How Conventional Is a Promise? (Edinburgh)

How Conventional Is a Promise?
(Edinburgh version)

Toshihiko Ise

To be presented in July 1998 at the University of Edinburgh

INTRODUCTION Artificial virtues depend on human conventions or agreement. This is well-known as one of Hume's principal findings in the theory of morals. A convenient way of expressing this finding would be to say that artificial virtues are conventional. I am afraid, however, we are tempted to imply views that can not be justifiably attributed to Hume when we use this expression. The tendency to confuse conventions and rules and to talk as if conventions were a kind of rules, is an example. Thus, David Lewis thinks he is following the Humean line when he formulates his definition of a convention, which is obviously a kind of rule.* As it is indicated by the fact that I call this a confusion, I do not think Hume regards conventions as rules. In reality, when Hume refers to a convention he always refers to an act or a series of acts of entering into agreement, and he never uses the word "convention" interchangeably with "rule" or "law." A straightforward way to show this will be to examine how Hume uses the words in question when he discusses artificial virtues. Another way might at first seem more roundabout, but it is the way in which I aim at showing not only that Hume does distinguish conventions from rules, but also that he has good reason to do so. The key lies in Hume's treatment of promises in the Treatise, which I am going to deal with in the present paper.

* Convention, Harvard University Press, 1969, pp. 3f.

The examination of Hume's account of promises will be especially useful for making clear his concept of conventions, because, paradoxically, that very account seems, at least prima facie, to induce us to make the identification of conventions with rules, which I am warning against. More specifically, it seemingly suggests the view that the binding power of promises depend on the rules that govern the conventional expression of promises and these rules are "conventions of promises." I will show that this view is not, and can not be that of Hume by way of asking, "How conventional is a promise in Hume's view?"

In his treatment of promises, Hume seems to assign an important role to the verbal expression "I promise".* We often use a sentence beginnig with "I promise" in making a promise. It would not sound unnatural to call a sentence of that form "the conventional formula" of a promise, even if we do not specify what we mean by the word "conventional". So Hume seems to think promises are conventional in the sense that they depend on the conventional formula. This way of understanding Hume's argument naturally induces us to attribute to him the view that the conventions on which promises depend are the rules that govern the use of the conventional formula. To show that this attribution is mistaken, I will first compare with Hume's view John Austin's suggestion that the use of a conventional formula is essentially involved in the definition of a promise. (section I) Next I will examine and finally reject two lines of interpretation that try to find in Hume a version of the view that promises are rule-governed. Promises can not be completely analyzed in terms of rules because fully rule-governed acts have also rules that fix the context in which they take place, while the context of a promise is embedded in and inseparable from the context of common life. Promises are not conventional in the Austinian sense of the word. (section II)

* "If promises be natural and intelligible, there must be some act of the mind attending these words, I promise." (T 516) See also the passage from T 522, cited below.

If promises can be called conventional, it will be in the sense that the context of a promise is provided by conventions, that is, human agreement. I will present an alternative reading of Hume's argument that showes that the binding power of a promise derives from the communication of an intention, depending upon conventions as agreement. (section III) The conclusion will be that Humean conventions constitute the basic agreement that supports the moral and legal edifice of justice, rights, and obligations, which are more or less codified, but this agreement itself can not be written into rules. (section IV)

I THE AUSTINIAN PICTURE OF A PROMISE The view that people do not only say but also do things with words has become popular in these several decades owing to works by John Austin and his followers. They distinguish, among what people do with words, illocutionary acts and perlocutionary acts. This is conveniently explained by contrasting what we do in saying something and what we do by saying it, though we do not always use these prepositions conforming to the distinction. Illocutionary acts, it is said, are what we do in saying something, and perlocutionary acts are what we do by saying something. For example, arguing is an illocutionary act, while convincing is not an illocutionary but a perlocutionary act, because my speech can in itself constitute an argument, while it can be characterized as convincing the audience only if they are convinced, but their being convinced is an effect of my argument and its connection with the speech is only contingent.*

* How to do things with Words, 2nd edition, edited by J. O. Urmson and Marina Sbisà, Oxford University Press, 1975, p. 102.

Austin says that an illocutionary act may "be said to be conventional, in the sense that at least it could be made explicit by the performative formula."* A performative formula is a sentence whose verb is in the first person, present tense, indicative mood and names the illocutionary act performed by means of the sentence. For example, I use a performative formula when I start my argument with saying "I argue that" and so on.

* Op. cit., p. 103.

Promising is one of the examples of illocutionary acts that are most often referred to by philosophers of this school. Thus, if Austin were right, a promise would be conventional in the sense that it could be made explicit by saying "I promise that" and so on.

Hume may seem to think much the same way as Austin, when he says, in the third Book of the Treatise, that a certain "form of words constitutes what we call a promise," and that when "a man says he promises any thing, he in effect expresses a resolution of performing it; and along with that, by making use of this form of words, subjects himself to the penalty of never being trusted again in case of failure." (T 522)

It appears then that Hume agrees with Austin on two things. First, he says that the use of the performative formula constitutes a promise, namely, that it is the use of the formula "I promise that ..." that makes a promise what it is. Second, he would agree to call a promise conventional in one sense or another.

We must further ask, though, whether these apparent agreements are genuine, and how much light this comparison throws on Hume's concept of conventions. We will set out to answer these questions in the next section.

II TWO VERSIONS OF AUSTINIAN CONVENTIONALISM It is, of course, far from certain that Hume and Austin would call a promise conventional in the same sense. We may attribute to Hume the view that it is conventional in the sense its obligation depends on conventions since it follows from his main thesis concerning all artificial virtues. This is, though, not the sense in which Austin calls an illocutionary act conventional. Austin's point is that it is one of the defining characteristics of an illocutionary act that it can be performed by a conventional device, and that the performative formula is the conventional device. This second sense of conventionality is not automatically included in the first. It is true that Hume's insistence on the role of the performative formula makes his resemblance to Austin look significant, but Hume is not very clear about why the performative formula is so important, and how its use is related to the conventionality of a promise.

According to a way of thinking that many philosophers of language of mid-twentieth century share, the importance of the conventional formula for promising might be explained in terms of rules. The idea is that promising is a rule-governed practice and the rules governing the practice defines the conventional formula as something that makes a promise what it is. If we could extract from Hume's text a form of argument along this line, we could justifiably say that Hume regards promises as conventional also in the Austinian sense, though the relation between the two senses of conventionality would then still have to be explained. Hume's text seems to suggest two possibilities, which we will shortly discuss.

Before going into the discussion, it will be convenient to cite two paragraphs towards the end of Treatise III, ii, 5 that are highly pertinent to our discussion. These are presented as additional reflections that confirm that the obligation of promises is not natural. Let us name them (R1) and (R2), respectively.

(R1) focuses on the complex relationship between the speaker's intentions and the obligation of a promise, while (R2) compares promises to religious ceremonies like transubstantion and holy orders. I show the text of these two paragraphs in full in the tnote.*

* Since (R1) is structured as the presentation of a general rule, followed by the qualification of the rule in three steps, it will be convenient to present it schematically, by separating those steps.

(R1) [Introduction to the paragraph] To confirm this, we may subjoin some other reflexions concerning that will, which is suppos'd to enter into a promise, and to cause its obligation. [Presentation of the general rule] The expression being once brought in as subservient to the will, soon becomes the principal part of the promise; nor will a man be less bound by his word, tho' he secretly give a different direction to his intention, and with-hold himself both from a resolution, and from willing an obligation.

[Qualification step 1] But tho' the expression makes on most occasions the whole of the promise, yet it does not always so; and one who shou'd make use of any expression of which he knows not the meaning, and which he uses without any intention of binding himself, wou'd not certainly be bound by it.

[Qualification step 2] Nay, tho' he knows its meaning, yet if he uses it in jest only, and with such signs as shew evidently he has no serious intention of binding himself, he wou'd not lie under any obligation of performance; but 'tis necessary that the words be a perfect expression of the will, without any contrary signs.'

[Qualification step 3] Nay, even this we must not carry so far as to imagine, that one, whom, by our quickness of understanding, we conjecture, from certain signs, to have an intention of deceiving us, is not bound by his expression or verbal promise, if we accept of it; but must limit this conclusion to those cases where the signs are of a different kind from those of deceit.

[Conclusion of the paragraph] All these contradictions are easily accounted for, if the obligation of promises be merely a human invention for the convenience of society; but will never be explain'd, if it be something real and natural, arising from any action of the mind or body. (T 523f.)

(R2) I shall further observe, that, since every new promise imposes a new obligation of morality on the person who promises, and since this new obligation arises from his will; 'tis one of the most mysterious and incomprehensible operations that can possibly be imagin'd, and may even be compar'd to transubstantiation or holy orders, where a certain form of words, along with a certain intention, changes entirely the nature of an external object, and even of a human creature. But tho' these mysteries be so far alike, 'tis very remarkable that they differ widely in other particulars, and that this difference may be regarded as a strong proof of the difference of their origins. As the obligation of promises is an invention for the interest of society, 'tis warp'd into as many different forms as that interest requires, and even runs into direct contradictions, rather than lose sight of its object. But as those other monstrous doctrines are mere priestly inventions, and have no public interest in view, they are less disturb'd in their progress by new obstacles; and it must be own'd, that, after the first absurdity, they follow more directly the current of reason and good sense. Theologians clearly perceiv'd that the external form of words, being mere sound, require an intention to make them have any efficacy; and that this intention being once consider'd as a requisite circumstance, its absence must equally prevent the effect, whether avow'd or conceal'd, whether sincere or deceitful. Accordingly they have commonly determin'd, that the intention of the priest makes the sacrament, and that when he secretly withdraws his intention, he is highly criminal in himself; but still destroys the baptism, or communion, or holy orders. The terrible consequences of this doctrine were not able to hinder its taking place; as the inconvenience of a similar doctrine, with regard to promises, have prevented that doctrine from establishing itself. Men are always more concern'd about the present life than the future; and are apt to think the smallest evil which regards the former more important than the greatest which regards the latter. (T 524f.)

Let us now resume the discussion. The two possibilities I mentioned above are the following. First, we might regard a promise as comparable to a move in a game. Second, one might argue that the rules of language give the conventional formula a privilege over other forms of words. I will briefly explain each idea and then see whether they give a sufficient support to the interpretation of Hume as a precursor of Austinian conventionalism.

First, assimilation to a game. In (R2) Hume compares promises to religious ceremonies like transubstantiation and holy orders. Such ceremonies are supposed to bring changes in the nature of things and persons, but the presumed changes have no significance outside the religious practice of which such ceremonies are parts. Thus, we may say that the supposed changes are the results of moves in a game and they are unintelligible independently from the rules of the game. In such a game, the use of a certain form of words constitutes a supposed change by virtue of the rules. Why, then, does Hume compare promises to such moves in a game? Because, according to him, the obligation of promises does not arise from any natural act of the mind, or natural motive. Thus, the conventional formula of a promise represents nothing that is naturally intelligible. If it can constitute a promise, then it must be by virtue of a set of rules that give it its significance as a move in the game.

Second, the rules of language. In (R1), Hume discusses the circumstances in which the utterance of words fails to constitute a promise, and mentions the knowledge of the meaning of words and the showing of a serious intention to bind oneself as necessary conditions for a promise. Though Hume's discussion is too brief and unsystematic to extract from it a set of necessary and sufficient conditions for promising, the rough picture seems not unlike John Searle's analysis of promises as a type of illocutionary acts.* Searle provides his analysis by way of giving the necessary and sufficient conditions for the utterance of a sentence to constitute a promise. Of the nine conditions of successful promising that he proposes, the first condition that mentions linguistic competence of the parties involved, and the sixth, the seventh, and the eighth conditions that mention several sorts of intentions, seem to have counterparts in Hume's (R1). The most crucial, though, to his argument that connects by means of the rules of language the meaning of an expression to the illocutionary force of the act performed in its utterance, is the ninth condition. It says, "The semantical rules of the dialect spoken by [the speaker] S and [the hearer] H are such that [the sentence] T is correctly uttered if and only if [the other eight] conditions [...] obtain." Thus, it is the semantical rules governing the verbal expression that play the central role in making its utterance a promise. The other conditions serve to supply the appropriate context for the application of the rules.

* Speech Acts, Cambridge University Press, 1969, ch. 3, especially sects. 1 and 2. In this book Searle regarded the performative formula as the typical device to perform directly the illocutionary act named by its main verb (See op. cit., pp. 30f.), but later he came to consider its use rather as an indirect speech act. See Searle and Daniel Vanderveken, Foundations of Illocutionary Logic, Cambridge University Press, 1985, pp. 2f.

These two suggestions, though, are both inadequate to convince us that Hume was an Austinian conventionalist.

As to the first suggestion, the analogy between a promise and a move in a game is only partial. This we can understand by looking closely at Hume's own comparison of promises and religious ceremonies. In (R2) Hume contrasts the complexity of the conditions that determine the obligation of promises with the neatness of the rules that govern the alleged effect of a rite. The sacrament requires the intention of the priest who makes it and the absence of the intention prevents the effect. The relation between the speaker's intention and the effect of the utterance is more complex in the case of promises, where deceitful promises equally bind, though therein the intention of carrying them out is absent, as Hume observes in (R1). The difference, then, seems to be defined in terms of the connection between the effect and the act of mind that is supposed to produce it. The presence of the mental act is the necessary and sufficient condition for the sacrament to take place, but this is not the case with regard to promises.

If this is all Hume wanted to say in (R2), though, typical examples of games like chess or baseball seem to be on the same side with promises rather than that of religious ceremonies. If one of the players moves a piece to a certain relative position with regard to the opponent's king, it is a checkmate because of the publicly observable positions of pieces, rather than because of the player's secret intention. Still, games like chess, as well as religious rites, contrast with promises by the neatness and explicitness of their rules. Then, the cause of the difference must lie somewhere else. Let us now turn to the account of the "many different forms" that the obligation of promises is said to be "warp'd into" that is given in (R1).

As we have schematically shown above, the "contradictions" accompanying the obligation of promises are described by way of formulating prima facie general rules and presenting their exceptions, which are in their turns accommodated in additional general rules, to which exceptions are again found. What Hume refers to when he talks about "contradictions" is the fact that every time a rule is proposed we find an exception to it.

As far as we look at the first formulation of the rule in (R1), the analogy between games and promises seems pretty close. One might think of putting the rule in the "count-as" form proposed by Searle.* Thus, a person's utterance of a certain verbal expression counts as a promise. Similarly, we can put many rules of a game in the form "X counts as Y." For example, "An umpire's raising the right hand counts as judging a runner out." This formulation is, of course, quite rough and incomplete. At the very least, the rule of judging a runner out does not really make sense unless it is supplied with the qualifying phrase, "in a baseball game." To generalize this point, a rule of the form "X counts as Y" must be attended with a phrase that describes the appropriate situation in which X counts as Y. The general form of count-as rules would then be "X counts as Y in C," where C stands for the appropriate context for X to count as Y.

* Speech Acts, ch. 2, sect. 5.

At this point, the disanalogy between games and promises becomes visible. That is, when we think of the way to supply the context in which a certain utterance counts as a promise, we can no more rely on the concept of games and their rules, whereas in cases of standard games the conditions for a rule to be applicable are stated by a set of further rules. For example, in baseball, the rules of the game define the situation in which a certain bodily movement counts as a significant step in the game, and they also give explanation to the concepts that appear in the rules, by way of defining what it is to be a batter, a runner, a fielder, an umpire and so on. Any such simple definition of the relevant context is unavailable in regard to the obligation of promises. This difficulty in determining the applicability of the general rule is, I believe, what Hume mainly has in mind when he talks of "contradictions" surrounding promises. How do we know when and when not to take a person seriously if she utters the words "I promise"? The answer depends much more on our general cognitive and communicative skills than on a set of neatly stated rules. Thus, we have to decide whether the person before us is linguistically competent (step 1), whether she intends to do what she says she promises (step 2), and if not, whether she intends her words as a joke or a lie (step 3). Games have the rules that determine when they start and end. To that extent the situations in which games are played are distinguishable from the course of events in real life. By contrast, promises are inseparable from the context of common life, and there is no rule to determine whether the "promising game"* is being played. I think this is enough to show that a promise is not quite a move in a game.

* R. M. Hare uses this phrase as the title of his paper. See 'The Promising Game,' 1964, repr. in Philippa Foot (ed.), Theories of Ethics, Oxford University Press, 1967, pp. 115-127.

The examination of the first suggestion so far also saves us from much of the work in dealing with the second suggestion. The second suggestion is based on the apparent similarity of the argument in (R1) to Searle's analysis of illocutionary acts. We have already looked at (R1), and the result testifies against this reading of Hume.

As we have seen above, Hume's argument in (R1) implies that a promise will not be completely analyzed in terms of a rule of the form "X counts as Y in C", because no complete specification of C is available. In Searle's idea, the rule of promising is a constitutive, "count-as" rule. It is roughly of the form, "the utterance of T count as a promise in the context where such and such conditions obtain." Searle presupposes the existence of a linguistic community in which there is a fixed relationship between utterances and their contexts, but in (R1) Hume pays attention to indeterminacy of that very relationship. The analysis of a promise along Searle's line, therefore, is anything but satisfactory from the Humean point of view. Thus, we can not read into Hume neither of the two versions of Austinian conventionalism.

III MEANING AND HUMEAN CONVENTIONS One might object, though, that I have not shown that promises are not rule-governed, but only that the system of rules that govern them is open-ended. There is the central general rule to the effect that a person's utterance of a certain verbal expression counts as a promise, after all, even if we can not completely specify the context in which the rule applies.

Well, yes and no. Certainly we can make a promise only by meaning something, in most cases with words, and to suppose there is a count-as rule would be a convenient way to describe how meaning gives rise to obligation. We must, though, beware of being misled by the idea of a rule. When we talk of a rule, we are easily tempted into thinking that making a promise is more a matter of using a fixed form of words required by a predetermined code than a matter of conveying certain meaning by one or another of indefinitely many possible ways. Let me explain. If you are the foreman of the jury in court, you can bring in a verdict by saying "guilty"* at the right moment, but can not do this by saying "criminal," "responsible," and so on. By contrast, I can promise to give a watch to someone, by saying "I will give you this watch" or maybe even "This watch will be yours" instead of using the so-called conventional formula. What makes my utterance of one of these sentences a promise, rather than a threat, or a simple statement of a fact, is that my giving the watch to the hearer is in the future, that it is within my power to do so, that it is to her advantage and so on. This list can be continued indefinitely. In short, it is the existence of an appropriate situation or context, and not a rule governing the conventional formula that makes the utterance of a sentence a promise. Granted, my utterance of any one of these sentences can fail to constitute a promise, for example, if I have already promised her to give the watch and then confirm by the present utterance my obligation that already exists. In that case, though, I am no more capable of making a new promise by the utterance of the conventional formula, than by the utterance of one of the other sentences. Besides, as even Searle is well aware, we can use the formula "I promise," and make a threat rather than a promise, depending upon the context.** Thus, we do not go any further in explaining the function of words in promise-making by supposing that there are rules that govern the so-called conventional formula.

* The example is from P. F. Strawson, 'Intention and Convention in Speech Acts,' 1964, repr. in Searle (ed.), The Philosophy of Language, Oxford University Press, 1971, p. 26.
** Speech Acts, pp. 58f.

What, then, is the function of words in promises? Páll S. Árdall once argued that promises are statements of intentions.* He was I believe quite right in this. When we make a statement of our intention to do something in the right context, we therein promise to do it. What we then actually say may be "I will do such and such," "I will do such and such. And that's a promise," "I promise to do such and such," or it may be put in still other words. If, as Hume seems to suppose, we can convey by words only what can be at the moment of the utterance in our mind, (T 516) then we can mean no more than that we have a certain intention in uttering a verbal expression of a promise. An intention, though, by itself can not give rise to any obligation. Then there must be something that gives the one who promises reason to perform the act he promises, and gives the one to whom the promise is made reason to expect the performance, and to blame the promisor in case of failure. That something is I believe what Hume calls conventions. I think this is what Hume means when he says, "A resolution is the natural act of the mind, which promises express;** but were there no more than a resolution in the case, promises would only declare our former motives, and would not create any new motive or obligation. They are the conventions of men, which create a new motive." (T 522)

* 'And That's a Promise,' Philosophical Quarterly, 18, 1968, pp. 224-237.
** In some contexts, we may use 'express' in the sense of revealing, as opposed to stating or describing. Hume's use or the word, though, is not restricted in this way. It would not be pertinent here if one should object that expression is not statement.

There is a possible objection to this exegesis, and it seems to have some textual evidence to support it. In the citation I have just made, I stopped deliberately in the middle of Hume's sentence. The full sentence goes as follows: "They are the conventions of men, which create a new motive, when experience has taught us that human affairs would be conducted much more for mutual advantage, were there certain symbols or signs instituted, by which we might give each other security of our conduct in any particular incident." (ibid.) Hume continues, "After these signs are instituted, whoever uses them is immediately bound by his interest to execute his engagements, and must never expect to be trusted any more if he refuse to perform what he promised." (ibid.) Isn't it then natural to understand Hume as saying that the conventions that create the new motive are the rules that govern the newly instituted symbols or signs of promises, after all?

This alternative interpretation, though, goes against Hume's description of the convention that establishes the scheme of promise making and keeping, which appears in the next paragraph. "All of them, by concert, enter into a scheme of actions, calculated for common benefit, and agree to be true to their word; nor is there any thing requisite to form this concert or convention,* but that every one have a sense of interest in the faithful fulfilling of engagements, and express that sense to other members of the society." (T 522f.) Here Hume says that the convention is a concert by which people enter into a scheme of actions, and this concept of a convention does not imply a rule that governs a certain form of words. Thus, it is an act of entering into agreement that establishes a system, but is distinct from the system itself or one of its rules.

* My italics.

It remains to be explained, though, why Hume concentrates his attention to the expression "I promise" and seemingly neglects other verbal expressions equally suitable for making a promises. (T 516, 522) There are I believe two reasons. First, Hume did not make a sharp distinction between a contract and a promise. Thus, in explaining the sense of interest that establishes the scheme of promise making, he speaks of promises, contracts, and engagements without making any distinction between them. (T 522)* A contract is, however, a institutionalized form of agreement and requires a form that is determined by a preexisting code.** Such a codified form is not necessary for a promise, but we who live in a society where both promises and contracts exist, it is not always easy to keep this distinction in mind.

* The Oxford English Dictionary says indicates that the word 'contract' has been employed in a legal sense since it first appeared in the English language.
** See above the example of 'guilty'.

Second, Hume did not have a clear idea of the intentional character of verbal signs. When he says the verbal expression of promises expresses no natural act of the mind because no such act can give rise to obligation, (T 516f.) he seems to presuppose that as far as words are signs of natural acts of the mind, they are functionally redundant accompaniments. If this were the case, we would have to invent new words that express no act of the mind, in order to bind ourselves by words. Words, though, are not mere accompaniments of the things signified, even when they express or signify things that can exist independently from and precedently to them. Thus, we can use words in the absence of the things signified and do so intentionally. They are not natural accompaniments but intentional signs, and insofar as they are intentional they are non-natural.* The substance of what Hume says about the verbal expression of promises, for example in (R1), apply to forms other than the so-called conventional formula, as far as they are non-natural signs of resolutions or intentions. Thus, when we express our intention by an intentional sign, that expression constitutes a promise.

* Cf. H. P. Grice's discussion of natural and non-natural meaning. 'Meaning', 1957, repr. in Studies in the Way of Words, Harvard University Press, 1989.

Let us now summarize the result of discussion in this section. In making a promise, we "in effect express a resolution of performing it," to borrow Hume's words. (T 522) The expression of the resolution or intention constitutes a promise, when it is done with an intentional sign in an appropriate context. What, then, supplies the context and gives us reason to keep our own word and count on other people's word? The answer is, conventions. In this sense we may call a promise conventional, and it is the only sense in which Hume would call it so.

IV CONCLUSION "[The convention] is only a general sense of common interest; which sense all the members of the society express to one another, and which induces them to regulate their conduct by certain rules," says Hume, when he first explains the convention that establishes the rules of justice. (T 490) Both here and in the description of the convention that makes promises possible cited above, Hume refers explicitly to the essential point of conventions, that is, the expression of the sense of interest that is common to every member of society. Thus, it must be clear by now that in Hume's view, conventions are acts of entering into agreement and not a kind of rules or laws. Rule-governed pattern of behaviour may develop on the basis of conventions, but conventions themselves are not rule-governed and can not be written into rules.

Before concluding, let us look at a further reason conventions resist attempts to write them into rules. The reason is that non-natural communication is involved in conventions. In a convention, people express to each other the sense of interest common to them all. This communication of the sense of interest must have the intentional and non-natural character that is typically shown in linguistic communication. Because, as Hume says, "a man, unacquainted with society, could never enter into any engagements with another, even tho' they could perceive each other's thoughts by intuition," (T 516) and if he is capable of entering into conventions, he will be no longer "unacquainted with society," the knowledge of other minds necessary for conventions must be acquired not by intuition or other non-social means but by non-natural, intentional communication which is essentially social.

It might sound strange to say that conventions resist writing into rules because of intentional communication involved, since intentional communication is typically done by means of language, and language is often considered quintessentially rule-governed behaviour. We have, though, seen above how hopeless an attempt is to analyze in terms of rules even such speech acts as promises, which some philosophers regard as obvious examples of rule-governedness of language. The outcome of such an attempt will be irreducibly open-ended, and at the core of this open-endedness lies the essential feature of conventions, that is, the very intentionality of social communication.

In conventions, people intentionally express their sense of interest to each other. Here, though, any person has no guarantee that her expression will be understood and accepted by others, nor does she know whether she understands properly the others' sense of interest. Since there is no predetemined code that ensures the success of communication, entering into conventions is always, to some extent, a leap in the dark. Once a scheme of actions is established by conventions, people try to make it stable by means of codes of morality and laws. These moral and legal edifice, though, is built on the network of intentions of people who keep trying to understand, and make themselves understood to, each other. Promises, though they are supported by previous conventions, remind us of the fallibility of trust in the rules, since they involve trust in future actions that are by nature not yet realized.