REVISIONS OF AUPR

AUPR considered in passing the suggestion that if Y fails to do Ay he thereby ceases to undertake to do Ay, in which case (O2)(b) ceases to generate the performance obligation on X. Even if this is true, it does not solve the problem of Interdependence where X defaults, for (O1) imposes on Y a performance obligation that is not conditional on any undertaking by X. A solution is needed that applies to a default by either party. There are two possibilities: one is to modify the clauses of the models of an agreement, the other to modify the principles of obligation.

A proposal of the former kind is to combine the approaches of the two models, so that one party's undertaking is conditional on the other's undertaking, while the other's is conditional on the first party's action: for example, Y undertakes to X that if X will do Ax Y will do Ay, and X undertakes to Y that if Y will give that undertaking X will do Ax. Interconditional models will be discussed more fully in the next section. The present proposal solves the problem where X defaults, for nothing in the principles of obligation implies in that case that Y has an obligation to do Ay; but it leaves the problem where Y gives his undertaking but fails to do Ay, for in that case (O2)(b) imposes on X an obligation to do Ax.

A second proposal is to keep the two models separate but in each case to expand clause (a) and/or clause (b). Take first the simpler model, (M6). As already noted, this substantially meets Interdependence where Y defaults, so a modification is needed only to deal with default by X. An appealing suggestion is that (b) be expanded to:

• (b1) Y undertakes to X that:

• (i) subject to (ii), Y will do Ay;
• (ii) if X fails to do Ax Y will not do Ay.

((ii) might be amplified to cover the situation where it becomes clear that X is going to fail to do Ax, but the present version is accurate enough.) The undertaking in (b1) is equivalent to an undertaking that if and only if X will do Ax Y will do Ay. But X's undertaking in (M6)(a) is that if Y will do Ay X will do Ax. We are therefore back with interconditional undertakings and the objection that both parties will comply if neither does anything.

Now consider (M4). (b) changes to (b1) and (a) is modified to reflect that change:

• (a1) X undertakes to Y that if Y will give the undertaking in (b1) X will do Ax.

• (a1) leaves the problem of Interdependence where Y gives his undertaking but fails to do Ay; for, as in the case of the proposal to combine the two models, (O2)(b) implies in that case that X has an obligation to do Ax. So it may now be suggested that (a1) be expanded in the way that (b) was expanded to (b1):

• (a2) X undertakes to Y that:
• (i) subject to (ii), if Y will give the undertaking in (b1) X will do Ax;
• (ii) if Y fails to do Ay X will not do Ax.

But the combination of (a2) and (b1) again faces the objection that both parties will comply with the undertakings if X fails to do Ax and Y fails to do Ay.

The result is that the proposed modifications of the models' clauses solve the problem of Interdependence at the price of creating interconditional undertakings. So far as I can see, any other such modification leads to an outcome at least as bad.

The other approach is to modify the principles of obligation. It can be conceded that (O1) and (O2) may have exceptions, for example where X becomes unable to φ or where φing would be a wicked action; but the modifications needed—if they are needed—to accommodate such cases will not solve the problem of Interdependence. It may be objected, however, that (O2) is fundamentally misconceived. Suppose that X, a driver, undertakes that, if his passenger Y asks him to stop the car, he will do so. By (O2)(b), if Y asks X to stop, X has an obligation to stop. Y now asks X to stop. By (O2)(b), X has an unconditional obligation to stop. But, if it is unconditional, the following is false:

• (OS) If X does not stop and Y does not ask X to stop, X does not breach his obligation.

But (OS) is true. So (O2) is false.

The objection compounds two confusions, between internally and externally conditional obligations and between an obligation's being unconditional and its having a condition that is met. X's obligation is not internally conditional: for example, there is no proposition (expressed by “P”) such that, by virtue of (O2)(b), X has this obligation: to stop if (or to stop only if) P. But, as (O2)(b) makes clear, X's obligation is externally conditional: if Y asks X to stop, X has an obligation to stop. Of course it has been assumed that the condition is met, but that does not make the obligation unconditional. Suppose that Y's asking or not asking X to stop is the sole factor determining the question whether X has an obligation to stop. Then Y's asking X to stop is a necessary as well as a sufficient external condition of X's having the obligation. It follows that, if Y does not ask X to stop, X does not have the obligation. Given the point just made about the sole determining factor, it also follows that, if X does not stop and Y does not ask X to stop, X does not have the obligation. But X cannot breach an obligation he does not have. So (OS) is true. Contrary to the objection, therefore, X's having the obligation generated by (O2)(b) is consistent with (OS)'s truth.

Where X defaults on his performance obligation, the problem of Interdependence arises by virtue of (O1). This principle might be denied on the ground that an undertaking only creates an obligation on the giver if the recipient relies on the giver's fulfilling the undertaking.⁵

On this theory the relevant substitution-instance of (O1) must be expanded to:

• (O1R) If:

• (a) Y gives Uy; and
• (b) X acts in reliance on Uy,
• Y has an obligation to do Ay.

((b) might be amplified to specify that the reliance is to X's detriment, but the present version is good enough.) Suppose that, in a situation to which (M4) or (M6) applies, the only candidate for a possible such action in reliance by X is Ax. Then, if X in fact fails to do Ax, (O1R) does not generate an obligation on Y to do Ay. Assume that the situation contains no other ground for such an obligation. Then it substantially meets Interdependence. Now suppose that, in a situation to which (M4) applies, Y defaults on his performance obligation. (As already noted, (M6) does not substantially breach Interdependence where Y defaults, even if (O1) and (O2) are assumed.) In this case the problem of Interdependence arises by virtue of (O2). On the reliance theory of obligation the relevant substitution-instance of (O2), so far as relevant, must be expanded to:

• (O2R) If:
• (a) X gives Ux in (M4); and
• (b) Y acts in reliance on Ux in (M4),
• then, if Y gives Uy, X has an obligation to do Ax.

Suppose that the only candidate for such a possible action in reliance by Y is Ay: then, if Y fails to do Ay, (O2R) does not generate an obligation on X to do Ax. If it is assumed that the situation contains no other ground for such an obligation, it again substantially meets Interdependence.

This response faces difficulties. First, as to the argument concerning default by Y, the most likely candidate for a possible action by Y in reliance on Ux in (M4) is not Ay but Uy; for Ux in (M4) is the undertaking not that if Y will do Ay, but that if Y will give Uy, X will do Ax. Second, there is a large literature attacking the theory that reliance is the ground of the obligations associated with promises: many of the objections extend to undertakings other than promises. Third, I have argued elsewhere that there is no interesting relation of conditionality between reliance and obligation,⁶

in which case (O1R) and (O2R) are unsound. Fourth, neither model of an agreement mentions action in reliance: this is as it should be, for, whether the making or the carrying out of an agreement is in question, it is doubtful that either party need be relying on the other.⁷ Assume this model of reliance:

• X, in doing Ax, relies on Y to do Ay where:
• (I) X does Ax;
• (II) X has the goal Gx;
• (III) If X does Ax, Gx will be achieved if and only if Y does Ay;
• (IV) X believes (III);
• (V) X believes that Y will do Ay;
• (VI) (I) is true because (II), (IV) and (V) are true.⁸

  This model is used in id. at ch. 5. As explained there, a model of “strong” reliance is produced by adding: (VII) X can do Ax'; (VIII) if X does Ax', Gx will be achieved even if Y does not do Ay; (IX) if X does Ax, X does not do Ax'; (X) X believes (VII), (VIII), and (IX). Black, Reliance and Obligation, supra note 6, uses the model of strong reliance.

The discussion can be restricted to reliance in the carrying out of an agreement, for that is at issue in the arguments based on (O1R) and (O2R). (As already noted, this is a weakness of the argument based on (O2R): the best candidate for Y's action in reliance is Uy, which forms part of the making, not of the carrying out, of the agreement.) Imagine that X and Y are flatmates who agree that X will make the beds and Y will wash the dishes. It may be argued as follows that X, in performing his part of the agreement, need not be relying on Y to perform Y's part. Suppose that X's goal is merely that the beds get made and the dishes washed: it does not matter to him who does the chores. Suppose also that, if Y fails to wash the dishes, the cleaner will wash them. Then it is not the case that, if X makes the beds, his goal will be achieved if and only if Y washes the dishes. Hence the relevant substitution-instance of (III) is false. Likewise X may believe that, if Y defaults, the cleaner will wash the dishes: if so, and if X is rational, he will not believe the relevant substitution-instance of (III), in which case the relevant substitution-instance of (IV) is false. X may also not believe that Y will wash the dishes: he knows Y to be lazy and unreliable. In that case the relevant substitution-instance of (V) is false. X may have been going to make the beds anyway, in which case, if “because” in (VI) means “only because,” the relevant substitution-instance of (VI) is false. If “because” means “at least partly because,” the relevant substitution-instance of (VI) is still false if that of (IV) or (V) is false.

These counterexamples are controversial. As to (III) and (IV), it may be replied that, even if X's main goal is just that the beds get made and the dishes washed, X must also have the more specific goal that X make the beds and Y wash the dishes: otherwise he would have no reason to make the agreement with Y. But, if this goal is taken to be Gx, the relevant substitution-instance of (III) is a tautology and hence true. Given that it is a fairly simple one, X is likely to believe it, at least in the dispositional sense of “believe,” in which case the relevant substitution-instance of (IV) is also true. If X is bright enough to make the agreement, he is bright enough to believe the tautology; indeed, an occurrent belief in the tautology is likely to have formed part of the process of reasoning that led him to make the agreement. It might be objected that the nature of belief is such that it cannot have a tautology as content; but even if this is true it may be better to qualify (IV) to accommodate the exceptional case where (III) is a tautology than to concede that the case constitutes a counterexample to (IV). As to (V), if X does not believe that Y will wash the dishes, why did X make the agreement? Perhaps he merely hopes that Y will wash them: but in that case X is relying on Y according to a weaker, but still plausible, model of reliance that expands (V) to include mental states that fall short of belief. A more persuasive counterexample is one in which the agreement is legally enforceable (hardly likely in this domestic situation), X does not even hope that Y will perform, but X entered into the agreement in reliance on the law to provide a remedy for Y's default. Here there is still reliance by X, but not on Y to perform.

There is no need to pursue the dialectic further. If, as seems likely, the conclusion is that the carrying out of an agreement need not involve the parties' relying on each other to perform, (O1R) and (O2R) fail to impose obligations on the parties to an agreement where such reliance is missing. Absent any other relevant principles of obligation, this has the counterintuitive consequence that an agreement of this kind creates no performance obligations on the parties. Such an agreement does not satisfy Interdependence: the question of interdependence does not arise.

If the reliance theory is rejected, it might be proposed that specific exceptions should be built into (O1) and (O2) to cover situations where the undertaking in question is given as part of an agreement. The relevant substitution-instances of (O1) and (O2) are expanded to:

• (O1A) (a) Subject to (b), if Y gives Uy, Y has an obligation to do Ay.
• (b) If clauses (a), (c), and (d) of (M4), or the corresponding clauses of (M6), are satisfied, Y's obligation is cancelled if X fails to do Ax.

• (O2A) (a) Subject to (b), if X gives Ux in (M4), then:
• (i) X has an obligation, to do Ax if Y gives Uy; and
• (ii) if Y gives Uy, X has an obligation to do Ax.
• (b) If (b)–(d) of (M4) are satisfied, X's obligation is cancelled if Y fails to do Ay.

These revisions clash with the strong intuitive judgment that, unless an undertaking is conditional on its recipient's doing a certain thing, the obligation created by the undertaking is, normally at least, not cancelled merely by the recipient's failure to do that thing. (It is frivolous to reply that the obligation is cancelled by the recipient's releasing the giver and that a release is a failure to fail to release.) Absent any independent reason to reject that judgment, the revisions are ad hoc devices to avoid the problem of Interdependence.

It might be suggested that an independent ground for (O1A) is the fact that, as clause (c) of the models of an agreement provides, Y's reason for giving Uy is that X gives Ux: from this, it may be said, it follows that Y's obligation to do Ay is cancelled if X fails to do Ax. But, first, it is obscure how this is supposed to follow and, second, the defense cannot be extended to (O2A), for (M4) does not provide that X's reason for giving Ux is that Y gives Uy. It is not a promising option to try to remove the second problem by adding to (M4) a clause to the effect that X gives Ux for the reason that Y gives Uy, for that would substitute a problem of bootstrapping: the reason-relation includes a causal chain, as argued in AUPR; causes normally precede their effects; hence, if each party's reason for giving his undertaking is that the other gives his undertaking, each gives his undertaking before the other, quod non.

Modification of the principles of obligation has been no more successful than modification of the clauses of (M4) and (M6) as a means of avoiding the problem of Interdependence. This motivates the pursuit of the third and fourth options listed at the start.

INTERCONDITIONAL UNDERTAKINGS

Option three is to produce a quite different undertaking-based model of an agreement. One thought is that accounts in terms of interconditional undertakings have been dismissed too quickly. The objection was that if X and Y give interconditional undertakings respectively to do Ax and Ay, both will comply if neither does anything, whereas they must do those actions if they are to comply with an agreement to do so. It may be added that the latter point follows from the main purpose of making agreements, which is to enable the parties to enjoy the benefits that flow from their performing the agreed actions. Normally those benefits will not arise if the parties do nothing: if they did arise, an agreement would be unnecessary.

The additional point does little to support the objection. That the making of agreements has this purpose does not entail that the purpose is fulfilled in every case. In any event, an interconditional account can explain how the making of agreements promotes the purpose. Suppose that X and Y give the undertakings mentioned at the beginning of the paper: X undertakes to Y that if Y will do Ay X will do Ax, and Y undertakes to X that if X will do Ax Y will do Ay. Even if X may comply with his undertaking without doing Ax, he thereby runs the risk of breaching it: this will happen if Y does Ay. Provided, then, that X wants to comply and is not certain that Y will not do Ay, X is motivated to do Ax—and conversely for Y. In that case their giving these undertakings increases the probability that they will respectively do Ax and Ay and enjoy the ensuing benefits. Hence, if their giving the undertakings constitutes their making an agreement, the making of the agreement promotes the purpose. Moreover their giving the undertakings serves a subsidiary purpose of at least some agreements, which is to give each party an assurance as to what the other will do in specified circumstances.

Stripped of the additional point, the objection to interconditional accounts expands to this argument: X and Y comply with the undertakings if X fails to do Ax and Y fails to do Ay; they do not comply with the agreement if they so fail; if the agreement consists solely of the undertakings, the parties comply with the agreement if and only if they comply with the undertakings; so the agreement does not consist solely of the undertakings. The first step is the most contentious.

The notion of compliance needs scrutiny here. Roughly at least, if X and Y agree that X will do Ax and Y will do Ay, they comply with the agreement if and only if X does Ax and Y does Ay. (Arguably a more accurate formulation would provide that they act with certain intentions, but the question of intention can be set aside in this discussion.) Thus each party complies if and only if he acts in such a way that the part of the agreement's content that concerns him is true. Likewise X complies with his undertaking if and only if he acts in such a way that the content (P) of the undertaking is true. There is a weak and a strong sense of this biconditional's right-hand side: in the weak sense X acts in such a way that P is true if and only if X performs some action and P is true; the strong sense adds the condition that X's performing the action makes P true. We may thus talk of weak and strong compliance. “Action” includes omissions. “Makes P true” can be left unanalyzed: it covers causation and constitution, possibly among other things.

There are various ways of modeling an agreement in terms of interconditional undertakings: the conditionality may be external or internal; the conditions specified may be necessary, sufficient, or something more complex; one party's undertaking may be conditional on the other's giving an undertaking, as in (M4), or on the other's doing the action he undertakes to do, as in (M6). The most attractive model uses the internally conditional undertakings just mentioned:

• (M7) X and Y agree that X will do Ax and Y will do Ay where:
• (a) X undertakes to Y that if Y will do Ay X will do Ax; and
• (b) Y undertakes to X that if X will do Ax Y will do Ay.

(The conditional in clause (a) is equivalent to “X will do Ax unless Y will not do Ay,” for “P unless Q” is equivalent to “P if not-Q”; likewise for clause (b).) Assume for simplicity that the “if”s here are truth-functional: each conditional is true if either its antecedent is false or its consequent is true. Suppose now that X fails to do Ax and Y fails to do Ay. Then each party weakly complies with his undertaking in that he omits to perform and the content of his undertaking is true. But neither party strongly complies, for his omission does not make the content true: the content of X's undertaking is true by virtue of the falsity of its antecedent, and what makes the antecedent false is not X's but Y's omission—and conversely for Y's undertaking. So, in the case of (M7), the first step of the argument against interconditional accounts is true for weak but false for strong compliance. Assuming that the other steps are sound, the objection therefore succeeds for weak but fails for strong compliance.

In further defense of (M7) it might now be argued that the objection's success as regards weak compliance is Pyrrhic, as the concepts of weak and strong compliance are artifices that fail to reflect the intuitive notion of compliance. X strongly complies with the content P of his undertaking if and only if X performs an action that makes P true. But, the argument goes, the truth of the right-hand side of this biconditional is not sufficient for compliance according to the intuitive notion. A fortiori the right-hand side of the biconditional characterizing weak compliance (“… if and only if X performs some action and P is true”) is not sufficient. To see the insufficiency in the case of strong compliance, suppose that X undertakes to Y that if X hears a noise X will shout. X surely does not comply with his undertaking if he plugs his ears, even if by doing so he makes false the antecedent of the undertaking's content and thus makes the content true. The reason is that by plugging his ears X undermines the undertaking's point.

This argument confuses the words X utters with the undertaking he gives. The example is thinly described, but it is natural to imagine that X and Y are hiding in the jungle and are surrounded by enemies. When X says “If I hear a noise, I'll shout,” he knows that Y takes him to intend (probably via a Gricean mechanism) to make Y believe that X will warn Y of danger. Recognizing that the notion of “point” (and of “undermining” the point) can be explicated in various ways, we can say that making Y believe this is the point of X's utterance. The point will be achieved only if Y takes X to mean more than he says, that is, to be giving the conjunctive undertaking (Ux1) that X will listen out for a noise and that if he hears one he will shout. If X plugs his ears, the first conjunct, and hence the whole content, of Ux1 is false. In that case X neither strongly nor weakly complies. Nor does X comply according to the intuitive notion. So the example, expanded thus, does not show that the concepts of strong and weak compliance conflict with the intuitive notion. If, however, X merely undertakes (Ux2) that if he hears a noise he will shout, and if this “if” is truth-functional, he strongly and weakly complies by plugging his ears, but he also complies according to the intuitive notion. So again there is no conflict.

Of course the circumstances in which both parties take X to be giving Ux2 would be odd: we might imagine that Y is a doctor giving some sort of motor coordination test to X. A scene more readily suggested by the argument is one in which X and Y are back in the jungle but X is one of the enemy: he makes his utterance in order to mislead Y. Not wanting to utter a falsehood, he plugs his ears and jesuitically maintains that he intended his words literally. It might be suggested that X here breaches an implied undertaking not to mislead Y: if that undertaking is confused with Ux2, the mistaken result emerges that by plugging his ears X breaches Ux2. But again the result rests on a confusion, as also on the debatable claim that there is any such implied undertaking. (There is also the question of what “implied” means here, and in particular whether it is an alienans adjective, so that an implied undertaking is not an undertaking.)⁹

The defender of (M7) may now reply that the truth-functional reading of “if”, in X's utterance and in Ux1 and Ux2, is another artifice: “if” normally signifies not merely that the antecedent is false or the consequent true but that there is good reason to infer the consequent from the antecedent.¹⁰

If the two undertakings are understood thus, the result for Ux1 stays the same: regardless of the conditional second conjunct of Ux1's content, X's plugging his ears makes false the first conjunct: X thereby fails to comply strongly, weakly, or according to the intuitive notion. But, the defense runs, the result for Ux2 changes. Assume that X's making his utterance provides a good reason to infer that X will shout from the proposition that he hears a noise. If X plugs his ears, our intuition is that he does not comply with Ux2. Nor does he strongly comply, for his plugging his ears does not make it true that there is good reason to make that inference. But he weakly complies, for he performs some action and the content of Ux2 is true. Weak compliance thus diverges from intuitive compliance.

There are two flaws in this defense. First, if Ux2's “if” is given the proposed strong reading, it is not intuitively clear that X fails to comply with Ux2: my intuition, at least, is uncertain. Second, it may be retorted that it is the strong, not the truth-functional, reading of “if” that is an artifice. This raises the old debate over the relation between “if” and “⊃,” which I shall not pursue. The standard argument for their equivalence invokes rules of implicature.¹¹

(Such rules may be invoked to justify Y in taking X to be giving Ux1 in making his utterance.) If that or some other argument works, the defense of (M7) fails and no persuasive reason has yet been given for rejecting the concept of weak—or strong—compliance. In that case the earlier conclusion stands: as regards (M7), the objection to interconditional accounts succeeds for weak but fails for strong compliance.

Irrespective of compliance, and thus of that objection's success, there is reason to reject (M7), for it seems that (M7) represents at best an attenuated and peripheral concept of agreement. If X and Y give their undertakings, one or other is likely to press to make the position clear: X might say, “So, do we have an agreement that I'll do Ax and you'll do Ay?” If Y replies “Yes,” they have an agreement in a sense different from that represented by (M7). That sense is captured by the offer-acceptance model presented in the next section.

Before that, it is worth briefly reviewing some other interconditional models. For this purpose I continue to use the truth-functional “if” and the concepts of weak and strong compliance. One proposal is to reverse the direction of conditionality in (M7):

• (M8) X and Y agree … where:
• (a) X undertakes to Y that only if Y will do Ay will X do Ax; and
• (b) Y undertakes to X that only if X will do Ax will Y do Ay.

Suppose again that the parties fail to perform. Then, as before, each weakly complies with his undertaking. But in this case each party also strongly complies. Consider X's undertaking. Its content is equivalent to “If X will do Ax, Y will do Ay.” If X fails to do Ax, he makes this conditional true by making its antecedent false. Hence, in the case of (M8), the first step of the argument against interconditional accounts is true for both forms of compliance. Assuming that the other steps are sound, (M8) should be rejected outright.

The next proposal is to amend (M7) so that each party's undertaking is internally conditional on the other's giving an undertaking rather than acting:

• (M9) X and Y agree … where:
• (a) X undertakes to Y that, if Y will undertake to X that Y will do Ay, X will do Ax; and
• (b) Y undertakes to X that, if X will undertake to Y that X will do Ax, Y will do Ay.

Clause (a) is the same as in (M4). (M9), like (M4), breaches Interdependence, for by (O2)(b) each party has an obligation to perform whether or not the other performs.

Standing to (M8) as (M9) stands to (M7) is:

• (M10) X and Y agree … where:
• (a) X undertakes to Y that only if Y will undertake to X that Y will do Ay will X do Ax; and
• (b) Y undertakes to X that only if X will undertake to Y that X will do Ax will Y do Ay.

This model fails because each party weakly and strongly complies with her undertaking if neither performs.

Corresponding to (M7)–(M10) are externally conditional undertakings. The counterpart of (M7) is:

• (M11) X and Y agree … where:
• (a) if Y does Ay, X undertakes to Y that X will do Ax; and
• (b) if X does Ax, Y undertakes to X that Y will do Ay.

(M12), the counterpart of (M8), reverses the direction of conditionality in (M11). (M11) and (M12) are implausible because, assuming (as above) that the parties' giving the undertakings constitutes their making the agreement, the models entail that the making of the agreement depends on its performance.

The counterpart of (M9) is:

• (M13) X and Y agree … where:
• (a) If Y undertakes to X that Y will do Ay, X undertakes to Y that X will do Ax.
• (b) If X undertakes to Y that X will do Ax, Y undertakes to X that Y will do Ay.

(M14), the counterpart of (M10), is the same as (M9) but with (a) and (b) reversed. (M13) and (M14) breach Interdependence. Suppose that Y gives his undertaking. By (a), X gives his. By (O1), each has an obligation to perform whether or not the other does.

It would be tedious and fruitless to examine further interconditional models. None of (M8)–(M14) is an improvement on (M7) and (M7) is unsatisfactory. I can think of no other kind of undertaking-based model that would avoid the objections.

THE OFFER-ACCEPTANCE MODEL

The fourth option is to model an agreement otherwise than in terms of undertakings. A model that will commend itself to lawyers uses concepts familiar from the law of contract:¹²

X and Y have an agreement where X makes an offer to Y that Y accepts. English law also includes consideration as an element of a contract, but consideration is not necessary for there to be an agreement. The content of the offer and acceptance, and thus of the agreement, is normally a conjunctive, not a conditional, proposition of the form “X will do Ax and Y will do Ay.”¹³ To accommodate such a content, “offer” and “accept” must be understood broadly: perhaps “propose” and “assent” would be happier terms.¹⁴ The relevant concepts of offer and acceptance are tolerably clear at an intuitive level, and in developing the model we can help ourselves to—without necessarily accepting all of—the detailed doctrines of offer and acceptance that contract law provides: hence my preference for this terminology. Neither the offer nor the acceptance by itself imposes an obligation on either party to perform his part of the agreement.¹⁵ (In certain circumstances an offer may, in the absence of an acceptance, impose some other obligation on the offeror, e.g., to keep the offer open for a certain period.) But normally, if an agreement is made, each party has a cancelable obligation: roughly, X's obligation is to do Ax and it is cancelled if Y fails, or it becomes clear that Y is going to fail, to do Ay—and conversely for Y.

The parties' obligations can be specified more accurately if times are assigned to the offer, the acceptance, and the actions that X and Y agree to perform.¹⁶

Suppose that at T1 X makes an offer which at T2 Y accepts. The agreement thus comes into being at T2. The content of the agreement is that at T3 X will do Ax and at T4 Y will do Ay. Normally T1 precedes T2, which precedes T3 and T4. T3 may be earlier or later than or simultaneous with T4. At T2 X gets an obligation to do Ax at T3. Unless the obligation is cancelled, it continues up to and including T3. If at T3 X does Ax, the obligation ceases. If at T3 X fails to do Ax, the obligation likewise ceases, but is replaced by an obligation—also cancelable—on X to compensate Y for the failure. At any time from T2 to T3, the obligation is cancelled if Y releases X from it or if Y fails, or it becomes clear that he is going to fail, to do Ay at T4. (Y only fails during that time to do Ay at T4 if T4 is no later than T3.) Y's obligation is symmetric with X's. Just as the law of contract amplifies the concepts of offer and acceptance, so its doctrines of performance and breach can be used to refine further this account of the parties' obligations.

It might be objected that this account is only terminologically different from one based on undertakings: what it is for X to offer that X will do Ax and Y will do Ay is merely for X to undertake to Y that if Y will do Ay X will do Ax, and what it is for Y to accept the offer is merely for Y, in response, to undertake that Y will do Ay. In that case we are back with (M6) and the problem of Interdependence.

“Undertaking” is an elastic term and it can be accepted that there may be an extended sense in which offer and acceptance are undertakings;¹⁷

but they are not undertakings in the sense discussed earlier and they are not of the forms just described. One difference between an undertaking and an offer is that normally an undertaking cannot be withdrawn whereas (this is the general rule in English contract law)¹⁸ an offer can be withdrawn at any time before it is accepted. This reflects the difference, already noted, between the relations that offer and acceptance, on the one hand, and undertakings, on the other, bear to obligations: an undertaking by itself normally creates an obligation, for example, of a kind covered by (O1) or (O2), but both an offer and an acceptance are normally needed to create an obligation.¹⁹ If that is not obvious, it may help to revert to the terminology of proposal and assent: if X says to Y “I propose this: I'll make the beds and you'll wash the dishes,” he surely does not thereby put himself under an obligation to make the beds. He might do in odd circumstances, for example where X has previously told Y that if X says those words he intends to commit himself to making the beds; but in that case X's “proposal” amounts to an undertaking. This response is not an ad hoc device to elude the objections to undertaking-based models, for the concepts of offer and acceptance, and an understanding of their relations to obligation, are established in both our informal practice and contract law. A historical account of the concepts' development would reinforce the point.²⁰

Given that each party complies with the agreement if and only if he acts in such a way that the part of the agreement's content that concerns him is true, and given that, according to the offer-acceptance model, the content of the agreement is that X will do Ax and Y will do Ay, each party complies only if he performs the action specified for him by the agreement: this is true whether “acts in such a way that … is true” is given the strong or the weak sense. The model thus is not exposed to the objection to models using interconditional undertakings, that the parties will comply if neither performs. But it may be argued that the offer-acceptance model falls to an analogue of that objection, which applies to the obligations associated with the agreement: the new objection is that if both parties fail to perform they will not breach any such obligation. X's obligation to do Ax is cancelled, it has been said, if Y fails, or it becomes clear that Y is going to fail, to do Ay. This entails that X's obligation is externally conditional: if Y does not do Ay, X does not have the obligation and a fortiori does not breach it. Absent any other obligation associated with the agreement, X does not breach any such obligation if Y fails to perform. The same goes for Y.

This objection ignores the temporality of the obligations and thus fails to distinguish between a party's never having an obligation and his ceasing to have one. It has already been granted that X ceases to have his obligation at T3 (the time for X to do Ax) and at, any time between T2 (the time Y accepts and the agreement is made) and T3, if Y releases X from it or if Y fails, or it becomes clear that he is going to fail, to do Ay at T4 (the time for Y to do Ay). But normally, from T2 to the earliest time any of these events occurs, X has the obligation to do Ax. Suppose that T1 (the time X makes his offer) precedes T2, which precedes T3, which precedes T4, and that at T3 X fails to do Ax. X thereby breaches his obligation and cancels Y's. Y, on that basis, decides not to do Ay. In that case neither party performs, but one of them is in breach of an obligation associated with the agreement.

The unusual case might arise in which neither party gets an obligation, or at most each gets one that exists for a scintilla of time. This will happen if at T2 an event occurs that cancels the obligations; for example it becomes clear that both X and Y are going to fail to perform. If neither performs, it can here be said that the parties breach the agreement without breaching any obligation associated with it. The same applies to the more usual case where the parties fail to perform and their obligations have been cancelled at some time later than T2 but before the time for performance. If the objector resists the idea that the parties can breach an agreement without breaching obligations, the situation can be redescribed in either of two ways. First, it might rather artificially be said that in both these cases there are obligations that are breached—evanescent obligations in the former case—but that at the time of breach the obligations have ceased. Second, a distinction might be drawn between breaching an agreement and not complying with it: a party breaches an agreement if and only if he does not comply with it and his noncompliance constitutes a breach of an obligation associated with the agreement. But there seems no good reason to deny that an agreement can be breached without a breach of obligation.

Specifying times does not provide a way of modifying the principles of obligation to save the models in AUPR from the objection that they violate Interdependence. Consider Y's undertaking to X (in clause (b) of (M4) and (M6)) that Y will do Ay. If times are introduced, (b) becomes:

• (bt) At T5 Y undertakes to X that Y will do Ay at T6.

Most of the things true of X's accepted offer are true of Y's undertaking: at T5 Y gets an obligation to do Ay at T6; unless the obligation is cancelled, it continues up to and including T6; if at T6 Y does Ay, the obligation ceases; if at T6 Y fails to do Ay, the obligation also ceases but is replaced by an obligation to compensate; at any time from T5 to T6, the obligation is cancelled if X releases Y from it. But, unlike the case of the accepted offer, it is not true that Y's obligation is cancelled if X fails, or it becomes clear that X is going to fail, to do Ax. So there is still a breach of Interdependence. Again this distinction is not ad hoc: it reflects our understanding of the nature of undertakings and their relations to obligations.

Granted that normally where there is an agreement the parties have obligations of the kind specified in the offer-acceptance account, it is a further question what the grounds are of those obligations—in other words, what the states of affairs are in virtue of which those obligations obtain (“ground” and “in virtue of” will be left unanalyzed here).²¹

A simple thought is that the ground of each party's obligation is the fact that the parties have made the agreement. But that might not be so: it could be, for example, that their making the agreement is a cause of the grounds of the obligations associated with the agreement. That view might be taken by someone who holds that the ground of each party's obligation is the fact that the other party relies on him to perform. This reliance theory (an analogue of the reliance theory, discussed earlier, of the obligations associated with undertakings) may seem to have the attractive feature that it explains the cancelability of each party's obligation where the other party fails to perform: if Ax would be X's action in reliance, but X fails to do Ax, Y's obligation to do Ay is cancelled. But this is a mistake, for on the reliance theory Y's obligation is not cancelled: it never arises. The theory in fact fails to make sense of the obligations described by the offer-acceptance account: they arise at T2, the time at which the agreement is made, but the obligation attributed to Y by the reliance theory normally arises, if at all, at T3, the time at which X is to do Ax. T2, as already noted, precedes T3.

There is a large literature on the grounds of the obligations associated with contracts and much of it can be extended to apply to those associated with noncontractual agreements. Take again the obligation on Y: if its ground is not the fact that X relies on Y to do Ay, it might be the fact that Y intends or believes or reasonably believes that X will thus rely on him; or that Y authorizes X to rely on him; or that X expects or reasonably expects, or that Y intends (etc.) that X will, or authorizes X to, expect or reasonably expect, that Y will do Ay; or that X trusts Y to do Ay (here and in some of the following options there are the same variations as before, in terms of intention, belief, and reasonableness); or that X suffers harm if Y does not do Ay; or that Y benefits if X does Ax; or that Y assures X that Y will do Ay; or that Y voluntarily assumes his obligation; or that Y exercises a normative power; or that Y intends to convey a truth about his future behavior, so that his obligation is one of veracity; or that if Y fails to do Ay he lacks honor, integrity, fidelity, or some other such desirable property. Then there are theories that purport to identify aspects of the whole agreement, for example the fact that it represents the will of, or a consensus between, the parties, or a meeting of their minds;²²

or that it constitutes a bargain or an exchange; or that it permits cooperation between the parties; or that it allocates risk between them in a way that reflects their abilities to bear it; or that it is economically efficient; or that it transfers rights between the parties; or that it is an instance of an institution or practice that has constitutive obligation-generating rules; or that there is a “categorical” rule²³ of compliance with agreements. More broadly the ground of the obligation might be characterized in the terms of a general moral theory, for example that compliance maximizes utility.

I shall not adjudicate between these accounts or consider whether some other one might be better; it may be that full generality is unattainable and that different accounts are needed for different cases.²⁴

An account in terms of some feature of the whole agreement may be more attractive, as it distances the obligations associated with agreements from those associated with undertakings: many of the features mentioned in the first half of the previous paragraph have in fact been invoked to explain promissory obligation. As already noted, the relations borne to obligations by offer and acceptance and by undertakings are different. In any event, there is no reason to think it harder to account for the ground of the obligations associated with a noncontractual agreement than for those associated with a contract.

Does the offer-acceptance model meet the four criteria of adequacy? The idea of Symmetry is that both parties do the same thing. This criterion is indeterminate, for sameness is relative to description: X and Y do the same thing in making an agreement that X will do Ax and Y will do Ay, but different things in that X offers and Y accepts. If this difference amounts to a violation of Symmetry, the criterion should be discarded, for the difference concerns only the process leading to the agreement: as noted in AUPR, it is a normal part of this process that one party “goes first.” Once the agreement is made, there is symmetry between X and Y both in the content of the offer and the acceptance and in the obligations to which the parties are subject.

Gilbert expresses Obligation by saying that “the agreement directly generates the relevant performance obligation for each of the parties.”²⁵

She does not say what “directly generates” means, but intuitively the model meets the criterion: normally, from the moment the agreement is made, each party is subject to a performance obligation. By the same token, the model meets Simultaneity, which requires that the parties acquire their obligations at the same time. It has already been seen that the model meets Interdependence.

The offer-acceptance model thus meets the criteria as well as does Gilbert's theory of agreements as “joint decisions,” but it avoids the obscurities of Gilbert's account, notably her claim that the parties constitute a “plural subject.”²⁶

It also avoids the objections to undertaking-based models. But it might be claimed that these are superior to the offer-acceptance model in the respect that the relevant concepts of offer and acceptance are, but that of an undertaking is not, dependent on the concept of an agreement, in this sense: the concept of an undertaking can be fully grasped by someone who has no grasp of the concept of an agreement, whereas a grasp of the latter is presupposed by a full grasp of the concepts of offer and acceptance. Substitution of “agree” for “contract” in a standard definition of a contractual offer²⁷ yields: an offer is an expression of willingness to agree on specified terms, made with the intention that it shall become binding as soon as it is accepted by the person to whom it is addressed. Agreement is modeled in terms of offer and acceptance, which in turn are understood in terms of agreement: this, it seems, is a vicious circle.

One reply would be to say that the concept of an undertaking can likewise only be fully grasped by someone who grasps the concept of an agreement: but this would merely expose undertaking-based models to the objection of circularity. Another reply would be to argue that the concepts of offer and acceptance do not depend on that of an agreement. The best reply is that it is no objection to a model that the concepts it employs (offer and acceptance) depend on the concept it represents (agreement). Where this is the case, the model may still be illuminating if the employed concepts can be sufficiently, albeit incompletely, grasped without a grasp of the represented concept. A sufficient grasp is often possible where connections can in turn be traced between the employed concepts on the one hand and, on the other, further concepts which themselves can be sufficiently grasped without a grasp of the concept represented. This is the case with the offer-acceptance model. In the first place, contract law's highly developed doctrines of offer and acceptance—for example, on offer or acceptance by conduct, the distinction between offers and invitations to treat, continuing negotiations, communication of acceptance, acceptance in ignorance of an offer and termination of an offer—connect the concepts of offer and acceptance with various other concepts that do not depend on the concept of an agreement. Second, it is arguable that the concepts employed in the model are affiliated to other, simpler, concepts of offer and acceptance which are themselves independent of the concept of an agreement. If one caveman holds an arrowhead out to another and the other one takes it, there may be said to be an offer and an acceptance, but it might also be said that there is no agreement, nor even a protoagreement, between them to the effect that, say, the offeror will allow the offeree to keep the object. The point is debatable—if there is primitive offer and acceptance, why is there not also a primitive agreement?—but need not be settled, for the point about the doctrines of offer and acceptance is enough to defend the offer-acceptance model against the charge of circularity.

A final objection to the offer-acceptance model is that it does not fit the case in which an agreement is brought about by a third party. Suppose that X and Y by themselves cannot agree what to do. In order to settle things, Z suggests to both of them that X do Ax and Y do Ay. X and Y say OK. It seems that there is then an agreement between X and Y but that neither has made an offer which the other has accepted. One response, as noted, is to concede that this is a different case or concept of agreement from the one the model is intended to capture. The response is not ad hoc, for the example is unusual. But it may also be argued, in either of two ways, that the model fits the example. First it may be said that Z makes an offer which X and Y accept, and so the agreement has three parties. This strains the normal meaning of “offer,” but it has already been acknowledged that “offer” in this context has a broad sense close to “propose”: there is nothing odd in describing Z as making a proposal. The example still diverges from the standard case in that the content of the offer, the acceptances and the agreement does not specify an action by the offeror; but again it has already been allowed that such an action is normally, not always, specified by the content.

The second argument is that X and Y have an agreement to abide by Z's suggestion and that this agreement fits the model: it might be, for example, that X and Y are a management and a trade union in dispute, that X proposes that they settle the issue by referring it to an arbitrator, Z, and that Y accepts the proposal. Of course the situation need not be like this; Z might simply come along and make his suggestion. But even here there is likely to be an offer by X that is accepted by Y: X may say to Y “What do you think of Z's idea? Shall we go along with it?” and Y may say “Yes,” or they may achieve the same result less articulately, say by significant eye contact. In the odd situation where X and Y merely respond to Z and do not interact at all with each other, it is intuitively doubtful that there is an agreement between them.