with whom
Justice Marshall, Justice Blackmun, and Justice Stevens join, dissenting.The critical question in this case is whether Adamson ever breached his plea agreement. Only by demonstrating that such a breach occurred can it plausibly be argued that Adamson waived his rights under the Double Jeopardy Clause. By simply assuming that such a breach occurred, the Court ignores the only important issue in this case.
I begin by demonstrating that, even if one defers to the Arizona Supreme Court’s construction of the plea agreement, one must conclude that Adamson never breached that agree*13ment. I then show that, absent a conscious decision by Adamson to breach his agreement, our cases provide no support for the Court’s conclusion that he has waived his rights under the Double Jeopardy Clause.
I
At the heart of this case is a plea bargain, an agreement to be interpreted in a constitutional context. We are asked to define the constitutional rights and responsibilities that arise from the language of that agreement, from the Due Process Clause of the Fourteenth Amendment, and from the Double Jeopardy Clause of the Fifth Amendment. The Court correctly observes that it must “assess independently the plea agreement’s effect on respondent’s double jeopardy rights.” Ante, at 6, n. 3. I think that the Court errs, however, in concluding that its assessment can proceed without an independent examination, informed by due process principles, of Adamson’s actions under that agreement. Ibid. Deferring to the Arizona Supreme Court’s construction of the agreement cannot relieve the Court of its responsibility to determine whether, in light of that construction, Adamson can be held to have lost his federal constitutional protection against being placed twice in jeopardy. The requirements of due process have guided this Court in evaluating the promises and conduct of state prosecutors in securing a guilty plea. Santobello v. New York, 404 U. S. 257 (1971). There is no reason to ignore those requirements here.
A
Without disturbing the conclusions of the Arizona Supreme Court as to the proper construction of the plea agreement,1 *14one may make two observations central to the resolution of this case. First, the agreement does not contain an explicit waiver of all double jeopardy protection.2 Instead, the Arizona Supreme Court found in the language of ¶¶ 5 and 15 of the agreement only an implicit waiver of double jeopardy protection which was conditional on an act by Adamson that breached the agreement, such as refusing to testify as it required. Adamson v. Superior Court of Arizona, 125 Ariz. 579, 584, 611 P. 2d 932, 937 (1980). Therefore, any finding that Adamson lost his protection against double jeopardy must be predicated on a finding that Adamson breached his agreement.
Second, Adamson’s interpretation of the agreement — that he was not required to testify at the retrials of Max Dunlap and James Robison — was reasonable. Nothing in the plea agreement explicitly stated that Adamson was required to provide testimony should retrials prove necessary. Moreover, the agreement specifically referred in two separate paragraphs to events that would occur only after the conclusion of all testimony that Adamson would be required to give. Paragraph 8 stated that Adamson “will be sentenced at the conclusion of his testimony in all of the cases referred to in this agreement and Exhibits A and B, which accompany it.” 789 F. 2d 722, 732 (CA9 1986) (emphasis added). At the *15time that the State demanded that Adamson testify in the retrials, he had been sentenced. Paragraph 18 stated that “[t]he defendant is to remain in the custody of the Pima County Sheriff from the date of the entry of his plea until the conclusion of his testimony in all of the cases in which the defendant agrees to testify as a result of this agreement.” Ibid. At the time the State demanded that Adamson testify in the retrials, Adamson had been transferred from the custody of the Pima County Sheriff. Adamson therefore could reasonably conclude that he had provided all the testimony required by the agreement, and that, as he communicated to the State by letter of April 3, 1980, the testimony demanded by the State went beyond his duties under the agreement.3 The Arizona Supreme Court rejected Adamson’s construction. But even deferring to the state court’s view that Adamson’s interpretation was erroneous, one must also agree with the en banc Court of Appeals that Adamson’s interpretation of the agreement was “reasonable],” and was supported by the plain language of the agreement, “[l]ogic, and common sense.” Id., at 729.
In sum, Adamson could lose his protection against double jeopardy only by breaching his agreement, and Adamson’s interpretation of his responsibilities under the agreement, though erroneous, was reasonable. The next step in the analysis is to determine whether Adamson ever breached his agreement.4
*16B
This Court has yet to address in any comprehensive way the rules of construction appropriate for disputes involving plea agreements. Nevertheless, it seems clear that the law of commercial contract may in some cases prove useful as an analogy or point of departure in construing a plea agreement, or in framing the terms of the debate. E. g., Blackledge v. Allison, 431 U. S. 63, 75, n. 6 (1977). It is also clear, however, that commercial contract law can do no more than this, because plea agreements are constitutional contracts. The values that underlie commercial contract law, and that govern the relations between economic actors, are not coextensive with those that underlie the Due Process Clause, and that govern relations between criminal defendants and the State. Unlike some commercial contracts, plea agreements must be construed in light of the rights and obligations created by the Constitution.
The State argues and the Arizona Supreme Court seems to imply that a breach occurred when Adamson sent his letter of April 3, 1980, to the prosecutor in response to the State’s demand for his testimony at the retrials of Dunlap and Robison. See ante, at 5. In this letter, Adamson stated that, under his interpretation of the agreement, he was no longer obligated to testify, and demanded additional consideration for any additional testimony. Ante, at 4-5, n. 2.
Neither the State, the state courts, nor this Court has attempted to explain why this letter constituted a breach of the agreement.5 Of course, it could not plausibly be argued that *17merely sending such a letter constituted a breach by nonperformance, for nothing in the plea agreement states that Adamson shall not disagree with the State’s interpretation of the plea agreement, or that Adamson shall not send the State a letter to that effect.6 But one might argue that, in the language of commercial contract law, the letter constituted a breach by anticipatory repudiation. See Tr. of Oral Arg. 32-33. Such a breach occurs when one party unequivocally informs the other that it no longer intends to honor their contract. “[W]here the contract is renounced before performance is due, and the renunciation goes to the whole contract, is absolute and unequivocal, the injured party may treat the breach as complete and bring his action at once.” Roehm v. Horst, 178 U. S. 1, 7 (1900).7 The reason for the rule is plain: “announcing [one’s] purpose to default” destroys the assurance of future performance that is central to a commercial contract.8
*18In the conventional case of anticipatory repudiation, therefore, the announcement of an intention to default on the contract constitutes a breach.9 In his letter of April 3, however, Adamson did not announce such an intention. To the contrary, Adamson invoked the integrity of that agreement as a defense to what he perceived to be an unwarranted demand by the prosecutor that he testify at the retrials of Dunlap and Robison. And in insisting that he had no obligation to perform as the State demanded, Adamson advanced an objectively reasonable interpretation of his contract.
We have held in the commercial sphere that a letter of the sort that Adamson sent does not constitute anticipatory repudiation. In New York Life Ins. Co. v. Viglas, 297 U. S. 672 (1936), the Court addressed the question whether an insurance company’s notification to a policyholder that it would henceforth refuse to continue paying disability benefits constituted a breach of the contract. The Court ultimately found that the company’s subsequent action to stop payment constituted a breach of the agreement, noting that the insurance company’s refusal was based on unfounded facts. Id., at 678. But the Court held that the notification alone did not constitute a breach by repudiation. As Justice Cardozo explained, for a unanimous Court:
“Repudiation there was none as the term is known to the law. Petitioner did not disclaim the intention or the *19duty to shape its conduct in accordance with the provisions of the contract. Far from repudiating those provisions, it appealed to their authority and endeavored to apply them. . . . There is nothing to show that the insurer was not acting in good faith in giving notice of its contention that the disability was over.” Id., at 676.
The law has been settled since Vigías that “[a]n offer to perform in accordance with the promisor’s interpretation of the contract although erroneous, if made in good faith, is not such a clear and unequivocal refusal to perform as amounts to a renunciation giving rise to an anticipatory breach.” Kimel v. Missouri State Life Ins. Co., 71 F. 2d 921, 923 (CA10 1934).10 As the court in Kimel explained:
“ Tf this were not the law, it would be a dangerous thing to stand upon a controverted construction of a contract. Every man would act at his peril in such cases, and be subjected to the alternative of acquiescing in the interpretation adopted by his opponent, or putting to hazard his entire interest in the contract. The courts have never imposed terms so harsh, or burdens of such weight. It would amount to a virtual denial of the right to insist upon an honest, but erroneous, interpretation.’” Ibid, (citation omitted).
Adamson has done no more here to repudiate his plea agreement than did the New York Life Insurance Company in Vigías, or the Missouri State Life Insurance Company in Kimel. After his lawyers were informed, by telephone, of the State’s view that his plea agreement obligated him to testify, he responded with a letter advancing his own reasonable interpretation of the agreement. Although the area of *20breach by repudiation, like other areas of commercial contract law, is not free from ambiguity,11 it seems plain that even under commercial contract principles Adamson did not breach his agreement.
Of course, far from being a commercial actor, Adamson is an individual whose "contractual” relation with the State is governed by the Constitution. The determination of Adam-son’s rights and responsibilities under the plea agreement is controlled by the principles of fundamental fairness imposed *21by the Due Process Clause. To grant to one party — here, the State — the unilateral and exclusive right to define the meaning of a plea agreement is patently unfair. Moreover, such a grant is at odds with the basic premises' that underlie the constitutionality of the plea-bargaining system. Guilty pleas are enforceable only if taken voluntarily and intelligently. E. g., Boykin v. Alabama, 395 U. S. 238 (1969). It would be flatly inconsistent with these requirements to uphold as intelligently made a plea agreement which provided that, in the future, the agreement would mean whatever the State interpreted it to mean. Yet the Court upholds today the equivalent of such an agreement. The logic of the plea-bargaining system requires acknowledgment and protection of the defendant’s right to advance against the State a reasonable interpretation of the plea agreement.
This right requires no exotic apparatus for enforcement. Indeed, it requires nothing more than common civility. If the defendant offers an interpretation of a plea agreement at odds with that of the State, the State should notify the defendant of this fact, particularly if the State is of the view that continued adherence to defendant’s view would result in breach of the agreement. If the State and the defendant are then unable to resolve their dispute through further discussion, a ready solution exists — either party may seek to have the agreement construed by the court in which the plea was entered. By following these steps the State would have placed far fewer demands on the judicial process than were in fact imposed here, and would have fulfilled its constitutional obligation to treat all persons with due respect.
C
The unfairness of the Court’s decision does not end here. Even if one assumes, arguendo, that Adamson breached his plea agreement by offering an erroneous interpretation of that agreement, it still does not follow that the State was entitled to retry Adamson on charges of first-degree murder. *22As the Court acknowledges, ante, at 7, immediately following the decision of the Arizona Supreme Court adopting the State’s construction of the plea agreement, Adamson sent a letter to the State stating that he was ready and willing to testify.12 At this point, there was no obstacle to proceeding with the retrials of Dunlap and Robison; each case had been dismissed without prejudice to refiling, and only about one month’s delay had resulted from the dispute over the scope of the plea agreement. Thus, what the State sought from Adamson — testimony in the Dunlap and Robison trials — was available to it.
The State decided instead to abandon the prosecution of Dunlap and Robison, and to capitalize on what it regarded as Adamson’s breach by seeking the death penalty against him. No doubt it seemed easier to proceed against Adamson at that point, since the State had the benefit of his exhaustive testimony about his role in the murder of Don Bolles. But even in the world of commercial contracts it has long been settled that the party injured by a breach must nevertheless take all reasonable steps to minimize the consequent damage. One prominent commentator has explained the rule in this way:
“If the victim of a breach can protect himself from its consequences he must do so. He has a duty to mitigate damages. . . . This is a duty, a kind of altruistic duty, toward’s one’s contractual partner, the more altruistic that it is directed to a partner in the wrong. But it is a duty without cost, since the victim of the breach is never worse off for having mitigated. Rather it is a duty that recognizes that contractual duties are onerous enough that they should not be needlessly exacerbated.” C. Fried, Contract as Promise 131 (1981) (footnote omitted).
*23Here it is macabre understatement to observe that the State needlessly exacerbated the liability of its contractual partner. The State suffered a 1-month delay in beginning the retrial of Dunlap and Robison, and incurred litigation costs. For these “losses,” the State chose to make Adamson pay, not with a longer sentence, but with his life. A comparable result in commercial law, if one could be imagined, would not be enforced. The fundamental unfairness in the State’s course of conduct here is even less acceptable under the Constitution.13
II
In addition to abdicating its responsibility to consider carefully the contractual and due process elements of this case, the Court does violence to the only area of constitutional law that it does address, double jeopardy. The Double Jeopardy Clause states that “No person shall ... be subject for the same offence to be twice put in jeopardy of life or limb. ” The Court’s explanation of how Adamson has waived this protection is unsupported by case law or logic.
“A waiver is ordinarily an intentional relinquishment or abandonment of a known right or privilege.” Johnson v. Zerbst, 304 U. S. 458, 464 (1938). Because we “‘indulge every reasonable presumption against waiver’ of fundamental constitutional rights,” ibid., we generally will enforce only those waivers that are knowing, intelligent, and voluntary. In certain circumstances, however, the Court has enforced waivers of the double jeopardy rights that would not meet this standard. For example, the Double Jeopardy Clause has been held not to bar retrial of a defendant who *24successfully moves for a mistrial. United States v. Dinitz, 424 U. S. 600 (1976). In Dinitz, the Court reasoned that “[t]he important consideration, for purposes of the Double Jeopardy Clause, is that the defendant retain primary control over the course to be followed in the event of [prejudicial prosecutorial or judicial] error.” Id., at 609. In such circumstances, “a defendant might well consider an immediate new trial a preferable alternative to the prospect of a probable conviction followed by an appeal, a reversal of the conviction, and a later retrial.” Id., at 610.
In United States v. Scott, 437 U. S. 82 (1978), the Court extended the logic of Dinitz to cases in which the defendant successfully moved to dismiss the indictment “on a basis unrelated to factual guilt or innocence of the offense of which he is accused.” 437 U. S., at 98-99. Two reasons supported the judgment. First, as in Dinitz, the defendant, in choosing to move to dismiss, retained control over the proceedings. 437 U. S., at 93-94, 98-99. Second, even though dismissal, unlike a mistrial, resulted in a final judgment normally held to bar reprosecution, the Court found it crucial that the proceedings had ended in midtrial, hence “without any submission to either judge or jury as to [defendant’s] guilt or innocence.” Id., at 101.
“[In this situation, the defendant] has not been ‘deprived’ of his valued right to go to the first jury; only the public has been deprived of its valued right to ‘one complete opportunity to convict those who have violated its laws.’ Arizona v. Washington, 434 U. S. [497, 509 (1978)].” Id., at 100.
The Court today relies exclusively on the first rationale of United States v. Scott. It argues that because Adamson fully understood the implications of breaching his agreement and made a voluntary choice to breach that agreement, he may be held to the consequences of his choice.
Scott alone cannot support the decision here. First, Adamson obviously did not retain control over the course of *25the proceedings against him. The unexamined assumption of the Court’s claim that he did, of course, is that Adamson made a voluntary decision to breach his agreement. For Adamson to have retained control comparable to the control evident in moving for a mistrial or a dismissal, he would have had to have deliberately chosen to breach his agreement. But he never made such a choice. Indeed, as discussed in Part I, supra, Adamson never breached his agreement at all. But even assuming that his actions could, in hindsight, be strictly construed to constitute a breach, it is plain that Adamson never took any act that he knew or realized would constitute a breach of the agreement. As a result, the Court’s argument that Adamson waived the protection of the Double Jeopardy Clause is untenable. Even under Scott, such protection cannot be lost through strict liability.
Second, this case does not involve a midtrial decision by a defendant to terminate the trial. It is therefore not a case in which the public has been deprived of its valued right to one complete opportunity to convict someone charged with breaking the law. Unlike Dinitz, and unlike Scott, Adamson had his guilt determined by a court prior to the alleged waiver of double jeopardy. As the Court reiterated in Scott, “the primary purpose of the Double Jeopardy Clause was to protect the integrity of a final judgment.” 437 U. S., at 92. The comparatively limited extent to which Scott violated the integrity of a final judgment is itself unique in double jeopardy jurisprudence. See id., at 109, n. 6 (Brennan, J., dissenting). But in carving out a limited exception for certain final judgments (those entered in midtrial on grounds other than factual guilt or innocence), the Court in Scott offered no reasoning that could be used to undercut the integrity of final judgments as to guilt.
Adamson’s interest in protecting the final judgment as to his guilt was substantial. That interest could be protected without compromising society’s right to one complete opportunity to obtain a conviction. Adamson did not consciously *26take any action that would undermine the integrity of that judgment — he did not deliberately choose to breach his plea agreement. Therefore, even if we construe his agreement to contain an implied waiver of double jeopardy protection in the event of a breach, Adamson cannot be held to have waived that protection.
Ill
The Court’s decision flouts the law of contract, due process, and double jeopardy. It reflects a world where individuals enter agreements with the State only at their peril, where the Constitution does not demand of the State the minimal good faith and responsibility that the common law imposes on commercial enterprises, and where, in blind deference to state courts and prosecutors, this Court abdicates its duty to uphold the Constitution. I dissent.
Although in text my argument proceeds on the assumption that deference to the Arizona Supreme Court’s construction is appropriate, I note here my view that its construction is premised on an interpretive method that is obviously biased and unfair. In rejecting Adamson’s interpretation of the agreement, the Arizona Supreme Court relied not on the plain language of the agreement, which offers the State only modest support, but *14rather on a colloquy that occurred at the time Adamson’s plea was taken. See ante, at 5-7, n. 3. Yet at the same time that the court went outside “the four corners of the document” in order to uphold the State’s view, it denied Adamson’s request to introduce other evidence that he maintained would demonstrate that at the time of sentencing the State shared Adam-son’s understanding of the agreement. Ibid. In these circumstances,' the Court of Appeals would have been justified in remanding for the evidentiary hearing denied Adamson in state court, and thereafter independently construing the agreement.
Nowhere in the agreement do the words “double jeopardy” appear. Significantly, ¶17 of the agreement, which lists the “rights” which Adamson “understood] that he [gave] up ... by pleading guilty,” does not mention the right not to be placed twice in jeopardy. 789 F. 2d 722, 732 (CA9 1986) (reprinting agreement in full).
Prior to sentencing, Adamson had provided extensive testimony for the State. He testified that he had “made 14 court appearances ... on five separate cases consisting of approximately 31 days of testimony. . . . Of the 81 or so jurors who have heard my testimony all have returned guilty verdicts in each case resulting in seven convictions. I have been cross-examined under oath for approximately 190 hours ... by 22 different attorneys. ... I have cooperated in approximately 205 interrogative sessions. . . . Fifty-five of these have been formal face-to-face in-depth question and answer sessions, approximately.” App. 150-151.
It is important to recall that the Court only assumes that such a breach occurred. As I observed at the outset, there is no justification for such an assumption — only by examining whether the alleged breach occurred can *16the Court “assess independently the plea agreement’s effect on respondent’s double jeopardy rights.” Ante, at 6, n. 3. See Part II, infra.
The Arizona Supreme Court stated only that “[t]he record before us is replete with indications of petitioner’s refusal to testify further in the Bolles murder cases.” Adamson v. Superior Court of Arizona, 125 Ariz. 579, 582, 611 P. 2d 932, 935 (1980). Although the court did not identify what those “indications” were, there appears to be only one other event (besides Adamson’s letter of April 3) to which it could have referred.' On April 18, 1980, Adamson was called to testify in proceedings prior to the retrial of Dunlap and Robison. Ante, at 5. He did not testify, but instead *17invoked his Fifth Amendment privilege. As this Court recounts, Adam-son invoked this privilege because the prosecutor had informed him, by letter of April 9, 1980, that the State already considered him in breach of his plea agreement and therefore vulnerable to reprosecution. Ibid. At the pretrial hearing, the trial judge, apprised of the plea agreement and of the State’s letter, refused to grant the State’s motion to compel testimony. The trial judge ruled correctly. Once the State declared that Adamson had breached his agreement, and that the State no longer was bound by the agreement, it relinquished any right it otherwise would have had to demand that Adamson continue to adhere to that agreement, i. e., to testify. Therefore, while Adamson did indeed refuse to testify on April 18, he did not thereby breach his agreement.
Indeed, at oral argument the United States, arguing as Amicus Curiae on behalf of the State, conceded that the agreement did not bar a good-faith challenge. Tr. of Oral Arg. 19.
See Restatement (Second) of Contracts § 250 (1981); Uniform Commercial Code § 2-610, 1A U. L. A. 321 (1976 and Supp. 1987); J. White & R. Summers, Uniform Commercial Code 212-214 (1980); 4 A. Corbin, Contracts § 973 (1951); 2 S. Williston, Contracts §§ 1322, 1323 (3d ed. 1968).
Equitable Trust Co. v. Western Pacific R. Co., 244 F. 485, 502 (SDNY 1917) (L. Hand, J.), aff’d, 250 F. 327 (CA2), cert. denied, 246 U. S. 672 (1918).
The classic case is Hochster v. De la Tour, 2 El. & Bl. 678, 118 Eng. Rep. 922 (Q.B. 1853), from which the doctrine of breach by anticipatory repudiation evolved. In that case, De la Tour first contracted to hire Hochster, then prior to the starting date of employment sent Hochster a letter stating that his services would not be needed. The court held that the letter constituted a breach of the contract, and that Hochster did not need to wait until after the starting date to bring suit. In Roehm v. Horst, this Court discussed Hochster at length, and concluded that it provided “a reasonable and proper rule to be applied in this case and in many others.” 178 U. S., at 20. Commentators continue to draw on Hochster to illustrate the principle. E. g., C. Fried, Contract as Promise 128-130, and n. 25 (1981).
See, e. g., Williston, supra, §§ 1322, 1323, pp. 132-133, 136-138 (“[A]n erroneous interpretation, asserted in good faith, will not amount to a breach”); Corbin, supra, § 973, p. 911 (“Where the two contracting parties differ as to the interpretation of the contract or as to its legal effects, an offer to perform in accordance with his own interpretation made by one of the parties is not in itself an anticipatory breach”).
Since New York Life Ins. Co. v. Viglas, 297 U. S. 672 (1936), courts and commentators have attempted to refine the distinction between advancing a reasonable (but erroneous) interpretation of a contract and repudiating a contract. For example, one court has held that repudiation occurs when a party has “persistently demanded an unwarranted condition precedent to its required performance” and thereby evinces a lack of good faith. Pacific Coast Engineering Co. v. Merritt-Chapman & Scott Corp., 411 F. 2d 889, 895-896 (CA9 1969) (emphasis added). Others have interpreted the rule expansively. E. g., Bill’s Coal Co. v. Board of Public Utilities, 682 F. 2d 883, 886 (CA10 1982) (upholding party’s right to urge even a “bad-faith” interpretation of a clause without thereby causing breach).
Alluding to the difficulty of determining when an erroneous interpretation of a contract has been offered in “good faith,” one commentator has recommended abandoning Viglas in favor of a standard of strict liability, under which any person who advances an interpretation that ultimately proves erroneous may be held to have repudiated the contract. See E. Farnsworth, Contracts 634-636 (1982). By contrast, another commentator who has acknowledged that same difficulty has nevertheless recognized that “the parties must communicate to clarify or modify the agreement to compensate for defects in the agreement process,” and has therefore recommended that “[t]he law ... be structured to encourage the parties to work out [disagreements over the meaning of the contract] whenever possible through good-faith renegotiation and modification; it cannot hope to achieve this goal if good-faith requests for modification are treated as repudiation.” Rosett, Partial, Qualified, and Equivocal Repudiation of Contract, 81 Colum. L. Rev. 93, 108 (1981) (footnote omitted). Of course, even if a policy of strict liability were thought meritorious in the commercial sphere, such a policy would be intolerable in the constitutional context, where the demands of due process, discussed infra, require a State to honor a defendant’s right to advance a reasonable and good-faith interpretation of an ambiguous plea agreement.
Conversely, if Adamson had refused to testify at this point — after an authoritative construction of the agreement had been rendered — then he could be deemed to have breached his agreement.
The curious and as yet unexplained decision of the State to abandon prosecution of Dunlap and Robison in favor of Adamson is not unrelated to the question whether the State’s actions in this ease amount to prosecutorial or judicial vindictiveness. Cf. North Carolina v. Pearce, 395 U. S. 711 (1969). This question, along with several others, was presented but not decided below, 789 F. 2d, at 725, and should be decided by the Court of Appeals on remand.