Conventions and Constitutive Norms

1 Assertion Conventionalism

Stalnaker (2014, p. 36–7) contrasts two different ways of thinking about speech acts. Austin (1962) advocates thinking of them as social practices constituted by social norms, established and maintained by conventions; this is entailed by his proposing a framework for their characterization that assumes that “there must exist an accepted conventional procedure having a certain conventional effect” (ibid., 14). Grice (1957) takes them instead to be definable in natural, psychological terms, in terms of a peculiar kind of reflexive intention. As a result of Strawson’s (1964) forceful criticism of Austin’s (1962) social account of speech acts, and in spite of the important work of proponents of such accounts like Searle (1969) and Alston (2000), until recently the Gricean psychological account supported by Strawson has been the default in contemporary philosophy; Stalnaker’s own work is a good illustration. This situation has been changing in the past years, in part through the deserved impact of Williamson’s (1996/2000) account of assertion, which has brought back into the philosophical landscape normative accounts on which assertion is defined by constitutive norms.

Austin and Dummett (1981, p. 298; 1993, p. 21–223) conjoined normativism and conventionalism in their views of assertion. On their view, speech acts are “social constructs” in a clear-cut sense of the notion, as elaborated below in §3. Williamson, however, is keen to dash any conventionalist hopes that his account might encourage, invoking for this purpose the modal disparity argument I will critically examine in the rest of the paper:

Constitutive rules are not conventions. If it is a convention that one must φ, then it is contingent that one must φ; conventions are arbitrary, and can be replaced by alternative conventions. In contrast, if it is a constitutive rule that one must φ, then it is necessary that one must φ … a rule will count as constitutive of an act only if it is essential to that act: necessarily, the rule governs every performance of the act (op. cit., 239).

Here is a reconstruction of this argument, AMD (Argument of Modal Disparity):

1. Assertion is a kind defined by constitutive norms.

2. Constitutive norms impose their obligations in all possible worlds.

3. Conventions are arbitrary, hence contingent, hence practices instituted by them do not obligate in all worlds.

   ∴ The assertoric practice is not conventional, for it exists and imposes obligations in worlds in which no conventionally instituted norm imposes obligations.

There is an interpretation of the conclusion of Williamson’s argument – that constitutive rules are not conventions, as he states it – on which nobody should reject it. ^[1] Taken as abstract entities – blueprints for practices, as Rawls (1955, p. 26 fn.) aptly puts it – conventions (i.e. institutions or practices established by conventions) and constitutive norms (i.e. institutions or practices defined by constitutive norms) are different entities, different types or kinds. ^[2] Nonetheless, Williamson’s argument is either misleading, or a straightforward non-sequitur. For constitutive norms thus understood are not really normative, ^[3] while the assertion-conventionalism defended by the philosophers who had previously espoused the view that Williamson’s paper objects to, like Austin or Dummett, assumed that assertion is normative. AMD thus fails, because the conclusion, as just interpreted, is perfectly compatible with any sensible conventionalist claim that can be ascribed to such conventionalist philosophers.

This is thus what I will argue: the main consideration invoked in AMD – the modal difference between conventions and constitutive norms – is inapt to cast doubt on forms of conventionalism worth considering. When taken as addressed to reject a sensible conventionalism about assertion worth taking seriously, the second premise in AMD is false. In arguing for this, I will emphasize a distinction that is needed to properly understand these issues – to have a clear conception of kinds defined by constitutive rules like games and, perhaps, speech acts: the distinction between Rawlsian blueprints, mere putative normative kinds, and truly normative kinds, whose defining norms have been enforced. In this way, and beyond the refutation of the unfortunately influential modal argument, I hope to contribute to get a better understanding of what social constructs conceived as defined by constitutive norms are.

Let me articulate such a sensible form of C(onventionalism) about A(ssertion) that, although wrong also in my view, remains untouched by Williamson’s argument:

CA The practice of assertion exists (so that speakers are in fact bound by its constitutive norms) only because a convention instituting and preserving it is in place.

Here I am assuming the following characterization of what a convention is – a minimal common core to the accounts by Lewis (1975), Bach and Harnish (1979, p. 120–134), Davis (2003, p. 204–219), Marmor (2009) and Geurts (2018): a convention is a social regularity in the behavior of a group, which serves a common interest (it solves a “coordination problem”, on Lewis’s account) and is arbitrary in that there is an alternative which would have solved it just as well. I want to emphasize that I do not at all intend to argue for CA; on the contrary, I will end up suggesting that CA is not true, to that extent agreeing with Williamson. My claim will rather be that the debate about whether core illocutionary forces are conventional should address CA, putting aside the irrelevant modal considerations that Williamson’s argument deploys.

Searle (1995, p. 28) also contends that constitutive rules are not conventions, on the basis of considerations very similar to Williamson’s: “I am discussing rules and not conventions. It is a rule of chess that we win the game by checkmating the king. It is a convention of chess that the king is larger than a pawn. “Convention” implies arbitrariness, but constitutive rules in general are not in that sense arbitrary”. Searle’s argument is not as developed as Williamson’s, because of that, I’ll only discuss the latter. What I say below will show however why, although the constitutive rules of chess are not themselves conventions, chess is reasonably taken to be a conventional practice, in the very sense that philosophers like Austin or Dummett (wrongly, in this case, in my view) took assertion to be. ^[4]

Williamson’s anticonventionalist argument is influential in its apparent simplicity, which justifies critically addressing it. Thus, referring to it Ball (2014a, p. 341) produces a straightforward version of AMD, arguing that assertion is not conventional on that basis: “constitutive norms govern the action types of which they are constitutive as a matter of necessity. So, constitutive norms are not conventional”. Jankovic (2018, p. 376, fn. 11) also assumes the argument: “Constitutive rules and conventions should be distinguished in general. It is a rule of chess that a pawn moves a certain way, and one is not playing chess if one is not acting in accord with this rule. It is a convention that thus-and-such a piece of wood serves as a pawn – it is arbitrary and we could use any object as a pawn as long as we agree on the same one”. ^[5] The first goal of the paper is to show why the argument is a non sequitur. In so doing, I will highlight a central distinction needed to understand what normatively constituted kinds are, which appears to be easily overlooked in this debate. This gives an additional interest to the ensuing discussion. The paper should thus help to understand better a central debate in social ontology: what normatively constituted kinds are, and what it would take for them to be conventional; and thereby, as said, to get a better understanding of what social constructs defined by constitutive norms are.

I will take up these issues in §3. Before I need to briefly outline the debate on whether speech acts like assertion have a normative nature, or rather a psychological, “natural” one. That is independently needed to frame the discussion, but it will also help us to understand how the crucial distinction I make in §3 is so easily overlooked, in spite of being so obvious once it is pointed out. The reason is that the debate takes for granted that the kind at stake (assertion) is already in place, which invites disregarding the issue of how that has come to be.

2 Assertion: Natural or Normative?

In this section I present the notion of assertion that we will be discussing. As I’ll point out, the assumptions on assertion articulated here are shared with Williamson. Debates on this issue assume that we have a pre-theoretical notion of assertion (or stating, claiming, affirming, which I take to be other common words for the same phenomenon) on which this is an act that we perform in central cases by uttering declarative sentences: “In natural language, the default use of declarative sentences is to make assertions”, Williamson (op. cit., 258). I take such central cases (Williamson’s “flat-out” assertions, ibid. 246) to be those in which we aim to be “taken at our word”, such as using literally a declarative sentence to answer a request for information, or to tell friends how our day went. I’ll assume further that the distinction that the Kripke-Putnam arguments about natural kinds have made salient, between nominal essence and real nature, applies also in this case. ^[6] Here are three features of the nominal essence of the kind I aim to pick out: (i) It is a kind of intentional act, whose point is to produce outright belief. (ii) In performing it, speakers present themselves as believing what they say. (iii) It has a “word-to-world” direction of fit – when its content does not fit the world, there is a mistake in the act, unlike in the case of the central acts we make with interrogatives and imperatives.

These features of the nominal essence of assertion, which help us picking out central cases, manifest themselves in our normative practices: in our finding appropriate to criticize, or required to excuse, assertions when what is said is false, when the speaker does not believe it, or the audience already has the belief. Conversely, there are uses of declarative sentences that fail to be central because they clearly lack some of these features. Thus, e.g. cases in which the sentences occur in a fiction, or in which the speaker adds an appositive “I guess”, “I think”, “I assume”, “I conjecture”, or “I promise”; cases in which the sentence occurs embedded as the antecedent of a conditional, a disjunct, or as a prejacent in a modal sentence; explicit performatives such as “I hereby promise that …”; or, indeed, cases in which the speaker manifestly does not aim to be believed merely because she is saying so, like many of the claims literally made by means of declarative sentences in this paper.

As indicated, this is only intended to elucidate the intuitive features of a pre-theoretically familiar kind. It is part of the idea that the kind in question is natural in that it is presumed to have an essence or nature “hidden” in not being immediately accessible to intuition, and available only, if at all, after theoretical scrutiny. The Gricean and Williamsonian proposals to be summed up presently are attempts at theoretically specifying it; ^[7] such research might well conclude that there is after all no kind shared by central cases, but merely a disunified motley (Cappelen 2011). If, however, the assumption is confirmed, it might turn out that, given the true nature of assertion, we also make it in non-central cases.

As pointed out, conventionalism about illocutionary forces usually comes together with the nonetheless conceptually independent claim that such forces have a normative character. This, in its turn, is sometimes motivated by the observation we have just made for the case of assertion, that we evaluate illocutions relative to different norms; for instance, in the case of assertions relative to whether they are justified or otherwise, true or informative. As Hindriks (2007) notices, however, we also evaluate assertions relative to (invoking Rawls’s (1955) well-known distinction, influentially picked up by Searle (1969, p. 33–34)) merely regulative norms, norms that regulate, relative to certain purposes, acts that in themselves are constitutively non-normative – for instance, as witty, polite or well-phrased, in the case of speech acts.

Some writers question the distinction between constitutive and regulative rules; see Hindriks (2009) for discussion. ^[8] But there cannot be anything conceptually problematic about the distinction itself, at least on the Aristotelian essentialist assumption I am making that the kinds of practices we are talking about have constitutive natures. One of the many merits of Williamson’s proposal, which partly explains its deserved influence, lies in the way it clear-headedly combines the view that practices like assertions have a nature specifiable by a definition, with the straightforwardly normative character of that nature.

Unlike what Searle (1969, p. 36, 41) has led some to think, the distinction does not have to do with the logical form of norms: whether they are genuine imperative, or rather have a “A counts as B in C” shape. ^[9] All norms, regulative or constitutive alike, are semantically directives – although for the sake of convenience I will state them in declarative form, using modals such as “should” or “must”. All norms here relevant, regulative or constitutive, have contents that obtain under some conditions and fail to obtain under others. The difference between regulative and constitutive norms lies just in that only the latter specify the constitutive nature of a practice: A norm N is constitutive of a practice π iff N specifies the constitutive nature of π, along the lines that Williamson’s KR below – or any of the alternatives – is supposed to define the identity of the kind assertion, whose nominal essence we characterized above. N is instead merely regulative for π if π has an independent constitutive nature – for instance, one specified in psychological terms by the Gricean account GA below, in the case of assertion. ^[10]

Let us now compare a paradigm Gricean account of the speech act we have independently identified – the influential one to be found in Bach and Harnish (1979, p. 42) – with Williamson’s account. “R-intending” here is to be explicated in terms of Gricean communicative intentions:

(GA) To assert p is to make an utterance thereby R-intending the hearer to take it as a reason to think that the speaker believes p and intends the hearer to believe it.

Bach & Harnish’s GA is a descriptive account, not a normative one: unlike normative accounts like those presented below, by itself it does not mention norms, but only certain psychological states of speakers and their intended audiences. ^[11] As Hindriks (2007) notes, although it is indeed a feature of our assertoric practices that we criticize performances that violate rules like those mentioned in the normative accounts below – for instance those that are false – these facts about our practices of appraising assertions are insufficient to justify normative accounts. For we also evaluate assertions – say, as witty, polite or well-phrased – relative to merely regulative norms, norms that regulate, relative to certain purposes, acts in themselves not constituted by such rules. All norms we apply to assertion are merely regulative of a constitutively non-normative practice defined by GA. The regulative norms in question could be derived from an ultimately moral sincerity rule such as SR: ^[12]

(SR) In situations of normal trust, one ought to be sincere.

Thus, for instance, the appraisal of assertions relative to a truth rule could be explained as merely regulative, on the assumption that GA characterizes their nature, as derived from SR given the further assumption that the speaker’s belief that his assertion is supposed to give the audience reasons to ascribe to him is itself regulated by a truth rule.

In contrast with descriptive accounts such as GA, Williamson claims that the following norm or rule (the knowledge rule) is constitutive of assertion, and individuates it:

(KR) One must ((assert p) only if one knows p).

In the course of the debate that this proposal has generated, other writers have accepted the view that assertion is defined by constitutive rules, but have proposed alternative norms; thus, Weiner (2005) proposes a truth rule, (TR), Lackey (2007) a reasonableness rule, (RBR), and I and others – García-Carpintero (2004, forthcoming), Hinchman (2013), Pelling (2013) a knowledge provision rule, (KPR):

(TR) One must ((assert p) only if p).

(RBR) One must ((assert p) only if it is reasonable for one to believe p).

(KPR) One must ((assert p) only if one’s audience gets thereby to be in a position to know p).

The obligations these rules impose are sui generis, like those constitutive of games, the model on which Williamson bases his account: they do not have their source in norms of morality, rationality, prudence or etiquette. They are not all things considered, but pro tanto; in any particular case, they can be overruled by stronger obligations imposed by other norms. They are intended to characterize what is essential or constitutive of assertion (and not, as it may seem at first glance, of correct assertion). The view is that assertion is an act essentially constituted by its being beholden to the relevant norm. On Williamson’s view, assertion is the unique representational act such that, in performing it, one is committed to knowing the represented proposition; i.e. the propositional act such that, if one performs it without knowing the intended proposition, one is thereby contravening its constitutive obligation. There are additional rules contributing to a full characterization of assertion, as in Searle’s (1969) well-known account or in Alston’s (2000) elaboration, i.e. “sincerity” or “preparatory” conditions. The rules are intended to characterize what an act must “count as” for it to be an assertion, i.e. what Searle describes as its “essential rule”.

As indicated, it is common ground among participants in these debates that assertion is the actually existing practice for which we gave an independent characterization above: what is done by default (i.e. unless the utterance is ironic, fictive, includes canceling parenthetical remarks such as “I conjecture”, etc.) by uttering declarative sentences. This gives us an independent specification of the phenomenon that we aim to characterize, and hence gives us a grip that allows us to evaluate them: it is the act, whatever its proper definition is, that is in fact associated with the indicative mood in natural languages as used on central cases, and which speakers intentionally purport to make by such means on such occasions.

The crucial difference between normative accounts along the lines of those just outlined and descriptive accounts such as GA lies in the question of whether all norms we invoke to appraise assertions are merely regulative (as on the latter view), or some of them (the truth rule, some knowledge rule, some reasonableness rule) are instead essential or constitutive. Williamson presents his proposal as a hypothesis for which he gives abductive support. We do not need to go into the reasons he has offered for this. What we will use from this section for our discussion in the next is the point that assertion is an already existing practice whose instances we can independently specify in core cases, which is assumed to have a nature worth theorizing about, and which, on worth considering proposals, is in fact a specific normative kind.

3 The Alleged Modal Disparity between Convention and Assertion

Now, as suggested above, one might have thought that a normative view of speech acts along the lines of Williamson’s for asserting could provide some support for a conventionalist claim like CA. Of course, Gricean conceptions of forces, such as GA for assertion, reject such conventionalist claims, following Strawson’s (1964) influential criticism of Austin. ^[13] On these views, assertion is a non-arbitrary natural phenomenon, not at all dependent on convention for its institution and preservation. Williamson, however, argues as we have seen that in fact an account of assertion in terms of constitutive rules makes it impossible for assertion to be conventional in any significant sense. ^[14] Readers knowing the philosophical context in which the paper was produced would assume that his argument is addressed at conventionalist views of assertion like Austin’s or Dummett’s – philosophers who had conjoined before normative views of assertion with conventionalist claims about it. I will now show why the argument is fallacious.

As we saw in §1, Williamson’s argument AMD invokes a modal disparity between acts that are essentially defined by a constitutive norm, and conventions. Conventions are arbitrary; they have alternatives, and hence may not have existed, even in the presence of the needs motivating their implementation. The obligations they impose are thereby equally contingent: “if it is a convention that one must φ, then it is contingent that one must φ” (op. cit., 239). Obligations deriving from constitutive rules, in contrast, are necessary: “if it is a constitutive rule that one must φ, then it is necessary that one must φ” (ibid).

This argument should be immediately suspicious. Consider the illocutionary forces that Searle (1969) classifies as declarations, such as redoubling at bridge, calling out a player at baseball or marrying a couple. Intuitively, such forces are conventional; ^[15] philosophers with little sympathy for conventionalism about assertion agree on this (cf. Strawson 1964, p. 456–457). However, if acts like assertions are practices defined by constitutive norms, declarations are likely also similarly defined, and hence, Williamson’s modal disparity argument would establish that they cannot be conventional either. The point also applies to games, the very model on which Williamson bases his account: while they are defined by constitutive norms, intuitively they are as conventional as the clearest paradigms of conventional activities, such as using particular expressions to convey particular meanings or driving on the right.

To see where the fallacy in AMD lies, we will reflect on a point that Williamson notices. He admits that his account of constitutive rules involves some idealization, which can be seen by considering the case of real games. Let us take the case of soccer, association football. Throughout its history, there have been changes in the constitutive rules that define that game, without intuitively a change in its identity. How does this fit with the account in terms of constitutive rules? Williamson appeals here to Lewis (1975) related discussion of languages and language: one and the same historically continuous language in use might be properly characterized at different stages by different “grammars” or theoretically articulated abstract languages: “a population that at one time has the convention of speaking a language L may later change to a convention of speaking a distinct language L*, constituted by slightly different rules. Likewise, in the present technical sense of “speech act”, the rules of a speech act are essential to it. A population that at one time has the convention of using a certain device to perform a speech act A may later change to a convention of using that device to perform a distinct speech act A*, governed by slightly different rules. “Game” can receive a similar sense” (op. cit., 239).

So, in the intended “technical sense”, the speech acts defined by norms such as KR, TR, RBR and KPR are like Lewis’s abstract languages, and we should distinguish them from the actual speech acts that people perform, which are rather like Lewis’s languages in use; for the latter can be characterized by different instances of the former at different stages in their history. There is perhaps actually used English, which at some time in a history originating from a language brought to England by Germanic settlers in the fifth to seventh centuries might be properly characterized by abstract language AE₁ – and hence had as a proper temporal part actually used AE₁ – and at a different stage by AE₂ – and hence had as another proper temporal part actually used AE₂. ^[16] Similarly, there is actually played association football, with a history going back to games played at public schools in England in the 18th century and earlier, and characterized by different set of rules as played at different times and places, first partially codified in the mid-nineteenth century. ^[17] Likewise, there is actually occurring assertion, which in the quotation above Williamson suggests we understand as we have neutrally characterized it in §2, as the act done by default when uttering declarative sentences. This act might have had a history such that, say, TR initially defined it, and now Williamson is right that it is rather KR that defines it.

Note also that abstract languages might fail to characterize any actually used language. Two linguists might dispute whether it is AE₁, or rather AE₂, that properly characterizes English at a certain point; perhaps both are wrong, and neither set of rules properly does the job. As Schiffer (1993) notes, it would be inapt to say in such a case that neither AE₁ nor AE₂ “exist”, because being abstract entities they exist in all worlds. ^[18] We will say instead that they are not used or in force. The same could apply, mutatis mutandis, to the acts “in the technical sense” whose constitutive rules TR, RBR, KR and KPR specify. To be perspicuous, let us use “assertion” to refer to the naturally existing act made by default by uttering declarative sentences, and “assertion-tr”, “assertion-kr”, and so on, to the abstract acts that those rules specify – what Rawls (1955, p. 26 fn.) takes to be mere blueprints for practices. The debate thus concerns whether assertion is characterized by one such rule (as opposed to being a merely psychological type of entity, say, one defined by GA), and, if so, by which, assertion-kr, assertion-tr, etc; or perhaps, as pluralists such as Levin (2008) or Goldberg (2015) suggest, in fact several rules define it in different contexts. Some of those blueprints may thus not be in force at all in the actual world.

Williamson is thus aware of the distinction between abstract normative types, and normative types that are in force, and in fact chooses to talk about the former. In more recent work, he is more forthcoming and explicit about this: “I will follow the usual practice of sometimes using the word ‘norm’ for candidate norms, putative norms that may or may not be genuine norms (may or may not yield some sort of genuine ‘ought’ or ‘should’), when no confusion is likely to arise” (Williamson, forthcoming). We can now see that Williamson’s claim, “if it is a constitutive rule that one must φ, then it is necessary that one must φ” is just trivially true. For this is a claim about speech acts in his “technical sense”, i.e. about speech acts as merely blueprints, or as merely putative norm-defined kinds, such as assertion-kr, defined by constitutive rules perhaps nowhere in force, giving nobody reasons to act, entailing no genuine “ought”s or “should”s. ^[19] To be sure, these abstract entities are not conventional; but nobody properly understanding the issues would ever deny this, certainly not philosophers holding conventionalist views about assertion such as Austin or Dummett. They where making claims about a kind, assertion, which they take to be really normative, really giving assertors and their audiences reasons to act – not about mere “candidate” or “putative” norms, mere blueprints for normative kinds.

To illustrate, let us consider a traditional game before anybody tried to codify its constitutive rules, such as what was later called association football as it was played in a given public school in England in 1800. Let us imagine two different plausible candidates to fully and properly characterize their constitutive rules, 1800F₁ and 1800F₂. Imagine also that the first is correct but not the second. They both describe abstract games by their constitutive rules, so, as Williamson notes, to the extent that one plays either of them, necessarily one is obligated by its rules. But note firstly that this is a rather Pickwickian sense of obligation; for it applies both to the rules codified by 1800F₁, but also to those codified by 1800F₂ – a game that we might assume nobody has ever played, so that no actual person has ever been beholden to its norms. Secondly and more important, it is a matter of convention, and hence arbitrary, that people playing association football at that time and place were under the obligations codified by 1800F₁. For these people could have instead agreed to be under the obligations imposed by 1800F₂: presumably this would have served equally well the motivations they had to engage in that activity. ^[20] Hence, although the game is defined by constitutive rules, which necessarily oblige in the Pickwickian sense, it was conventional in this clear-cut sense: it was a matter of convention that a game defined by those rules, 1800F₁, was in force, as opposed to one defined by 1800F₂.

This shows that, although the trivial points that abstract putative norms are not conventions, and do not come to be by convention, are true, entities essentially constituted by norms might be conventional – as should also intuitively be the case. For the former are claims about blueprints, while the latter is one about normative kinds that are in force – about what it is that makes them be in force. Hence, as an argument against the conventionality of assertion AMD is at best a non sequitur, betraying a misunderstanding of conventionalist claims. If taken as addressed to serious conventionalist claims such as CA, AMD fails because its second premise is false: for a constitutive norm to really impose any obligation, it must be in force; and its being in force might well be the result of a convention. If that is the case (as in the case of games), the practice, albeit defined by constitutive norms, will fail to impose its obligations in all worlds.

4 Concluding Remarks

For my purposes, there is no point in speculating about the actual intentions of philosophers deploying AMD. ^[21] The point I have been stressing is this. There are serious conventionalist claims that have been made in philosophy, in particular about assertion by philosophers like Austin and Dummett not long before AMD was formulated. It is natural to take it that a perspicuous and compelling normative view of assertion, like the one Williamson (1996) puts forward, offers a good background to promote them. And it is natural by the same token to take the argument one finds there, AMD, to be aimed to crush such hopes. ^[22] As I showed at the outset, there is good evidence that many people take it that way. But in fact AMD does not thwart them, because conventionalist philosophers were not claiming that merely putative norms are conventional – an obviously wrongheaded view. They were rather claiming that the actual practice of assertion we in fact have imposes its constitutive obligations on assertors exactly on the same grounds that the practice of soccer we in fact have imposes its own on players – to wit, by convention. AMD does not address that claim. ^[23]

Unlike games and declarations, like Strawson (1964) I do not believe it is sensible to think that fundamental speech acts like assertions, requests or promises are conventional, even though I take all those types to be defined by constitutive norms. I thus ultimately agree with Williamson. But the true reason has nothing to do with modal disparities, although non-arbitrariness has something to do with it. To properly investigate this issue, we should ask ourselves why it is that the norms of assertion are in force, assuming that some of them define assertion. It is difficult to believe that assertion is conventional: the being in force of a norm whose point is ultimately to allow for testimonial knowledge is hardly arbitrary, hardly something for which there are alternatives that might have served the purpose equally well. The explanation seems more to do with teleology and social norms, than with convention. ^[24] I thus think that, unlike games and declarations, fundamental speech acts like assertion, promising or requesting are not conventional, but I cannot elaborate on my reasons here.

In this paper, I have offered a reply to a popular modal disparity argument against assertion-conventionalism, made by Searle, Williamson and many others. I have highlighted the relevance of Rawls’ distinction between mere blueprints for practices, candidate norms, and norms to which people are truly beholden. The view that practices like assertion or promising are defined by constitutive norms is controversial, but it has many sympathizers since Rawls formulated it. Here I have emphasized that, on that view, such practices should be understood as social constructs, put in force by social activities, whether or not such activities are just conventions. When they are thus understood, the modal disparity considerations leave entirely open that they are. I have ended up suggesting however that in fact it is not plausible to take them to be conventional, in a sense philosophically significant for this debate.