Ethics 95 (105) 604-612

Can two promises add up to an agreement? Not according to Margaret Gilbert.1 She has forcefully challenged the orthodox view that an agreement is an exchange of promises. She works through an intricate series of examples of promise-exchanges and argues that none qualifies as an agreement. Assuming that she has not overlooked any plausible candidates, she concludes that agreements are essentially different. It seems, however, that her examples are all exchanges of promises only in an attenuated sense of "exchange." I propose to defend the orthodox view by considering what it is for promises to be exchanged in a fuller sense and what sorts of promises thus exchanged can comprise an agreement.

Gilbert begins with some sample agreements designed to illustrate what is essential to the obligations created by agreements. Her focus is on two-person "agreements on acts," in which each party undertakes a "performance obligation." As illustrated by an example in which Rita and Peter agree that she is to walk Fido and he is to groom Tibbles,2 the two obligations are created simultaneously and are interdependent. Then Gilbert argues, in light of her series of progressively modified examples, that no sort of promise-exchange can yield the obligations of an agreement and yet satisfy both the simultaneity and the interdependence criteria. In some cases the obligations are not interdependent, because each would exist even if the other did not; in particular, default on one would not nullify the other. In certain other cases, where the promises are so framed as to create only conditional obligations, the result is too weak. In the extreme case, the parties find themselves in a kind of Alphonse-and-Gaston situation in which the conditions for performance never get fulfilled ("After you," "No, after you, "No, after you," etc.). Other things go wrong in other cases. So, Gilbert concludes, a mere sequence of promises cannot do what is done by an agreement, whose "simultaneously generated unconditional obligations are interdependent" (p. 645).3 In her view the reason the promise-exchange model fails is that it does not capture the element of "joint decision" and "joint commitment" that characterizes agreements (p. 646).

I do not contest Gilbert's analysis of her series of examples, but I do question her implicit assumption that an exchange of promises is a sequence of acts performed in response to one another, like an exchange of blows or of pleasantries. With just two parties, as in her examples, a promise-exchange is simply a sequence of two promises, one made in response to the other.4 Her contention is that no such sequence, whether the promises are unconditional, conditional, or mutually conditional, can add up to an agreement. But is such a sequence of promises really an exchange of promises?

Forget promises and agreements for the moment and compare gifts and property-exchanges. Consider simple purchases or trades, made face-to-face and consummated on the spot.5 I give you $10 for a CD. You give me a CD player for my camcorder. These are exchanges, not pairs of gifts. I don't really give you the $10 or the camcorder, and you don't really give me the CD or the CD player. Of course these items change hands and in that attenuated sense we "give" them to each other, but they are not gifts. Giving is unilateral. At Christmas we might, as we say, "exchange" gifts, but this is not really an exchange. If you give me a CD player and I give you a camcorder, my coming to own the CD player does not depend on your coming to own the camcorder, nor vice versa. But in a genuine exchange, a reciprocal transfer of ownership, each party gains ownership only if and when the other does.

Now no one has ever suggested that a trade is an exchange of gifts. No one supposes, just because we use the word "give" as we do, that a trade consists of a pair of gifts (mutually conditional gifts?) or, just because we describe what we do on certain holidays as "exchanging gifts," that symmetrical gift-giving is an exchange in the (full) sense of making a trade. In short, no one would seriously ask "Is a trade an exchange of gifts?" in the way Gilbert asks "Is an agreement an exchange of promises?" Of course she answers her question negatively, and asks it only because the orthodox view is intuitively plausible and widely held. Unfortunately, in focussing on the word "promise" she neglects the word "exchange," at least as understood in its full sense. In any kind of exchange in that sense, the parties each do something for something. I pay you $10 for your CD (you sell me your CD for $10). I exchange my camcorder for your CD player. An exchange is inherently reciprocal. Now if in a trade the parties exchange goods, what do they exchange in an agreement? They would prefer to exchange future actions directly, but they can do only the next best thing, namely, exchange commitments to act.6

It is ultimately a verbal question whether the ordinary notion of a promise does or does not preclude the possibility of reciprocal promises, as implicated in the promise-exchange model of agreements. There is no sense lapsing into a verbal dispute on this question, and Gilbert is surely right to insist that what is important is not what we call things but the range of possibilities we recognize. Even so, she stipulates that "one of the rules of procedure here is to take the exchange in question at face value. In particular, each person makes a promise, and no more than a promise" (p. 634).7 But she also prefers to "construe the notion of an exchange of promises broadly," i.e., in a weak sense of "exchange," and thus does not consider any promise-exchanges in the narrower but richer sense of the word. As a result, she excludes from consideration the possibility that what the promise-exchange model requires is not an extended notion of promising but a restricted notion of exchange, as in exchanges of property.

In order to appreciate the parallel between an exchange of property and an exchange of promises, we should recognize a distinction that may be obscured by the act-object ambiguity of the noun "promise," a kind of ambiguity shared by a great many nominalized verbs. It can mean either the act of promising (making a commitment) or what is promised (the content of the commitment). Acts of promising can be exchanged only in the attenuated sense of Gilbert's promise-exchanges, one in response to the other, but things promised, much like property, can be exchanged in the full sense. In an exchange of property each party acquires the right to a certain item belonging to the other, and in an exchange of promises each party acquires the right to the other's performance of a certain action.

There is a further parallel between (simple) trades and agreements. Ordinarily, they are not made in one stroke. Rather, somebody goes first, by making an offer, and counter-offers may follow. A trade or an agreement is consummated and takes effect when an offer is accepted. Now what is an offer? The offer in a trade is a conditional commitment to give up one thing for another, where the condition is the other's acceptance. Acceptance includes a reciprocal commitment. When the offer is accepted (if it is a bona fide offer and not a mere suggestion), ownership of the two items changes hands. In the case of an agreement on acts, an offer is not merely the suggestion for one party to do one thing and the other to do another. An offer is a conditional promise to do one's part in such a arrangement.8 Its condition is the other's acceptance. And acceptance of an offer is an unconditional promise by the other to do his part in the arrangement.

So Gilbert is perfectly correct to say that an agreement is a joint decision that yields a joint commitment. However, this does not show, or even suggest, that an offer and its acceptance are not promises in the ordinary sense. Gilbert's dilemma for the promise-exchange model, that the result of a promise-exchange is either too strong, if the requisite promises are unconditional, or too weak, if they are conditional, does not take into account the nature of offers and acceptances. Both are promises. An offer is a conditional promise, but its acceptance by the other party turns the conditional obligation it creates into an unconditional one.9 Acceptance removes the condition by satisfying it. On the "offer-acceptance" model, as we may call it, the combination of offer and acceptance creates the pair of reciprocal unconditional obligations that comprise an agreement.10 These are the promises (the things promised) which are exchanged in an agreement.

It might be objected that even if a promise-exchange of the sort just described does result in an agreement, the agreement itself can be effected only by a joint action, like shaking hands. In that case it is not the offer-cum-acceptance but the joint action, the handshake, that creates the commitments made by the two parties, and, so the objection continues, engaging in this joint action is not making a promise (in any ordinary sense). However, in these circumstances the handshake does not effect the agreement at all but merely acknowledges it. For when the parties shake hands, they are not making new commitments but only acknowledging the ones they have already made. When an offer has been already made and accepted, shaking hands functions merely as a conventional means for both parties to acknowledge their obligations (their own and each other's). The situation is different if the offer has not yet been accepted. In that case, to shake hands is to accept (shaking hands serves as a conventional means for the second party to accept the offer, as well for the offering party to acknowledge both obligations). The offer-acceptance model still applies.

There is a third sort of case, however, which does not fit the offer-acceptance model. Two people can enter into an agreement without either party having made an offer. Though not common, this can happen if a third party suggests a plan and engineers its joint acceptance: "I have an idea: you do A and you do B. OK? Shake hands." The two parties shake hands and the agreement is thereby effected. This case does fit Gilbert's conception of an agreement, since the two parties are not making promises in any straightforward sense and clearly are engaging in a joint action. Even if there is some sense in which they are making promises, the offer-acceptance model cannot explain why. The symmetry of the case prevents it from fitting that model. However, this is an exceptional case and, even if it does show that not every agreement is an exchange of promises, it hardly shows that none is.

None of Gilbert's examples of promise-exchanges consists of an offer and an acceptance, but several come fairly close. Let us take up two of them and see why they do not fit the offer-acceptance model. Consider her Promise-exchange 8.

Peter: "I promise to groom Tibbles, if you promise to walk Fido."11

Rita: "I promise to walk Fido."

This does not yield an agreement, according to Gilbert, because "the parties are in the situation they would have been in had they issued simple unconditional promises in the first place" (p. 642). One reason she says this is that "Peter incurs a promissory obligation at once." This is true, but it is consistent with Promise-exchange 8's being an agreement, inasmuch as Peter's obligation is merely conditional (until Rita removes its condition). The real problem here is that Rita's obligation is independent of Peter's. How, then, does the obligation created by acceptance of an offer differ from that? If Rita were to accept an offer by Peter, namely for him to groom Tibbles and for her to walk Fido, then she would be obligated to walk Fido as part of an arrangement in which he is also to groom Tibbles. Her obligation would not be independent.

After dismissing two embellishments on Promise-exchange 8, which also fail to meet the interdependence criterion, Gilbert takes up two "exchanges of unconditional promises plus special understandings" (pp. 643-4). On the more plausible one, there are two special understandings designed, respectively, to incorporate the simultaneity and the interdependence conditions: (a) the first promise is not to take effect until the second has been given, and (b) if (once both promises are in effect) one promise is broken, the other promise is nullified, and no further obligation arises out of it.12 Gilbert argues that not only does this understanding leave the interdependence criterion unmet, it "fails to respect the nature of promising," by "fail[ing] to respect the fact that, if you have made an unconditional promise that still stands, you are obligated by that promise irrespective of whatever other promises are kept or broken" (p. 644). In her view, the present proposal "is an attempt to capture a different notion that has certain things in common with that of a promise-exchange" (p. 644). In my view its defect is that it fails to capture what is essential to an exchange of promises. The special undertandings attempt to build features of exchanges in general into the contents of the particular promises being exchanged. These are features common to exchanges of promises and of property. In both cases, at least ordinarily, the person making the offer proposes a plan involving both parties and commits himself to one part in that arrangement; the person accepting the offer commits himself to the other part in it.

Now Gilbert would object that the efficacy of an offer and its acceptance cannot be accounted for merely by appealing to the general nature of promises. That is true, for our explanation appeals also to general features of exchanges. However, she would further object that this is enough to show that the offer-acceptance model does not really account for the force of agreements, at least not insofar as it purports to be a special case of the promise-exchange model. Recall that Gilbert's procedure is to ask, regarding each promise-exchange, "If precisely these promises had (in effect) been exchanged, would the stated criteria [of an agreement] have been met?" (pp. 634-5), where it is stipulated that "each person makes a promise, and no more than a promise" (p. 634). As Gilbert applies it, this procedure presupposes that the terms of a putative agreement are to be given entirely by the contents of the two promises that make it up. It is clear from the wording of her examples that Gilbert expects the terms of agreement to be fully expressed once the promises are spelled out in full. As a result, her procedure effectively requires the generic conditions on agreements to be incorporated into the terms of particular agreements. Such a requirement is surely excessive.13 If we adopted Gilbert's procedure, then not only could we not explain the force of exchanges of promises as agreements, we could not even explain the force of simple promises.

To see why, suppose we demanded that the general conditions on promises be incorporated into the contents of each particular promise. Two such conditions are that a promise doesn't go into effect if the promisee rejects it and that the promisee has no right to place obstacles in the way of the promisor. Should we then suppose that an ordinary promise is conditional on its not being rejected and is qualified by a no-obstruction clause? It would be silly to suppose that an ordinary promise, when made fully explicit, takes the form, "On condition that you do not object, I promise to do A-unless you obstruct me." That would make all promises both externally and internally conditional (in the sense of note 11). Just as it would be unreasonable to expect specific promises to encode the conditions on promises in general, so it is unreasonable to demand that fully explicit agreements specify the general conditions on agreements. Just as we cannot expect to account for the force of promises without invoking the general conditions on promises, so we cannot to expect to account for the force of promise-exchanges without invoking general conditions on exchanges (as well as the conditions on promises). Yet this is what Gilbert does in regard to the examples we just considered. Both Promise-exchange 8 and the promise-exchange with the "special understanding" attempt to incorporate general conditions on agreements.

Thus it is unreasonable for Gilbert's rules of procedure to prohibit invoking general conditions to explain how, on the offer-acceptance model, an exchange of promises could have the force of an agreement. It is no objection to that model that the wordings of the offer and its acceptance do not capture the general conditions on agreements. Those conditions are accounted for by the fact that the two promises are being exchanged. In the case of Peter and Rita, for example, the dialogue would have to take something like the following complex form:

Peter: "I offer to groom Tibbles, as part of a plan in which I am to groom

Tibbles and you are to walk Fido, provided that you accept my offer."

Rita: "I accept your offer: I promise to walk Fido as part of a plan in

which you are to groom Tibbles and I am to walk Fido."

The following simple form is explicit enough:

Peter: "I offer to groom Tibbles in exchange for your walking Fido."

Rita: "I accept."

Peter does not have to say that his offer is conditional on acceptance. That is in the nature of an offer. Nor does Rita have to say that her acceptance constitutes a promise, much less make explicit that it is being given in exchange for Peter's promise. That is in the nature of an acceptance. It is no objection to the offer-acceptance model that the two promises do not incorporate, either singly or jointly, the general features of exchanges.

The offer-acceptance model differs substantively from the generic promise-exchange model that Gilbert applies to her examples. In particular, it puts the simultaneity and the interdependence criteria in a different light. First, on the offer-acceptance model the obligations created by the two promises that ordinarily make up an agreement are not really created simultaneously. The offer creates an obligation before its acceptance creates one. Thus the offer-acceptance model violates the letter of Gilbert's simultaneity criterion, but it does fulfill the spirit of that criterion, for the condition on the first obligation, the obligation created by the offer, is removed when the offer is accepted, and that occurs precisely when the accepting party's obligation is created.14

Second, what is the status of interdependence on the offer-acceptance model? Can that model explain how default on the part of either party nullifies the other's obligation to do his part? Here is one suggestion. On the offer-acceptance model, what is agreed upon is a pair of actions. The agreement is not for each to do one thing iff the other party does the other, for then it would be fulfilled even if neither party did anything. Rather, since what is agreed upon is a pair of actions, one for each to perform, each party has an unconditional obligation to do his part. Then if one party defaults, the other cannot do his part in that pair. He can perform the same action he would have performed,15 but he cannot perform it as a member of the pair. It was his obligation to perform the action not unilaterally but as a member of the pair, and this is precluded by the other's default. On the present suggestion, then, the interdependence condition is met by virtue of how the two actions are described under the agreement. Each party is obligated to perform an action under a description to the effect that the action is a member of the understood pair of actions.

This suggested explanation of interdependence is obviously superficial. It does not exploit the central feature of the offer-acceptance model. On that model the obligations undertaken by the two parties are interdependent because they are exchanged. They are exchanged one for the other, as in an exchange of property. In a trade, each party does not gain ownership until the other does. Moreover, an exchange of property can be nullified only by mutual agreement, not unilaterally (one party may offer to undo an exchange by offering to return the item he received, and the other can accept and return the item he received). Similarly, a performance agreement can be nullified, but only by mutual agreement. Neither party can unilaterally rescind an agreement, and to break an agreement is not to annul it. In these ways, on the offer-acceptance model, the obligations created by an agreement are interdependent, and their interdependence is explained by the fact that they are exchanged.

Although I have appealed to the fuller notion of exchange, I have taken the practice of exchange for granted and have not tried to explain its foundations. Nor have I suggested, in connection with Gilbert's suggestion that "agreements have a good title to represent-even if they do not constitute-the basis of human social union" (p. 647), that exchanges play a more fundamental role. All I have argued is that Gilbert's dilemma for the promise-exchange model of agreements operates with an attenuated notion of exchange, not the fuller notion as exemplified in trades. Promises of the right sorts, an offer and its acceptance, can and do create the obligations essential to an agreement, and they are interdependent because exchanged. Gilbert is right to insist that an agreement involves a joint commitment-that is why an agreement is not a promise-exchange in the weak sense-but this is precisely the sort of commitment yielded by the two promises that constitute an agreement.

Notes

1. "Is an agreement an exchange of promises?," Journal of Philosophy, XC, 12 (December 1993): 627-649. All page references are to this article.

2. Gilbert's other example, in which the parties agree to meet at a certain place and time, is a solution to a coordination problem. As such it introduces an inessential feature into the discussion, in that each action is pointless unless the other is performed as well. A different but equally inessential feature is present if performance of one action is necessary for the performance of the other, e.g., where the agreement is for one to cook dinner and the other to wash the dishes. Our discussion will be limited to agreements like the one between Mavis and Peter, where the two actions are independent and so are their purposes.

3. As she also puts it, "agreements combine security and motivational force in a way that no exchange of promises is capable of doing" (p. 645). This way of putting it could mislead, for the issue here is not distrust or unenforceability. Gilbert is not concerned with such questions as whether the other party is trustworthy or whether one can count on the other trusting oneself. The question here is what the relevant obligations are, not whether the parties will fulfill them. As I understand the above passage, one has enough security if the other party is obligated to perform without waiting to see what one does, and one has enough motivation if one is obligated to perform without waiting to see what the other does.

4. The "problem of noise," as Gilbert describes it, prevents two (spoken) promises from being made simultaneously. One party must go first, but "we do not take [one] person to be committed before the other is" (p. 635). Later I will suggest that in ordinary agreements there is a certain way in which one party is committed first. In any case, the simultaneity criterion applies not to the promises but to the obligations they create.

5. More complex cases, in which the transfers are to be made later, would involve agreements giving rise to performance obligations. Such cases would thus defeat the purpose of the parallel I wish to draw here between agreements and trades.

6. Here it might be claimed, considering that a trade is not an exchange of gifts, that the very idea of a trade of promises does as much violence to the notion of a promise as the idea of a trade of gifts does to the notion of a gift. The idea is that just as gifts are unilateral and not conditional on reciprocation, so it is with promises. However, this claim, if meant as an argument, just begs the question. Besides, it seems that if promises can be conditional, they can be reciprocal, and gifts can't even be conditional (a conditional "gift" is just a conditional promise of a gift).

7. Then "the question is: If precisely these promises had (in effect) been exchanged, would the stated criteria [of an agreement] have been met?" (pp. 634-5).

8. Of course an agreement can begin with such a suggestion ("How about if I do A and you do B?"). Since the first party has not yet committed himself, the other party's approval of the suggestion ("OK") in effect constitutes the offer, which the first party can then accept ("Done!").

9. More precisely, acceptance turns it into an obligation that is not conditional on acceptance. So although Gilbert describes the obligations created by an agreement as unconditional, in fact they can be conditional, but only in ways irrelevant to the present discussion. For example, two parties might agree that one is to bring an umbrella if it rains and the other is to bring sunblock if it shines, and each would be conditionally obligated accordingly. For simplicity's sake, where it is clear what is meant I will use the term "unconditional" as short for "not conditional on acceptance."

10. Interestingly, at the end of her article Gilbert suggests that even ordinary promises involve joint commitment, in that the promisee "in some sense 'accepts' the promise" and "the promisor and the promisee become partners to a degree" (p. 648). For example, the promisee has no right to obstruct the fulfillment of the promise or to complain when it is fulfilled. So it is puzzling that she does not take into account the role of acceptance in the forging of an agreement.

11. As Gilbert points out, the conditional obligation, in this and several other examples, can be "internally" or "externally" conditional. Whereas internally conditional promises create conditional obligations, externally conditional promises are not made-they do not count as promises and the obligation is not created-until the relevant condition is met (p. 633). In describing them in this way, Gilbert has good reason to be skeptical about them and to consider them "only for purposes of argument." For she assumes that a promise is not made until the obligation it creates goes into effect. But we can distinguish the time of undertaking of an obligation from the time the obligation comes into being, in which case it is not really the promise but the obligation that is externally conditional-the obligation does not exist until the condition is met (with an internally conditional obligation, performance is not required until the condition is met). A possible source of confusion here is the act-object ambiguity of the noun "promise."

12. This formulation is weaker than Gilbert's, which needlessly stipulates that "if (once both promises are in effect) one promise is broken, both promises are nullified and no further obligations arise out of them for either party" (p. 644). This wrongly suggests that one can nullify one's obligation in an agreement by defaulting on it. Yet this seems to be Gilbert's view. Earlier she had written, "an unrescinded agreement generates its obligations for the parties until either these obligations have been fulfilled or one party has defaulted," so that, for example, "if Hal violates the agreement, he nullifies both his own and Mavis's obligation under it" (p. 631). This seems tantamount to claiming that to break an agreement is to nullify it.

13. It would be almost like requiring that for an argument to be valid, i.e., for its conclusion to follow from its premises, it needs the further premise that its conclusion follows from its (other) premises. See Lewis Carroll, "What the tortoise said to Achilles," Mind, IV, 14 (April 1895): 278-80.

14. If the offer is rejected, then the conditional obligation it created is effectively nullified. Notice, however, that it is not essential to the notion of an offer that rejection automatically nullify an offer. It is possible, though unusual, for an offer to be good even after being rejected ("Let me know if you change your mind"), though ordinarily it is understood to be withdrawn.

15. This is true only if the action is not part of a joint action, like lifting a heavy trunk, and only if its performance is not dependent on the performance of the other action. These special cases were excluded from discussion in note 2.