranted without supporting evidence. This is a not a problem for the pragmatic antirealist, who takes the speech acts of competent speakers typically to accord with what those speakers intend and appear to make. However, Burgess adds:

Word-meaning and speaker-meaning, though distinct, are not independent. It would be impossible for words to mean what they do if everyone always used them to mean something else, and difficult for them to mean what they do unless most people most of the time use them to mean that, so that a randomly chosen person at a randomly chosen time probably means what he or she says. (26)

Could Burgess argue that since both figurative fictionalism and pragmatic antirealism propose that the relevant predicates of the target discourse are not used literally, this objection apply to both theories?

Not so: the pragmatic antirealist’s position is that the disputed predicates are being used literally. Burgess’ argument rests on the assumption that the literal meaning of a word should be one of the standard ways in which it is used. This is a problem for the figurative fictionalist, who argues that the standard use of the disputed class of expressions is metaphorical (or in some other way non-literal). The pragmatic antirealist, however, rejects the assumption: it is normal for concepts in literal usage to undergo pragmatic modifications in context. Moreover, there seems to us to be a tension between Burgess’ first and second objections. On the one hand, he criticises Yablo for attributing to users of mathematical discourse speech acts that they are not aware that they are performing; on the other hand, he attributes to speakers a metaphysical content to what they mean of which they are not aware nor intend to communicate. Pragmatic antirealism is able to do justice to speakers’ understanding both of the speech acts they are making and of their content.


Scepticism concerning these discourses has been advanced on various grounds, of which the following is merely a sampling. Mathematical entities are thought to be problematic because, if they are abstract objects then it is hard to account for the accuracy of beliefs about them: what effect can non-spatial, acausal entities have on the mental states of mathematicians? (Field 1989). Moral properties have been thought to combine objectivity and prescriptivity in a metaphysically dubious way (Mackie 1977); or to be superfluous to explaining moral thought, given debunking evolutionary explanations (Joyce 2006). Propositional attitudes have been argued to be posits in a radically defective folk-theory of the mind, which will have no place in completed neuroscience (Churchland 1981).


See e.g., Fodor (1975) on propositional attitudes; Brink (1989) on ethics.


What counts as ‘realism’ in these debates is to some extent a matter of stipulation by individual authors. Note that on our use of the expression, theories that respond to the sceptical problems by denying that the allegedly dubious entities are mind-independent will count as realist. Response-dependence theories may fit this bill (for an example of these theories in ethics see Wiggins (2002)).


See e.g., Churchland (1981) on propositional attitudes; Garner (2010) on ethics.


See e.g., Dennett (1987) on propositional attitudes; Joyce (2001) on ethical discourse.


It should be noted that not all the theories we sketch below have been applied to the D-discourses. For instance, subjectivism has not been applied to mathematical discourse.


Some expressivists allow for truth-apt content that does not represent the distinctive subject matter of the target discourse. See, for example, Ridge (2006).


Some ethical constructivists construe moral truth as the satisfaction of certain procedures of moral scrutiny (Korsgaard 1996; Street 2008), albeit ones that are requisite for practical decision making. This can be understood as a form of minimalism.


See Schroeder (2008) for a useful roundup of the various problems facing expressivist semantics and attempts to resolve them.


These and other problems with reductionism are discussed in Blackburn (1984, pp. 151–166).


See Schroeder (2008, pp. 16–17).


See Tappolet (2000).


As with semantics, there are different ways of developing the pragmatic position (see Huang 2007 for an overview). The view we present here is most closely aligned with Recanati (2004), Carston (2002) and Sperber and Wilson (1995) but could, with minor modifications, be brought in line with any of the leading forms of pragmatics.


Carston usefully distinguishes (a) logical information: the inference rules and analytical implications of the concept; (b) encyclopaedic information, which involves commonplace beliefs about the object of the expression; (c) lexical information about the phonetic and syntactic characteristics of the expression that encodes the concept. In the examples we are considering here, it is logical information that is being modified.


We will leave aside the question of whether the truth-conditions of sentences can be specified by semantic interpretation (as Sperber and Wilson allow), or whether additional pragmatic information is required to give a propositionally complete content (Bach 1994).


Note that a further subdivision is required here to accommodate two other positions that adopt a pragmatic framework. First, a pragmatic realist might accept that D-utterances are pragmatically modified but also maintain that D-sentences are true (even if they are rarely asserted with their semantic content intact). Second, a different kind of pragmatic antirealist could maintain that (some) D-sentences are true, but that pragmatic modification renders them false.


This is often known as the ‘Quine-Putnam indispensability argument’. The locus classicus is perhaps Putnam (1979).


While realism can be construed as a purely metaphysical thesis (Miller, 2009), it is difficult to see how (M) could be defended without (S). If ethical judgements do not represent ethical facts, then no argument in favour of believing the truth of an ethical claim would lend support to the existence of ethical facts. (See FitzPatrick 2009, p. 747).


See Dreier (2004) for a review.



Thanks to David Liggins and Michael Ridge for comments on an earlier draft of this paper.

Copyright information

© Springer Science+Business Media B.V. 2011