1 Cultural Evolution: Basic Concepts

The core idea of cultural evolution (or CE in what follows), as traditionally propounded, is that behavior patterns in a social group may be accounted for in non-genetic terms as being adaptations to that group’s environment.Footnote 1 Such patterns are transmitted by means of learning rather than genetically. As such the transmission process may run through parents (vertical), but also through teachers, mentors, and other high-status community members (oblique), as well as through peers (horizontal), as opposed to the entirely vertical transmission found genetic evolution (Creanza et al. 2017). Both the “teaching” and “learning” processes emphasized in CE might be done implicitly, so that the “teacher” might not be intending to convey information,Footnote 2 and the “learner” might acquire new information or skills without trying to or realizing that she is doing so. Also, although the explanation is in non-genetic terms, the phenomena to be explained might interact in interesting ways with genetic changes. That is part of the story about, for instance, the evolution of lactose tolerance among Western adults (Ibid, p. 7783).

To be explained in CE terms, the behaviors thus transmitted must give the community in which they propagate a survival advantage over other communities that are otherwise similar including earlier versions of that same community. CE would accordingly offer explanations of such human practices as sophisticated hunting techniques and the construction of tools. It could also provide explanations of patterns of behavior not essentially bound up with artifacts such as incest taboos and conversational turn-taking. Either way it might account for why anatomically modern humans survived over the last 100,000 years while, say, Neanderthals did not. It might also account for why certain human groups have been more successful than others as measured by population, material culture, or robustness against disease or environmental disruption.Footnote 3

The notion of behavior at issue in discussion of CE should be construed broadly so as to include institutions such as marriage, double-entry bookkeeping, and coming-of-age rites. It is not restricted to behaviors in the sense of bodily movements. Heyes (2018) argues that not just behaviors but also cognitive mechanisms should be accounted for as results of cultural evolution. We do not need to take a stand on whether Heyes is right about this, since neither behavior nor cognitive mechanism is exactly the right category for what we wish to explain. Social institutions presumably supervene both upon patterns of observable behavior and psychological phenomena such as cognitive and affective states.

Cultural evolution tends to be cumulative, so that one innovation can build on others that have preceded and enabled it, and such an innovation might occur in the same generation as the one on which it builds. The result is that innovations might spread fast, leading to abrupt changes over relatively short periods of time. (Such changes are also possible in genetic evolution but relatively rare.) From the point of view of CE, substantial changes can occur in just a few generations.Footnote 4

In this light, we may observe that CE offers explanations of behavior patterns that are nonconventional due to being superior to other feasible alternatives. Stone tools are outperformed by bronze, which is in turn outperformed by iron, and so forth. However, CE is also equipped to offer at least partial explanations of behaviors and institutions that are conventional. We have room for at least partial explanations of conventions in light of the fact that many behavior patterns may be seen as Nash equilibria in situations in which more than one such equilibrium is available. When we have multiple Nash equilibria, CE might explain the fact that one of these needs to be chosen. However, it won’t by itself account for how one equilibrium is in fact chosen. Instead, it will need to appeal to an extraneous posit such as an intelligent agent realizing that one of these options needs to be chosen, and settling on one for reasons not pertaining to its intrinsic superiority. That choice might be due to a natural event that by chance leads an agent to do things in one way rather than another that might have been equally viable.

2 Cultural Evolution and Speech Act Theory

As I will use the notion here, a speech act is characterized partly in terms of the notion of speaker meaning introduced by Grice (1989). Grice propounded an analysis of speaker meaning for the sake of elucidating the phenomenon of a speaker meaning something by performing a certain action; with modifications his approach is useful for understanding speech acts as well.Footnote 5 In Grice’s formulation, an agent A who speaker means that P does something with an objective of producing a cognitive effect on an addressee by means at least in part of the addressee’s recognition of A’s objective.Footnote 6 The necessity of these conditions has been challenged in light of cases in which A acts overtly but without any aim that her addressee come to the intended psychological state by means of recognition of A’s intention (Schiffer 1972, p. 42). For while some speech acts such as telling might require that this condition be met, it is doubtful that all speech acts require this.Footnote 7 For instance, a judge in a court of law who declares an attorney’s objection ‘over-ruled’ need not concern herself with what others in the courtroom come to believe about the attorney or how they do so: the attorney’s objection is simply over-ruled by virtue of the judge’s words. Instead, as we understand the notion here, then, an agent A who speaker means that P does something with an objective of making an aspect of her commitment overt: intended to be recognizable (but not necessarily recognized) as intended. Distinct speech acts may then be differentiated from one another in terms of the kind of commitment made overt.Footnote 8

It has also been observed that everyday conversation is characterized by collaboration and coordination between speakers and their addressees. This occurs at a more mundane level as an addressee helps finish a speaker’s sentence or provides a word that the speaker was struggling to recall (Geurts 2022). A speaker might likewise sense an addressee’s reaction and accordingly modulate her utterance into a conjecture rather than an assertion or a promise rather than a statement of intention to perform a future action. One might be tempted to see these phenomena as undermining an approach to speech acts in terms of objectives. Yet the fact that an agent might modify what she does in light of what she observes others intending or doing, does not make any less intentional her resulting behavior. As a driver I switch lanes in a way that is sensitive to what other drivers are doing; this does not make what I do any less subject to my objectives. So too, as a speaker, what I say is usually sensitive in part to what I think will work in the particular social milieu in which I find myself. The resulting utterance is no less intentional for all that.Footnote 9

With the notion of speaker meaning clarified, let us say that a speech (or illocutionary) act is any act that may (but need not) be performed by a speaker saying and speaker meaning that she is doing so. ‘May (but need not)’ because one can assert without saying that one is doing so, just as one can warn someone of danger without saying that one is doing so.Footnote 10 (One might just draw their attention to the danger with a gesture and a terrified face.) By contrast, an act of speech will be an act of uttering a meaningful sentence or phrase. Accordingly, an act of speech (equivalent to Austin’s notion of a locutionary act) is not as such a speech act. One might utter a meaningful sentence in one’s sleep or in the course of rehearsing lines from a play without performing any speech act.Footnote 11 So too, the effect that a speech act might adventitiously or characteristically have on an audience member need not itself be a speech act. Thus, convincing someone of a proposition might result from a series of assertions I make in the course of an argument, but convincing is not a speech act according to our elucidation above. Likewise for embarrassing, shaming, boring, or irritating an addressee.

Let us note also that what speech acts are available to speakers will likely vary from one culture to another. Peaceful cultures may have no need for declarations of war, and cultures that do not fuss over the nuances of argumentation may have no need for presumptions or suppositions for the sake of argument. More broadly, our approach will attempt to see speech act norms as products of cultural evolution, with the understanding that different cultures may produce different speech act practices as ways of coping with their various ecological, political, etc., situations. Adopting this approach will enable us to see speech act norms as having an adaptive role to play in and for the cultures in which they occur. The approach also enables us to see such norms as propagated over time through learning (whether implicit or more self-conscious) rather than by genetic means. In this way, speech acts (or at least their non-syntactic aspects) are accounted for in terms quite different from the dominant “Universal Grammar” approach to the evolution of syntactical competence.

What sort of adaptive function might speech acts play? Answering this question requires considering distinct levels of communicative sophistication. Even before language enters the scene, organisms engage in communication in the sense of designed transmission of information, where the design at issue may but need not be connected to intention.Footnote 12 Accordingly, just as we may say that the mammalian heart is designed to pump blood, and human skin is designed inter alia for thermoregulation, some behaviors and adaptations are designed for the transmission of information. For instance, many organisms are toxic, and often that toxicity functions to deter predation. In some cases, known as aposematism, toxic organisms advertise their toxicity by possessing traits making them perceptually salient. For instance, the bright coloration of the Amazonian tree frog is likely designed for the job of transmitting the information that it is toxic (Ruxton et al. 2018).

A frog might be born into a population of brightly colored, toxic kin with a mutation causing it to share that bright coloration with them but lack their toxicity. The bright coloration is still designed to convey the information that its possessor is toxic. We might say that the organism presents itself as toxic although it is not, just as an analog clock that has stopped at 3 pm presents the time as being 3 pm even when it is not 3 pm. The clock is correct when it is 3 pm, and incorrect at all other times. By contrast the mutant tree frog is chronically incorrect on the issue of its toxicity. More generally, we may formulate a principle relating information, design, and correctness:

Information, Design, and Correctness (IDC) Principle

When an artifact, biological trait, or behavior D is designed (by natural selection, cultural evolution, or an agent’s intentions) to convey the information that P, and P holds, then D is correct; otherwise D is incorrect.

The IDC Principle rules the mutant tree frog as chronically incorrect even though its mutation gives it an advantage over its kin. This is due to the fact that poison is costly to produce and requires internal defenses to prevent self-poisoning. These advantages make the mutant more likely to outlive and thus out-reproduce its siblings. Over time, its offspring, which may inherit its bright coloration and non-toxicity, stand a chance of swamping the population of tree frogs in the area. If this occurs, it is also likely that predators such as snakes and birds will start to ignore coloration and incorporate a new kind of reptile into their diet. This will in turn result in the signaling system collapsing: bright coloration in this species of tree frogs will no longer mean, “I’m toxic”.

Theorizing about communication from a (not necessarily cultural) evolutionary point of view helps us to raise questions about the stability over time of signaling systems. We will see below that some of these questions carry over to discussion of communication in the context of cultural evolution as well. Also, we would expect greater communicative sophistication when behavior is subject to agential control. To illustrate the situation, let us imagine a communicative system, Vervish, based loosely upon the well-studied system of alarm calls among vervet monkeys (Cheney and Seyfarth 1996). These creatures produce different calls for different types of predator such as leopards, snakes and martial eagles. In response to a vervet producing one of these alarm calls, other vervets will respond appropriately, such as by scampering to high branches in a nearby tree for the “leopard” call, or hiding in the underbrush in response to a “martial eagle” call.

One approach to interpreting the above system of communication is to idealize it in such a way as to suppose that the vervets have a limited vocabulary and syntax permitting only monadic predication. Their communication system does not have the means for recursive operations that would yield the complexity that we find in the great majority of human natural languages. We may nevertheless imagine that in addition to four noun-like terms for different kinds of predator a (raptor), b (leopard), and c (snake), and d (also meaning raptor, and thus a synonym of a), Vervish also possesses four predicate-like terms for characterizing them: F (large), G (small), H (near) and K (far). Concatenations of such expressions will produce sixteen possible Vervish expressions possessing truth conditions such as the following:

‘Fa’ is true iff there is a large raptor;

‘Kc’ is true iff there is a distant snake;

‘Gb’ is true iff there is a small leopard.

Vervish contains no connectives, tense markings, evidentials, or other grammatical accoutrements. It is thus an impoverished system for communication, and we may refuse to call Vervish a language due its lack of recursive syntax.Footnote 13 Also, while our experience with many modern languages might lead us to expect different lexical items such as ‘F’ and ‘c’ to be tokened sequentially when ‘Fc’ is tokened, that expectation need not be realized in Vervish. Instead, these different lexical items might appear as distinct aspects of an utterance rather than distinct components thereof.Footnote 14

In spite of its limitations I propose Vervish as a candidate for what is sometimes called a “protolanguage” (Maynard Smith and Harper 2004). It will serve our purposes by enabling us to pose questions about the evolution of speech acts. Imagine, then, an ancient primate uttering a Vervish sentence such as ‘Hb’ with an objective of conveying the information corresponding to the truth conditions it encodes. This utterance, that is, is designed to convey the information encoded by ‘Hb’.Footnote 15 Because, further, ‘Hb’ encodes information that may be formulated in propositional terms, a tokening of that expression stands to be true or false depending on how things are, namely depending on whether there is a nearby leopard. Accordingly, and given the IDC Principle, in uttering ‘Hb’ the primate is liable to be correct or incorrect on the issue of the presence of a nearby leopard depending on whether there is indeed a leopard nearby.

Even though the speaker will be correct or incorrect on the issue that her utterance’s content represents, this does not entail that her utterance is also an assertion. The reason is that at least as the term is used in contemporary philosophy of language and epistemology, assertion requires more sophistication than our speakers of Vervish are likely to be able to muster. For instance, assertions are expected to be sincere in the sense that one who asserts that P is to believe that P. However, while it is true that a speaker of Vervish is likely to believe that Hb when it utters ‘Hb’, that is not to say that she either is or takes herself to be subject to any norm relating utterance to belief.

3 Potential Precursors to Modern Speech Acts

The speaker’s utterance of ‘Hb’ is more than a locutionary act, however. After all, the object of the utterance is to convey information, and thereby perhaps as well to affect the behavior of conspecifics. These are not features we find in locutionary acts as such, since a locutionary act is defined only as the utterance of a semantically contentful expression. How then might we conceptualize a typical utterance of ‘Hb’ as produced by speakers of Vervish? I suggest that we may do so by disentangling different strands running through the concept of commitment as it is used in speech act theory. As Green argues (2016, 2020a), commitment as it applies to assertion and like acts may be broken down into the distinct notions of liability, frankness, and fidelity. The first of these is the concept we have encountered previously. The second is just another term for sincerity, while the third, fidelity, concerns a speaker’s readiness to back up a claim if challenged to give reasons on its behalf. To elaborate on this third condition, observe that a consensus has coalesced around the view that one who asserts that P is expected to provide reasons in support of that claim if presented with an appropriate challenge, where such reasons might involve deferring to the authority of a person or institution from whom the speaker got the information (“I read it in the Times.”).Footnote 16 We may accept that this is part of the normative structure of assertion without being committed to the conclusion that the justification that the defender of a claim provides rises to the level of knowledge; as such we may remain neutral on the question whether assertion is governed by a so-called knowledge norm.

These three strands making up the commitments associated with assertion provide a starting point for speculating about what its evolutionary precursors might have looked like. One such precursor would involve the concept of liability but not those of frankness and fidelity. Let us call this urassertion, or ursertion for short. Perhaps the earliest utterances of Vervish are ursertions. Such a comparatively impoverished speech act may still play a useful role in a community. For one, ursertions may be correlated highly enough with worldly affairs to be better-than-chance sources not just of information, but of true information. Further, communities in which ursertors are reliable sources of true information likely have better prospects of survival than those in which they are not. If these two conjectures are correct, then we may also reasonably hypothesize that speakers of Vervish would have good reason to heed one another’s ursertions. What is more, practices might grow up to assess and support the reliability of speaker’s utterances. For instance, if conspecifics track a speaker’s record of being right or wrong in what they say, that may be enough to enable them to mete out rewards for those with strong records. Such rewards may come in the form of higher position in grooming, mating, and feeding hierarchies. This would in turn incentivize speakers to get things right when they produce utterances of Vervish.

Speakers who seek the rewards of being reliable ursertors do well to organize their utterances to follow their beliefs, that is, to ursert that P only when they believe that P. The reason is that ursertions are plausibly more likely to track the truth if subject to both a liability and a frankness norm, rather than just a liability norm. After all, it’s a reasonable hypothesis that at least for creatures whose sensory and cognitive apparatus produces beliefs that are on the whole accurate, harnessing utterance to belief serves as a further constraint on what can be said tending toward accuracy. Thus while ursertion is not conceptually tied to frankness, it is plausible to suppose that until a successor practice becomes so tied, it will be unstable in the following sense: once a group of speakers have a practice of ursertion, members of that community will on the whole do better by, and thus tend toward, urserting only what they take to be the case, that is, only what they believe. This would result in a new linguistic practice, which we may call semisertion.

In light of the foregoing we may now hypothesize two precursors to our modern practice of assertion and one analogue thereto:

  1. 1.

    Ursertion is governed only by the liability norm, in that one who urserts that P is liable to being correct or not depending on whether P is true; but the speaker is not expected to believe P or have any other psychological state we have called a frankness condition; and the speaker does not need to be prepared to defend P if challenged (fidelity). Can there be a practice governed by liability but not frankness or fidelity? We have already seen that inorganic artifacts such as machines can be correct or not when they are designed to convey information. Such machines, as well as organisms that are also designed inter alia to convey information to one another, might do so without being able to track one another’s beliefs. Accordingly there would be no way in which a norm of frankness could emerge among either type of communicator.Footnote 17 Similarly, a community of inorganic or organic ursertors might lack the lexical resources to ask for or give reasons for what they say to one another; but even if they have those resources they may simply not be interested in putting one another up to the challenge of seeking and providing such justification.

  2. 2.

    Semisertion is governed by the liability and frankness, but not the fidelity norm. Semisertion is stronger than is ursertion in involving two commitments rather than one (assuming that these two commitments would not jointly conspire to make things easier on agents governed by them both); however, one who semiserts P still is not obliged to defend P if challenged. A contemporary analogue of semisertion is opining, in which a speaker might express her point of view but not expect to respond to requests for justification for the view expressed. Similarly, a community of semisertors may not yet have hit upon the idea of challenging one another for justification: for all they know, speakers of Vervish manifest their states of mind and are liable to be either right or wrong in what they say, but nothing more.

  3. 3.

    Assertion* is governed by all three of the liability, frankness and fidelity norms. We suggest that assertion* is similar in many respects to our modern practice of assertion, but do not here need to take a stand on whether it is identical with it. Further, the distinctive norms of ursertion, semisertion, and assertion* correspond to conditions 6, 4, and 5, respectively, in a discussion by Price (1998) addressing assertion from a genealogical point of view.Footnote 18 Price also points out that what we here call semisertion (close to his concept of a “MOA”) leaves something to be desired in a community that depends on individual members being sources of true information for one another. As he puts it, practices of semisertion lack the “friction” that comes of speakers being able to disagree with and on that basis argue with one another.Footnote 19 Such interaction will be difficult in an impoverished communicative system such as Vervish. It is not impossible: I might wordlessly reply to your quizzical look in response to my utterance by pointing to or gazing overtly at the situation that prompted it—an approaching snake for instance. However, if speakers of Vervish were to expand its vocabulary to facilitate the process of giving and asking for reasons, that would help them achieve the aforementioned friction, which could in turn help their utterances of indicative sentences become not just sources of information but of knowledge.

4 A New Perspective on Conventionalism About Force

Our approach to assertion and related communicative practices through the lens of cultural evolution suggests a genealogy for this practice that could find empirical support through either archeological discovery or computer modeling. It may be hoped that other researchers will pursue one or both of these avenues.Footnote 20 In the nearer term, we may find support for the present approach by noting how it helps shed light on two problems, one concerning the conventional nature of speech acts (this section), and the other concerning the way in which illocutionary norms may be abused (Sect. 5).

A longstanding debate in pragmatics concerns the question whether—and if so, in what way–speech acts, including members of the assertive family,Footnote 21 are conventional. Observe first that just because important elements of language are conventional, we cannot infer that speech acts are as well. Lexical meaning (the fact that ‘dog’ refers in English to dogs, and so on) is conventional, but given the popularity of the thesis of Universal Grammar, many will doubt that basic aspects of syntax are. Let us use semantic conventionalism to denote the claim that words have the meanings they do in virtue of conventions,Footnote 22 and use force-conventionalism to denote the claim that speech acts are conventional in a way that goes beyond the conventional meanings of the words that are typically used in their performance.Footnote 23,Footnote 24

Since it is unlikely that factors creating speech acts are underpinned by genetics alone, we may be tempted to class speech acts alongside lexical phenomena and thus see them as conventional as well.

Before rushing to that conclusion, however, it will be helpful to clarify the notion of a convention.

To describe a particular convention perspicuously we may do so with sentences of the form, “Group G’s conventional way of doing X is to Y.” Thus: our conventional way of driving on roads in North America is to drive on the right, English speakers’ conventional way of referring to dogs is with the word ‘dog’, and among Italian mafioso a conventional way of signaling respect to someone is to kiss them on the cheek. These formulations have the advantage of specifying both a behavior and the end it serves.

As the notion is used here, conventions are (a) regularities in behavior that are (b) normative, and (c) arbitrary (Green 2021b). The norms in question need not be moral norms, and may instead be founded upon practical rationality. (In this sense, driving on the right side of the road is normative even if it is not enforced by law: if most other drivers drive on the right, then it is prudent for me to do so as well.) To be arbitrary, a pattern of behavior must have an alternative that can achieve approximately the same result and with approximately the same efficacy. Driving on the right is arbitrary because driving on the left side of the road is such an approximately equally efficacious alternative.Footnote 25

Whether a pattern of behavior has an approximately equally efficacious alternative cannot always be settled in the armchair or by appeal to common sense: as Burge (1975) points out, a society might believe, incorrectly, that theirs is the only possible way of doing something even when equally efficacious alternatives are hiding in plain sight. Also, because there is no firm line separating a regularity in behavior from a pattern that is common but not dominant, this analysis predicts that there will be some degree of indeterminacy as to what patterns of behavior are conventional. That is as it should be, since there are many practices about which we are inclined to feel it is indeterminate whether they are conventional or not.

Even with the above elucidation, we still encounter unclarity about what it is to say that a speech-act type such as assertion, promise, bequeathal, or retraction is conventional. The reason is that discussions of convention are often bound up with so-called constitutive rules, which are generally explained by way of contrast with regulative rules. A regulative rule provides norms for the proper doing of a thing—at least what some agents consider to be proper—but where that thing can be done without any rules. Eating can be done without adhering to any regulative rules governing it even if doing so may be thought uncouth. By contract, a constitutive rule governs a pattern of behavior that can only be carried out in the presence of that rule. Check-mating can only be carried out when rules of chess are in force; in the absence of such rules all people can do is move objects in geometrically describable ways across boards with variously-colored squares.Footnote 26

Corresponding to the constitutive-rule way of conceiving matters as they pertain to speech acts is strong force-conventionalism, on which speech acts can only be performed by invoking a convention of a certain sort. If strong force-conventionalism is true, then illocuting is like check-mating, which can only be performed in a social milieu in which the rules of chess are in force. On another understanding of the thesis, weak force-conventionalism, it is to say that there may be conventional ways of performing speech acts, but carries no commitment to the claim that these are the only ways of doing them. Greeting, for instance, can be performed with a handshake or a “Hello!”, but we can greet others in non-conventional ways as well. In addition to theses about speech acts generally, we may also contemplate forms of strong (weak) force conventionalism about particular speech act types such as promising, asserting, excommunicating and bequeathing.Footnote 27

Austin appears to have held something close to strong force-conventionalism, although his position is not entirely limpid.Footnote 28 Strawson (1964) famously challenged Austin’s position, arguing that while strong force-conventionalism is true of some speech acts, it is doubtful that it is true of all speech acts. For instance, while accepting is a speech act (one can accept something by saying, “I accept,”), one can also accept something without calling upon anything that could plausibly be termed a convention. Instead, one need only overtly manifest one’s willingness to take a thing that is in someone else’s power to give. Strawson puts the point with the example of entreaty:

I do not want to deny that there may be conventional postures or procedures for entreating: one can, for example, kneel down, raise one’s arms, and say, “I entreat you.” But I do want to deny that an act of entreaty can be performed only as conforming to such conventions….[T]o suppose that there is always and necessarily a convention conformed to would be like supposing that there could be no love affairs which did not proceed on lines laid down in the Roman de la Rose or that every dispute between men must follow the pattern specified in Touchstone’s speech about the countercheck quarrelsome and the lie direct. (Strawson 1964, p. 444)

Since an overt manifestation of an intention can occur without the invocation of any conventions, it would appear that entreating is not essentially conventional, and more precisely that strong force conventionalism about entreating is not true; similar remarks may be made about accepting. Further, for those cases of entreating and accepting that do not depend on (extra-semantic) conventions, Strawson (1964) suggests that they be understood as generated by speaker-meaning as he understands that notion, namely in terms of intentions to produce psychological effects in an addressee by means (at least in part) of that addressee’s recognition of one’s intention to do so.Footnote 29

What may we say about assertion in light of the above? Green (2020a) provides a case in which a speaker invokes nothing more than semantic conventions but makes what is evidently an assertion. If this case is possible, then strong force-conventionalism as applied to assertion is untrue. Might weak force-conventionalism about assertion fare better? Such plausibility as it has would appear to stem from the fact that the indicative grammatical mood is a conventional device used for indicating that a content is being put forth as having word-to-world direction of fit. However, this device only indicates an utterance’s membership in the assertive family rather that its being an assertion. If so, then the indicative mood is not a conventional indicator of assertoric force, but rather is a conventional indicator of a force-genus of which assertoric force is a particular species.Footnote 30

Even if speakers’ ways of indicating that their utterances have assertoric force is not essentially conventional, it might be replied that assertion is itself a communicative practice that is conventional. Put differently, it might be suggested that keeping semantic conventions fixed, there are still other ways of carrying out the practice of assertion besides the one that is currently in vogue and which would be approximately equally efficacious. In support of this version of conventionalism about assertion, we might recall that assertion is a human practice governed by complex norms, and there is no guarantee that a heretofore undiscovered linguistic group will adhere to all such norms. That is, one might suggest in support of force-conventionalism concerning assertion that certain linguistic communities have conventional ways of transferring information by making assertions.

However, that claim will in turn be true only if at least one viable alternative to this method of information transfer, using different norms from those governing assertion as we know it, would do approximately as well in those communities.

If our discussion in Sect. 3 was on the right lines, however, we should suspect that plausible precursors to assertion such as ursertion and semisertion would be inferior to it as social practices underwriting information transfer. Neither ursertion nor semisertion, then, would be a witness to the conventional nature of assertion. Furthermore, it is not easy to see what non-assertoric practice might be a viable alternative to assertion that is neither substantively more nor less effective as a means of information transfer within a linguistic community. Of course, my failure to discern such an alternative may just reveal a lack of imagination or ingenuity on my part. But to my knowledge, no researcher writing on assertion has argued for its conventionality by providing an alternate practice that would be comparably viable. Such a proposal would be required for the support of even a weak form of force conventionalism about assertion.

Iron weapons are better than bronze, and crop-rotation is generally a superior method of farming to monoculture. So too, aside from the conventional nature of the symbols used, there is nothing conventional about double-entry bookkeeping. Similarly, while a linguistic community need not have a practice of assertion, if it does have such a practice there is no reason to think that this practice will be a conventional one. What is more, a speaker may make an assertion either with or without employing a device that conventionally indicates that they are doing so. Assertion, therefore, is at best a practice of which weak force conventionalism is true.

5 Puffery: A Tonk-Like Abuse of Speech Act Norms

In addition to clarifying the issue of force-conventionalism as it applies to assertion, our genealogical approach also helps to illuminate ways in which illocutionary norms may be abused, particularly in the arena of public discourse. To appreciate why, note first that inferentialists hold that the most informative way in which to characterize the meaning of expressions in natural or formal languages is in terms of the inferential patterns in which these expressions participate. We see into the essence of ‘and’ for instance, by virtue of knowing that if ‘A’ and ‘B’ have been established, then we may infer ‘A and B’, and that if ‘A and B’ has been established we may infer both ‘A’ and ‘B’.Footnote 31

It is well known that inferentialist approaches to the meaning of logical constants need to take precautions against an objection raised by Prior (1960). He argued that if one could define a logical constant entirely in terms of introduction and elimination rules such as we just mentioned for ‘and’, there would be nothing to prevent the introduction of a logical constant, ‘tonk’, governed by the following introduction and elimination rules:

A entails A tonk B A tonk B entails B

These two rules together enable us to infer anything whatsoever so long as at least one proposition has been established. Prior concludes that the possibility of introducing a constant such as ‘tonk’ shows that an inferentialist approach to logical constants must be misguided.

Belnap (1962) explains how Prior’s concern may be allayed by attending to the independently motivated requirement that new expressions may only be introduced into a language in a manner that is a conservative extension of that language. An extension of a language L is conservative just in case, after that extension has been introduced, it is not possible to prove anything exclusively in the old vocabulary of that language that was not provable prior that extension. In this light, it is easy to see that introduction of the constant ‘tonk’ is not a conservative extension of the kind of language that Prior had in mind. Further, it is a reasonable demand on the introduction of any expression into a language that its introduction be a conservative extension of that language. This reply does not show that inferentialism about logical constants or other expressions is correct, but it removes one objection to the approach.

The Prior-Belnap interchange provides useful background for consideration of how speech act norms may be abused in public discourse and elsewhere. The reason is not that abusers of speech act norms attempt to introduce non-conservative extensions of the languages in which they speak. Rather, such speakers can willfully misrepresent the force of their speech act, thereby prompting others to treat as known material that does not merit that epistemic status. In so doing they, like the user of ‘tonk’, can establish things they should not be able to establish. This power is particularly nefarious in the public domain, where millions or even billions of people might come to take as common ground propositions that are either untrue, or insufficiently well-supported by evidence to merit that acceptance.

To appreciate more fully the maneuver we are trying to isolate, suppose that in uttering an indicative sentence a politician is semiserting that P. (Recall that semiserting is analogous to our modern practice of opining.) Nonetheless, by means of his confident tone of voice and solemn facial expression, he allows his audience to understand him as asserting that P even though he does not take himself to be prepared to respond to appropriate challenges calling for justification for belief in P. Ceteris paribus, when a speaker asserts that P among interlocutors who have confidence in his credibility, P will be added into conversational common ground (hereafter CG), defined as that set of propositions that all parties to the conversation accept, and recognize one another as accepting.Footnote 32 A similar process will occur if interlocutors take the speaker to be asserting that P even if he is not actually doing so but merely semiserting. Notice that while such a speaker behaves in a way that is misleading, he is not lying. Further, if he were to make clear that he is merely opining, the most reasonable response among interlocutors would be to add to common ground the proposition that this speaker is of the opinion that P as opposed to P itself.Footnote 33 However, in semiserting P while presenting himself as asserting it, a speaker provides us with one way in which someone can have undue epistemic influence on his audience.

The pufferfish is well known to respond to danger of being attacked by expanding itself by intake of water (Wainright and Turrigan 1997). After thus puffing itself up, it appears to potential predators to be larger than it in fact is, and together with the toxic spines that now project from the animal’s skin, thereby deters predation. With this phenomenon as inspiration, let us use the term puffery to refer to the act of putting forth a content as meeting higher epistemic standards than it in fact does. A puffer is one who engages in puffery, and may be an individual speaker such as a politician running for office or social media influencer, or a government, multinational corporation, or other institution prone to make utterances representing its position as a whole.

As we have noted, the puffer does not lie. Indeed he may well believe what he says. But by presenting what he says as something known, he invites his addressees to accept that content into CG and thereby treat it as known. In so doing, the puffer does something tonk-like: the user of tonk will be empowered to establish things that she is in no position to establish (even if they are in fact true). Similarly the puffer will be apt to place into CG information that, even though he may believe to be true, and which may in fact be true, he cannot claim to know.

The behavior is especially pernicious in public discourse. Talk radio hosts, social media influencers, speakers on television and the like are often protected from challenge by audience members simply due to the format of their platforms. We see this even before puffery comes into consideration, for even if these speakers take themselves to be asserting what they say rather than opining, it is difficult for audience members to challenge their claims.

If I have accurately described puffery and its pernicious effects, we have uncovered a kind of illocutionary infelicity that although not envisioned by Austin, can be accommodated into his taxonomy of infelicities. For puffery as described here would constitute a misfire according to that taxonomy (1975, p. 18). Just as in classic cases of presupposition-failure, one who purports to assert that P fails to do so; so too, the puffer purports to but does not assert that P. However, unlike what we find in typical presupposition-failure cases, the puffer leads others to think that he is asserting that P, and may in fact be treated as having done so. In puffery, then, we have cases of misfire that may still be communicatively fecund.Footnote 34

Also, puffery as described here is not to be subsumed by concepts articulated in more recent discussions of abusive language. Recent work on illocutionary silencing, dogwhistles, and figleaves does not address this kind of case. No one need be silenced by virtue of having a content illicitly entered into common ground. So too, a puffer could but need not be covertly signaling to any part of his audience that what he is proffering for entry into CG meets a lower standard than it appears to.Footnote 35 Nor is a puffer attempting to hide a nefarious attitude by covering it with a disavowal of dubious motives.

A challenge to the viability of the concept of puffery here suggested might be thought to come from those who hold that an audience, by taking a speaker to be speaking with force F, can thereby constitute its being the case that the speaker’s utterance has force F even if that is not what the speaker intended. For instance, some of the literature on illocutionary silencing might be understood as arguing that if an audience takes a speaker to be performing speech act S, then that is the speech act she has performed even if she intended otherwise. (Kukla 2014) If this view is correct, then the puffer as imagined earlier in this section has indeed asserted the propositions making up the examples we contemplated above, due to the fact that his audience takes him to be doing so. The defect that remains is that his assertion is improper on account of being made on insufficient grounds.

However, this approach fails to take the measure of the fact, as noted in Sect. 2, that the force of a speech act that is not essentially conventional is determined primarily by speaker’s intentions. Once we remember this, then we will be able to appreciate that it is only when speakers are cognizant of how their addressees construe the force of their (non-conventional) speech acts, that this might affect that force. The reason stems from a conceptual connection between intention and belief: one cannot intend to do something one believes herself unable to do.Footnote 36 For instance, if I believe that none of my students will take my utterance as a command that they turn off their cell phones and start taking notes, it will be difficult for me to form the intentions needed to for my utterance to be a command. Instead, I may only be able to muster an intention that they take what I say as a suggestion or even a request.

6 Safeguarding Against Puffery

How might we protect ourselves against puffers and puffery? I suggest that the most effective path begins with attention to the ways in which assertions can be costly (Green 2009) in well-regulated epistemic communities. In spite of the commonly repeated dictum that talk is cheap, reminding ourselves of the ways in which speakers can lose their reputations by making incorrect or ill-considered claims, enables us to recover a sense of how we stick our necks out when making assertions. Of course, on a busy downtown street corner, conversations among strangers unlikely to encounter each other again are not held to strict epistemic standards. However, in situations in which speakers are likely to have repeated interactions with one another, and each values her reputation as a reliable source of true information, assertions are rightly accorded a special (although not unique) status: in such situations speakers should, and realize that they should, only assert what they know to be true. Since, however, knowledge is often difficult to come by, assertion carries a risk not only of misleading others, but of a speaker’s having her own reputation tarnished.

Public discourse complicates this picture because it can be difficult to hold those who address large audiences to account for what they say. Even if there are experts able to tell when a multinational corporation is “greenwashing” its environmental policies, for instance, that expert analysis may lack the platform to which the multinational has access. As a result the expert analysis may be drowned out by the multinational corporation’s engaging in what we have called puffery: what the corporation says may not be a lie, and may even be correct, but cannot in good faith be said to be something the corporation knows to be true (De Freitas et al. 2020). The move toward Corporate Social Responsibility is a small step toward holding corporations accountable for such behavior (Ibid).

So too, politicians have a strong incentive to engage in puffery, concerning their careers prior to entry into politics, their success with voters, the efficacy of their past or suggested policies, and so on. My suggestion concerning how to address it is not complicated: wherever possible we should insist that rather than permitting them to control the conversation, actual and would-be holders of public office should be required to respond to questions from the public including those who are particularly well-versed in the topics under discussion. These questions should wherever possible include challenges calling for justification of what has been asserted; or, barring that, clarifying that what was said was not an assertion but was instead the voicing of an opinion (close to our notion of semisertion). This is one way in which voters may hold politicians accountable for what they say. It may be hoped that in so doing, members of the public will also learn to better discern utterances by public figures that merit acceptance from those that should be taken seriously but not accepted until backed by appropriate evidence.