Context

This book is a development of Stalnaker's dominant research program (going back to Stalnaker 1970), defending the “autonomy of pragmatics” thesis that “it is possible and fruitful to theorize about the structure and function of discourse independently of a specific theory about the mechanisms that languages use to serve those functions” (1). The program is broadly Gricean, explaining the contents of utterances and attitudes independently of how they are encoded in language or thought.The central notion remains the common ground (CG), a body of information presumed shared by interlocutors, defined in terms of mutual acceptance: in its simplest form it is a set of propositions everyone in the conversation accepts, accepts that everyone accepts, and so on, ad infinitum. CG determines a context: a set of worlds compatible with it. The basic goal of inquiry is to shrink this set, eliminating uncertainty about how things are.Also key is the notion of a derived context, a local context that can be operative locally, at various stages of interpretation, but is derived from the CG-context (CGC) of the utterance. The interaction between these two is employed in accounting for attitude ascriptions (chap. 2), presupposition and projection (chaps. 3 and 4), self-location (chap. 5), epistemic modality (chap. 6), indicative and subjunctive conditionals (chap. 7), and disagreement (chap. 8). I begin with the role of these two notions.Stalnaker distinguishes CGC from a K-context (KC), familiar from Kaplan, Lewis, and Montague, which represents aspects of an utterance situation relevant for interpreting what is said. A KC can be modeled by a centered possible world (a world, time, and agent sequence), initiated by the utterance's spatio-temporal location. Both CGC and KC are theoretical notions, aiming to explain phenomena described by the pretheoretic notion of context: for example, how agents exploit context in determining what is said, or change it with their speech acts. So, a theoretical notion of context should at least identify the features of the utterance situation on which an interpretation depends, and account for the effects of speech acts on context. Providing a list of features on which interpretation depends, KC seems ideal for the first role. However, Stalnaker argues, KC doesn't explain how agents can exploit its features in determining what is said, as they may lack information about these features. One can, however, think of a CGC as a set of KCs, as the interlocutors' attitudes can be represented by a set of multicentered possible worlds, capturing how subjects locate themselves, and one another, in the world as they take it to be (26). Thus, CGC becomes an “infinitely iterated attitude...that is our representation of a context in which a discourse takes place” (121).Taking the CGC as central to the dynamics of communication opens a way for an autonomously pragmatic account: supervening on mental states of the agents, CGC provides a resource for explaining how they convey information, and coordinate attitudes, independently of substantive linguistic assumptions. The changes an utterance imposes on the CGC, as well as the ways in which it depends on it for interpretation, are explained in terms of what agents do: what they presuppose and intend with those utterances. But the defense of autonomy requires a further generalization of the Gricean framework: pragmatic reasoning does not begin after semantic interpretation is complete; it is present in all stages of interpretation, constantly affecting the CGC. To capture pervasive pragmatic effects, relying on a single—even if constantly changing—CGC is insufficient; this is where a derived context plays the key role. To understand this notion, start with an example:(1) If Mary has a sister, her sister is away.(2) If Mary's sister is away, Mary must be sad.An utterance of “Mary's sister is away” normally requires a presupposition that Mary has a sister. The requirement carries over to (2) as a whole, but not (1). Why? This is the presupposition projection problem.As Stalnaker explains (1), its consequent is interpreted relative to a derived context, created by the supposition that Mary has a sister. This context is derived from the CGC, by (hypothetically) adding this supposition to it. Hence, this context includes this information, and so the presupposition of the consequent is satisfied in it. Since the consequent is only evaluated relative to this derived context, there is no need for the CGC to accommodate the presupposition.While dynamic accounts offer similar explanations, they posit substantive linguistics mechanisms determining derived contexts and presuppositions. Stalnaker, by contrast, argues derived context is best understood pragmatically, and presupposition is best understood as about what speakers tend to presuppose, rather than what expressions presuppose. Thus, his explanation is autonomously pragmatic: the derived context in (1) is derivative from the CGC, and even if antecedents of conditionals signal that a supposition is in place, speakers are making the suppositions, signaling a shift from the CGC to a hypothetical one.But how exactly should we understand pragmatic autonomy? Is it that for any derived context, we can point to agents' mutual attitudes, say, mutually interpreting a constituent with respect to the common ground enriched with mutual suppositions, in a way that lets us understand local interpretation of embedded clauses continuously with the pragmatic interpretation of whole utterances? So, throughout a discourse, we can understand both CGC and derived contexts as an attitude, understood independently of linguistic practice. A more ambitious thesis is that at each point such attitudes are determined by interlocutors' cooperative reasoning about the purposeful use of utterances and so are best theorized about pragmatically. The dynamics of context-change should be theorized in pragmatic terms, so that manipulations of context throughout the course of interpretation are understood as rational responses to manifest evidence. Stalnaker defends the modest thesis in arguing that context is pragmatic, but the ambitious one is often in play in his arguing that dynamics of the discourse is autonomously pragmatic. However, as we shall see, there are reasons for skepticism regarding this more ambitious thesis.Just as an expression can be interpreted relative to a derived context, the pragmatic force of an utterance can affect which context is relevant for its interpretation. So, Stalnaker argues epistemic modal (and indicative conditional) utterances are associated with a pragmatic force that is prospective: unlike assertion, which adds information to CGC, it changes the space of possibilities in the CGC, expanding it so as to make the modal utterance true. The force of an utterance of “It might be raining” adjusts the CGC to allow for some raining possibilities, if none are already allowed (otherwise, just highlighting such possibilities). The context in which the claim is evaluated is the prospective one resulting from this force-induced change.This is where the autonomously pragmatic explanation seems ambitious: like any rational cooperative activity, communication happens against a body of shared information, which guides a conversation, as it guides any other joint activity. We adjust and manipulate it—and negotiate how to do so—as a rational response to manifest events, and available evidence. The manipulation of the CGC—adding new information in light of new evidence, shifting to a derived context as a result of a supposition, or changing the context by adjusting epistemic preferences through a speech act force—is a special case of this general capacity for rational cooperation. While space prevents discussing all complexities and applications of the account, I want to focus on this ambitious autonomy of Stalnaker's pragmatic account.There is reason for skepticism that the manipulation of context can fully be explained via joint rational activity, as described. To see why, we examine supposition more closely. While Stalnaker allows that the compositional semantics of an if-clause can signal a shift to a derived context, he argues that a supposition is primarily a speech act that, however made, can “go beyond a single sentence, including a sequence of statements, proposals, questions” (91). Intuitively, (3) is a case of the kind just described:(3) A wolf might walk in. It would eat you. (Roberts 1989)While Stalnaker doesn't directly discuss modals signaling suppositions, it is crucial for the correct interpretation of the second sentence that it be interpreted against a derived context including the supposition that a wolf walks in. But while the felicity of (3) prima facie fits nicely with ambitious autonomy, how shifts between suppositional environments are navigated across discourse crucially invokes linguistic rules that go beyond, and cross-cut, our capacity for cooperation. Consider (4):(4) A wolf might walk in. But, then again one probably won't. It would eat Harvey. (Stojnić 2017)Why is (4) bad? Its first two sentences contrast possibilities, one in which a wolf walks in and one in which it doesn't. Its final sentence requires a suppositional context in which a wolf walks in. This possibility has just been mentioned and it is evident the speaker has to suppose it for the final sentence to be felicitous. That (4) is nevertheless infelicitous suggests the dynamics of context are not fully explained by what speakers do, but require linguistic knowledge to explain how they do it. This is even clearer when we contrast (4) with (5), differing only in the order in which possibilities are mentioned:(5) A wolf probably won't walk in. But, then again, one might. It would eat Harvey.Moreover, the smoothness of (5), contrasted with (4), cannot merely be a matter of the proximity of the relevant supposition, as possibilities can affect the interpretation of modals at a distance:(6) A wolf might walk in. It might eat Harvey. But it might not eat anyone. Harvey would be relieved.In (6), the modal in the third utterances has to be interpreted against the whole suppositional context in which a wolf walks in, not just the subcontext in which it eats Harvey. Moreover, Stalnaker maintains that the suppositional context is created prior to the clause being interpreted, as the interpretation depends on it. But it is hard to see how nested suppositional structures like (4)–(6) can be navigated pragmatically, without linguistic manipulations of context, and without relying on the prior interpretation of each clause to guide disambiguation. Similarly, consider (7) (Yalcin 2007):(7) If it is not raining and it might be raining, then I am uninformed.The statement in (7) seems incoherent, regardless of the context. Stalnaker's explanation is that the antecedent requires a suppositional context within which it's both not raining and left open that it is—an impossible context. However, it is crucial for this explanation that both conjuncts in (7) are interpreted in the same derived context (or, that the second conjunct is interpreted in the context to which the supposition of the first has been added). But (7) cannot fully be explained in terms of what speakers do: there is overwhelming evidence the speaker is not signaling that the modal should be interpreted relative to the context in which it is not raining—that delivers inconsistency! Since CGC is available after a shift to the suppositional context (as witnessed by its availability for the interpretation of indexicals), reasoning about the cooperative move of a rational speaker would interpret (7) as (8):(8) If it is not raining and CGC leaves raining open, we are uninformed.That (7) is not interpreted thusly suggests linguistic mechanisms are driving the dynamics of context-shifting (see Stojnić 2017 for discussion).Even if Stalnaker is right that the dynamics of conversation proceeds by manipulating informational context, those manipulations and the resulting interpretation often rely on linguistic mechanisms. One could accommodate this within an autonomously pragmatic account by arguing the facts about compositional semantics are one among many pieces of evidence agents rationally respond to by adjusting the context, so that linguistic manipulations of context are subsumed under rational response to evidence. But, semantic facts aren't just a piece of evidence on a par with others: they trump other evidence even when that results in an inconsistent interpretation. Of course, conversation is a rational cooperative activity, and we can approach it by looking at what agents do: speakers say what they say for a reason, and the audience assumes a level of cooperativeness and rationality. But if we are interested in what's specific about linguistic communication, or how meaning gets determined, it seems we cannot abstract away from how the interlocutors draw on specifically linguistic knowledge, rather than general reasoning, in contributing meaning to conversation, or navigating the dynamics of context, even if we allow reasoning to intervene at all stages of interpretation.So, how strongly should we understand the “autonomy of pragmatics,” and how to understand semantic/pragmatic interface within this framework? The weaker understanding—that we can characterize context through agents' attitudes—while compatible with above examples, does not point to a separation of semantics and pragmatics. It is in principle compatible with all manipulations of context being explained grammatically. But, the more ambitious claim that the dynamics of context-change can be explained by Gricean reasoning seems challenged by the above examples.That said, it is impossible to overstate the importance and fruitfulness of the Stalnakerian framework in studying semantic/pragmatic interface. This latest book is another invaluable development of that framework, tackling an impressive range of cutting-edge problems in both novel and insightful ways

Contemporary philosophy · Context (archaeology) · Epistemology · Geography · Language, Discourse, Communication Strategies · Linguistic research and analysis · Philosophy · Syntax, Semantics, Linguistic Variation

• Modal subordination and pronominal anaphora in discourse

  Open Access•Craige Roberts•Linguistics and Philosophy•1989

  Add to list