1. Introduction
Expressivists about normative vocabulary typically want to contrast descriptive declarative speech, which has a straightforwardly representational function, with normative discourse, where declarative sentences are used in other, non-representational ways. Thus, normative expressivism tends to be a local form of anti-representationalism (see Blackburn Reference Blackburn and Price2013). By contrast, pragmatists like Price (Reference Price2011, Reference Price, Johnson and Smith2015) recommend embracing anti-representationalism all across the board. Global anti-representationalism is motivated to a large extent by minimalist views about representational notions, according to which such notions cannot play a substantial role in our ultimate explanations of why sentences mean what they mean. Adopting minimalism is often taken to be a natural move for expressivists, perhaps even a compulsory one. Therefore, it seems that local expressivists find themselves in an unstable position: once one becomes a representational minimalist for some specific domain of declarative speech, it is difficult not to endorse minimalism globally (see Chrisman Reference Chrisman2008; Dreier Reference Dreier2004; Price Reference Price2011).
My aim is to examine whether the sort of demarcation that local expressivists have in mind can be drawn in a way that respects the representational minimalism endorsed by pragmatists. In other words, I will try to reconstruct the expressivist contrast without appealing to representational notions, but by identifying relevant differences in the practical effects of declarative speech acts in the domains taken to be non-descriptive by expressivists. More specifically, I will propose that assertoric speech acts in those domains behave in a relativistic way. The distinction between relativistic and non-relativistic assertion, which can be specified in non-representational terms, would underlain the contrast pursued by expressivists about normative matters.
I start by introducing in Section 2 the challenge faced by local anti-representationalist views. In Section 3, I characterize in more detail global pragmatist approaches, and, in Section 4, I show how normative and evaluative speech can be integrated within a pragmatist framework. In Section 5 I discuss the limitations of trying to capture the alleged non-descriptive nature of normative speech by appeal to the involvement of higher-order modal operators. In Section 6 I turn to my positive proposal. I first sketch a model of relativistic assertion and then I explain how the contrast between descriptive and non-descriptive speech can be reformulated in terms of the distinction between relativistic and non-relativistic assertoric practices.
2. Expressivism and the bifurcation thesis
Expressivist theories, as I will understand them here, aim to explain why sentences have the meaning they have in terms of the mental states conventionally associated with their use – more broadly, in terms of the characteristic function of the speech acts conventionally performed by uttering those sentences. By contrast, representationalist theories of meaning explain what it is for a sentence to have a certain meaning by appeal to some representational relation between the sentence, or its use, and the world.
I will regard both theories as metasemantic proposals, in other words, as proposals about what it takes for some sentence to have certain semantic features (Charlow Reference Charlow2015; Chrisman Reference Chrisman2015; Pérez Carballo Reference Pérez Carballo, Burgess and Sherman2014; Silk Reference Silk2015). In a way, expressivist and representationalist metasemantics can be made compatible: there can be representationalist versions of expressivism. This will happen when the characteristic function of the relevant speech acts (or the mental states associated with them) is specified in representationalist terms. So, one could say that the function of assertions of p is to express the belief that p, and then add that believing that p is a matter of representing the world in a certain way (in short, the function of assertions of p would be to represent the world as being a p-world). This account of why p means what it means would be both expressivist and representationalist.
However, expressivist often want to claim that there are areas of discourse where this representationalist approach is inadequate, in particular normative and evaluative speech (and also epistemic modals or indicative conditionals). In such areas of discourse, the relevant speech acts would have a non-representational function – they would be associated with non-representational mental states. For example, normative statements tend to be associated with non-doxastic mental states with a motivational dimension (e.g. desires, plans, preferences or states of endorsing some norm). In this way, expressivism provides an attractive framework for those with antirealist intuitions about some domain. Expressivism allows one to grant that the vocabulary in the domain in question can be used to make meaningful assertoric speech acts, while denying that these speech acts are meaningful (and sometimes accurate) in virtue of there existing referents for that vocabulary, or there being facts that correspond to the contents asserted. Thus, expressivism makes it possible for the antirealist to account for our patterns of speech in the relevant domain without undertaking undesired metaphysical commitments.
Now, it seems that expressivism can only play this role in the antirealism debate if it is contrasted with an alternative realist characterization of the target area of speech. Such realist characterizations will typically be couched in representational terms: sentences in the target domain mean what they do because they are used to refer to certain entities in the world. Indeed, expressivists often want to argue that some target domain of discourse works differently than paradigmatically descriptive areas of speech. In contrast to such paradigmatically descriptive speech (e.g. speech about middle-sized dry goods), the characteristic functions of our discourse in the target area would be non-representational. The antirealist intuition behind many expressivist proposals is precisely that talking about chairs and tables is not the same as talking about, say, norms or values. In particular, a realist interpretation of the second, but not the first type of talk would involve problematic metaphysical commitments. So, those sharing these antirealist intuitions will be prone to reject that our speech about norms and values is descriptive in the same way that our speech about chairs and tables is. This distinction between non-descriptive and genuinely descriptive areas of declarative speech is what authors like Huw Price call the bifurcation thesis (Price Reference Price2011, Reference Price, Johnson and Smith2015). Expressivists who embrace the bifurcation thesis are only local anti-representationalists (Blackburn Reference Blackburn2010, Reference Blackburn and Price2013).
There are authors, however, that endorse a global anti-representationalist stance (Brandom Reference Brandom1994, Reference Brandom2000; Horwich Reference Horwich1998; Price Reference Price2011; Williams Reference Williams2010, Reference Williams and Price2013). According to this approach, which I will call pragmatist, the explanation of why an expression means what it means should be made, in all areas of speech, in terms of its use, of its characteristic function. And, crucially, for these authors our ultimate account of the use of the relevant expressions cannot rely on representational notions. It would not be good enough to say that ‘red’ is used to describe things as being red. We would have to further explain what we do when describing something as red, the pragmatic impact or function of such a speech act (for instance, by specifying the norms governing that speech act in our practices).
This sort of pragmatist position can be motivated in several ways. On the one hand, some may be reluctant to accept the existence of word-world representational relations, perhaps on naturalistic grounds. In particular, it may be doubted that there is any specific, informative relation between our words and the world that amounts to a representational relation (beyond the stuttering relation specified by saying, for instance, that ‘water’ refers to water). On the other hand, one can argue that there are internal pressures for anti-representationalist to become global. This is because it seems natural for expressivists to accept minimalism about representational notions such as truth, reference, belief or description. I will understand minimalism as the view that whether the relevant notions apply to some area of discourse is merely a matter of whether expressions in that area of discourse are amenable to take part in certain patterns of use (without further requiring that such expressions engage in specific relations with the world). So, on a possible minimialist view of truth, that a sentence p is truth apt just means that it is assertable and it is appropriate to say things like ‘It is true that p.’ Likewise, according to a minimalist account of belief, believing that murder is wrong would just require that one's moral attitude is properly expressed by asserting ‘Murder is wrong.’
Expressivists have good reasons to adopt representational minimalism: this allows them to account for the superficial similarity between the target area of discourse and paradigmatic descriptive speech (e.g. it will be possible to recognize the truth aptness of normative declarative sentences). The problem of opening the door to minimalism is that we stop having a clear contrast between descriptive and non-descriptive speech. At least, such a contrast cannot be traced in representationalist terms any more, for instance saying that in descriptive domains words have referents, sentences are truth apt and facts are stated. On a minimalist approach, these things will also happen in the domains targeted by the expressivist. After becoming a minimalist, thus, the natural step is to abandon local anti-representationalism and accept global pragmatism (or perhaps endorse some form of quietism). Dreier (Reference Dreier2004) has called this the problem of creeping minimalism (for discussion, see Asay Reference Asay2013; Chrisman Reference Chrisman2008; Golub Reference Golub2017).
In a way, global pragmatism can be seen to bring victory to the expressivist field. From a pragmatist perspective, all meaning is a matter of use, and the variety of functions of language is acknowledged: representation loses its central position in our explanations of what it is for an expression to have a certain meaning (see Price Reference Price2011). Nevertheless, some expressivist may want to preserve a genuine distinction between paradigmatically descriptive speech and other forms of declarative discourse, in particular normative discourse (see for instance Blackburn Reference Blackburn2010, Reference Blackburn and Price2013). Can this demarcation be maintained within a global pragmatist framework? Maybe, but it will have to be reconstructed in non-representational terms. The relevant contrast will be a contrast in the characteristic uses and functions of declarative speech in different domains. In this paper, I explore how this contrast can be drawn. More specifically, I will investigate whether the broadly antirealist intuitions motivating many versions of expressivism can be captured by pointing to certain features of the use of declarative speech in the domains traditionally targeted by expressivists.
In Section 3 I outline a general pragmatist account of assertion. Then, I explore whether we can delineate the desired distinctions within this general pragmatist framework.
3. Global pragmatism
The view I am calling global pragmatism is a generalized use-based theory of meaning. According to this view, the meaning of all expressions is to be explained in terms of their use (of their pragmatic effects), ultimately characterized without appealing to representational notions (Brandom Reference Brandom1994; Price Reference Price2011; Williams Reference Williams2010, Reference Williams and Price2013; for discussion, MacFarlane Reference MacFarlane, Weiss and Wanderer2010). So, semantic features are attributed to expressions in order to codify their conventional use. Moreover, having certain semantic features is just a question, at bottom, of being used in certain ways and having certain characteristic functions. If the mental state involved in accepting a sentence is taken to be a feature of its use, pragmatism becomes a generalization of expressivism.
Note, however, that the connection between semantic features and use does not need to be direct. In particular, semantic values do not need to be identified with characteristic functions, mental states expressed or other features of use (see Charlow Reference Charlow2015). It is only necessary that the relevant features of the use of an expression can be derived, in a context, from its semantic value.Footnote 1 So, pragmatism understood in this way is a metasemantic thesis, connecting the attribution of semantic values to an expression with its use. The picture I have in mind is as follows (for proposals in a similar spirit, see Charlow Reference Charlow2015; Chrisman Reference Chrisman2015; Yalcin Reference Yalcin, Egan and Weatherson2011). Semantics is in the business of ascribing (compositional) semantic values to expressions (e.g. truth values at a context and index). The semantic value of a sentence determines, in a context, what can be called the informational upshot of its utterance, its contribution to the conversation – for instance, what proposition is put forward when asserting the sentence.Footnote 2 Finally, pragmatic theory studies the conversational effects of the use of a sentence, its characteristic function (e.g. the conversational effects of making an assertion with a certain content). Non-representational pragmatism requires that there is a suitable relation between the semantic value of a sentence and its characteristic use (specified in non-representational terms).
In principle, pragmatist metasemantics may be compatible with different semantic frameworks. To be sure, certain semantic theories make the connections with use more perspicuous. For instance, in dynamic semantics, the characteristic conversational effects of utterances of a sentence can be directly read off from the semantic features assigned to the sentence (see Charlow Reference Charlow2015; Starr Reference Starr, Charlow and Chrisman2016; Veltman Reference Veltman1996). However, at least in many cases, the same phenomena can be described by combining a static semantic theory with dynamics at the pragmatic level – that is by assigning static semantic values and adding bridge principles that determine the conversational effects of uttering sentences with such semantic features (Rothschild and Yalcin Reference Rothschild and Yalcin2017).
For my purposes here, I do not need to choose a specific semantic theory. I will only impose the pragmatist metasemantic requirement that the attribution of semantic features is suitably connected with an explanation of the use of sentences. My focus will be on examining the characteristic conversational effects of speech acts performed by uttering the relevant sentences (more specifically, those conventional, public effects that are directly determined by the linguistic rules governing the use of words). My ultimate goal is to explore whether it is possible to specify, without appealing to representational notions, a contrast between the characteristic effects of paradigmatically descriptive speech and of evaluative discourse (and the other types of discourses targeted by expressivists).Footnote 3
Let us begin by considering a simple, Stalnakerian model of the conversational effects of assertions (Stalnaker Reference Stalnaker2014). In this sort of model, the state of the conversation at a certain time is characterized by the common ground of the conversation at that moment, that is by the information mutually presupposed by the participants of the conversation. An assertion would be a proposal to update the common ground by adding a proposition to it. It is customary to characterize the information in the common ground in terms of the worlds compatible with such information (i.e. the possible worlds left open in the conversation). Adding propositions to the common ground would have the effect of ruling out those worlds incompatible with the proposition asserted.
Taking this general model of assertion as a starting point, one may try to develop a non-representationalist characterization of the pragmatics of assertoric discourse, for instance along the lines of Brandom's inferentialism (Reference Brandom1994, Reference Brandom2000). According to Brandom, when a speaker makes an assertion, she becomes committed to the proposition asserted and to the propositions inferable from it (using as potential collateral premises the rest of the propositions she is committed to). Similarly, if the speaker is entitled to undertaking a commitment to some proposition, she also counts as entitled to a commitment to the propositions following from it. Moreover, two propositions count as incompatible when an entitled commitment to one of them automatically precludes being entitled to the other one. The state of the conversation would be specified by a conversational scoreboard that keeps track of the commitments and entitlements acquired by the speakers; the conversational score registers the propositions accepted in the context of the conversation and whether there is entitlement for such an acceptance.Footnote 4 An assertion would be a proposal to update the conversational score: by taking an assertion on board, the audience revises their commitments and entitlements.Footnote 5
It is natural to think of assertoric updates in a Brandomian model as additive: when accepting a proposition, speakers acquire commitments (and, if entitled, also entitlements), which are added to the conversational score. Nevertheless, appealing to the notion of incompatibility, this updating process can also be interpreted in an eliminative sense. The idea would be that the acceptance of a proposition potentially reduces the set of propositions compatible with the speakers’ commitments. This sort of inferentialist view can also be formulated in terms of worlds. We just have to say that a proposition p is associated with the set of p-worlds, the set of worlds compatible with p.Footnote 6 So, if q follows from p, then a commitment to taking the world to be a p-world would bring a commitment to taking it to be a q-world as well. Thus, the inferential role of the proposition p would explain what it is for a world to be a p-world.
In accordance with the pragmatist project, Brandom aims to describe the conversational impact of assertions without appealing to representational notions. In order to do so, he characterizes the dynamics of assertoric speech in terms of our normative practices of assessing each other and of attributing entitlements and commitments – what he calls scorekeeping, following Lewis (Brandom Reference Brandom1994, ch. 3; Lewis Reference Lewis1979).Footnote 7 For instance, if you assert p, and then you assert something incompatible with it, the other participants in the practice (acting as scorekeepers) are entitled to treat you as doing something inappropriate, something you had no entitlement for. In this way, the pragmatic effects of assertions are determined by the social norms governing scorekeeping, which are specified without the involvement of representational concepts.
In principle, Brandom's pragmatist characterization applies to assertoric speech in all domains, including paradigmatic descriptive areas of discourse. Brandom's proposal, thus, is an example of a non-representationalist account of the conversational effects of assertoric speech in general. It is not my purpose here to argue in favor of Brandom's theory. I will just assume for the sake of argument that such a theory, or something in a similar pragmatist spirit, can be made to work. What I want to do is to see whether it is possible to provide a pragmatist reconstruction of the distinction between descriptive and normative speech, once this sort of general non-representational account of assertoric discourse is in play. I will start by consider how normative and evaluative speech can be integrated in the model of assertion I have been sketching.
4. The pragmatics of normative discourse
Think of normative and evaluative sentences like:
(1) One ought to help one's friends
(2) Venice is a beautiful city
A pragmatist can deal with such sentences in a straightforward way. It can simply be said that, when one of these sentences is asserted, the proposition expressed is added to the common ground. In an inferentialist model, this would involve acquiring a commitment to such a proposition and to those inferable from it. For instance, if you are committed to the claim that beautiful cities should be preserved, then accepting (2) commits you to accepting that Venice should be preserved. The pragmatist could even say that, when (2) is added to the common ground, worlds in which Venice is not a beautiful city are ruled out by the presuppositions of the conversation (i.e. the possibility that Venice is not beautiful becomes incompatible with the commitments undertaken by the participants in the conversation).
It should be stressed that the pragmatist's appeals to the notions of proposition, truth or fact do not carry metaphysical weight. The idea of correspondence to the facts does not play a substantive role in the pragmatist's ultimate story about why sentences mean what they mean. Indeed, given the pragmatist's minimalist understanding of the notions of truth and fact, it is a trivial observation that assertable sentences are truth apt and are in the business of making factual claims about the world.Footnote 8 Insofar as one is prepared to assert (2) one should be prepared to assert ‘It is a fact that Venice is a beautiful city’ or ‘It is true that Venice is beautiful.’ Note, in particular, that the notion of proposition appealed to in this pragmatist account would also be thin or deflationary. Whether the utterance of a sentence expresses a proposition would just be a matter of whether the speech act has the characteristic effects of an assertion and is evaluable for truth.
It may be useful for certain purposes to refine the characterization of the conversational state, so that it does not just reflect generally the possible worlds compatible with the common ground. For instance, one can also keep track explicitly of the (moral, aesthetic, gastronomical) preferences or values of the speakers (see Silk Reference Silk2015). Similarly, one can distinguish a normatively neutral specification of possibilities (i.e. a specification in which no information about values, norms or preferences is included) and a classification of such possibilities in accordance with their normative or evaluative features (e.g. whether they are permissible possibilities). Instead of just saying that there is a possible world in which Mary is wrong in not helping her friend Peter, we would first specify a possibility in which Mary does not help her friend Peter, and then we would classify such a possibility as impermissible. So, the state of the conversation would be characterized by means of two separate parameters, the first one tracking a set of open possibilities specified in a normatively neutral way (call them ‘worlds’) and the second one providing a classification of these open possibilities as permissible or impermissible (see Charlow Reference Charlow2015; Starr Reference Starr, Charlow and Chrisman2016).Footnote 9
Once these finer distinctions are introduced, one might be tempted to think that descriptive sentences can be differentiated as those whose acceptance leads to updates that rule out worlds (specified in a normatively neutral way), but do not affect the selection function that classifies worlds according to their permissibility. Non-descriptive sentences would be associated with other types of updates, in particular updates that modify the permissibility selection function (Starr Reference Starr, Charlow and Chrisman2016, 387). Now, introducing these distinctions may be illuminating for some purposes but, by pragmatist lights, it is not enough to capture the purported difference between genuinely descriptive and non-descriptive discourse. It will not do just to label the information registered by the first parameter of the scoreboard as ‘descriptive’ or ‘factual’ and the information in the second parameter as ‘non-descriptive’ or ‘normative’. These distinctions presuppose the relevant bifurcation between descriptive and non-descriptive information, rather than explain it. The pragmatist still needs to show that these labels are not merely arbitrary, but reflect pragmatically significant differences in the conversational effects of using each type of sentence (and, crucially, it should be possible to characterize such differences in non-representational terms, without presupposing the contrast between descriptive and non-descriptive information).
A possible way of reconstructing the bifurcation thesis within an inferentialist, pragmatist framework is by appeal to the Humean intuition that normative thought and discourse is distinctively linked with the motivation and justification of actions. In this way, Chrisman suggest drawing the relevant contrast by focusing on the role played by normative discourse (in particular, ought-claims) in practical reasoning (Chrisman Reference Chrisman2008; for discussion, see Tiefensee Reference Tiefensee2016). The idea is that only normative claims (e.g. ‘I ought to cook dinner’), and not merely descriptive ones, can directly provide entitlement to perform some action (e.g. to cook dinner). Moreover, according to Chrisman's suggested view, normative claims can only be inferred from premises including further normative or practical claims. Thus, from a set of merely descriptive premises, a normative conclusion would not be inferable. This would capture the common idea that normative attitudes are required for motivating and justifying action.
This is the sort of difference pragmatists should be after. What distinguishes normative claims, on Chrisman's (Reference Chrisman2008) view, is their connection with action (and with other normative claims). We do not need to invoke representational notions to characterize this connection. So, this is an appealing proposal for the pragmatist. However, this way of drawing the bifurcation thesis is arguably too limited (as Chrisman Reference Chrisman2015 has come to acknowledge) and potentially problematic (see Tiefensee Reference Tiefensee2016). I think, therefore, that it is worthwhile to explore further ways of characterizing the contrast between descriptive and non-descriptive speech. One first reason for this is that it is controversial whether actions can only be motivated and justified by explicitly normative considerations. Several authors argue that standard descriptive considerations often justify, and also motivate actions (for instance, Alvarez Reference Alvarez2010; Dancy Reference Dancy2000). For example, the fact that a car is approaching may justify and motivate one not to cross the road. Actually, Brandom himself (Reference Brandom2000, 89–92) maintains that the premises of non-enthymematic pieces of practical reasoning may be entirely descriptive, and not involve explicitly normative concepts.
Certainly, many expressivists will share the Humean intuition that merely descriptive attitudes cannot do all the work in motivating and justifying actions (after all, this sort of intuition is often offered as support for expressivism about normative speech). Nevertheless, I take it to be preferable to remain as neutral as possible on this debate. Moreover, as Chrisman himself points out (Reference Chrisman2015, 179–182), it is not clear that the same direct connection with the motivation and justification of actions is observed in relation to all types of ought-claims. Consider, for instance, ought-claims about the past. Arguably, such claims do not play a direct role in motivating the agent's current actions – although they may be appealed to in order to justify past actions. It is also questionable whether a direct connection with motivation and justification of action always takes place in other forms of evaluative discourse (e.g. ‘Her paintings are very elegant’). At any rate, it seems that the link with action is far less prominent in other areas of discourse that many expressivist will want to situate in the non-descriptive side of the divide, for instance, speech involving epistemic modals or indicative conditionals (Chrisman Reference Chrisman2015, ch. 7).
In what follows I examine other ways of demarcating the division between genuinely descriptive discourse and other forms of assertoric speech. I start by discussing whether this division can be drawn by appeal to the distinctive impact of modal claims on the conversational state.
5. Test dynamics
Normative speech often involves modal operators, such as ‘ought’ and ‘may’. Indeed, modal operators seem to play a central role in many of the areas of discourse targeted by expressivists (e.g. epistemic modals). It may be argued that modal sentences behave in a characteristic way, which distinguishes them from basic declarative sentence. Roughly, the intuition is that the correctness of modal claims does not depend (or not only) on how the actual world is, but on features of certain possible situations. So, ‘You may visit the museum for free’ is true if some permissible world is such that you visit the museum for free (even if in the actual world perhaps you decide to go to the cinema instead).
Is it possible to appeal to the distinctive behavior of modal operators in order to develop a pragmatist-friendly bifurcation thesis? It seems that one could try to contrast genuinely descriptive sentences with declarative modal sentences, which do not provide information about the way the actual world is, but do something else instead (see Charlow Reference Charlow2015, 18–19; Chrisman Reference Chrisman2015; Veltman Reference Veltman1996). In this way, Chrisman (Reference Chrisman2015) argues that modal operators are associated with ‘metaconceptual operations’ in which more basic contents are manipulated. Following Sellars and Brandom, Chrisman fleshes out this idea from an inferentialist perspective, suggesting that sentences involving modal operators have as their characteristic function to allow us to make inferential connections explicit: on Chrisman's proposal, modal sentences get their ‘content from being usable to acknowledge inferential connections between more basic items rather than to refer to things in the world’ (Reference Chrisman2015, 197).
The distinctive behavior of modal sentences can be characterized in a more precise way within the sort of model of assertion sketched above. According to simple accounts of assertion, the assertion of standard declarative sentences introduces information in the conversation, by adding a proposition to the common ground (or, equivalently, by ruling out worlds). Several authors argue that modal sentences are associated with a different type of update instruction (Charlow Reference Charlow2015; Starr Reference Starr, Charlow and Chrisman2016; Veltman Reference Veltman1996). Assertions of such sentences would not introduce new information (or a least not only), but rather impose a test on the conversational state. Performing a test amounts to checking whether the conversational state has certain features. If the test is passed, the conversational state remains as it was (it is not updated). Otherwise, we get a defective conversational state. A defective conversational state calls for modification, if the conversation is to proceed. However, the modal sentence does not specify which particular modification has to be implemented in order to overcome the defectiveness of the conversational state.
Take as an example ‘Mary ought to help Peter’. On the view under consideration, accepting this sentence involves checking whether all the open worlds that are classified as permissible in the conversational state are worlds in which Mary helps Peter. If this is the case, the test is passed. Likewise, ‘The keys must be in the kitchen’ tests that all open worlds in the conversational state are such that the keys are in the kitchen.
The crucial distinction here is between enforcing a property on the conversational state and testing whether the state has such a property (Charlow Reference Charlow2015, 36–37). For instance, when ‘The keys are in the kitchen’ is accepted, all worlds in which the keys are not in the kitchen are ruled out from the common ground. Thus, the property that all open worlds are such that the keys are in the kitchen is enforced on the conversational state: the conversational state is directly modified in a way that ensures that it has that property. By contrast, an assertion of ‘The keys must be in the kitchen’ leads to checking whether the conversational state has the relevant property (i.e. being such that in all open worlds the keys are in the kitchen). So, in tests the property in question is not directly enforced, but checked.
A natural reaction to a failed test is to accommodate, that is to modify the conversational state in a way that makes it pass the test. In the example of ‘The keys must be in the kitchen’, this will typically involve eliminating those open worlds in which the keys are not in the kitchen. Thus, the conversational state ultimately derived from the acceptance of the modal sentence may be the same as the state directly resulting from updating on the sentence ‘The keys are in the kitchen’. It must be stressed, however, that the path to this final state is different in each case. The updating rule associated with the non-modal sentence directly instructs one to eliminate the worlds incompatible with the keys being in the kitchen. In other words, the semantic value of the sentence determines that the conversational state has to be updated in this way. By contrast, the semantic information of the modal sentence only specifies what test is to be performed, but not how the conversational state is to be modified after the failure of the test. When a test fails, the subsequent revision of the conversational state proceeds via accommodation: speakers modify the conversational state so that the sentence uttered can be accepted without rendering such conversational state defective (or else, they reject the sentence uttered). This process of accommodation is not directly guided by the semantic features of the sentence uttered, but by broader conversational rules concerning rational attitude-revision.
So, as long as there is a pragmatically significant difference between changes to the conversational state introduced via updating instructions encoded semantically and changes introduced via accommodation, the pragmatic profile of modal sentences will be clearly distinguishable from that of other declarative sentences (Charlow Reference Charlow2015, 36–38). It seems, therefore, that by embracing test semantics for modals the pragmatist is in a position to differentiate the characteristic uses of paradigmatically descriptive discourse and discourse involving modal operators (in particular, speech about obligations and permissions).
Can the contrast between descriptive and non-descriptive speech be recast in terms of this distinction between tests and eliminative updates? I think that there are reasons to remain cautious about the scope of this proposal. The problem is that not all instances of declarative speech typically taken to be non-descriptive involve modal operators. More specifically, normative and evaluative speech does not always include modal, intensional operators. There are many evaluative sentences in which the relevant evaluative expressions seem to constitute ordinary first-order predicates, rather than expressing higher-order operators (this is acknowledged in Chrisman Reference Chrisman2015, ch. 7). Arguably, this is the case with ‘cruel’ in ‘Mary is very cruel’, ‘tasty’ in ‘Artichokes are tasty’, or ‘beautiful’ in ‘Venice is a beautiful city.’ Treating these sentences as involving higher-order operators seems to be unmotivated and ad hoc. Therefore, appealing to the distinctive conversational effects of modals would not allow us to differentiate these evaluative sentences from other declarative sentences involving ordinary first-order predicates. A possible option is to argue that only normative sentences that include modal operators (e.g. ‘ought’) should actually be seen as contrasting with paradigmatic descriptive speech. Evaluative discourse about goodness, tastiness or beauty would be, according to this view, on the descriptivist side of the divide. While this is a possibility, it seems that it will not be appealing to those with expressivist inclinations – after all, ‘beauty’ and ‘good’ are among the standard targets of expressivist proposals.
The distinction between tests and eliminative updates, therefore, is not enough to draw a general division between paradigmatic descriptive speech and other forms of assertoric discourse, including evaluative claims with no modal operators. In the next section, I explore the possibility of reconstructing the bifurcation thesis in terms of the contrast between relativistic and non-relativistic assertoric practices. This way of reformulating the bifurcation thesis allows one both to deal with first order evaluative predicates and to capture the anti-realist intuitions behind expressivism.
6. Relativism
According to the simple model of assertoric speech sketched above, assertions are proposals to add a proposition to the conversational common ground. From this point of view, we can say that two assertions are incompatible when, for any possible non-defective conversational state, the simultaneous addition to the common ground of the two propositions asserted makes the conversational state defective. In this way, when two speakers make incompatible assertions they would be proposing that the common ground is updated in incompatible ways. The two updating proposals cannot be both accepted in relation to the same common ground, on pains on getting a defective conversational state.
In standard assertoric practices, only one of two incompatible assertions can be correctly accepted in relation to a given conversational state – that is, only one of two incompatible propositions can be correctly added to the common ground. In Brandom's terminology, when two propositions are incompatible speakers cannot be actually entitled to accept both of them. This idea can be expressed in terms of a truth correctness-standard governing assertoric practices: asserting p is correct only if p is true (Kölbel Reference Kölbel, Garcia-Carpintero and Kölbel2008, 10). In general, out of two incompatible propositions only one can be true, so only one can be correctly asserted.
Let us make the plausible assumption that correctness-standards have normative force, so that an incorrect performance is impermissible or inappropriate (Whiting Reference Whiting2009). If this is so, then at least one of two speakers asserting incompatible things is doing something impermissible. This is why assertoric disagreement generates a normative tension (see Price Reference Price2003). When one makes an assertion that is incompatible with a previous assertion by another speaker, one is challenging that speaker's entitlement to make her assertion (Brandom Reference Brandom1994). So, disagreeing assertions lead to an unstable conversational state – there are normative pressures to revise such a state in order to move back to a situation of agreement.
The resulting picture is that of assertion as a practice governed by norms, in particular a norm that prohibits asserting something false. Arguably, this basic norm gives rise to the requirement to retract an assertion that is false, given that asserting falsities is impermissible (MacFarlane Reference MacFarlane2014, ch. 5; Price Reference Price2003; Whiting Reference Whiting2013). Furthermore, in principle it will be appropriate to challenge, criticize or reject a false assertion – more broadly, it will be appropriate to treat asserting something false as impermissible (Brandom Reference Brandom1994; MacFarlane Reference MacFarlane2005, Reference MacFarlane2007).Footnote 10 These sorts of normative constraints (and perhaps others) characterize the speech act of assertion.
Assuming this general framework for assertion, we can now describe specifically the properties of relativistic assertoric speech. I will follow here the main ideas of MacFarlane's account of relativism (Reference MacFarlane2005, Reference MacFarlane2007, Reference MacFarlane2014).Footnote 11 In accordance with pragmatist metasemantics, MacFarlane thinks that the attribution of relativist semantic features to a sentence is only vindicated if it is suitably connected with relevant aspects of the use of that sentence (MacFarlane Reference MacFarlane2007, Reference MacFarlane2014, ch. 5; an alternative pragmatist rendering of relativism can be found in Shapiro Reference Shapiro2014). Footnote 12 Thus, we need to specify how relativistic assertoric practices work.
The basic norm of standard assertion applies as well in relativistic discourse: it is impermissible to assert something false. The main difference in relativistic assertion is that the truth values of propositions are not perspective-independent; therefore, the correctness or permissibility of an assertion is not perspective-independent either. In non-relativistic domains, if a proposition is actually false as evaluated from a certain perspective, it will also be false from any other perspective or context of evaluation. So, a false assertion will be properly assessed as impermissible from any perspective in the practice. By contrast, in relativistic discourse, an assertion that is actually true as evaluated from a certain perspective can be actually false in relation to a further perspective of evaluation. As a result, there will be no perspective-independent answer to the question of whether an assertion is false, and thereby impermissible. Assume, for instance, that discourse about personal taste is relativistic. Then, the proposition ‘Artichokes are tasty’ can be true in relation to the perspective of evaluation of someone that likes artichokes, but false in relation to the perspective of someone who does not like them.
Appealing to the notion of truth in relation to a context of assessment, the basic norm of assertion can be reformulated as follows: asserting p is permissible, as assessed from context c’, only if p is true in relation to the context of assessment c’ (and the context of utterance c).Footnote 13
In a similar way, we can formulate norms for relativistic retraction (MacFarlane Reference MacFarlane2014, ch. 5) and challenges or rejections (MacFarlane Reference MacFarlane2007). The idea is that a speaker in a context c’ ought to retract an assertion that is false in relation to that context of assessment c’. Likewise, a speaker in a context c’ may permissibly challenge or reject an assertion that is false as assessed from context c’. These norms are a generalization of the norms for retraction and challenges in non-relativistic domains.Footnote 14
What we find in relativistic domains is that there does not need to be a perspective-independent way of determining which of two disagreeing agents is right. It may happen that, from the perspective of a first agent, it is actually permissible to make an assertion that, from the perspective of a second agent, is actually permissible to reject or challenge.Footnote 15 So, in relativistic domains it can be the case that, from the perspective of some agent, it is permissible to add a certain proposition to the common ground, whereas it is impermissible to do so from the perspective of other participants in the conversation. These sorts of disputes in relativistic domains can only be settled by adopting some evaluative perspective: there is no neutral way of solving relativistic disagreements.
6.1. Relativism and expressivism
My proposal is that, in general, assertions in the areas of discourse traditionally targeted by expressivists behave in a relativistic way. Thus, in the target domains agents still put forward their views in the form of assertions (and often express their beliefs in doing so), but these assertions are relativistic. This association between relativist assertion and the areas of discourse addressed by expressivists opens a clear path for a pragmatist reconstruction of the bifurcation thesis. The distinction between non-relativist (i.e. absolutist) and relativist assertoric practices allows the pragmatist to recover the contrast between declarative sentences used in a genuinely descriptive way and declarative sentences receiving non-descriptive uses (for instance, in normative discourse). This proposal has at least some initial plausibility, given that most areas of discourse targeted by expressivists have also received a relativistic treatment (see MacFarlane Reference MacFarlane2014 for a detailed review). In particular, it should be noted that modal speech – which, as discussed in Section 4, has been taken to be a source of non-descriptive assertions – lends itself naturally to a relativistic analysis (MacFarlane Reference MacFarlane2014, ch. 10–11).
Indeed, relativistic assertion captures to a large extent the anti-realist, non-descriptivist motivation behind most expressivist proposals. As we have seen above, relativistic disputes cannot be adjudicated in a neutral way, but only from inside one of the perspectives potentially engaged in the debate – there is no external viewpoint from which it can be determined what side of the dispute is right. So, if normative discourse worked relativistically, then one could only give a verdict about some first-order normative controversy from the perspective of some endorsed system of norms of values.Footnote 16 There is a sense, therefore, in which relativistic disputes are not about how the world is, but rather about what evaluative perspective to adopt (e.g. what values to endorse). It is in this sense that assertoric discourse in the domains targeted by expressivists can be said to be ‘non-factual’ or ‘non-descriptive’. Of course, in keeping with the minimalism favored by pragmatists, we will still be in a position to talk about (monadic) truth, descriptions and facts in relation to such domains, but these will be relativistic facts: what the facts actually are will depend on the perspective of evaluation. Thus, the debate between realist and anti-realist accounts of a given domain of discourse would become a debate about whether assertions in such a domain are governed by relativistic norms (rather than about whether fact-talk is warranted in that domain).
One may wonder what is the point of relativistic speech. Why would we use assertoric speech in areas of discourse where our primary aims are not representational or descriptive? A plausible answer is that we do so in order to take advantage of the normative mechanisms afforded by assertoric practices. Assertoric challenges tend to impose normative pressures on the addressee, insofar as they express disapproval (the agent making the challenge presents herself as taking the addressee to be doing something impermissible). In this way, assertoric speech is a systematic and effective way of creating normative frictions that promote the coordination and alignment of our attitudes (see MacFarlane Reference MacFarlane2007, Reference MacFarlane2014). As Macfarlane (Reference MacFarlane2007, 30) puts it: ‘Perhaps, then, the point of using controversy-inducing assessment-sensitive vocabulary is to foster coordination of contexts.’
The idea, therefore, is that in domains where there is a point in attuning our attitudes, it will make sense to resort to assertoric speech. In this way, it is not so much that we can use assertions in areas of discourse that aim to describe the world, but rather that when we want to introduce the normative friction associated with assertoric disagreement, we will tend to make use of assertion (and thereby, the relevant area of discourse will automatically count as descriptive in a minimalist way). So, as Price argues, it is the application of the norms of assertoric practice ‘which creates the disagreement, where initially there was mere difference’ (Price Reference Price2003, 17). Of course, speakers are not always obliged to become subject to the normative pressures of assertoric speech. It is often possible to elude assertoric disagreements by retreating to reports of one's attitudes or claims that are explicitly relativized to perspectives (e.g. ‘I like artichokes’, ‘Artichokes are tasty for me’).Footnote 17
One of the main virtues of the relativistic approach I am sketching is its theoretical parsimony: it does not require a radically different account of assertion and the semantics of declarative sentences specifically tailored for the domains of discourse targeted by expressivists. Speech in these domains would co-opt the normative mechanisms of regular assertion, so we can retain much of the standard analysis of assertoric discourse (rather than having to appeal to additional normative mechanisms, such as disagreement in non-doxastic attitudes or metalinguistic negotiation). In particular, we can keep understanding assertoric speech acts in these domains as proposals to add propositions to the common ground. For instance, by asserting ‘It is wrong to steal books from the library’, I propose to rule out from the common ground those worlds in which it is permissible for you to steal books from the library (or, equivalently, I propose to rule out from the permissibility set those worlds in which you steal books from the library). In this way, the attitude directly expressed in typical instances of relativistic assertions would just be a (relativistic) belief. Perhaps the ultimate goal of the speech act is to express some further attitude, such us the endorsement of some system of norms and values, but this would be done indirectly, rather than as part of the characteristic function of the speech act. Presumably, by presenting yourself as believing that the world is such that it is impermissible for you to steal, you present yourself as endorsing a system of norms that make that action impermissible. So, relativistic assertion may be an indirect way of expressing non-doxastic attitudes, but this would be done just by resorting to the standard mechanisms of assertoric practices, without having to introduce further linguistic machinery.Footnote 18 Expressivist models, therefore, can be reformulated in a simple way within a relativistic framework. In other words, relativism can be seen as an implementation of the expressivist project without high theoretical costs.Footnote 19 By contrast, other proposals would require more radically revisionary accounts of our linguistic practices – for instance, treating sentences with first-order evaluative predicates as actually involving modal operators (as discussed in Section 5).
6.2. Identifying relativistic practices
Of course, even if relativist assertion is governed by the same types of norms as standard assertoric speech, the point I am trying to make is precisely that there is a significant practical difference between both forms of assertoric discourse, and that this difference may ground a pragmatist reconstruction of the bifurcation thesis. The relevant practical difference is that, as explained above, in relativistic assertion disputes cannot be settled in a perspective-independent way. So, an external observer studying a relativistic practice would find disagreements in which there seems to be no potential fact or piece of evidence that would adjudicate the dispute in a manner acceptable to all (rational) debaters.
To be sure, it is not clear that one can assess a relativistic dispute from a genuinely external point of view. Arguably, evaluating the merits of some position in a relativistic debate can only be done by adopting oneself a particular assessing perspective. And from the inside of a practice it is difficult to ascertain whether such a practice is relativistic or rather absolutist (Cappelen and Hawthorne Reference Cappelen and Hawthorne2011, 460–461; MacFarlane Reference MacFarlane2014, 199–200; Shapiro Reference Shapiro2014, 145). After all, participants in a relativistic practice will assess some assertions as correct because, from their perspective, they actually seem to be so. This is this same that happens in a non-relativistic, absolutist practice: agents treat as correct what seems correct from their perspective. The subjective experience of agents, then, will be very similar when participating in a relativistic and a non-relativistic practice. However, this does not need to be problematic for my proposed reconstruction of the bifurcation thesis; indeed, it is as it should be, in view of the entrenched, recalcitrant metaethical debates about realism. A relativistic approach allows us to do justice both to anti-realist and absolutist intuitions – and to explain why the limits between descriptive and non-descriptive discourse are often unclear.
All this does not mean that we cannot decide, on reflection, that relativism offers the best model for characterizing an assertoric practice we are part of. This will happen, for instance, if we reach the conclusion that the sorts of debates participants in the practice engage in are unlikely to find a perspective-independent resolution, and not due to some epistemological limitation, but rather because there is no potential ‘smoking gun’ – that is, there is no potential decisive piece of evidence that would be recognized as such by all rational participants in the debate. In those areas of discourse less prone to a relativistic analysis, there tend to be possible tests and sources of evidence whose force and significance is acknowledged by all relevant agents. So, if we are discussing the location of your keys, all parties will presuppose that direct perception of the keys on the kitchen table would provide (at least in normal circumstances) a definitive answer to the issue discussed. In this way, in order to determine whether some assertoric practice is relativistic, we can investigate whether it is suitably related to some such baseline, agreed-upon standards of correctness. Non-relativistic assertoric speech will in general be inferentially connected with practices in which there are sources of evidence that are presupposed to provide perspective-independent verdicts, such as empirical observation (see Chrisman Reference Chrisman and Brady2011). Depending on what types of sources of evidence we are willing to count as playing this arbitrage role, we will include more or less areas of discourse in one side or the other of the divide established by the bifurcation thesis.
7. Conclusions
Traditional expressivist projects highlight the distinction between domains of discourse where declarative sentences are used to describe the world, and domains where they are used in other ways. This distinction becomes blurred when one adopts a minimalist perspective about representational notions, according to which all instances of assertoric speech automatically count as purporting to describe how things are. My proposal has been to reconstruct the distinction between descriptive and non-descriptive assertoric speech by appeal to the contrast between relativistic and non-relativistic assertion. This contrast can be specified in non-representational terms, so it offers a way of reconstructing the expressivist distinction that is available for pragmatists embracing minimalism about representational notions. Moreover, this way of reconstructing the relevant distinction works both for sentences involving first-order evaluative predicates (such as ‘good’ or ‘beautiful’) and also for higher order operators (such as ‘ought’). In general, the areas of discourse where expressivism has some initial plausibility are also amenable to a relativistic analysis.
Funding
This work was supported by Ministerio de Economía y Competitividad [grant number FFI2014-57258-P].
Disclosure statement
No potential conflict of interest was reported by the author.
Acknowledgements
Thanks to an anonymous referee for this journal and audiences at the Granada Workshop on Epistemic Expressivism and the 2017 EPISOC summer school for their feedback.
