The Yale Law Journal


Treaty Denunciation and "Withdrawal" from Customary International Law: An Erroneous Analogy with Dangerous Consequences

05 Jan 2011

In their recent article in The Yale Law Journal, Professors Curtis Bradley and Mitu Gulati argue for a sweeping reformulation of international law relating to the legal force of customary norms. The model that they propose (the “Default View”) has the highly counterintuitive feature that states would be entitled to unilaterally opt out of customary norms with which they disagree, even long after the norm in question became an accepted part of international customary law and even if they had approved of the norm originally.

Among the grounds that they provide for this result are an analogy between customary and treaty law and the model’s supposed functional advantages. Bradley and Gulati’s analogy to treaties, however, involves a serious distortion of the existing law of international agreements, which does not in fact grant a right of unilateral withdrawal. Their claim about functional advantages is also problematic: it overlooks the already adequate flexibility of the current approach (which they call the “Mandatory View”). Finally, the authors acknowledge that a state exercising the proposed right to revoke would elicit international disapproval—an apparent recognition of the costs of their model to the international community. All in all, the case for the “Default View” has not been made.

I. The Supposed “Unilateral Right” To Revoke Treaties

Bradley and Gulati base their claim that customary international law (CIL) ought to be unilaterally revocable on an analogy to treaties, as to which the right to revoke is (supposedly) well-nigh universal. Alluding in a rather cursory way to qualifications to this supposed right, the authors argue that a comparable right should exist for international custom. Given the centrality of this claim to their argument as a whole, a closer examination of the international law of treaties is warranted.

In fact, the right to revoke or withdraw from a treaty—in those limited circumstances where such a right exists—almost always arises because of what both of the parties agreed to observe, either when the treaty was signed or subsequently. The right to opt out is mutual, not unilateral. The few exceptions that are created by operation of law (rather than by agreement of the parties) are narrow and cannot begin to support the authors’ expansive claim of a unilateral right to revoke.

A. The Vienna Convention and the Presumption of Continuance in Force

Both Part V of the Vienna Convention on the Law of Treaties and the Convention’s drafting history demonstrate that Bradley and Gulati’s premise of a general right to unilateral withdrawal from treaties is mistaken. The Convention, which is widely understood to reflect CIL, does not recognize any general unilateral right to revoke or withdraw. Articles 42, 54, and 56 establish a presumption that, absent one of these circumstances, a treaty will continue in force. Withdrawal is appropriate only in the limited situations that the Convention specifies.

The Vienna Convention recognizes several different circumstances in which a party to a treaty may withdraw. The first is where the treaty itself provides the right to withdraw; in other words, the withdrawal provision was intended or agreed to by all of the signatories. The intention may be either explicit or implicit in the agreement. The second occurs where the parties, at some point subsequent to a treaty’s entry into force, agree to terminate or suspend the operation of the treaty. The third is where the Vienna Convention itself recognizes a right of withdrawal by operation of law—in effect, as a result of a legal rule and independently of the parties’ intentions.

Article 42, entitled “Validity and continuance in force of treaties,” recognizes that it is the normal state of affairs for treaties to continue in force; a party that seeks to withdraw from or terminate a treaty bears the burden of showing that the conditions for withdrawal exist. Paragraph 2 of Article 42 provides: “The termination of a treaty, its denunciation or the withdrawal of a party, may take place only as a result of the application of the provisions of the treaty or of the present Convention . . . .” Concerning the rationale for Article 42, the official Commentary of the International Law Commission observed:

  [It] considered it desirable, as a safeguard for the stability of treaties, to underline in a general provision at the beginning of this part that the validity and continuance in force of a treaty is the normal state of things which may be set aside only on the grounds and under the conditions provided for in the present articles.  

Article 54, entitled “Termination of or withdrawal from a treaty under its provisions or by consent of the parties,” reiterates the importance of mutuality; the parties may agree, either during the treaty’s drafting or subsequently, upon the conditions for termination or withdrawal. It states:

  The termination of a treaty or the withdrawal of a party may take place:  
  (a) in conformity with the provisions of the treaty; or  
  (b) at any time by consent of all the parties after consultation with the other contracting States.  

Article 56(1), entitled “Denunciation of or withdrawal from a treaty containing no provision regarding termination, denunciation or withdrawal,” likewise explains that withdrawal from a treaty is generally not possible unless the treaty contains a provision for its termination. It states that there are circumstances under which a revocation provision that is not expressly stated in the treaty may nonetheless be implied:

  A treaty which contains no provision regarding its termination and which does not provide for denunciation or withdrawal is not subject to denunciation or withdrawal unless:  
  (a) it is established that the parties intended to admit the possibility of denunciation or withdrawal; or  
  (b) a right of denunciation or withdrawal may be implied by the nature of the treaty.  

By stating that the basic rule is that, except in certain limited and specified circumstances, “[a] treaty which contains no provision regarding its termination and which does not provide for denunciation or withdrawal is not subject to denunciation or withdrawal,” Article 56 places the burden on the party seeking to withdraw.

Article 42(2) of the Convention recognizes that there are situations where a right to withdraw is imposed by operation of law—in other words, according to some legal rule and regardless of the other party’s consent. These situations are detailed in Articles 46 through 62. None of these articles supports the claim made by Bradley and Gulati. To the contrary, the detailed and specific nature of these exceptions underscores that no general right of unilateral withdrawal exists. Such detailed and specific provisions would not be necessary if a general unilateral right of withdrawal existed.

Article 46 deals with inconsistency between state internal law and the treaty; such inconsistency does not suffice to invalidate the treaty unless the violation was “manifest and concerned a rule of its internal law of fundamental importance.” Article 48 concerns the possibility of error in the treaty; Article 49 concerns fraud in the inducement; Article 50 raises the issue of corruption of a state’s representative; Articles 51 and 52 concern coercion of a state’s representative; and Article 53 deals with inconsistency between the treaty in question and jus cogens. Article 60 sets out the consequences of a breach by one of the parties; under limited circumstances, a breach by one gives the other party the right to withdraw. Articles 61 and 62 deal with impossibility of performance and fundamental change of circumstances.

These are very specific and limited circumstances; none of them supports a general “unilateral right of withdrawal,” and Bradley and Gulati do not really claim that they do. Bradley and Gulati do mention the provision regarding “fundamental change of circumstances,” but they reveal in a footnote their awareness that this basis for revocation is extremely limited. These provisions no more demonstrate a general unilateral right of withdrawal from treaties than the contract law doctrines of change of circumstances or fraud in the inducement demonstrate a general unilateral right to revoke a contract.

B. The “Right To Invoke Certain Grounds as a Basis for Withdrawal or Termination”

The drafters of the Convention attempted in still another way to clarify that there was no general right to withdraw from a treaty. During the proceedings of the International Law Commission (ILC) and the Vienna Conference, an important distinction was made between the right to terminate (or right to withdraw) and the right to invoke certain grounds as a basis for withdrawal or termination. The ILC’s official commentary explained its reasons for choosing the latter of these verbal formulations: “The formula ‘invoke as a ground’ is intended to underline that the right arising under the article is not a right arbitrarily to pronounce the treaty terminated.” Evidently, the ILC did not mean to provide states with a unilateral right of revocation at will.

This verbal formulation was introduced so as to provide stability of commitment in states’ international relations. Herbert W. Briggs observes:

  That the distinction was understood at the Vienna Conference on the Law of Treaties is clearly demonstrated by the rejection, by votes 52-4-34 and 51-3-38, of Venezuelan proposals to substitute a right “to terminate” for the right to invoke a breach as a ground for termination or suspension set forth in the International Law Commission’s draft. After amending the Commission’s text of Article 57(2)(c) [Art. 60(2)(c), Vienna] to substitute a right “to invoke the breach as ground” for suspension for the right “to suspend” which had crept into the Commission’s text, the Vienna Conference adopted the article by a vote of 88-0-7.  

If one takes seriously the argument that customary and treaty law are analogous, then one does not come to the conclusion that Bradley and Gulati reach. Accepting their analogy would entail that the customary norm would be “unilaterally revocable” only if: (1) at the time it was formed, revocability was the parties’ intent, either explicitly or by implication; (2) the parties later in fact came to an agreement that the customary norm in question should be revoked; or (3) some other exception existed as a matter of operation of law and not simply as a matter of unilateral preference.

None of these conditions is met in the context that Bradley and Gulati address. There is no reason to believe that the states whose conduct brought CIL into being silently intended that it should be revocable at will. To the contrary: the opinio juris requirement for CIL provides that even to qualify as customary law in the first place a norm must be perceived as binding. A norm that is intended to be unilaterally revocable would not qualify under this test as it is currently understood. All in all, the analogy between customary and treaty law does not have the consequences that Bradley and Gulati intend.

II. Functional Advantages and the Avoidance of Sovereignty Costs

Accepting that the case for their model cannot be made solely by resorting to legal history or judicial precedent, the authors devote much of their article to detailing the Default View’s supposed “functional” advantages. Among these claimed advantages are ways that their model obviates the need for states to undertake avoidance maneuvers in order to sidestep customary international legal obligations. But Bradley and Gulati’s functional analysis tends to overlook the ways in which the Mandatory View already permits avoidance. The foundation on which their model rests is not, in reality, functional; it is a value judgment that a state’s freedom from “sovereignty costs” is more important than the benefits foregone because other states are similarly liberated.

A. Examples of Erroneous Functional Reasoning

The problem with Bradley and Gulati’s functional analysis is clear in their reasoning about how states should be expected to act under the two different models. One of their examples purports to demonstrate that “[a]llowing for lawful exit rights under CIL might also enhance treaty-making”:

  [M]ultilateral treaties are often invoked today as evidence of the content of CIL. This creates an anomaly, however, whereby nations frequently have a right to exit from the underlying treaties but not from the similar CIL derived from the treaties. This anomaly in turn increases the “sovereignty costs” associated with establishing and joining multilateral treaties. By contrast, allowing for exit rights under CIL would reduce these costs and thereby potentially encourage more treaty-making.  

The argument is incorrect in its claim that nations cannot already avoid the CIL obligation that might be triggered by the widespread adoption of the treaty. First, a state intending to sign a widely adopted treaty or convention may help protect against its incorporation into CIL by making clear that it does not believe that the norm in question is supported by opinio juris—a “subjective sense of legal obligation.” Second, and more importantly, the persistent objector doctrine (which the authors accept as a genuine norm of international law) would allow states an opportunity to opt out of the CIL norms in this example, as the norms had not yet crystallized.

A second, comparably flawed example of supposed functional superiority deals with states suspecting that a new CIL rule may be in the offing. The authors posit that these states might fear that the rule “will turn out to have unforeseen negative consequences” so that they will immediately rally against it. This avoidance maneuver is unnecessary under their proposed model; they claim:

  Under the Mandatory View, when there are suggestions of a new rule, some nations might be concerned that the rules will turn out to have unforeseen negative consequences. If so, these nations with concerns will work hard to prevent new CIL from forming out of a fear that, once it forms, it will be binding and hard to change. By contrast, if a right of future withdrawal is permitted, it provides nations with a form of insurance, in that they can experiment with how the rule works for them and then back out of it if its negative effects outweigh the benefits.  

The authors fail to consider that, under the persistent objector doctrine, states in such a position need not be subjected unwillingly to the new rule. As in the previous example, they can protect themselves by repeatedly declaring themselves in opposition or by denying the subjective element of opinio juris.

B. Value Judgments as Opposed to Functional Reasoning

Both of these examples concern the effect of the two competing models on future behavior: the signing of treaties and the formation of CIL in the first case, and working to prevent formation of new CIL that is feared will be undesirable in the second. The examples’ future orientation is important, because it is in cases where the norm is not yet crystallized that the doctrine of the persistent objector is most protective. At the point in time on which these examples focus—during the state’s strategic analysis of the likely future effects of treaty-signing or the merits of the looming change in law—it is not too late to start representing consistently that the norms embedded in the treaty do not constitute existing international custom. If one focuses exclusively on the future consequences for state strategic behavior of choosing between the two models, the difference between them is not great and Bradley and Gulati’s functional arguments (as a result) are not particularly convincing.

The differences between the Mandatory and the Default Views are greatest in relation to the consequences of events that have already taken place. Where customary law already has been formed, it will be too late for the persistent objector doctrine to release states from their obligations under the Mandatory View but not too late for withdrawal under the Default View. In such cases, the two models point in opposite directions. But to make their analysis retroactively applicable to state decisionmaking and norm formation that has already occurred, their argument needs to be grounded in something other than restructuring state behavior for the future. The argument would have to be something like the unfairness or unreasonableness of holding a state to CIL simply because it failed to object, or was not in any position to object, at the time at which the norm crystallized. This is not, strictly speaking, a functional argument, and it rests on value judgments somewhat foreign to the article’s basic orientation. The chief of these is avoidance of sovereignty costs, a topic to which we will return below.

III. “Reputational Incentives” Not To Exercise the “Right” of Withdrawal

At the same time that Bradley and Gulati propose a unilateral right of withdrawal from CIL, they anticipate that withdrawals would harm a state’s reputation enough that states would be disinclined to exercise the right. They never explain why withdrawal, with all the policy advantages that it supposedly provides, should trigger international opprobrium. Importantly, a state’s exercise of a right of withdrawal provided by treaty—which the authors represent as analogous to withdrawal from CIL—does not have these adverse “reputational incentives.” The fact that withdrawal from CIL would adversely affect a state’s reputation, in the authors’ view, suggests their unexpressed awareness that, in the eyes of the international community, withdrawal is the evasion of a legal commitment.

A. “Reputational Incentives” Under the Default View

The authors’ discussion of reputational incentives is part of their attempt to show that their proposed “Default View” would make it easier to differentiate between “good” and “bad” actors. The capacity to differentiate, which is important (they say) because it enhances a state’s ability to determine which states would make reliable partners, is depicted as limited under the traditional Mandatory View. Bradley and Gulati explain that under the Mandatory View, a good actor that disagrees with a norm may have to violate it because there is no other way of registering its objection. Since both good and bad actors engage in violations of CIL, albeit for different reasons, it will not be clear to other states whether the motive underlying a particular violation is defensible (to change the law) or indefensible (to renege on a commitment). Bad actors, to put it another way, can effectively disguise themselves as good ones.

If the authors’ Default View is adopted, they claim, a good actor will have the option of withdrawing and will do so when it wishes to show disapproval of the norm in question. Bradley and Gulati claim that it will therefore be possible to differentiate the good actors from the bad. Their explanation of the reasons for this conclusion is short, and it is not completely clear how they think states will make this differentiation. But one thing is apparent: for Bradley and Gulati’s argument to work, bad actors must behave differently somehow from good actors. The point that they are making, after all, is that under their model states will be better able to tell which other states are reliable. Thus, if they believe that good actors would withdraw, the authors seemingly have to argue that bad actors would not.

But why not? One would think that for bad actors, the withdrawal strategy would allow them easily to mimic the behavior of the good actors (as they do under the Mandatory View, Bradley and Gulati claim). The authors’ explanation of why the bad actors would not simply follow the lead of the good actors—withdrawal—is that “there are reputational incentives not to withdraw”:

  One might ask why, in a world with the Default View, nations would not avoid this reputational pressure to comply by simply withdrawing from a CIL rule before violating it. Part of the answer is that, to the extent that CIL rules are viewed as legitimate and beneficial, there are reputational incentives not to withdraw from them.  

This argument raises a host of questions. The most important of these for present purposes is: why should exercise of a supposed right of unilateral withdrawal give a state a bad reputation? If withdrawal were really the eminently defensible act that the authors make it out to be, there would be no reason for the international community to object to other states withdrawing. Under treaty law an exercise of a right of withdrawal would not cause harm to a state’s reputation. Obviously, under treaty law, there is no reason why withdrawal should cause harm; it is the exercise of a right that the parties themselves contemplated and deliberately included in the treaty.

The explanation (one suspects) is that the authors realize that withdrawal will not be seen as the innocent act that they would characterize it as; withdrawal has a bad odor because it would be interpreted by the international community as the abandonment of a commitment. The fact that the authors posit reputational consequences for withdrawing reveals their awareness that withdrawal (in the eyes of the international community) is opportunistic and unacceptable.

B. The Unilateral Right To Revoke and the Avoidance of “Sovereignty Costs”

Bradley and Gulati attach great importance to avoiding “sovereignty costs,” meaning “the symbolic and material costs of diminished national autonomy.” In their examples it is repeatedly assumed that states will go to some lengths to avoid undertaking binding obligations. The practical benefits of flexibility are given primacy; little practical importance is attached to the main reason that states in fact make international agreements, namely, to obtain the commitment of other states. States are thought of first and foremost as promisors with interests in minimizing the extent of their obligations, and only secondarily, if at all, as promisees who benefit from these commitments.

This is not necessarily the way in which states themselves see things. Surely there is a trade-off involved. States presumably enter into treaties because it is of some value to them to obtain the commitments of other states. That commitment, clearly, is less valuable if it is less reliable, and it is surely less reliable if other state parties have a right of unilateral withdrawal. Treaty law is, in this respect, truly analogous to CIL; the strength of one’s own commitment varies along with the strength of the commitment of other parties. Maintaining one’s own autonomy comes at the cost of creating a right of exit for one’s partners. You get what you pay for.


Bradley and Gulati’s article may be illuminating as a barometer of American views and interests: international law is viewed with suspicion by some American politicians, as an infringement on national prerogatives. It is clear that powerful countries such as the United States do not depend so much on international law to protect themselves—their defenses are primarily military and economic.

It seems quite likely, however, that other states, especially smaller and less powerful ones, are more friendly to the idea of legal commitment and less exclusively focused on “the symbolic and material costs of diminished national autonomy.” International law is one of the few weapons that small and powerless states have against the United States. The challenges facing international law are difficult enough as things stand without further diminishing its force by depriving international custom and practice of their ability to generate binding obligations.

Lea Brilmayer is Howard Holztmann Professor of International Law, Yale Law School. Isaias Yemane Tesfalidet received an LL.B. from the University of Asmara, Eritrea and an LL.M. from Yale Law School and is a fellow at the Forum for International Criminal and Humanitarian Law. The authors wish to thank Rebecca Crootof, Yale Law School, Class of 2011, for her contributions to this Essay.

Preferred citation: Lea Brilmayer & Isasias Yemane Tesfalidet, Treaty Denunciation and “Withdrawal” from Customary International Law: An Erroneous Analogy with Dangerous Consequences, 120 Yale L.J. Online 217 (2011),