Prepragmatics Widening the Semantics/Pragmatics Boundary

One of the most important and, at the same time, most controversial issues in metasemantics is the question of what semantics is, and what distinguishes semantic elements (features, properties, phenomena, mechanisms, processes, or whatever) from the rest. The issue is tightly linked with the debate over the semantics-pragmatics distinction, which has been vibrant for a decade or two, but seems to be reaching an impasse. I suggest that this impasse may be due to the failure to recognize a distinct realm that should be subsumed neither under semantics nor pragmatics, but may be labeled "prepragmatics". My ultimate goal is to put forward and defend a novel picture of our language architecture, according to which: semantic content is strictly poorer than the lexically encoded content (and therefore does not involve any contextually determined material – not even the reference of demonstratives); pragmatics, as widely held, does not reach into truth-conditions and does not affect truthvalue, while its mechanisms require the capacity of reasoning about one's beliefs and intentions; and, fnally, there is a distinct prepragmatic level at which sentences and/or utterances get evaluated for their truth value (but also for other properties, such as their modal status or assertoric content), and which takes into account various kinds of contextual information. In the frst half of the paper, I turn to a topic that has long been of interest to philosophers of language and has been one of the "stumbling stones" in the discussions of the semantics/pragmatics distinction, namely demonstratives. In the mainstream, direct-referentialist view, the semantic contribution of a (demonstrative or indexical) pronoun is some contextually determined object or individual. I argue that the mainstream view has diffculties in maintaining such an approach to the semantics of demonstratives and, at the same time, Prepragmatics – April 2013 1 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 drawing a principled line of demarcation between semantics and pragmatics. In the second half of the paper, I put forward my own proposal, which, in a nutshell, views semantic mechanisms as being intimately linked with stable lexical meaning and takes pragmatic processes to require a full-fedged capacity to reason about the speaker's mental states and to deploy general principles akin to Gricean maxims. While the resolution of reference may, in certain cases, require the latter, it normally does not. The upshot of the paper is to show that if there is indeed room for a family of linguistic phenomena that are neither semantic nor yet fully pragmatic, then the resolution of demonstrative reference is a candidate par excellence to belong there. 1. Semantics vs. pragmatics: some preliminaries One of the central tasks for metasemantics is to characterize what semantics is. The more specifc problem of where to draw the line between semantics and pragmatics has received considerable attention among philosophers of language in the past decade.1 The source of the problem is that there seem to be several equally plausible criteria for drawing the distinction that converge in many cases, but not in all. Those cases in which the different criteria fail to converge have been of greatest interest to the different parties in the debate, but equally well to those linguists who are seen as working at the semanticspragmatics interface. It is believed that the frst attempt of formally distinguishing semantics from pragmatics goes back to Morris (1936), who took the former to be the study of “the relation of signs to objects which they denote and whose properties they truly state” and the latter, the study of “language as a type of communicative activity, social in origin and nature, by which members of a 1 A number of collections of articles specifcally on this issue may be mentioned: Turner (ed.) 1999, Bianchi (ed.) 2005, Szabó (ed.) 2006, Stojanovic (ed.) 2008, Ezcurdia and Stainton (eds.) 2011, to mention only a few. Prepragmatics – April 2013 2 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 social groups are able to meet more satisfactorily their individual and common needs” (p. 10).2 Since then, this broad distinction between semantics and pragmatics has been seconded by more refned ones, and from the literature, there seem to emerge four criteria that may be roughly formulated along the following lines.3 (i) The semantic stuff is lexically encoded in the linguistic expressions themselves; the pragmatic stuff need not be (and typically is not) lexically encoded. (ii) The pragmatic stuff depends on various contextual factors; the semantic stuff remains stable from one context to another. (iii) The semantic stuff determines the truth conditions; the pragmatic stuff is truth-conditionally inert. (iv) The semantic stuff obeys the principle of compositionality, closely mirroring syntactic structure; the pragmatic stuff need not be compositional. To get a better understanding of the motivations behind the four criteria, suppose that in reference to Aisha, I tell you: (1) She is obnoxious. Suppose, however, that Aisha's behavior makes it obvious that she is obnoxious, so that by telling you that she is obnoxious, I am not telling you something informative. Presumably, then, the reason for telling you (1) is not to inform you that Aisha is obnoxious. Rather, I may be saying (1), for instance, in order to convey something along the lines of: (2) I suggest that we avoid Aisha for the rest of the evening. What I would thus convey with (1) – which is what I would have expressed 2 I am borrowing the quotation from McNally (forthcoming), p. 3. 3 I shall formulate the criteria as distinguishing semantic "stuff" from pragmatic "stuff": the reason for choosing such a jargon term is that at this stage, I would like to stay neutral on what it is precisely that the distinction bears upon, and in particular, whether it is abstract entities such as elements, features or properties, or rather, more concrete entities such as interpretation mechanisms and processes. I shall return to this question at the beginning of sect. 3. Prepragmatics – April 2013 3 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 had I uttered the sentence in (2) – is uncontroversially something that is only pragmatically associated with my utterance of (1). And indeed, it falls on the pragmatic side according to all four criteria: (i) the suggestion that we avoid Aisha for the rest of the evening is not lexically encoded in the meaning of the sentence uttered in (1); (ii) in order to convey that suggestion, I must rely on various contextual factors; my interlocutor must reason about what my intentions were in uttering (1), etc.; (iii) the suggestion has no bearing either on the truth conditions or on the truth value of (1): (1) is true if Aisha is obnoxious and false if she isn't;4 and this is so regardless of how we feel about the question of whether we'd better avoid her for the rest of the evening; (iv) the suggestion does not enter the compositional derivations that one can perform on the sentence in (1); for instance, “It's not the case that she is obnoxious” need not convey that I do not suggest that we avoid her for the rest of the evening; "Everyone thinks that she is obnoxious" does not convey that everyone thinks that I suggest that we avoid her; etc. So far so good: implicatures fall out as uncontroversially pragmatic. But now, what would be uncontroversially semantic in our example? Let t be the time at which (1) is uttered. One might think (as Grice himself did) that the proposition that Aisha is obnoxious at t is what semantics delivers; the semantic content, or 'what is said'. After all, the implicature that we'd better avoid Aisha is presumably derived from the proposition that she is obnoxious, together with the general assumption that obnoxious people had better be avoided. Indeed, the mainstream view, frmly established in semantics and philosophy of language since David Kaplan's Demonstratives, holds that the pronoun 'she' in (1) semantically contributes a contextully determined referent 4 I am ignoring here the fact that "obnoxious" may be an evaluative predicate, hence that the truth value of (1) may depend not only on whether Aisha is obnoxious simpliciter, but also on from whose point of view her obnoxiousness is being judged. Prepragmatics – April 2013 4 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 (in our case, Aisha herfself) to the content of (1). The view further holds that 'obnoxious' contributes the property of being obnoxious, that the present tense in the copula 'is' contributes t (i.e. the time at which (1) is uttered) and that the copula itself contributes predicate application. Putting all this together, the proposition that Aisha is obnoxious at t is what, according to the mainstream view, falls out as the output of the semantic machinery. My main goal in the next section will be to argue that the situation is more complex than this. 2. Semantics or pragmatics? The challenge from demonstratives Recall that the mainstream view holds that the semantic content associated with (1) (i.e. with the sentence "She is obnoxious" uttered in reference to Aisha at time t) is the proposition that Aisha is obnoxious at t. I shall now show that this proposition falls on the semantic rather than pragmatic side with respect to only two of the four criteria laid out in the previous section. The case at point is demonstrative reference and, at the linguistic level, the third person pronoun 'she'.5 What is at issue, then, is whether Aisha, qua the person about whom the speaker is talking in (1), pertains to the semantic or, rather, to the pragmatic level associated with (1). According to criterion (i), taken at face value, the person to whom I am referring with the help of the 3rd person pronoun does not belong to semantics but rather, to pragmatics. This is because the lexical meaning of 'she' does not encode the information that the word should stand precisely for Aisha. If it did, then every time I used the pronoun 'she' I would be talking of Aisha, 5 It should be noted that temporal reference and, at the linguistic level, the present tense are just as problematic as demonstrative reference, and that issues that have to do with temporal reference and with the contribution of tenses and, more generally, temporal expressions to the semantic content are very complex in their own right. For the sake of simplicity, let me set those aside for the rest of the paper, and focus on personal pronouns. Prepragmatics – April 2013 5 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 which is absurd. Now, one could tamper with the idea of lexical or linguistic encoding. It is typically assumed that the mere fact that there is a word, 'she', that appears to stand for Aisha and that the lexical meaning of this word "invites" the interpreter to search for a (female) referent would be suffcient to render Aisha "linguistically encoded" in the sentence in (1). Be this as it may, what remains uncontroversial is that Aisha herself is not part of the lexical meaning of 'she'. Moreover, the dependence of (1)'s truth value on Aisha is not to be dealt with at the level of semantics according to criterion (ii) either. One clearly needs context in order to select Aisha, rather than some other female, as the person relevant to the truth of (1). So, on criterion (ii), Aisha is, at best, pragmatically associated with my utterance of (1). The mainstream view's strategy for rescuing the semantic status of the reference of demonstratives is to give up criteria (i) and (ii) taken at face value, and reintroduce suitable variants of those. I have already pointed out how one can reinterpret the idea of lexical or linguistic encoding in such a way that Aisha ends up being "encoded" in (1). As for criterion (ii), one might want to allow for "semantic contextuality"; that is, for the possibility of appealing to the context in the course of semantic interpretation. Indeed, indexicals are often taken to be those expressions that by defnition contribute contextually determined referents to semantic content. Regardless of whether this rescue strategy can work for so-called "pure" indexicals such as the frst person pronoun 'I', what I would like to point out is that the strategy leads to tensions when applied to 3rd person pronouns such as 'she', or to demonstratives such as 'this' and 'that'. The main problem, in a nutshell, is that by allowing such contextually determined items to enter the level of semantic content, the view will fnd it diffcult to prevent many other types of contextual information from doing so, to the point that it will end up allowing into semantic content a lot of stuff that is generally believed to belong Prepragmatics – April 2013 6 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 to the level of pragmatics.6 Let me try to make the point on an example. Consider a situation in which Byeong is doing restoration works in a house, and is both the one painting all the rooms and rewiring electricity in all of the rooms. Now suppose that I say: (3) Byeong hasn't fnished the guestroom yet. In such a situation, whether (3) is true or false is not yet determined by how things are: said in certain contexts, (3) will be true, yet in other contexts, false. For, suppose that what is at issue in the conversation is how far the painting work has advanced, and that he has indeed fnished painting the guestroom (and even the whole house). Then (3) is false. On the other hand, if what is at issue is how far all of the restoration work has gone, then given that he hasn't fnished rewiring the electricity in the guestroom yet, (3) is true. This shows that even if we hold fxed the world, the time of utterance, and the reference of 'the guestroom', we still need more context in order to evaluate (3) for a truth value.7 The case of (3) raises the following dilemma for the mainstream view: (Option I) Maintain a boundary between the sort of contextual information that may be appealed to within semantics and the sort of contextual information to which only pragmatics has access in such a way that Aisha, qua the referent of 'she', belongs to the semantic content of (1), but the information that Byeong 6 It has been argued, e.g. in Cappelen and Lepore (2005), that once we start letting context into semantics, we get onto a slippery slope: there is no way to prevent letting more and more context in. However, their argument assumes that the slippery slope only begins after demonstratives: maybe with quantifer domains, maybe with gradable adjectives, maybe with the location argument involved in predicates such as 'rain'. Somewhat ironically, the same argument can be turned against their own view (to the extent that they are endorsing the Kaplanian picture): as soon as one lets context in, as in the case of demonstrative reference, one will have already stepped onto a slippery slope. 7 The danger of arguing by way of example is that there is always a risk that one may fnd the example at stake wanting. What is more, examples of so-called coercion, as in (3), have been discussed in the linguistic literature, although this has had relatively little impact on the philosophical literature. Prepragmatics – April 2013 7 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 hasn't fnished either painting or rewiring the electricity does not belong to the semantic content of (3). (Option II) Bite the bullet and accept that in one context, the semantic content of (3) is the proposition that Byeong hasn't fnished painting the guestroom, in another context, it is the proposition that he hasn't fnished rewiring the electricity in the guestroom, in yet another context, it is that he hasn't fnished either painting or rewiring the electricity, and so on. Now, option (II) is a viable option: it is precisely the option endorsed by socalled radical contextualists (e.g. Travis 1985, Recanati 2004) who argue that pragmatics freely "intrudes" into semantics. It also appears to be the option favored by those linguists who have discussed the phenomenon of coercion (e.g. Pustejovsky 1995, Egg 2003).8 Note, however, that this option departs signifcantly from the mainstream view, which takes the effects of the context on the semantic content to be much more limited and constrained and tries to preserve a semantics-pragmatics boundary that does not tolerate pragmatic intrusion. What I would like to argue now is that one who goes for option (I) must do so at the cost of postulating an arbitrary divide between contextual effects that may be dealt with within semantics vs. those that must remain confned to the realm of pragmatics. Option (I) presupposes that we can distinguish the way in which 'she' is context-dependent from the way in which "fnish the 8 There are important differences among the proposals made in the linguistic literature. Caricaturing to some extent, we may see Pustejovsky (1995)'s proposal as one that keeps the semantic contribution of the verb intact and locates the contextual variations in the contributions of the noun-phrase (e.g. in one context, "the guestroom" would be enriched into "painting the guestroom"; in another, into "painting or rewiring the guestroom"), while Egg (2003)'s proposal takes the meaning of the verb to be semantically underspecifed and thus locates the contextual enrichment in the verb itself. Prepragmatics – April 2013 8 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 guestroom" is (or, if you prefer, isn't) context-dependent. To see if we can distinguish them indeed, let us frst see how our four criteria for the semantics-pragmatics distinction settle the question of whether the action that hasn't been fnished, such as painting and/or rewiring electricity, belongs to the semantic content of (3). So let's frst ask whether there is a linguistic element in the sentence uttered that invites the action (e.g. painting vs. painting and rewiring electricity) into the semantic content. Well, why not: presumably the verb 'fnish' does, given that its meaning requires that there be some action (or process, as the case may be) that 'fnish' takes as its argument. Since no such action or process is contributed by any other linguistic item in the sentence, one might presume the verb to be the one that triggers a contextual search for a suitable action or process to take as its argument. Let's further ask whether the actions of painting and rewiring the electricity are anyhow "encoded" in the meaning of 'fnish'? Here, the answer will depend on what one takes linguistic "encoding" to amount to. I submit that if one is happy with the idea that in the case of (1), the way in which its truth value depends on Aisha is "linguistically encoded" in the sentence (viz. in the meaning of the pronoun 'she'), then one should be equally happy with the idea that if (3) is uttered in the painting context, then the action of painting is also "linguistically encoded" in the meaning of 'fnish', and that if (3) is uttered, say, in the context of cleaning the whole house, then another action, such as cleaning, would be similarly "linguistically encoded" in the meaning of 'fnish'. After all, just as the lexical meaning of 'she' merely constrains the referent to being a female, without encoding any additional Prepragmatics – April 2013 9 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 conditions that would make it possible to determine who the referent is (in a given context), the lexical meaning of 'fnish' can similarly be seen as constraining the event referred to with the verb phrase to being an event of a certain type; namely, one in which the agent completes or brings an end some process or action. Thus, in our example, both the event of Byeong's fnishing painting the guestroom and the event of Byeong's fnishing painting and rewiring the electricity in the guestroom comply with the constraint encoded in the lexical meaning of 'fnish'. In sum, if a view allows for "semantic contextuality" and assumes that for a contextual element to be admissible into semantics, all that it takes is that there be some suitable sentential constituent that "linguistically encodes" this element, with a notion of encoding weak enough for pronouns to "encode" their reference, then it is unclear how such a view could draw a boundary between semantics and pragmatics in any principled, non-arbitrary way.9 3. Disentangling Reference from Semantic Content Let me take stock. I started by presenting four criteria that may be seen as having emerged from the linguistic and philosophical literature and that are supposed to track the distinction between semantics and pragmatics. For a 9 To be sure, I did not show that there are no further criteria whatsoever that one could appeal to in order to get demonstrative reference on the semantic side but reference to an action, in a case in which a verb like 'fnish' takes an object as its complement, on the pragmatic side. The "Optionality Criterion" proposed by Recanati (2004) might be one such, but it would take me too far astray to argue that, at the end, that criterion, too, is problematic. At any rate, my claim is not that there can be no way in which a semantics-pragmatics distinction could be drawn; rather, it is that the mainstream view takes for granted something that is far from uncontroversial. Prepragmatics – April 2013 10 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 long time, it had been assumed that these criteria lined up, shaping up one and the same line of division between the two disciplines and, relatedly, between two classes of phenomena that are the objects of study of those disciplines. But as inquiry progressed and as various phenomena involving context-dependence, such as quantifer domain restriction, scalar implicatures, gradable adjectives etc., came to be studied in greater detail, the criteria started breaking apart. My goal in the previous section was to show that the case of demonstratives already brings out this divergeance clearly. The question now arises whether there are any interesting and substantive conclusions to be drawn from this. In other words, if the distinction between semantics and pragmatics turned out to be blurry and shaky, would that be a problem? What would be the reasons for assuming from the outset that there was any such clear and neat distinction to be captured? As a step towards answering these questions, I will start by discussing the ways in which one may interpret the very notion of the semantics-pragmatics distinction. I will then return to the case of demonstratives and will argue that the mainstream view misconstrues the phenomenon of direct reference. In the last two sections, I will draw further implications regarding, on the one hand, the semantic contribution of pronouns and, on the other, the semanticspragmatics interface. My ultimate goal is to propose a novel picture of our language architecture, on which demonstrative reference is neither semantic nor pragmatic, but pertains to a yet third area: prepragmatics. There are at least three ways of understanding the question of what the semantics-pragmatics distinction is. The frst would be to see it as a purely Prepragmatics – April 2013 11 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 terminological question. If so, when faced with the fact that our initial criteria fail to converge, it becomes a matter of terminological decision which criterion to give preference to (if any). Thus if one decides to focus on semantics as a study of how linguistic expressions relate to “objects whose properties they truly state” (cf. Morris 1936: 10), hence as having to do primarily with how language relates to the world and to truth, then one will locate demonstrative reference on the "semantic" side, but equally well many other contextual phenomena that affect truth value. On the other hand, if one focuses on semantics as being about the expressions' stable lexical meaning, then one will locate all those contextual phenomena on the "pragmatic" side. To the extent that it is a merely terminological choice, there would be hardly any point for the two parties to argue about who got it right. Although I believe that the debate on the semantics-pragmatics distinction may have been to a certain extent a terminological debate, I also believe that there is more to it. The second way of interpreting the idea of the distinction between semantics and pragmatics would construe it as a concrete cognitive distinction, namely between two types of cognitive processes that occur in our linguistic practice, or perhaps even as a distinction between two cognitive "modules". Putting the idea of modules aside to forestall any controversy, and focusing on the idea of two different types of cognitive processes, what would the divergence of the four criteria show? It would simply show that the binary distinction between "semantic" vs. "pragmatic" processes is too simplifcatory, and that the architecture of the different cognitive processes is more complex. What is more, it would come as little surprise that the processing of a given Prepragmatics – April 2013 12 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 expression, such as a pronoun like 'she', could trigger at the same time two different types of processes, such as, for instance, a "semantic" mechanism that deals with lexically encoded information and a "pragmatic" mechanism that deals with context-dependence. After all, it is taken for granted that such a pronoun is also processed phonologically and syntactically, which has never been seen as competing with its being also processed "semantically". The idea that there is a confict between an expression's requiring both semantics and pragmatics only arises if one assumes that once a given expression has been semantically processed, it can't require further pragmatic processing (only the semantic content, to which the expression has already contributed, can). The third way of understanding the question of the semantics-pragmatics distinction is the most relevant to the debate that has occupied philosophers for the past two or three decades. It starts from the assumption that semantics and pragmatics are two distinct and separate disciplines, with distinct objects of study and distinct theoretical sets of problems that they aim to resolve. Then the question of how to draw a line of division between those two disciplines becomes the question of what the primary objects of study for semantics and pragmatics are, and the related question of which theoretical and empirial questions they respectively aim to answer. These are metasemantic questions, and indeed, among the main questions in metasemantics. There is thus a neat contrast between this and the previous way of understanding the question of the semantics-pragmatics distinction: if seen as primarily a question about the cognitive mechanisms at play in processing and interpreting language, then the issue of deciding which type of Prepragmatics – April 2013 13 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 phenomena require "semantic" processing and which require "pragmatics" would rely for its answer on psycholinguistics (and more broadly on cognitive science and its recently emerged disciplines such as experimantal pragmatics); on the other hand, if seen as primarily a theoretical question, then it belongs to philosophy of semantics and linguistic theory. If we understand the question of the semantics-pragmatics distinction in this third way, should it come as a surprise that one and the same expression exhibits a behavior that according to some criteria pertains to semantics and according to other critaria to pragmatics? Presumably not. Take pronouns again. It is well-known that their linguistic behavior is of great interest to morphology and syntax, and this has never been thought to be incompatible with the idea that it should also be of interest to semantics. There is no prima facie reason why the relevance of pronouns to pragmatics should confict with their being an object of study also for semantics (as well as morphology and syntax). One might thus conclude that the reference of pronouns pertains to semantics to the extent that it affects the truth value and that it pertains to pragmatics to the extent that it involves context-dependence; to think that it pertains exclusively to the one or to the other must have been a wrong thought to begin with. I believe that a conclusion along those lines is on the right track, and even though my own proposal is that the context-dependence of demonstrative reference involves prepragmatics rather than pragmatics, it shares the spirit of such a conclusion. However, I'd like to point out that such a conclusion is still in tension with the mainstream view. The view holds that the reference of a Prepragmatics – April 2013 14 Isidora Stojanovic ijn _0 08 43 89 8, v er si on 1 12 J ul 2 01 3 pronoun is part of the semantic content of a sentence in which the pronoun occurs. But the view also embraces the Gricean assumption that the input to pragmatics are semantic contents (or 'what is said'). We thus get something of a circle: the semantic content needs to be determined for the pragmatics to get started, but at the same time, we need pragmatics in order to determine some of the constituents of the semantic content.10 One might think that the obvious way out of the circle is to reject the Gricean assumption. Regardless of whether or not we might want to reject it anyways, my goal in the remainder of this section is cast doubt on the assumption that demonstrative pronouns contribute their reference to semantic content. What I shall argue is that the mainstream view misconstrues the phenomenon of direct reference. To say that we should not think of reference as being part of the semantic content is not to say that we should remove reference from the entire picture. To the contrary, reference plays an important role in the proposal that I would like to put forward. Successful communication requires that people should be able to convey information about other people, about the things around them, about events and places. This, in turn, strongly suggests that a person should be able to refer to those things directly. Here are some examples of what I'd like to propose that we view as paradigmatic cases of direct reference. Imagine that we are at a football match Barça-Madrid and that we have just witnessed Messi scoring a goal. I say: