Directing Thought

1. Introduction

In An Inquiry into Meaning and Truth, Russell makes the following observation:¹

In adult life, all speech…is, in intention, in the imperative mood. When it seems to be a mere statement, it should be prefaced by the words ‘know that’. We know many things, and assert only some of them: those that we assert are those that we desire our hearers to know…

It follows that when, in adult life, you use a word, you do so, as a rule, not only because what the word ‘denotes’ is present to sense or imagination, but because you wish your hearer to do something about it…(Russell 1940: 26–27)

In effect, what Russell claims is that assertions—a ‘mere statement’ of fact—are essentially directives: speech acts aimed at changing the rational decision-making behavior of their audience.²

Russell is not remembered for this point. In his book Imperatives, Hamblin says of this passage that it does not represent Russell “at his philosophical best” (1987: 96). To be sure, if there is a version of the above statement—a version of what I will provocatively call Russell’s Thesis—that is defensible, it will clearly have to come in a highly modified form.

My aim in this paper is not to vindicate Russell, per se, but I do think that there is an important insight contained in this passage. In fact, I will argue that there has already been some (inadvertent) vindication of Russell’s claim by work undertaken in the second half of the 20th century, on speech act theory and pragmatics. Specifically, I aim to show how a version of Russell’s Thesis—a modified, but still significant version—is compatible with two widely-endorsed theses concerning language use: Intentionalism about speech acts, which is the view that speech acts are constitutively typed by the intentions with which they’re performed, and Context-first theories of communication, which hold that the function of assertion is to update a body of information shared among conversational participants. The modified version of Russell’s Thesis that I will defend can be more precisely stated as follows:

Russell’s Thesis For an act ϕ and class of characteristics Ω, if Ω makes it the case that ϕ is an assertion, then Ω makes it the case that ϕ is a directive.

Moreover, what I will argue is that Russell’s Thesis is an outcome of a plausible conjunction of Intentionalist and Context-first views that has been defended by Robert Stalnaker.³ Further, two interesting things follow from this: that directives are a more basic/fundamental form of speech than assertions, and that we can understand contributions to a discourse context as under a unified class of normative demands.

To give a quick preview of how this will go: for the Intentionalist, an assertion is an act performed with the intention of informing another agent, whereas a directive is an act performed with the intention of changing the practical, action-guiding states of mind (or directly modifying the actions) of another agent (Bach and Harnish 1979). These practical attitudes and actions are primarily responsive to norms governing practical reason. But the notion of information uptake that constitutes the success conditions of an assertion is controversial. One resolution to this controversy treats the relevant success state as a matter of change to a common body of information, of which conversational participants make use in joint inquiry: the common ground of a conversation (Stalnaker 2002). Joint inquiry does not require mutual belief, but instead something more like mutual acceptance of a body of information. This, I will argue, grounds the relevant notion of information uptake in attitudes and dispositions which are responsive to norms governing practical reason. Assertions are, effectively, directives to act as though something is true.

I would also like to give a quick preview of my theoretical motivations (I will take these up again in earnest in Section 6). While it is interesting and useful to think about what sorts of claims follow from which views, I also think that Russell’s Thesis gets something right about linguistic communication, at least when we think of it as a cooperative endeavor (as on the model of Clark 1996; Clark and Brennan 1991). One way that many have thought about communication is as a kind of joint doing—specifically, joint inquiry. But when we cooperate with one another—even when our aims are epistemic—we directly rely on one another as decision-makers, agents who act in particular ways. And specifically, as individuals who treat certain pieces of available information as true. There is thus an inherently practical dimension to language use, at least when we use it in a certain sort of cooperative conversational setting. Insofar as we are interested in taking such settings as central or fundamental to understanding language use—as many have been interested in doing—we should consider the practical aspect of assertion more carefully.

2. Russell’s Thesis

Russell’s claim in the quoted passage is that the speech we produce in making an assertion has—or is intended to have—an imperatival paraphrase.⁴ On his view, any assertion of the sentence ‘I’m going to the store’ can be interpreted as an order performed with the imperative sentence: ‘Know that I’m going to the store’. Russell’s comments are too brief to know exactly what notion of ‘should’ he had in mind when he says that a statement “should be prefaced” by ‘know that’, but I think that many considerations from natural language semantics tell against the most literal interpretation. (Utterances of) declarative sentences are not literal commands to think or know, and simply do not make the same semantic contributions as imperative sentences.⁵

I’d rather focus on a claim made in the second part of the quoted passage, namely that when you use a word, “you do so, as a rule… because you wish your hearer to do something about it” (emphasis mine). Russell’s Thesis, as I understand it, is that one speech act type can be reduced to another. Or to put this another way: that there is a way of understanding what assertions are, and a way of understanding what directives are, such that assertions are a kind of directive.

Assertions may be very roughly characterized as the type of act we (typically) perform by uttering a declarative sentence, and we might (again, roughly) characterize the point of an assertion as the transfer of information. Directives are the type of act we (typically) perform by uttering an imperative sentence, where the point is to modify an addressee’s rational decision-making behavior. But we can give fuller characterizations along a number of dimensions. These characterizations do not yet commit us to any particular account of either type of speech act.

Speech acts can be characterized along a number of cognitive, normative, and linguistic dimensions.⁶ Let us look at the ones that typically characterize directives and assertions. The cognitive characterization of assertion includes, on the part of the speaker, an intention that their addressee accept what they have said (Grice 1957), and the expression of a cognitive state with that content (Davis 2002); on the part of the addressee it is a kind of taking or treating as true (Stalnaker 2002). Normatively, assertions might be characterized by a demand to speak the truth (Williamson 1996), a commitment in speaking to the truth of what is said (Brandom 1994); and—on the part of the hearer—a warrant to treat what is said as true (Fricker and Cooper 1987), and to hold the speaker so accountable.⁷

Directives are cognitively characterized by an intention that their addressee do something (Bach and Harnish 1979), and the expression of a pro-attitude (Condoravdi and Lauer 2012), on the part of the speaker; on the part of the addressee it is the formation of an attitude directed at acting accordingly—cashed out in terms of things like preferences (Starr 2020), intentions (Harris 2021), and other practical attitudes. Normatively, directives might be characterized by authority relations (Lance and Kukla 2013; Lewis 1979b), such that you can only correctly command someone to do something if you occupy a certain social position, and given that you do, there is a demand on the hearer to do as they are told.⁸

Speech act types can also be characterized along more straightforwardly linguistic dimensions, owing to their common association with certain clause types. Assertions are canonically associated with utterances of declarative sentences, whereas directives are typically performed by utterances of imperative sentences.⁹

The version of Russell’s Thesis that I will consider holds that whatever class of characteristics makes it the case that some act ϕ is an act of assertion—whether this be attitude expression, speaker’s intention, normative features of the act, or some combination thereof—belongs to the class of characteristics that makes it the case that some act ϕ is a directive:¹⁰

Russell’s Thesis For an act ϕ and class of characteristics Ω, if Ω makes it the case that ϕ is an assertion, then Ω makes it the case that ϕ is a directive.¹¹

Given the aforementioned general characterizations of assertions and directives, it might seem puzzling that anyone should think of asserting as a particular kind of directing. How, for instance, could the expression of a belief be reduced to the expression of a preference, or intention? But such characterizations are just that: characterizations—they do not provide us with the constitutive features of these act types. It is not my intention in this paper to suggest that there is nothing meaningful about the distinction between asserting something p and directing someone to do something ϕ. Further, to say that there is some sense in which directives are more basic than assertions is not to deny that we should theorize about assertions qua their properties as assertoric acts.

Why, then, is it significant if Russell’s Thesis is true? Here are two reasons. First, it is very common to think of assertions—and the transfer of information—as basic, and to treat all other kinds of language use as derivative, in some sense, on assertions of declarative sentences; but if Russell’s Thesis is right, then we have to understand directives as more basic than other speech acts. What’s more, and as will become clear, we will have to think of the transfer of information not as the role or function of assertion, but as a kind of side effect of an activity that is meant to shape joint practical commitments. Whatever else might constitutively characterize assertion, it is significant if the point of asserting is to shape practical commitments and attitudes.

Second, the normative demands on assertion are generally thought to be epistemic, rather than practical (cf. Williamson 1996); but if assertions are a kind of directive, then we should perhaps understand these acts as being under a unified class of normative demands.¹² This can help us come to a better understanding of the social function of our assertoric practice. I will return to this in Section 6.

3. Intentionalism about Speech Acts

According to the Intentionalist, an act ϕ counts as a speech act of type T in virtue of the intentions with which ϕ was performed.¹³ Certain things—like normative demands, or expressive aspect—may come to be associated with any given speech act, but this will be because they are downstream consequences of the act’s constitutive features, or the result of social pressures placed on our performance of such acts in specific situations (Schiller 2021b).

Why think that speech acts should be classified according to the intentions with which they’re performed? One reason is that if you hold fixed what a speaker intends in saying something, but vary all other contextually-relevant features of their act, then you will not have changed the communicative significance of the act performed (Buchanan and Schiller 2022).¹⁴ Further, it is natural to think of communicative success as having occurred if and only if the speaker’s intentions have been appropriately recovered and acted on. And there is a powerful folk-theoretic notion of what we mean by uttering what we do that appears to be explicable in terms of judgments about speaker intentions and intentional action (Grice 1957). Intentional action also underwrites responsibility, and thus intentionalism offers a plausible connection between assertion and its norms (Buchanan and Schiller in press).

Intentionalism is also supported by empirical work concerning ordinary speakers’ judgments about, for example, linguistic commitments (Bonalumi et al. 2020), as well as a plausible story about the evolution of communication (Scott-Phillips et al. 2009; Scott-Phillips 2010), and various aspects of its psychological import (cf. Zawidzki 2008). But even if none of this were the case, Intentionalism would be worth discussing because of its widespread acceptance: Intentionalism about speech acts underwrites many mainstream theories of linguistic communication, and there are Intentionalist accounts of a variety of illocutionary phenomena including illocutionary silencing (Hesni 2018; Unnsteinsson 2019), lying (Saul 2012), and threatening (Schiller 2021a).

It is also worth noting that even though I cash out my proposal in terms of a particular theory of speech acts that takes speaker intentions as foundational, any theory will have to contend with the important role that speaker intention—and uptake—plays in linguistic communication. The upshots of this discussion are thus by no means limited to intentionalist theories about speech acts; rather, they ought to be considered on any theory that hopes to contend with how we use language to do things with others.¹⁵

The locus classicus for Intentionalist accounts of speech act types is Bach and Harnish (1979). Based on an account of meaning set out by Grice (1957) (and later developed in Schiffer 1972), and a taxonomical project undertaken by Austin (1962) and Searle (1969), Bach and Harnish provide a taxonomy for all the major speech act classifications. Here I will focus on assertions and directives.

On the view developed by Bach and Harnish, assertions are characteristically attempts to inform one’s addressee of something, where this attempt is constituted by an intention that one’s addressee update their information state (i.e., form a belief) on the basis of recognizing (a) the speaker’s very intention that they do so, and (b) some feature of the conversational context (here very broadly construed) that gives them a reason to make the intended update. Clause (b) is very often specified in terms of recognizing the speaker’s status as informed,¹⁶ but may even just consist in the recognition that they have asserted what they did; in this formulation, we might call this the expressive condition (I will discuss this again in Section 5.2).

Assertion S asserts that p to H, in uttering e iff S reflexively intends:

• (a) H to [information update] with p;

• (b) for (a) to be based in part on a contextually salient reason in support of [information update] with p.

Directives are given a disjunctive characterization by Bach and Harnish. Their class of directives includes requests, commands, questions, prohibitions, permissions, and advice (Bach and Harnish 1979: 47–48). I will, however, pursue the most general possible account.

Performing a directive requires “intending to bring about a non-cognitive, connotative, or action-guiding psychological state of some kind” (Harris 2014: 11–12). Directives, then, are typically characterized by an intention that the addressee adjust their behavior towards some prospective action, and that they do so because the speaker revealed this intention, in combination with background evidence.¹⁷ More specifically, directives are characteristically attempts to bring it about that one’s addressee does something ϕ, where this attempt is further characterized by an intention that one’s addressee’s basis for deciding to ϕ be (a*) recognizing the speaker’s very intention that they do so, and (b*) recognizing any relevant background conditions that favor their doing ϕ.

Direction S directs H to ϕ, in uttering e iff S reflexively intends:

• (a*) H to [do something] ϕ;

• (b*) on the basis of a salient background reason favoring ϕ.

I think we can already see what it would be for assertion to collapse into direction: first, if updating on the information that p falls under the general rubric of doing something ϕ. And second, but less centrally, if (b) can be explained in terms of (b*).¹⁸ What would this take? First we need to see that [doing something] is specified in terms that are specific enough to make Russell’s Thesis significant, and not simply trivial. For instance, if any mental process is a way of [doing something], then assertion will collapse into direction, but this would be a trivial thesis, and moreover it would be an account of directives that does not tell us much of anything about a specific type of speech act.

Under what specification of [doing something] is direction a significant and plausible thesis about a speech act type? On the view I’ll accept, doing something—in the relevant sense—is acting on the basis of practical rather than epistemic reasons.¹⁹ Practical reason refers to “rational standards that apply directly to conduct or to deliberation” (Korsgaard 1986: 5), and in this paper I will focus specifically on instrumental—or means-end—reasons for action. An agent gets a new instrumental reason to ϕ if they learn that ϕ-ing promotes some aim of theirs, or is a means to their ends. Learning that ϕ is permitted, required, or the course of action preferred by an authority, as the case may be, could serve as a reason for doing ϕ, given one’s background desires and preferences (to do only what is allowed, to do what is best, to avoid punishment, etc.).

By contrast, epistemic rationality refers to our capacity for our mental states to accurately reflect the world. As Joseph Raz explains the difference,

Epistemic reasons are governed by one concern: determination whether the belief for which they are reasons is or is not true. Reasons for a single action may, and typically are, governed by many concerns. A single action can serve or disserve a number of intrinsic values: It may be an act of friendship and of justice. (Raz 2009: 5–6)

The simplest construal, then, of the (putative) difference between assertions and directives is as follows: we can ask if a speaker aims to change their addressee’s behavior, or if they simply aim to change their beliefs. But here is a more sophisticated construal: does the speaker aim to put their addressee in a state that is correct to be in because it is accurate, or do they aim to put their addressee in a state that is correct to be in because being in that state furthers their ends in some way?

Consider the difference between having a reason to believe that p, and having a reason to act as though p is true.²⁰ These states are guided by different sorts of considerations: the former depends on reasons for thinking that p is true, whereas the latter does not. It might be that someone has a strong reason to act as though p is true, but the only way they are capable of doing that is to believe p. They thus have a (practical) reason to believe that p.²¹ There are practical reasons for belief, but belief is not characteristically responsive to practical reason: that believing p would be beneficial to you (perhaps it would result in you getting some much desired goods) is not an appropriate reason on which to base your belief that p.²² But such a reason does not create a rational basis for the belief that p—that is, what they apparently have reason to do is induce a belief that p. But the belief will not have a rational basis in, for instance, evidence (Feldman and Conee 1985) or a reliable belief-forming process (Burge 1993).

More generally, we can draw a distinction between reasons for attitudes qua the rational basis for an attitude, and the reasons (more broadly construed) for which an attitude might be formed. Forming attitudes on the basis of practical reasons might be justified in the course of inquiry that has epistemic aims. The clearest examples of this are making a supposition in the course of an argument by reductio. There might, in an attenuated sense, be an epistemic reason to accept the thing you suppose for reductio. This is because you are interested in learning something. But we need to distinguish such epistemic aims from the sorts of epistemic reasons we’re interested in here. When your aim is epistemic, your basis for acceptance might not be (and vice versa). This will be important for what comes later: it might be thought that inquiry always has an epistemic aim, but this does not mean that the attitudes fostered by inquiry are always epistemically rational ones (thanks to an anonymous referee for pushing me to say more here).

Another possibility is to think about what makes each of these states rationally apt. A belief is apt just in case it is true (i.e., just in case the representational content of the belief is true). The truth of something may be what makes an attitude of acceptance apt, just as the truth of something might be what makes a desire apt (the fact that p is good), but it is not the truth of the content p that is accepted, or the evaluation of the attitude as true, that makes it apt. A belief state is apt just in case that belief is true, but a state of accepting something p is not apt just in case p is true or that state of acceptance is true (if that is a distinction). We judge states of acceptance as appropriate on entirely different grounds.

Is this the only way of spelling out direction? No, I don’t think so. But if a theory of assertion should collapse into direction spelled out as such, that will be pretty significant.²³ That is, I take it that it would be significant to show that assertions can be conceived as acts undertaken to promote the formation of attitudes based on practical reason.

How specific do we have to be, at the outset, about the aims of these speech acts, qua the attitudes these acts are aimed at modifying? This will depend on the corresponding theory of communication that we accept. That is to say: the theory of illocutionary action depends on a corresponding theory of illocutionary success. We turn to such a theory now.

4. Communication and Acceptance

Russell’s Thesis is false if [information update] is a cognitive change characterized by responsiveness to epistemic reasons. I introduced the Intentionalist account of assertion in terms of the ambiguous notion of making someone ‘more informed’. But why not put this in terms of belief? After all, that’s what both Grice (1957) and Bach and Harnish (1979) do, not to mention many others. Call Griceanism the view just articulated, on which [information update] just is belief formation. Griceanism falsifies Russell’s Thesis because beliefs are characteristically subject to norms of epistemic rationality.

But there is reason for doubting that belief—rather than some practical attitude—is the target of assertion. One elaboration of the Gricean picture holds that the essential effect of an assertion is to change the context of a conversation.²⁴ Such accounts of communication originate with Stalnaker (1978; 2002; 2014) and Lewis (1979a), and the position that conversation is a collaborative process is defended in psycholinguistics as well (Clark 1996; Clark and Brennan 1991).²⁵ The basic idea is something like the following: conversations are goal-directed activities of joint inquiry, and the function of assertion is to contribute to an information register of which we avail ourselves in collaborating on a common question.²⁶ Call this set of information the conversational common ground.

Stalnaker points out that we can acknowledge two different kinds of change to the common ground (see Stalnaker 2014: ch. 1). On one hand, there are changes in response to a manifest event: a goat walks into the room, and we all notice (to borrow his example). On the other hand, there are changes in response to an utterance, corresponding to acknowledged linguistic rules and conventions. At some level, these changes are the same, insofar as they both consist in—or depend on—mutual awareness of a fact: for instance the fact that a goat walked into the room, and the fact that someone said something. But there is a difference between mutual awareness or recognition of p, and the further step of adopting p as part of the information relevant to our conversation.²⁷ Perhaps we are even mutually aware of things that we fail to take for granted for the purpose of our communicative exchange (more anon).

As Stalnaker notes, the way we repair defective conversations suggests that the common ground cannot just be participants’ mutual beliefs:

Alice, talking to Bob at a cocktail party, says, “the man drinking a martini is a philosopher”, intending to refer to a man that Bob knows is in fact drinking Perrier from a cocktail glass…[If] the false presupposition is irrelevant to the purposes of the conversation (The man’s drinking habits are not at issue—the reference to the alleged martini was just a way to identify him.), Bob might decide to ignore the matter, tacitly accepting what Alice is manifestly presupposing for the purpose of facilitating communication without disrupting the conversation with a distracting correction. That is, Bob accommodates, not by coming to believe the false proposition that Alice is presupposing, but by accepting it as part of the common ground. (2002: 717–18)

A proposition p is common ground among conversational participants just in case it is mutually accepted. Stalnaker defines mutual acceptance in two different ways: first, in terms of it being mutually believed that p is jointly accepted.²⁸ Belief is thus still implicated, but it is only implicated indirectly (note that belief is also implicated indirectly in ordinary practical reason). Later, he modifies this definition: something p is mutually accepted if it is mutually accepted that p is jointly accepted (accepted by all). A proposition p is common ground between a and b iff a accepts p, b accepts p, a accepts that b accepts p, b accepts that a accepts p, a accepts that b accepts that a accepts p, …

The point of any assertion is to change what is accepted.²⁹ This departure from the Gricean model is crucial to understanding the implications of the Stalnakerian account of assertion for Russell’s Thesis. This is because acceptance for the purpose of a conversation meets our basic characterization of [doing something]: it is a practical state that is correct to be in because being in that state furthers one’s ends in some way. The above example demonstrates this: that it will make their conversation go smoother is the basis on which Bob rationally accepts that the man in the corner is drinking a martini. This is not a proposition Bob has any (epistemic) reason to believe—rather it is a proposition that it is instrumentally rational for Bob to pretend (so to speak) is true. In other words, Stalnaker’s conception of context suggests that we should take conversational contexts to depend on—and linguistic interventions to be attempts at modifying—practical states of mind.

Even discourse that is aimed at the truth will be sustained by such practical attitudes, thus conceived. It is no mystery that practical reason should play some role in achieving epistemic aims. Sherlock Holmes’ aim is to figure out who committed the murder, but he knows that he can only think clearly if he takes some of his beloved snuff. Does Holmes thereby have an epistemic reason to get high? We could talk that way, but the point is that his doing it is something that makes practical, instrumental, sense because he has a goal which just happens to be associated with an epistemic achievement. Likewise, even in cases where the point of an inquiry is to learn something about the world, we might accept information that we believe to be false pursuant to that truth.

But this might be a bit quick: after all, Alice’s point in making her utterance was just to get across that he is a philosopher, and that is something Bob does come to believe. In other words, while it is true that Bob and Alice mutually presuppose something patently false, which neither may believe—that someone is drinking a martini—this is ultimately in service of conveying a belief. But is not my aim to deny that we often try to convey beliefs through speech. The point is that belief conveyance as such is not how we construct or sustain conversations.

Belief can play an important role in fixing what we accept. As Stalnaker notes, “the simplest reason to treat a proposition as true is that one believes that it is true” (Stalnaker 2002: 716). But as this quote suggests, we should think of acceptance as something we do over and above belief—something for which belief may be a reason.³⁰ Something’s being mutually believed is not sufficient for it to be part of the common ground: when a goat walks into the room we can choose to ignore that for the purpose of the conversation—we can choose to act as though the things we mutually recognize as true are not true. Even when a conversation is aimed at the truth, and we are aimed at building mutual awareness of the facts, the conversation itself is moved forward and sustained by practical attitudes:³¹

Given only what is common ground among a group of agents, one does not yet know how the agents of the context mutually regard the propositions in the common ground with respect to their other cognitive attitudes. To be given the common ground is only to be given a set of propositions mutually understood to be presupposed; it is not yet to be given that the agents also regard those presuppositions as knowledge, or as warranted belief, or conjecture, or fiction, or whatever. (Yalcin 2007: 1008)

That is to say: of course we often care very much what sorts of attitudes our interlocutors ultimately take to our assertions, and in many instances it may be mutually presupposed that the common ground—our mutual conversational presuppositions—represent doxastic commitments, what Yalcin calls ‘conversational tone’ (Yalcin 2007). But in such cases, what is accepted for the purpose of the conversation will simply be governed by the fact that it is instrumentally rational to accept only what is epistemically viable. In general, what we take on board for the purpose of a conversation is governed by such instrumental reasoning.

What determines the success of an assertion in a conversational context is, first and foremost, that our addressee operate under the assumption that what we’ve said is true. If I tell you “I’m disrespected by my colleagues” and you invite me to say more by asking what their disrespect consists in, then my assertion has succeeded, even if you privately believe that I am being dramatic. You go along with it out of pity for me: it seems prudent to you to do so. We sometimes don’t even particularly care if we’re believed: if you suggestively say to me “I’m quite hungry”, and I say in response, “Would you like a sandwich?”, and then go about making one for you, then your assertion has succeeded, and you’ve gotten what you wanted, whether or not I believe that you really are hungry. In fact suppose I don’t; I may nevertheless go along with your claim because I am deeply interested in what you will do with the sandwich I make you.³²

There are two things worth noting before we move on. The first is that this view makes no assumptions or claims about the cognitive significance of acceptance. Perhaps it is often the case that we accept something for the purpose of a conversation in virtue of the fact that we believe that thing—as Stalnaker (2002) notes, that you believe p is often a good reason to accept it. But we can add a substantial caveat to this: that you believe p is a good reason to accept it just given that the aim of your inquiry makes it so. There are many plausible instances in which believing something is not sufficient reason to accept it, such as when you are assuming a body of information you don’t believe in the course of a reductio, or—to borrow an example from Bratman 1992—when your belief that your squadron will not survive the night would get in the way of your planning tomorrow’s offensive maneuver.

The second thing worth noting is that shared bodies of information—however they are mediated, by whatever sorts of norms—are still informational structures, and are thus still subject to the structural constraints that govern any body of information. Insofar as our behavior can be characterized by appeal to a body of information, there will be some sense in which our behavior is only reasonable if the body of information that explains that behavior evolves in a rational way.

5. Russell’s Thesis: Revisited and Resisted

If we combine Intentionalism about speech acts with a Context-first account of communication, then we have the following account of assertion: an agent asserts that p just in case they act with the intention of bringing it about that p is accepted for the purpose of the conversation (and that it is done so by recognition of this very intention):

Assertion* S asserts that p to H, in uttering e, iff S reflexively intends:

• (a) H to accept p for the purpose of the conversation;

• (b) for (a) to be based in part on a contextually salient reason in support of accepting p.

Remember that the Intentionalist account of directives articulated above stated that:

Direction S directs H to ϕ, in uttering e, iff S reflexively intends:

• (a*) H to [do something] ϕ;

• (b*) on the basis of a salient background reason favoring ϕ.

We gave the following characterization of [doing something] ϕ: the object of a directive intention is that one’s addressee modify their behavior or mentation in a way that is practically rational, subject to the demands of practical reason. Russell’s Thesis was that the facts that ground an action’s status as an assertion ground its status as a directive as well. If assertion* is correct, then Russell’s Thesis is true.

6. What’s the Point of Assertion?

Context-first accounts of communication in general—and the theoretical notion of a common ground, specifically—can be understood as capturing a distinct feature of our communicative practice: how we use language to coordinate on a task. This task could be doing something, or it could be figuring something out. In either case, coordinated behavior, like all behavior, is rationalized by a body of information. Thus, coordinating on a task apparently demands of us that we share a set of beliefs, or even a body of knowledge, as the case may be. But what this ‘sharing’ ultimately amounts to in a situation where multiple agents are coordinated on some task is that we rely on one another to act in certain ways. I think it is the kind of behavioral intervention that we make on one another as cooperators that the Stalnakerian notion of assertion has something to tell us about.

Even if belief and knowledge are ultimately what underwrite the rationality of our joint action, we don’t directly rely on one another as believers and knowers: we rely on one another as actors. Consider a larger scale cooperative endeavor to see that this is true. A shell of rowers being piloted through dangerous waters may face a decision about whether to veer left or right to avoid an obstacle. Either action can only be accomplished if all the rowers work in unison. The rationality of this choice depends on knowledge of the world—on knowing the facts that make left or right the correct way to go—and so the rowers depend on the coxswain’s, or whoever is at the front of the boat’s, knowledge in jointly making this decision.

But in actually acting—in veering left to avoid some debris—the rowers rely on one another, and are relied on by the coxswain, to act as though it is true that there is some debris that makes veering right more dangerous. Knowledge—specifically the coxswain’s—underwrites the whole endeavor, to be sure, but that knowledge needn’t be shared by the rowers as knowledge in order for their cooperative action to be rational and to be rationally based on that knowledge. At a large enough scale, the expectation of common knowledge becomes unreasonable.

Further, that someone believes something p is no guarantee that you can rely on them as a believer of p. You need some extra bit of assurance: assurance that they not only believe p, but that they will act in this situation as though p. If I know the rower next to me believes the debris is spread out to the right, but that he plans to act as though it is spread out to the left—that is, plans to make himself dependable to others as a believer of the proposition that the debris is spread out to the left—then his actual belief is meaningless to our task.

Assertion is a way of signaling that you can be relied on to treat something as true. And the point of an assertion is to foster that reliability in one’s interlocutors as well. It may generally (or typically, or just often enough) only be rational to do this if that something is known by you to be true. But it’s clear that this will depend on the task itself. Maybe for certain types of tasks, like scientific inquiry, this demand will be general enough to even make salient a default expectation, or normative demand, of knowledge on assertion. Though as Mandelkern and Dorst (in press) note, such cases do not constitute anything like a general class. The point, however, is that on the view presented in this paper, there is no speech act—no natural kind—that can be appropriately generally characterized in this way.

That said, the Stalnakerian account does not give us the whole story about assertion. In fact what it tells us might be an extremely limited, perhaps quite provincial, story about one particular part of our communicative practice: namely, how to think about communicative settings that are genuinely instances of joint inquiry.³⁷ There are many who have wanted to treat conversation as a kind of activity—inquiry—and as Roberts (1996) observes, such discursive activity appears to be a human universal. We do learn a lot from thinking about the way we act so as to contribute to such social activities. This is true even if changing the context is “a secondary and inessential aim of communicative acts” (Harris 2020: 2–3).

That said, we do sometimes speak with the aim of getting others to believe what we have said.³⁸ I care very much if my friends believe me, for instance, when I tell them that I care for them. I may not, after all, be trying to get them to do something for me. Such an effect is perhaps not constitutive of what assertion is, but it is still an important one. Thus, there are plenty of distinctions still to be made between our capacity to inform and our capacity to direct action. It is just that these distinctions are not to be made at the level at which our intentions are revealed in order to participate in a cooperative activity.

Just to recap: I’ve argued that if we take seriously two claims about communication, a significant reductive thesis follows. The first claim is that speech acts can be identified by a (characteristic) intention to bring about a change in an addressee’s mental state. The second claim is about the mental states relevant to our communicative aims: namely, that these are mental states that come roughly under one set of rational demands. Maybe this is just reason to reject the Stalnakerian thesis.³⁹ I’m sympathetic to this to the extent that I think there are worthwhile taxonomical projects that do not display the reduction I have identified. But when it comes to the ebb and flow of human conversation, we engage with one another not just as language users but as rational facilitators of a joint practical endeavor, and this is something that the Stalnakerian picture may do a good job of capturing.

Acknowledgements

For feedback on earlier drafts that helped to improve the paper I thank Sam Berstler, Andy Egan, Matt Mandelkern, and two anonymous referees for this journal. I would also like to thank Ray Buchanan, Jake Quilty-Dunn, an audience at the 2023 Central APA (especially Daniel Harris, Peter van Elswyk, and Evan Westra), and participants in the Words Workshop (especially Nikki Ernst, Jonathan Jenkins Ichikawa, and Audrey Yap) for helpful discussion of this material.