HERBERT PAUL GRICE, "CONVERSATIONAL IMPLICATURE," THE OXFORD LECTURES, 1966, BANC MSS 90/135c

Speranza

Context is a foundational book. It articulates the core materials and methods Robert Stalnaker has been deploying for half a century, demonstrating how far a few simple principles go in the theory of meaning and philosophy of language and applying & Elisabeth Camp elisabeth.camp@rutgers.edu 1 Department of Philosophy, Rutgers University, 106 Somerset Street, Room 514, New Brunswick, NJ 08901, USA 123 Philos Stud (2017) 174:1617–1627 DOI 10.1007/s11098-016-0781-5 Author's personal copy them to cutting-edge debates. Its central thesis is ‘‘the autonomy of pragmatics’’: the claim that ‘‘it is possible and fruitful to theorize about the structure and function of discourse independently of specific theory about the mechanisms that languages use to serve those functions’’ (1). This Gricean strategy assumes that communication is a rational activity, and that by reconstructing the ‘‘tasks’’ language performs, we can clarify the ‘‘means’’ by which it accomplishes them (181). Pragmatics, broadly, is the language-independent study of communicative function, semantics the language-specific study of an implementing mechanism. The distinction is analytical rather than empirical: Stalnaker rejects the existence of ‘‘separable pragmatic and semantic processes’’; instead, ‘‘lexical and compositional semantics interact with pragmatic reasoning throughout the evolving process of interpretation’’ (78), in ways that are crucial for explaining general phenomena like presupposition and specific expressions like conditionals. If ‘‘autonomy’’ means a focus on pragmatics as a domain prior to and independent of language, the natural question is what becomes of semantics. I share Stalnaker’s broadly Gricean orientation. But I worry that the explanatory forces he unleashes may carry him beyond where he wants to go, forcing a choice between an even wider application for pragmatics or a more ecumenical conception of semantics. After briefly laying out Context’s basic methodology and materials in Sect. 2, I consider Stalnaker’s treatment of force in Sect. 3. In Sect. 4, I argue that the basic operative model of static truth-conditional semantics plus dynamic force marginalizes a range of interesting linguistic expressions and distorts the actual conventional function of language. 2 Common ground and context The central notion of Context is simple and widely influential from Stalnaker’s previous work: the common ground is ‘‘common background knowledge shared by the participants in a conversation (or perhaps more generally the participants in some cooperative activity)’’ (36). In any context, the common ground serves two roles: first, it provides a set of substantive background assumptions which can be exploited to coordinate efficiently and reliably; second, it distinguishes a set of possibilities that agents care about, and so contributes to specifying what effect an agent intends their action to have (36). The most familiar and focal case is of course speech; but as Stalnaker and Grice emphasize, the structure extends to any joint activity, including practical actions like assembling a TV (Clark 2005). The common ground is an entity, albeit a highly abstract one. It is determined by agents’ attitudes, but in an indirect, idealized way. First, we describe an individual’s belief state by specifying the set of possibilities compatible with what they believe: which possibilities are epistemically accessible to them. Next, we describe common belief by iterating this basic structure: specifying an accessibility relation that is the transitive closure of the accessibility relations among the individual agents’ belief states (44). Although the common ground is typically determined by what is mutually believed, actual belief is neither necessary nor sufficient; what matters is mutual acceptance for current purposes (45). Because it is defined in terms of group 1618 E. Camp 123 Author's personal copy acceptance, the common ground is a social phenomenon (38), in a way that carries over to the attitudes of individual agents. In particular, the attitude of presupposition drops out directly, as what an individual accepts as common ground; accommodation is accepting as common ground something that another agent reveals they take as common ground. Crucially, this social account of common ground, presupposition, and accommodation relies entirely on the logic of epistemic accessibility relations, without invoking conventional—let alone linguistic—rules at any point. Stalnaker’s framework for implementing the notion of common ground is also familiar: the ‘context set’, or set of possible worlds compatible with the common ground. Possible worlds are understood informationally, as ways the world might be. However, Stalnaker is ecumenical about what ‘ways’ are included in the space of possibilities: not just as what there is ‘‘a fact of the matter about’’ (199), but ‘‘individuated in whatever way is necessary to represent all the distinctions that may be relevant to what is at issue in the negotiation’’ (12). From this basic machinery, Context layers successively more complex relations among possibilities—derived contexts to model attitude reports and other temporary suppositions; multiple centering to deal with self- and other-locating attitudes; and parallel contexts to model subjunctive conditionals. Given a specification of the content of the common ground, we can also explain its dynamics: how it changes. A ‘‘manifest event’’ is ‘‘something that happens in the environment of the relevant parties that is obviously evident to all’’ (47), and thereby enters into the common ground. Some manifest events just happen, as when a goat walks into the room; but most are intentionally produced. The manifestness of an action U’s being produced in the context of the current common ground in turn makes manifest—‘‘at least when all goes well’’ (47)—that the agent intends to produce an effect E by U. In cases of joint practical activity, like assembling a TV, this effect E may apply directly to the world; but in the case of communication, it is a change to the common ground itself. In this way, we arrive at a notion of speaker meaning, grounded in common ground, that is ‘autonomous’ of language or any other convention. 3 Force and convention What becomes of semantics on this understanding of pragmatics? Language is a ‘‘conventional device used to communicate’’ (78): a set of shared presuppositions about the canonical intended effects of using certain signs, which facilitates the efficient production of informationally rich manifest events. As Stalnaker emphasizes, the common ground includes not just information about the world, but also ‘‘about the discourse itself, and about the beliefs, aims, intentions, and other attitudes of the parties to the discourse’’ (108). Given that agents manifestly use language to coordinate on all sorts of expectations about the background conditions and intended effects of words—including on things like social status, emotional affect, and rhetorical structure, and for enacting plans, preferences and possibilities—we might prima facie expect semantics to encompass any and all such rules. Pragmatic force in semantic context 1619 123 Author's personal copy However, Stalnaker restricts semantics to the compositional determination of truth-conditional contents. It is important for Stalnaker that semantic theory deliver a proposition, rather than just a truth-value, as output, because he takes the job of semantics to be explaining how agents use language to convey information; and a compositional theory that simply matches sentences with truth-values-in-contexts, like Lewis (1980), doesn’t accomplish this (23). But he assumes, with Lewis—and Grice, Montague, and many others—that the semantic machinery delivers an entity that is ultimately assessable for truth: ‘‘what we want from our semantic theory is a mechanism that takes as its input a sentence with a certain meaning together with a context and delivers, as its output, a proposition’’ (22). Moreover, this propositional, truth-conditional focus is embodied in the basic model of common ground as a body of information: assumptions about which (purported) possibilities do and don’t obtain. So, the basic framework involves a conventional, compositional, truth-conditional semantic core, embedded within a domain-general, rational pragmatics by which agents negotiate a common ground of information. Our current question is how to handle aspects of linguistic practice that are not straightforwardly truthconditional. The other crucial element of the model is illocutionary force, which lies at the periphery between semantics and pragmatics in an interesting way. Start with the basic case, of assertion. This is a clear instance where a functional separation of discourse ends from linguistic means is illuminating: the ‘‘essential effect’’ (52) of assertion—to add a proposition to the common ground—can be specified independently of any conventional practice or constitutive rule. But language conventionalizes this action; and so Stalnaker also includes ‘‘force rules’’ within a language game: To make an assertion is to say something, and not just to mean something…. A move in the game will be the production of a sentence with a certain conventionally indicated force, where force is explained in terms of the way that the move is intended to change the context. So the game will be defined by rules of two kinds: lexical and compositional rules that determine the content of what is said, and rules that determine the force with which the content is said (50-1). How tightly is the conventional ‘‘force rule’’ tied to the constitution of force? Stalnaker’s emphasis on understanding the dynamic effect of assertion independently of convention might suggest that he takes grammatical mood and other force markers to merely signal or make manifest what is essentially a pragmatic phenomenon. And he consistently classifies force rules as pragmatic. On this view, rhetorical questions like ‘What has Sam ever done for John?’ might well be assertions, because they function as such in the larger discourse, including in the sorts of hearer response they make felicitous. So might proposals to add information to the common ground that are proffered using representational systems that lack force markers, like maps or Carnap’s Aufbau. However, Stalnaker appears to take the ‘move’ of assertion to be at least partially constituted by conventional linguistic practice. Thus, he says that ‘‘it is because of the constitutive rules of the game that a certain act counts as an assertion (in 1620 E. Camp 123 Author's personal copy English) that /’’ (52). Whether the speaker of a declarative sentence genuinely attempts to make that move depends on the conversation: it may be manifest that they are merely pretending, or intending to contribute something other than U’s conventionally determined content. Or it may be manifest that they did assert something, but their proposed update may be blocked. So the actual effect of an assertion is independent of the linguistic rules. But what an assertion is depends at least in part on the conventional practice, and specifically on conventional indication of force. Further, this is not just a matter of labeling. A definition of assertion in terms of the ‘essential effect’ of context change is incomplete, Stalnaker says, because ‘‘a full characterization of what an assertion is would also involve norms and commitments’’ (89). He doesn’t specify what the status or effect of these norms and commitments might be. Many norms of assertion, such as saying only what is relevant and justified, can be derived from basic pragmatic principles. Some commitments can be specified in terms of dynamic effect: for instance, asserted content is to be added to the common ground for good, while conjectures are temporary. And others can be specified by their means of update: assertion presents information as new, while presupposition presents it as already established. But some are plausibly anchored specifically in convention. In particular, I would argue that while any change to the common ground is necessarily a social act, given the mutuality condition on common knowledge; assertion is a public act: staking a claim in a way that entails a commitment to justify the proferred content outside of the immediate conversational context (Peirce 1934; Brandom 1983). One consequence of this is that asserted contents are ‘‘detachable’’ from their current contexts: as Stalnaker emphasizes, assertions make contents available for later agreement and disagreement, and for testimonial report, in a way that other speech acts with the same essential effect don’t. On the one hand, the public commitment involved in assertion means that when a proposed update is risky, speakers may have good reason to implicate or insinuate, rather than assert it.1 On the other hand, the public status of assertion also affords speakers a kind of protection. That is, the social but non-conventional status of pragmatic presupposition and conversational implicature entails that they involve negotiation over the common ground that is informal, and thus vulnerable to power dynamics in potentially problematic ways. As a result, in some such contexts, assertion may be a speaker’s only recourse for effectively proposing an update to the common ground, precisely because its conventionality makes the speaker’s intention undeniably manifest in a way that informal forms of meaning don’t. Even overt assertion does not always suffice to get one’s attempted move enacted, or even registered (Langton 1993; Hornsby 1995; Maitra 2009). But conventional force rules lower the degree of context-specific cooperative coordination necessary for an agent to be heard. So far, we’ve seen that force is both pragmatic and conventional. In particular, we explain what assertion does in pragmatic terms; but we specify what it is by a 1 Pinker et al. (2008). Indeed, I argue (Camp forthcoming a) that speakers can mean and successfully communicate contents without adding them to the common ground. Pragmatic force in semantic context 1621 123 Author's personal copy conventional force rule, where this conventionality affects the downstream dynamics of discourse. In this way, we’ve added a layer of force to linguistic meaning on top of the truth-conditional machinery of semantics. But we haven’t seen any reason to challenge either the truth-conditional foundation or the separation of layers. Matters get more complicated with imperatival force. Stalnaker doesn’t directly address imperatives like ‘Wash the dishes!’, but he does explain ‘‘the distinctive character of a command or the issuing of permission’’ in sentences that contain deontic modals, like ‘You must wash the dishes’. Building on Lewis (1979), he posits both a semantic, truth-conditional content and a ‘‘force’’ rule, where the semantic value quantifies over a ‘‘sphere of permissibility,’’ while the force rule specifies that, when uttered by someone with appropriate authority and intent, a sentence containing a deontic modal updates the sphere of permissibility so that the proposition expressed is true. There are two important things to note about this strategy. First, the force rule integrates imperatival force into the basic update model without reducing imperatives to disguised assertions, as Lewis (1970) and Davidson (1979) do. Instead, assertions and imperatives accomplish the same end—of putting forward a proposition as true—by distinct means. Where an assertion expresses a proposition which is determined by the ‘‘prior context,’’ and proposes adding it to the common ground, the content of a command or permission is determined prospectively, by ‘‘the context as it is after the force rule is applied’’; its effect is to ‘‘change the context by changing a parameter of interpretation (the sphere of permissibility) so as to ensure that the sentence expresses a proposition that is true relative to the changed context’’ (136). Locating distinctively imperatival force in the timing and direction of fit between utterance and common ground elegantly maintains a unified account of discourse as ‘‘an enterprise of negotiating over the way that a body of information—the common ground—should evolve’’ (12). Second, by separating the force rule from semantic content, Stalnaker makes sense not just of sentences containing ‘must’ and ‘may’ with imperatival force, but also those used descriptively, to assert something about which worlds are already permissible. This also permits him to explain the effect of deontic modals (whether descriptive or imperatival) embedded within complex sentences: introducing an additional context parameter to the semantics accommodates cases where the operative sphere of permissibility is not that of the actual content, but of a subordinate or derived one (135). In the second half of Context, Stalnaker deploys this broad bipartite strategy, of positing distinct but coordinated semantic contents and pragmatic force rules, to a range of expressions including epistemic modals, indicative and subjunctive conditionals, and predicates of personal taste. The analyses differ in key details, but all embrace a kind of expressivism at the level of force—the intuition that ‘‘the expression is being used to do something other than to communicate an item of information’’ (149)—while preserving a static, truth-conditional semantics. Stalnaker invokes three central motivations for retaining a traditional semantics against a dynamic analysis which ‘‘merg[es] the effect of an assertion with the semantic content of the sentence used to make it’’ (65). The first is the familiar one of compositionality, specifically of explaining an expression’s contribution to complex 1622 E. Camp 123 Author's personal copy constructions like conditionals, disjunctions, and quantification. The second is to explain continuity between expressive and descriptive uses of these expressions, not just by subsuming them under a common formal analysis, but by acknowledging a continuum of cases, including ones where participants are uncertain or disagree about whether an utterance states a fact or proposes a standard or priority (162). Given such continuity, it would be syntactically and functionally unwarranted to segregate such utterances into two distinct classes; but so would shoehorning either class into the other mold. The third core motivation for resisting a dynamic semantics is the need to explain the ‘‘detachability’’ of some utterances’ effects (162). As we saw, a hallmark of assertion is to make asserted content available outside the immediate context. But the ‘update instructions’ proposed by dynamic analyses are inherently context-local: functions that take the current common ground as input and deliver a modified common ground as output. Thus, Stalnaker argues that insofar as utterances containing e.g. modals or conditionals can be assessed and employed by agents outside the local context, we need to posit a stable, truth-conditional propositional content for them (162). And conversely, to the extent that an utterance’s effect is ephemeral, or its content is ‘‘fragile’’ (146), this is evidence that essentially pragmatic, context-local factors involving the partitioning and accessibility of possibilities are playing a crucial role. 4 Expression and commitment Let’s take stock. The initial two-part model has now been augmented by an intermediate layer of force, which is conventional but pragmatic. Semantics and pragmatics interact intimately throughout interpretation, not just psychologically but formally. But we retain a strong division of labor between rules for determining propositional contents and those for updating the common ground. It might seem surprising that so much of the operation of language lives at the periphery between semantics and pragmatics; and that so much of pragmatics is conventional. But if this division of labor explains how language functions as a communicative device, the surprise is at most terminological. I worry, though, that the classification of pragmatics and semantics is more fundamentally problematic. On the pragmatic side, treating force rules as pragmatic, because they function to do something to the common ground, occludes the deep difference between pre-conventional, domain-general principles of rational coordination, versus expression-specific rules for modifying particular contextual parameters. While the core function of force rules is indeed illuminated by abstracting away from linguistic detail, those rules are not ‘autonomous’, since conventionalization makes a difference to what exactly they do. My bigger concern lies with the essential focus on information, as reflected in the truth-conditional semantics. I find most of Stalnaker’s specific arguments that modals and conditionals have descriptive as well as expressive functions compelling, as well as the strategy for integrating imperatival force within the informational framework. But modals and conditionals are just the tip of the Pragmatic force in semantic context 1623 123 Author's personal copy expressivist iceberg. Natural languages contain many, diverse words and constructions whose conventional functions are at least partly non-informational. An exclusively truth-conditional semantics doesn’t merely discourage attention to them; it distorts our understanding of what they do, and of the function of language more generally. One broad class of such words and structures comment on and regulate the structure of discourse itself (Roberts 1996; Asher and Lascarides 2003). These include illocutionary adverbs like ‘frankly’, ‘literally’, and ‘hereby’ (Bach 1999); overt markers of coherence and contrast like ‘actually’, ‘anyway, and ‘however’ (Fraser 1999) and prosodic intonation and stress (Bollinger 1972; Steedman 1991); and discourse particles like ‘Man,’ (McCready 2008) and hedging ‘like’ (Siegel 2002). Such words don’t contribute straightforwardly detachable information; and unlike modals and conditionals, they are not smoothly integrated into the core compositional machinery. So they are much more fully features of force than modals and conditionals. But they still admit of substantive formal analyses, and can have downstream discourse and truth-conditional effects. So at a minimum, providing an adequate analysis of them will further increase the explanatory scope of force, and draw it into more intimate interaction with truth-conditional semantics. Other words are expressive or force-laden in a way that is less easily accommodated within the model of static truth-conditional content plus dynamic force. The intuition that motivates classic expressivists is not just, or fundamentally, that certain utterances ‘do something other than communicate information,’ but that they function to manifest non-cognitive attitudes. Intentions, desires and emotions differ from belief partly in having internal motivational force; it is because expressive words signal such attitudes, in a way more direct than reporting or describing, that their communicative effects are likewise direct and motivational. Some words that are expressive in this stronger sense do have dynamic, contextlocal effects. In particular, Potts (2007) analyzes ‘pure’ expressives like ‘damn’ as updating a context’s ‘expressive setting’, which encodes an ‘‘ineffable’’ feeling that is formally segregated from the rest of the common ground; this substantive separation of affect and information is reflected syntactically by the word’s updating the expressive setting directly, never interacting with the compositional determination of descriptive, at-issue content. But not all non-cognitive attitudes are as fleeting as feelings, and not all expressive words are as dynamically local as ‘damn’. Plans and desires can be as robust and ‘detachable’ as beliefs; and classic expressivists argue that ethical discourse avows, and exhorts others to share, enduring commitments of just this sort. I take the smooth syntactic integration of thin ethical terms like ‘good’ within truth-conditional constructions to be evidence that they have at least partially truthconditional meaning, and I am not sure whether they are indeed conventionally motivational (Camp forthcoming b). But it is more plausible that other words—in particular slurs, epithets and thick terms—function to manifest distinctively noncognitive commitments (Jeshion forthcoming; Camp 2013; Gibbard 1992). Slurs in particular display a interesting mixture of independence from and involvement in truth-conditional machinery: typically, their expressive attitude projects out of the composition of at-issue content, but this is just a default propensity, with a robust 1624 E. Camp 123 Author's personal copy class of exceptions. I take this mixed behavior to suggest that the boundaries between expressive and informational contents (as well as between at-issue and peripheral contents) are more flexible than either Stalnaker’s or Potts’s models allow (Camp forthcoming c). Stalnaker’s framework imposes a choice between analyzing aspects of conventional meaning as truth-conditional, enduring content or as dynamic, context-local force. He is entirely correct that the choice need not be exclusive. But neither should it be exhaustive. Neo-Montogovian intensional semantics is well-developed and expressively powerful; and truth-conditional content does seem to play a privileged compositional role, insofar as many words with robustly expressive or dynamic functions resist full compositional integration. But many of the formal tools of intensional semantics can be exploited by models that supplement or replace possible worlds with entities like plans (Gibbard 2003) or attitudes like being-for (Schroeder 2008). These models have their own important limitations. But the crucial criterion is to capture the actual functions and effects of linguistic expressions, and ultimately of utterances; and it is not clear that any one unified model, especially not a uniformly truth-conditional model, achieves this. The choice between semantic content and dynamic force might not seem unduly constraining. After all, Stalnaker is admirably ecumenical about what goes into the common ground: not just information about the world, but also about the discourse, its participants, and their attitudes; and not just assumptions about robust facts, but also about ephemeral or illusory possibilities. Thus, he can assimilate any features and effects that participants care about within the common ground, by positing the appropriate propositions and parameters. The question is whether this appropriately captures the full, actual, conventional effects of the relevant utterance-types. Psychologically, there is a difference between endorsing the proposition that someone deserves respect and actually respecting them. In a similar way, robustly expressive utterances function—at least when all goes well, when the speaker is sincere and no one objects—to actually enact effects, not just on the common ground, but on the world, and each other. Such utterances make manifest, and directly perform, real-world relations of subordination, or honor, or obligation. And while some of these effects are local, many are more enduring. Rater than defining semantics as a compositional mechanism for determining propositional contents, I prefer to understand it as a system of conventions for creating commitments. Pragmatics exploits these conventions, along with other resources, to achieve coordination between agents in more or less specific contexts. Some of these pragmatic coordinations involve information; others expectations and obligations. But they all lack the public status that is conventionally engendered by explicit, overt claims, commands and promises. Such publicity matters—indeed, I think that it is the conventionality of lexical and illocutionary rules, rather than propositionality, that ultimately underwrites the cross-contextual ‘detachability’ of semantic meaning (Camp 2016). The fundamental model of discourse as ‘an enterprise of negotiating over the evolution of a body of information’ is a compelling one; in particular, it delivers an integrated conception of semantics and pragmatics which is amenable to precise formal analysis on the basis of minimal assumptions, and which thereby enables us Pragmatic force in semantic context 1625 123 Author's personal copy to explain crucial functional structures that are otherwise occluded in the murky muck of conversation. But like any abstract, idealized model, it ignores and distorts phenomena that don’t fit. One upshot of the ‘autonomy of pragmatics’ is to motivate ‘‘a broader conception of the general function of language,’’ which allows us to ‘‘tell larger story about the way the use of language determines the way in which a communicative situation evolves’’ (183). Stalnaker’s analyses go a long way toward accomplishing this goal for the expressions he considers. But the actual function and operation of natural language—and specifically, I would submit, of semantics—is more multi-dimensional, messier, and more directly engaged in the world than a uniformly informational model acknowledges. References Asher, N., & Lascarides, A. (2003). Logics of conversation. Cambridge, UK: Cambridge University Press. Bach, K. (1999). The myth of conventional implicature. Linguistics and Philosophy, 22, 367–421. Bollinger, D. (1972). Accent is predictable (if you’re a mind reader). Language, 48, 633–644. Brandom, R. (1983). Asserting. Nouˆs, 17, 637–650. Camp, E. (2013). Slurring perspectives. Analytic Philosophy, 54(3), 330–349. Camp, E. (2016). Conventions’ revenge: Davidson, derangement, and dormativity. Inquiry, 59(1), 113–138. Camp, E. (forthcoming a). Insinuation, indirection, and the conversational record. In D. Harris & M. Moss (Eds.), New work in speech acts. Oxford: Oxford University Press. Camp, E. (forthcoming b). Expressivism. In T. McPherson & D. Plunkett (Eds.), Routledge handbook of metaethics. New York: Routledge Press. Camp, E. (forthcoming c). Slurs as dual-act expressions. In D. Sosa (Ed.), Bad words. Oxford: Oxford University Press. Clark, H. (2005). Coordinating with each other in a material world. Discourse Studies, 7(4/5), 507–525. Davidson, D. (1979). Moods and performances. In A. Margalit (Ed.), Meaning and use (pp. 9–20). Dordrecht: D. Reidel. Fraser, B. (1999). What are discourse markers? Journal of Pragmatics, 31, 931–952. Gibbard, A. (1992). Thick concepts and warrant for feelings. Proceedings of The Aristotelian Society Supplementary, 66, 267–283. Gibbard, A. (2003). Thinking how to live. Cambridge, MA: Harvard University Press. Hornsby, J. (1995). Disempowered speech. Philosophical Topics, 23(2), 127–147. Jeshion, R. (forthcoming). Slurs, dehumanization, and the expression of contempt. In D. Sosa (Ed.), Bad words. Oxford: Oxford University Press. Langton, R. (1993). Speech acts and unspeakable acts. Philosophy and Public Affairs, 22(4), 293–330. Lewis, D. (1970). General semantics. Synthese, 22(1/2), 18–67. Lewis, D. (1979). A problem about permission. In E. Saarinen, et al. (Eds.), Essays in Honour of Jaakko Hintikka (pp. 163–175). Dordrecht: Reidel. Lewis, D. (1980). Index, context, and content. In S. Kanger & S. O¨ hman (Eds.), Philosophy and grammar (pp. 79–100). Dordrecht: Reidel. Maitra, I. (2009). Silencing speech. Canadian Journal Of Philosophy, 39(2), 309–338. McCready, E. (2008). What man does. Linguistics and Philosophy, 31(6), 671–724 Peirce, C. (1934). Belief and judgment. Cambridge, MA: Harvard University Press. Pinker, S., Nowak, M., & Lee, J. (2008). The logic of indirect speech. Proceedings of the National Academy of Sciences, 105(3), 833–838. Potts, C. (2007). The expressive dimension. Theoretical Linguistics, 33(2), 165–197. Roberts, C. (1996). Information structure: Towards an integrated formal theory of pragmatics. In J.-H. Yoon & A. Kathol (Eds.), OSU Working Papers in Linguistics (Vol. 49). Papers in Semantics, Columbus, OH: The Ohio State University Department of Linguistics. 1626 E. Camp 123 Author's personal copy Schroeder, M. (2008). Being for: Evaluating the semantic program of expressivism. Oxford: Oxford University Press. Siegel, M. (2002). Like: The discourse particle and semantics. Journal of Semantics, 19, 35–71. Steedman, M. (1991). Structure and intonation. Language, 67(2), 260–296.