1 Context as common ground

According to one central theme of Context, many of the notions employed in characterizing group inquiry are intelligible and applicable outside of language use.Footnote 1 In this respect Stalnaker is inspired by Grice, who aimed to elucidate speaker meaning, and on this basis linguistic meaning, in terms of intention and belief. Accordingly, linguistic meaning, implicature, presupposition, and the like are applications of a broader category of rational collaboration. Stalnaker expresses his own similarly-inspired approach as the doctrine of the Autonomy of Pragmatics: “it is possible and fruitful to theorize about the structure and function of discourse independently of specific theory about the mechanisms language uses to serve those functions.” (2014, p. 1) As we will see below, the notions of acceptance state and common ground may be characterized with no appeal to language, and we seem to be able conduct an inquiry in a non-linguistic environment. According to Isack and Reyer (1989), for instance, evidence from rock paintings suggests that humans and honeyguide birds (Indicator indicator) have been collaborating for some 20,000 years in search of beehives and the honey they can yield. The Boran people of Kenya will summon a honeyguide bird, who arrives in response to lead the humans with a characteristic call and by displaying its white tail feathers. When the hunt is successful, the humans get honey while the honeyguides get larvae and wax from the beehive. As a group of hunters follows such a bird, it may help them settle their inquiry into the beehive’s location in spite of the fact that the humans and the bird share no common language in any reasonable sense of that term.

Adherence to the Autonomy of Pragmatics also obliges us to seek not just characterizations, but also, where possible, explanations of linguistic phenomena in terms that are broader than the linguistic. Stalnaker presents accommodation as a case of this kind. Pending a more exact formulation of the notion of common ground below, a proposition is part of common ground (CG) among a set of interlocutors if it is presumed to be shared among them (2014, p. 2). Thus when a speaker overtly presupposes some proposition that is not yet part of common ground (CG), her doing so threatens to make the context defective in the sense that she presupposes something that the others do not. This can result in interlocutors talking past each other. On the hypothesis that defectiveness is not just a descriptive but also an explanatory concept in pragmatics, it is natural to wonder if interlocutors will aim to avoid defective contexts. This is just what they seem to do, by adjusting CG in such a way that it now includes the missing propositions. Even when CG does not contain the proposition that A owns a cat, soon after A’s utterance of, “Sorry I’m late, but my cat was spitting up hairballs,” interlocutors will in all likelihood see to it that CG does contain that proposition. Such a tendency would seem to be part of a larger pattern discernible in rational behavior: I will generally swerve to avoid hitting you if we are on courses that would otherwise collide. Even outside of language, we can accommodate each other in service to cooperative ends.

This conception of context in terms of common ground (what Stalnaker calls the CG-context model) possesses unquestionable explanatory power, shedding light on presupposition, accommodation, the behavior of certain types of conditionals, epistemic modals, and related phenomena.Footnote 2 The CG-context approach is also highly abstract, so merely pointing out that it fails to account for an aspect of communication is an inconclusive criticism. Instead our question should be whether it can be extended or modified to account for such a phenomenon while preserving its spirit. In what follows I will assess the CG-context approach’s prospects for making sense of the variety of ways in which interlocutors accept propositions (Sect. 2), some different types of conversation and the norms distinctive of these different types (3), some pre-illocutionary pragmatic phenomena (4), conversational injustice (5), and fictional discourse (6). Of course, the wider we can show the explanatory scope of the GC-context model to be, while preserving its spirit, the more that will redound to its credit.

2 Varieties of acceptance

Stalnaker characterizes an acceptance state as a psychological state in which an agent treats a proposition as true for some purpose (2014, p. 39). Belief, supposition for argument’s sake, imagining that something is so for the sake of a story, and conjecture are all acceptance states on this usage.Footnote 3 A degree of belief in, and sheer contemplation of a proposition are not acceptance states; similarly for a mental image of an object or event.Footnote 4 On this basis we may also say what it is for a bit of information to be common ground among a group of interlocutors. A proposition P is common ground between agents A and B just in case both A and B accept p, both accept that both accept p, both accept that both accept that both accept p, and so on (2014, p. 25); we’ll write p∈CGAB. This definition readily generalizes to more than two agents and to multiple propositions. Since ‘information’ is not being used as a factive here, and because the misguided may share common ground, it could happen that p∈CGAB while P is false. Since, further, we might accept P merely for the sake of argument, we might even know that it is false (2014, p. 232).Footnote 5

It will be useful to draw on a familiar distinction between the sentence or utterance used to perform a speech act, and the content of the speech act thus performed. An assertion that the bus arrives at noon takes a certain amount of time and has either acoustic or (if it is in sign language) visual properties. By contrast, what that assertion expresses is a proposition that is an abstraction from individual speech acts, and as such has neither spatial nor temporal location. Two distinct assertions, even when in different languages, might express one and the same proposition, and a proposition thus understood is available for expression whether or not anyone ever does express it in a speech act.Footnote 6

The distinction between the force and content of a speech act carries over to the act of asking a question and the question thus asked. The asking of a question will be an event with a location in space and time, whereas such an event will also express a question that may never have been asked before, and that someone else might find poignant or silly as the case may be. I will follow Hamblin 1971 and Bell 1975 in construing questions qua abstract objects as sets of propositions. The semantic object expressed by the sentence, “How many people attended the reception?” is thus {no one came to the reception, one person came to the reception,…, n people came to the reception}, for some n that is a plausible upper limit of the number of people the speaker asking the original question would have found conceivable, say one hundred.Footnote 7 With this set-theoretic notion of a question, we may now define a complete answer as any proposition that is a member of that set, and a partial answer as any non-empty proper subset of that set. Observe that it is possible to express a question without asking or raising it. Thus I might remark that Hitomi wonders where Mauritius is, without myself asking where that island country is located. So too, as semantic objects, one question Q1 may entail another question Q2, in that any complete answer to Q1 will also be a complete answer to Q2.

Like propositions, questions can be accepted, and can in turn become part of CG. With acceptance of a proposition understood in the CG-context framework as its bring true in all worlds in an agent’s context set, it is natural to follow Stalnaker’s suggestion (2014, p. 141) that we treat accepting a question as dividing that context set into partitions, each partition representing one complete answer to the question. My car keys are clearly missing: I’ve now banished from my context-set all worlds in which I know where they are. I now ask myself the yes–no question, did I leave them in my bike panniers? In taking this as a question worthy of investigation, I thereby subdivide my context-set into two cells, one of which contains only worlds in which I left the keys in the panniers, the other of which contains all the rest. Or I may ask myself the wh-question, where did I leave them?, which will express the set of propositions comprising such members as, I left them in the car, I left them in my panniers, I left them at the deli I stopped in for coffee this morning, and so on. In either case, we may understand the partitioning of one’s context set into sets of worlds, with each such set corresponding to one element of the question, as representing an agent’s undertaking to rule out all but one partition.

So understood, we may next understand a question’s becoming CG as follows: A accepts Q (that is, has his context set partitioned in a way demanded by that question), as does B; each accepts that the other does, and so on up the hierarchy. CGAB now comprises a question in addition to propositions, but its doing so is not a matter of its containing a further entity beyond the set of worlds corresponding to those propositions. It is instead a matter of CG’s structuring those worlds in a certain way. A group of interlocutors’ context set may be partitioned in a number of distinct ways corresponding to each question they have jointly accepted.Footnote 8

Propositions and questions may be part of CG while still being unmentionable due to considerations of etiquette, ethics, or prudence. When the room is elephant-free, however, we can felicitously and even overtly presuppose what is in CG; this in turn enables us to update it efficiently, with such updates in turn constraining the shape of future contributions to CG. For Stalnaker, the paradigm way in which speakers update CG is with assertions: a speaker makes an assertion, others have no reason to doubt her, and thereby the content of that assertion is now part of CG, as is the fact that the speaker has made that assertion (2014, p. 36). Speakers can also retract their claims, such as when one withdraws an earlier conjecture that P as new evidence comes in. Here CG contracts so as no longer to comprise P. It should now be clear that CG may also be updated with questions and their retractions. Further, a question is resolved when all but one answer it comprises have been excluded from the context set. To the yes–no question about car keys, I resolve it by looking in my panniers and finding nothing.

CG might also evolve due to events that are perceptibly salient to all relevant parties: as the lights go out in the conference room, it becomes GC among us all that this is so. These are manifest events (2014, p. 47), which will on occasion resolve questions as well. (A breeze lifts a pile of papers, revealing the missing keys.) Some contexts are defective: here at least one of the speakers is mistaken about what is CG among her interlocutors. A context is non-defective just in case it is not defective. Even in a non-defective context where p∈CG, it need not be CG that each party in the conversation accepts p. However, Stalnaker also defines an equilibrium context as one in which p∈CG iff it is CG that each party accepts P (p. 235).

There is a large space between the illocutionary and the manifest, and some of what occupies it can be important for the dynamics of conversation. Included here are forms of meaning within the remit of communication, but which lack the reflexive intentions characteristic of speaker meaning. In Sect. 4 I will argue that such cases can play a large role in a conversation. For now let us note that just as there are acceptance states beside belief, so too there are speech acts other than assertion in which speakers invite their audience to accept a proposition as true. Examples are conjectures, guesses, and those utterances in which we invite our audience to imagine that something is so. (Storytellers do a great deal of the latter.) The content of any such act may be added to CG. But given our definitions of CG and of defectiveness, a misunderstanding of the illocutionary force of a speaker’s utterances will not, per se, make a context defective. A might be telling B a tall tale about a recent mountaineering expedition, but B naively takes it to be a recounting of fact. B might then accept such a remark as, “Then I bivouacked at 14,000 feet,” as fact even though A only meant it as embroidery. Both accept it, both accept that both accept it, and so on in such a way as to make it part of CGAB. This context could even be an equilibrium, and yet there would still seem an important respect in which B has misunderstood A.

It is not clear whether the misunderstanding imagined in this last case justifies a revision of the CG-context model. However, we can readily imagine more serious problems arising from a similar source even when there is no mutual misunderstanding among interlocutors. Astrophysicists might be collaborating on the path of a planned space flight. One conjectures P, while another asserts that if P, then Q. Both illocutions are accepted into CG. These scientists are now committed to, and may well infer Q, but it would be irrational for them to take Q as established in the way that if P, then Q is. Instead, Q should be treated as having the status of a conjecture only. Similar dangers can arise for contents accepted with a variety of other modes of acceptance. To avoid such dangers, we would appear to need to posit a number of distinct common grounds for each set of interlocutors: one for what they accept as established fact, another for what they accept as a supposition, and so on.

Once we have posited multiple levels of acceptance for a single group of interlocutors, we may then ask whether there are any principles linking their different common grounds. One suggestion is that while anything in assertoric CG may be invoked into another GC, the reverse direction is not in general acceptable. Exceptions would be when a supposition of P for reductio produces a contradiction, thereby justifying accepting not-p assertorically; likewise for the conversational analogue of conditional-introduction. Further, a sheer guess that P might get “promoted” to an educated guess as new evidence comes in; or something we took ourselves to know now gets demoted to the class of educated guesses as we learn that what we thought was conclusive evidence on its behalf is only suggestive of its truth.

Summing up so far, I am suggesting that the CG-context model be extended to include the possibility of questions being accepted as well as being part of CG. I have also advocated recognition of different ways of accepting propositions, corresponding to many of the illocutionary forces with which we express them. One might worry that it is too cognitively demanding for interlocutors to keep track of multiple common grounds, but we should bear in mind that speakers often use external aids to help with such challenges: natural-deduction notational systems do just this, and we often offload the cognitive demands of such tasks by keeping notes on what we are presuming or guessing.

3 Conversational projects

The extensions of the CG-context model I am advocating bring into focus a view of conversations as projects that are richer and more nuanced than mere exchanges of information. To see why, recall that Stalnaker (1970) had distinguished two types of conversation, namely inquiries and deliberations. The former is a joint effort to ascertain information, while the latter is a joint effort to formulate a plan of action. Deliberations have tended to fade from the scene in Stalnaker’s subsequent writings, while inquiry has taken center stage.Footnote 9 Further, in characterizing the progress of inquiries, Stalnaker tends to describe agents as making assertions in the course of incrementing their common ground. Correlative with this is a view of the core function of language as the exchange of information: “Our general background story says that language is a device for conveying information.” (2014, p. 23) Yet while what I’ve elsewhere (Green 1999) called exchanges will involve a desultory sharing of information among interlocutors, conversations often have a teleological dimension: instead of just chatting, we frequently aim to answer a particular question or set of questions about what to do or what is the case (Green 1995, 1999). Such questions may but need not be formulated verbally, may at times be quite imprecise, and occasionally even confused. Yet just as non-linguistic collaborative activity is goal directed, so is collaborative linguistic activity, including conversation.

Even at a high level of abstraction, then, a view of language as a device for conveying information is at best half-true. A more accurate, yet still bird’s-eye view portrays language as a tool for solving problems both cognitive and practical, with information-exchange a crucial means to this end rather than an end in itself. In verbal inquiries, agents exchange information in order to solve theoretical problems, while in verbal deliberations, they exchange information and calibrate preferences in order to solve practical problems. What is more, as I’ll argue below, we can elaborate the CG-context framework to accommodate this perspective without doing it great violence.

This teleological perspective enables us to shed light on a problem raised by Thomason (1990). Discussing the

Cooperative Principle: Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk-exchange in which you are engaged. (Grice 1989, p. 26),

Thomason complains that while the CP tells us to make our contributions appropriate, this injunction seems empty. (“How should I dress for the party?” “Dress appropriately.”) But that complaint may be premature. If one knows what kind of party is being held (luau, hair metal, steampunk, etc.), that will do a great deal to direct one to the proper choice of attire. So too, if we can find a substantive taxonomy of types of conversation, that may give interlocutors some guidance. Now, it is not plausible that there is some one end toward which all conversations, or even all properly executed conversations, are directed. However I want to suggest that conversations are characteristically directed toward some end or other.Footnote 10 From this teleological perspective, conversations emerge as projects that might be spread over many years, continents, or journal issues.

The Cooperative Principle makes reference to the accepted purpose or direction of the talk-exchange, and in elucidating it Grice offers examples of collaborative projects–repairing a car and baking a cake (1989, p. 28)–that are both goal directed. Similarly, as two or more people engage in a conversation they typically work toward a common goal. When the next bus is arriving, why Abdullah is late for the meeting, how to measure wind speed, are all issues that can inspire and guide our verbal collaboration. In some cases interlocutors aim to answer a question about how things are (what causes the Northern Lights, how high those mountains are); in others we work together to formulate a way of getting something done (how to get to the train station, where to look for rain gear). Following Stalnaker (1970), let us call conversations aimed at formulating a course of action deliberations, and call those conversations aimed at answering a question about how things are inquiries.

We may also distinguish ostensible from ulterior conversational aims. The ostensible aim of a conversation between two neighbors chatting about the weather might be to answer the question whether snow is likely by the end of the week, while the ulterior aim of their conversation might be to strengthen their bond or for one to impress the other with her meteorological expertise. Our focus in what follows will be on ostensible conversational aims, which we should expect to be represented in some manner in CG.

In light of our discussion thus far, let us say that whereas a verbal exchange is any sequence of speech acts, a conversation is a sequence of such acts ostensibly aimed at answering either a theoretical or practical question.Footnote 11 Theoretical questions need not be terribly abstract: they include such questions as where one’s car keys are or what time the next bus arrives. What’s more, Grice’s Quality Maxim (“Make your contribution one that is true”) does not follow from the CP as a theorem: bullshitters, sensu Frankfurt (2005), might engage in a conversation, and even do what is required of them in, say, answering the question about the fish that got away, without holding themselves to a norm of truthfulness. For similar reasons conversations need not be about the actual world, but may instead concern worlds we construct: a story might lead us to an answer to the question what became of Madame Bovary after her affair described in the eponymous novel, or how a father and son made their way through the post-apocalyptic world of The Road. Such stories are typically told by a single author, but practices like cadavre exquis show this to be a contingent aspect of fiction only.

Conversationalists do often play symmetrical roles, while in other cases one speaker is doing a different conversational job from the others. For instance, in didactic conversations, one interlocutor aims to lead others to accept an answer to a practical or theoretical question about which the speaker may already have an opinion or plan, while in socratic conversations, an interlocutor aims to lead others to answers by helping them to formulate their own views or plans of action. Soliloquys, in which we carry on conversations with ourselves, might be either of the above kinds.Footnote 12

Some main types of conversation are displayed below. In symmetrical inquiries, speakers pool their information on the way to answering a question about how things are. In symmetrical deliberations we pool information and take into account our preferences as we work to formulate a plan of joint action. In didactic inquiries, one speaker answers a question for an audience; one familiar case of this is an academic lecture, while another is the telling of a story. A didactic deliberation is one in which a speaker tells others how to do something. In a socratic deliberation, one speaker elicits from another a plan for getting something done. Six different types of conversation are displayed in Table 1:

Table 1 A taxonomy of conversation-types
Full size table

Differentiating conversations into types helps to cash out the ceteris paribus character of some standard pragmatic explanations. For instance, it is widely held that speakers who assert P in a situation in which their making a logically stronger claim Q would also be relevant without infringing a conversational norm of brevity, thereby conversationally implicate either that Q is not true, or that they are not in a position to assert Q (Geurts 2010). Imagine that in the course of a symmetrical inquiry between A and B concerning post-Cold War global politics, A remarks,

  1. 1.

    Most former Soviet bloc countries are dictatorships.

    In so doing, A proffers (1) for entry into CGAB. But let us imagine further that B demurs, challenging A to support her claim. A might respond to this challenge by taking either a didactic or socratic route, each having its distinct advantages and pitfalls. Suppose that A takes a socratic strategy in which she asks B to recall what he knows. A asks,

  2. 2.

    Well, are Kyrgysztan and Tajikistan dictatorships?

    A replies,

  3. 3.

    Yes, those two certainly are.

    In so answering, A has made a weaker claim than he might have. For instance, he might instead have replied,

  4. 4.

    Yes, as is Turkmenistan.

(4) is relevant, and no less brief than (3). However, in answering A’s question with (3), B is not implicating either that (4) is not true, or that he is not in a position to assert (4). The reason is that if you are trying to convince me of something based on my own commitments, I should give up as little ground as possible while maintaining overall consistency as well as fidelity to those commitments. Conceding more ground than this would constitute conversational supererogation: acceptable but by no means enjoined upon the interlocutors’ participation in an asymmetrical socratic inquiry. What we implicate, and thus mean, is accordingly going to depend in part upon the kind of conversation in which we are engaging.

Thus far I have advocated an elaboration of the CG-context model to include multiple common grounds, a partitioning of CG to represent questions we seek to answer, and have suggested that for a more nuanced explanation of such phenomena as conversational implicature, we need also to recognize multiple kinds of conversational project.

4 Pre-illocutionary pragmatics

Developed along the lines we have just suggested, the CG-context apparatus is equipped to shed light not just on conversational implicata that are generated by the performance of a speech act, but also on some of the processes that create such speech acts in the first place. In an article in which I assessed attempts to invoke pragmatic phenomena in defense of Direct Reference theories of singular terms (Green 1998), I suggested that some highly generalized conversational implicata could make a difference for the truth conditions of the sentences in which they occur. This suggestion came in the form of an Embedded Implicature Hypothesis,Footnote 13 which I offered as a charitable interpretation of defenses of Direct Reference before criticizing the pragmatic strategy that was the focus of my article.

Recanati (2003) replied to this by arguing that the idea of a so-called embedded implicature may be ruled out on definitional grounds. Recall Grice’s account of conversational implicature,

A person who says (or makes as if to say) that p, has implicated that q, provided that (1) he is to be presumed to be observing the conversational maxim, or at least the Cooperative Principle, and (2) the supposition that he thinks that q is required to make his saying (or making as if to say) that P consistent with this presumption, and (3) the speaker thinks (and would expect the hearer to think he thinks) that it is within the competence of the hearer to work out, or grasp intuitively, that the supposition mentioned in (2) is required.

Invoking this account of conversational implicature, Recanati writes,

Implicatures are generated via an inference whose input is the fact that the speaker has said that p. Hence no implicature can be computed unless something has been said. (2003, p. 300)

(As Neale (1992) observes, on Grice’s idiosyncratic usage, ‘saying that P’ requires speaker-meaning P in a way that corresponds closely to assertion. Hence merely uttering a sentence while testing a microphone does not qualify for Grice as saying that P. This would appear contrary to a commonsense usage of ‘say’ and its cognates.) Recanati is correct that the characterization Grice gives of conversational implicature is formulated in terms of a speaker asserting that P or making as if to do so, and so strictly speaking only allows for implicata to be triggered under one of these two conditions. But it should be clear that this is too narrow even for Grice’s own purposes. First, he acknowledges the possibility of sub- or pre-illocutionary implicature, writing, in his ‘Retrospective Epilogue’, “It certainly does not seem reasonable to subscribe to an absolute ban on the possibility that an embedding locution may govern the standard nonconventional implicatum rather than the conventional import of the embedded sentence.” (Grice, 1989, p. 375).

Furthermore, Grice’s own account of implicature can with minimal violence be recast in order to accommodate this possibility. In parallel with the account that he gives, a first pass would be to allow that a person says that P so long as she has made an utterance admitting of that interpretation, and such that so interpreting her words would make her contribution one that, were it made, would best conform with conversational demands (Green 2008, p, 87). (This suggestion needs further generalizing to include the possibility of other illocutions than assertion; we return to this in a moment.)

In this same spirit, Stalnaker suggests some principles by which such sub-illocutionary phenomena might be explained:

1. If a statement made by a syntactically ambiguous sentence would be irrelevant or manifestly false on one of the readings, then that reading will be excluded, and the reason for excluding it will be that a conversational maxim will be violated if that were the intended reading;

2. If a statement would be appropriate only on one of a number of linguistically possible ways of interpreting a pronoun, then the pronoun will be interpreted that way. (p. 85)

These two dicta make a useful beginning, but are not going to help us with all types of disambiguation or pronoun-interpretation. Both disambiguations of a structurally ambiguous sentence containing multiple quantifiers might be relevant, and neither of them obviously false. Here, however, our taxonomy of conversation-types can help guide us toward a choice of one interpretation over the other. For in an asymmetrical socratic inquiry, the interlocutor who is the target of the socratic questioning will most reasonably assert the logically weaker disambiguation rather than the logically stronger one even when both are relevant to the question at issue, both plausible, both equally prolix, and so on. This is reason to think that is what a speaker in such a situation will in fact mean by her (ambiguous) words. By contrast, another type of conversation, such as a symmetrical inquiry, could well mandate a different interpretation of her words.

To accommodate speech acts other than assertions, and semantic contents other than propositions, we might recast Grice’s formulation of the explanandum associated with a case of implicature as follows: ‘In illocuting, or making as if to illocute, that P (or alternative semantic content), A implicated that Q (or alternative semantic content)'. Further, at least for those speakers participating in a conversation as that notion is understood here, an account of such phenomena might take the following form:

In uttering sentence S, speaker A illocutes (or makes as if to illocute) that P with force F (or alternative semantic content), and possibly also implicates that Q with force G (or alternative semantic content), just in case the supposition that A speaker-means P with force F (or alternative content), and possibly also speaker means without saying that Q with force G (or alternative semantic content), is required to make her utterance consistent with the presumption that she is doing what the conversation in which she is participating requires.

These ideas come together to suggest that Gricean explanations can get traction even if we do not yet have a speech act from which to generate implicata. They will allow us to explain how interlocutors impute illocutionary acts to one another (and sometimes even to themselves) on the basis just of utterances that are lexically or structurally ambiguous, or that contain non-words, are subsentences, or contain pronouns admitting a number of possible contextual fillings-in.

Once we have countenanced pragmatically relevant phenomena occurring below the level of the illocution, it is a natural next step to ask whether there are communicatively significant phenomena not resulting from the reflexive intentions widely held to be characteristic of speaker meaning. Stalnaker discusses two kinds of event that can change a conversation’s CG: speech acts, on the one hand, and manifest events, on the other, where the former but not the latter are instances of speaker-meaning. But meaning of a type germane to communication is not restricted to speaker meaning. Imagine (or recall) having a conversation with a teenager over whom you have some authority in which she rolls her eyes in response to something you say. Lisa Damour writes,

Adolescents usually hate being told what to do, and will reflexively resist even suggestions with which they agree. Imagine a girl who is planning to put on her warmest coat when her well-meaning mother urges her to bundle up. If the teenager is developing normally, not a cell in her body is inclined to respond with a sincere, “Great idea, Mom! I was just thinking the same thing.” (And her mom might be stunned, or at least wonder what her daughter was up to, if she did.) But the girl still wants to be warm. Enter the eye roll! One spin around the socket while donning the coat and the girl advertises her resistance while doing as she intended all along. (‘Why teenage girls roll their eyes,’ The New York Times, Feb. 17, 2016)

“Advertising one’s resistance” is not an illocutionary act; nor is it typically a form of speaker meaning. Like a contemptuous tone of voice, a spontaneous smile, a “miserable” smile, or a use of upspeak or vocal fry, it is a way of manifesting an aspect of one’s psychological state. Each of these behaviors is typically also designed to manifest the state that it does. But a designed manifestation of a psychological state is not, as such, a case of speaker meaning. Skin is designed for thermoregulation, but this does not imply that it was produced by an intelligent or sentient designer to do that job. So too, natural selection or cultural evolution might design our faces to express certain of our emotions. In a given case in which, say, a scowl, expresses one’s anger, an observer might describe the situation by saying that the person’s scowl means that he is angry. The notion of meaning invoked is not factive, and thus is not an instance of natural meaning. But the notion invoked is also not speaker meaning: spontaneous scowls are not generally underpinned by complex communicative intentions.Footnote 14

Stalnaker’s approach would see as eye-rolling, vocal fry, upspeak, spontaneous scowls, and the like as manifest events, like the crashing of thunder or flickering of the overhead lights. Manifest events tend to become CG quickly so long as interlocutors are paying attention to the situation and to one another. However, we lack an account taking us from overt behavior to the psychological state that it is sometimes designed to manifest.

This is where the notion of expression can do some work. Roughly, I construe expression as behavior that is designed to, and in fact succeeds in, conveying information about one’s psychological state (Green 2007, 2016b). Accordingly, while it can easily become CG that A rolled her eyes, what takes us to a psychological state—enabling the existence of that state to be part of CG–is the fact that the behavior expresses contempt. This is why it would be conversationally appropriate, though perhaps not politically expedient, to respond to the teenager’s eye-roll with a sentence presupposing its existence, such as, “Lose the attitude.”

5 Conversational injustice

In light of the interest that communicative injustice has generated among philosophers in recent years, it would be valuable to know whether the CG-context model can be conservatively expanded to address it. First of all, the phenomena that motivated the positing of illocutionary silencing look intelligible from the perspective of this framework. According to this “silencing” literature, some speech acts are important for a person’s ability to maintain her autonomy. The illocution of refusing is a case in point, given its utility in deflecting such things as unwanted romantic overtures. However, in a social milieu in which a person’s repeated attempts to resist romantic overtures are taken as veiled invitations, putative refusals of such overtures will often be ignored or minimized. Some writers hold that in such cases we should expect to find misfires, in the way in which my attempt to appoint someone to a position in an institution in which I have no authority will misfire (Langton 1993; Hornsby and Langton 1998). From the viewpoint of the CG-context approach, that would amount to a speaker’s performing an act of speech—and thus producing a manifest event—but no speech act.

Others have challenged this “silencing” approach on the ground that the speech act of refusing is more analogous to that of warning than it is to appointing: just as I can warn potential divers by posting a sign on a cliff—and have warned even those who ignore the sign—so too, I can refuse advances even from those who ignore me (Bird 2002). Yet even if these speech acts are not silenced in the relevant sense, they may nonetheless be muted. For the CG-context approach may observe that some speakers’ proffered contributions to CG might be put up to higher standards than are others’. CG depends on what interlocutors accept, so if certain parties to a conversation refuse to acknowledge that one speaker has performed a speech act, then neither its force nor its content will become part of CG, which will only register that this speaker has performed an act of speech. Or perhaps they acknowledge the performance of an illocution, but put the speaker up to impossibly high standards before the content of that illocution is absorbed into CG. In either case we see an asymmetry different from that discussed in our taxonomy of conversations, but helping to shed light on how certain speakers can be pushed to the margins of inquiry and deliberation.Footnote 15 Further, we can accommodate either such case with no need to revise the CG-context model of conversation and context.

6 Stories and conversations

On a taxonomy permitting asymmetrical-didactic conversation as a particular type, it is not difficult to see how the author of a history of the sinking of the Lusitania, such as Erik Larson in Dead Wake, is engaging in a conversation with his readers: he and his readers share a question (Why and how did this happen?) as well as background information as CG, which grows as the narrative progresses. Similarly, Antonio Damasio in Descartes’ Error reminds readers at numerous points in the text that he thinks of himself as engaging in conversation with them. We need not take him a speaking metaphorically. Of course, not all readers of these books will be full participants: some will demur from some of the author’s claims and thus have less CG with him than do other readers. This can also happen when a reader’s attention lapses or if the reader has trouble following some of the argument. Because of the potential isolation among readers from one another even for a given work, it seems best to think in terms of CGAR, where ‘A’ and ‘R’ refer to the author and reader respectively, with as many such relations as there are readers of a single work.

A natural next step is to wonder where in this framework fictional discourse fits in. In fact, as with the cases of silencing and muting, it does so with no further modifications required of the CG-context model. For we already have suppositions as permitted moves in any of the six conversation-types mentioned thus far. As a result, we do not need a new subcategory of conversation for works of fiction. Instead, let any such work simply be one of the six already given in which the author presents contents as suppositions, possibly but not necessarily for the sake of argument. We often speak of the “premise” of a novel, and this suggests that at least some novels may also be construed as thought experiments in which the author explores the consequences of what has been supposed: that people change gender on a regular basis (Ursula Le Guin’s The Left Hand of Darkness); that we live in a society exemplifying the principles of hedonic utilitarianism (Aldous Huxley’s Brave New World); that your negligence results in the drowning of a neighbor’s toddler (Jane Hamilton’s A Map of the World), that you fall in love someone who hides atrocities she has committed during wartime (Bernhard Schlink’s, The Reader), and so on). In some stories it can be challenging to discern what is being supposed, and the reader may need to struggle with the question for some time before getting her feet in the unfolding narrative (Green 2016a). By contrast, other stories, such as those written primarily for entertainment rather than instruction, may be seen as asymmetrical didactic inquiries in which most of what the author writes has the force of a supposition that is not for the sake of argument, and thus as not aiming to establish anything outside the scope of the fictional work.

Fictional and non-fictional conversations falling into the asymmetrical-didactic category do differ from one another in one principal respect. An author of a historical narrative is always entitled to “reiterate” a content that is widely acknowledged as fact outside of the CG that has been developed thus far with her audience. Thus Larson may draw on established medical information to explain how long it will be before victims of the Lusitania’s sinking start to experience hypothermia. By contrast, literary genres serve as filters on what can be reiterated, as well as what can be supposed. Stories involving the supernatural, such as Bram Stoker’s Dracula, disallow reiteration of naturalism or any of its implications, though they will usually permit the reiteration of, for instance, facts of human psychology. For a less extreme case, midcentury noir detective stories disallow such challenges as those that ask whether people are really as self-centered and cynical as depicted in novels of this kind. Such a question shows that the person asking it does not fully appreciate the genre.

A conversation driven by a fictional work may answer questions both internal and external to that work. Cormac McCarthy’s The Road asks us to imagine a post-apocalyptic planet in which only a handful of people survive to either scavenge what few scraps of food remain, or descend into cannibalism. Internal questions pursued by the conversation this work generates concern merely what happens in the story, such as whether the father and son who are its protagonists make it to the coast alive. Externally, matters may be more subtle. For while the author of The Road is engaging at a more superficial level in an asymmetrical didactic inquiry with his readers, at a deeper level he may be doing something else. One who reads this book is forced to ask herself whether she would resort to cannibalism to save her own life. As such, the work also functions as a means of asymmetric socratic inquiry. However, since communication between authors and their readers tends to be unidirectional, the answers that readers develop usually do not become part of CG. That limitation need not prevent the author from achieving her or his goal of provoking readers to face hard questions.

My aim in the foregoing has been to evaluate the CG-context approach by asking whether it can be extended with minimal modification to cases that an adequate account of group inquiry and deliberation should be able to illuminate. So long as we are prepared to countenance multiple common grounds for a given set of interlocutors, to partition context sets in ways corresponding to the semantic values of questions, and to distinguish conversations into types along the lines of Table 1 above, the model seems to fare well. It can then characterize while also illuminating such topics as fictional discourse, illocutionary silencing and muting, and pre-illocutionary pragmatic phenomena including disambiguation and embedded implicature. It can also support a more fine-grained explanation of the conditions under which conversational implicata are generated. Finally, although it does not itself inspire such a denial, the CG-context model is compatible with the denial that speaker meaning is the only communicative phenomenon that can make a contribution to CG. It may agree instead that expressive behavior can show an agent’s psychological state without being driven by reflexive communicative intentions, thereby making that state (and not just the behavior expressing it) as vivid as a crash of lightning.