delivered the opinion of the Court.

In Johnson v. New Jersey, 384 U. S. 719 (1966), we held that Miranda v. Arizona, 384 U. S. 436 (1966), “applies only to cases in which the trial began after the date of our [Miranda] decision . . . 384 U. S., at 721. In this case, we must decide whether Miranda’s, standards for determining the admissibility of in-custody statements apply to post-Miranda retrials1 of cases *214originally tried prior to that decision.2 We hold that they do not.

Petitioner was arrested on the morning of March 17, 1965, as a murder suspect, and was interrogated on three separate occasions, at 11:30 a. m., 2:50 p. m., and 7:05 p. m. Although indigent, he was not advised that he had the right to have an attorney present at the State’s expense. Approximately 10 minutes after the evening interrogation began, petitioner gave the police a statement in which he admitted struggling with the victim during a burglary the preceding evening.

Petitioner’s first trial commenced on January 13, 1966. He did not take the stand, but his incriminating statement was admitted into evidence. The jury found him guilty of murder in the first degree and burglary in the fourth degree. Disregarding the jury’s recommendation, the trial court sentenced him to death. During the pendency of petitioner’s appeal to the Supreme Court of Delaware, we decided Miranda and Johnson. In reversing petitioner’s conviction on various state grounds, the Delaware Supreme Court also determined, sua sponte, that under Johnson petitioner’s statement, which was obtained without fully advising him of his constitutional rights, would be admissible at his retrial. - Del. -, 230 A. 2d 262 (1967). It reasoned that the retrial would be a mere continuation of the case originally commenced prior to our decision in Miranda.

Petitioner’s second trial began on October 2, 1967. He was convicted of second degree murder and sentenced to life imprisonment. The Supreme Court of Delaware affirmed, again rejecting petitioner’s argument that under Johnson his incriminating statement was inadmissible at his retrial. - Del. -, 240 A. 2d 146 (1968). Because *215of a disagreement among state courts over this issue,3 we granted certiorari. 393 U. S. 950 (1968). For the reasons stated below, we affirm.

Petitioner and the decisions he relies upon4 emphasize our references in Johnson to “trials” commenced before the date Miranda was decided and our stated concern for the reliance placed on pre-Miranda standards by trial courts as well as by law enforcement officers. Peti*216tioner argues that this “studied” focus on the trial process indicates that we intended Miranda to be applied to retrials, which, he insists, begin that process anew. As Delaware correctly points out, however, more than once we stated our holding in Johnson in terms of “cases” commenced before the date of Miranda. See 384 U. S., at 733. Delaware and the authorities it relies upon5 argue that, since the word “case” usually incorporates all the judicial proceedings against an accused, a retrial is not the “commencement” of a case. Delaware also quotes our statement that only “[fjuture defendants will benefit fully from our new standards governing in-custody interrogation, while past defendants may still avail themselves of the voluntariness test.” Id., at 732. Delaware suggests that petitioner, who was tried six months before Miranda, cannot be regarded as a “future” defendant within the meaning of Johnson. That there is language in Johnson supporting the positions of both petitioner and respondent demonstrates what some courts and commentators have readily recognized: in that decision, we did not consider the applicability of Miranda to retrials.6 The issue simply was not presented.

Petitioner buttresses his interpretation of Johnson by arguing that Miranda must be applied to retrials in order to insure the uniform treatment of individuals similarly situated. If it is not applied, he points out, it is possible that different standards for the protection of constitutional rights could be applied to two defendants simultaneously tried in the same courthouse for similar *217offenses. This anomaly could result if one of the defendants had been previously tried for the same offense prior to Miranda. This identical result, however, is also possible under our more recent prospectivity decisions. Because both Desist v. United States, 394 U. S. 244 (1969), and Stovall v. Denno, 388 U. S. 293 (1967), selected the date on which the prohibited practice was engaged in, rather than the date the trial commenced, to determine the applicability of newly formulated constitutional standards, those standards do not apply to retrials of defendants originally tried prior to the dates the standards were announced. In fact, under those decisions, different rules could govern where neither defendant had been tried before, depending upon when the condemned practice was engaged in.

Moreover, as petitioner acknowledges, Johnson made it quite clear that Miranda need not be applied to trials commenced prior to that decision but not yet final when it was announced. On that date, petitioner’s case was in precisely that posture. The type of apparent incongruity petitioner urges us to avoid is equally present in refusing to apply Miranda to defendants whose cases, like petitioner’s, were not final on the date Miranda was decided, yet making an exception for petitioner simply because he was afforded a post -Miranda retrial for reasons wholly unrelated to the admissibility of his incriminating statement. Nor is petitioner’s hypothetical more disconcerting than applying the new standards for in-custody interrogation to Ernesto Miranda while denying them to other defendants whose cases, for wholly fortuitous reasons, simply reached this Court at a later date, although the defendants in those cases may have been both interrogated and tried after Ernesto Miranda.

In short, petitioner’s concern for what he refers to as “visible imperfection[s] in a judicial process” merely *218highlights the problem inherent in prospective decision-making, i. e., some defendants benefit from the new rule while others do not, solely because of the fortuities that determine the progress of their cases from initial investigation and arrest to final judgment. The resulting incongruities must be balanced against the impetus the technique provides for the implementation of long overdue reforms, which otherwise could not be practicably effected. Thus, raising the specter of potential anomalies does not further the difficult decision of selecting the precise event that should determine the prospective application of a newly formulated constitutional principle.

Once the need is established for applying the principle prospectively, as the Supreme Court of New Jersey has pointed out, “there is a large measure of judicial discretion involved in deciding ... the time from which the new principle is to be deemed controlling.” State v. Vigliano, 50 N. J. 51, 65-66, 232 A. 2d 129, 137 (1967). In our more recent decisions in this area, we have regarded as determinative the moment at which the discarded standards were first relied upon. See, e. g., Desist v. United States, supra; Stovall v. Denno, supra. The point of reliance is critical, not because of any constitutional compulsion, but because it determines the impact that newly articulated constitutional principles will have upon convictions obtained pursuant to investigatory and prose-cutorial practices not previously proscribed. See Johnson v. New Jersey, supra, at 733. See generally Schaefer, The Control of “Sunbursts”: Techniques of Prospective Overruling, 42 N. Y. U. L. Rev. 631 (1967).7

*219In Johnson, after considering the need to avoid unreasonably disrupting the administration of our criminal laws, we selected the commencement of trial as determinative. We of course could have applied Miranda to all judgments not yet final, although they were obtained in good-faith reliance upon constitutional standards then applicable. See Linkletter v. Walker, 381 U. S. 618 (1965). As we pointed out, however, that choice "would [have] impose[d] an unjustifiable burden on the administration of justice.” 384 U. S., at 733. On the other hand, we could have adopted the approach we took in Stovall and Desist and made the point of initial reliance, the moment the defendant is interrogated, the operative event. See Schaefer, supra, at 646. But in an effort to extend the protection of Miranda to as many defendants as was consistent with society’s legitimate concern that convictions already validly obtained not be needlessly aborted, we selected the commencement of the trial. Implicit in this choice was the assumption that, with few exceptions, the commission and investigation of a crime would be sufficiently proximate to the commencement of the defendant’s trial that no undue burden *220would be imposed upon prosecuting authorities by requiring them to find evidentiary substitutes for statements obtained in violation of the constitutional protections afforded by Miranda.

This same concern for the justifiable reliance of law enforcement officials upon pre-Miranda standards militates against applying Miranda to retrials, which would place a much heavier burden upon prosecutors to compensate for the inadmissibility of incriminating statements obtained and admitted into evidence pursuant to practices not previously proscribed. See, e. g., State v. Vigliano, supra; People v. Sayers, 22 N. Y. 2d 571, 240 N. E. 2d 540 (1968); Comment, The Applicability of Miranda to Retrials, 116 U. Pa. L. Rev. 316, 324-325 (1967). As we stated in Stovall, “[I]nquiry would be handicapped by the unavailability of witnesses and dim memories.” 388 U. S., at 300. The burden would be particularly onerous where an investigation was closed years prior to a retrial because law enforcement officials relied in good faith upon a strongly incriminating statement, admissible at the first trial, to provide the cornerstone of the prosecution’s case.8 Moreover, we cannot assume that applying Miranda to retrials would affect only a small number of cases. It could, for example, render significantly more difficult the prosecutions of defendants, some of whom may have been convicted many years ago, who are afforded retrials because their convictions were obtained in violation of recently articu*221lated constitutional principles that are fully retroactive. See, e. g., Berger v. California, 393 U. S. 314 (1969); Roberts v. Russell, 392 U. S. 293 (1968). Such a decision could also pose a serious obstacle to the successful prosecution of an undetermined number of defendants whose pr e-Miranda convictions are reversed because of errors under federal or state law that do not even constitute constitutional violations.9

In determining how much weight to give the increased evidentiary burden that would result if we were to insist that Miranda be applied to retrials, we must consider society’s interest in the effective prosecution of criminals in light of the protection our pr e-Miranda standards afford criminal defendants. As we pointed out in Johnson, an individual who cannot claim the benefits of Miranda may still resort to whatever state and federal procedures are available to insure that statements admitted against him were made voluntarily. Moreover, he may invoke a “substantive test of voluntariness which, because of the persistence of abusive practices, has become increasingly meticulous . . . , [taking] specific account of the failure to advise the accused of his privilege against self-incrimination or to allow him access to outside assistance.” 384 U. S., at 730. As a result, not applying Miranda to retrials will not preclude the invocation of “the same safeguards as part of an involuntariness claim.” Ibid. Thus, because of the increased evidentiary burden that would be placed unreasonably upon law enforcement officials by insisting that Miranda be applied to retrials, and for all the reasons we gave in Johnson for not applying Miranda retroactively, we hold *222that Miranda does not apply to any retrial10 of a defendant whose first trial commenced prior to June 13, 1966.

Accordingly, the judgment of the Supreme Court of Delaware is Affirmed.

dissenting.

As one who has never agreed with the Miranda case but nonetheless felt bound by it,* I now find myself in the uncomfortable position of having to dissent from a holding which actually serves to curtail the impact of that decision.

I feel compelled to dissent because I consider that the new “retroactivity” ruling which the Court makes today is indefensible. Were I free to do so, I would hold that this petitioner is entitled to the benefits of Miranda, this case being before us on direct review and being one which had not become final prior to the decision of Miranda. See my dissenting opinion in *223Desist v. United States, 394 U. S. 244, 256 (1969); Linkletter v. Walker, 381 U. S. 618 (1965). But since as to the retroactivity issue I am also bound by Johnson v. New Jersey, 384 U. S. 719 (1966), I must judge that issue within the confines of Johnson, which does not appear to have been overruled by what was done in Desist v. United States, supra.

In the Johnson case we held that the “guidelines” of Miranda should apply to all “persons whose trials had not begun as of June 13, 1966,” 384 U. S., at 734, the date on which Miranda was handed down. Today, however, the Court holds that Miranda does not apply to persons whose retrials have commenced after that date, if the original trial had begun before Miranda was decided. I find it quite impossible to discern in the rationale of Johnson any solid basis for the distinction now drawn.

The Court states that the retroactivity rule adopted in Johnson was “an effort to extend the protection of Miranda to as many defendants as was consistent with society’s legitimate concern that convictions already validly obtained not be needlessly aborted.” Ante, at 219. I too believe that a desire not to interfere with trials which were concluded or already under way at the time of Miranda lay at the core of what was done in Johnson. See 384 U. S., at 732-735. But that rationale would seem to require application of Miranda to subsequent retrials, rather than the contrary result mandated by the Court. When a defendant has had his pre-Miranda conviction set aside on other than Miranda grounds and is being retried, there is by hypothesis no “conviction . . . validly obtained” which might be “needlessly aborted” by application of the Miranda standards. There is no ongoing trial in which the prosecution’s strategy might have been premised on pre-Miranda confession rules.

*224I am also left wholly unpersuaded by the Court’s statement that application of Miranda to retrials would impose an intolerable “evidentiary burden” on prosecutors, for the Court ignores the fact that Miranda will impose a very similar burden whenever a defendant’s first trial has for one reason or another been substantially delayed and its commencement carried beyond the Johnson cut-off date.

Apart from the two propositions just discussed, the Court offers nothing in justification of its trial-retrial distinction beyond the general observation that the retro-activity “technique” necessarily entails “incongruities” which must be tolerated because of “the impetus the technique provides for the implementation of long overdue reforms, which otherwise could not be practicably effected.” Ante, at 218. But surely it is incumbent upon this Court to endeavor to keep such incongruities to a minimum. This in my opinion can only be done by turning our backs on the ad hoc approach that has so far characterized our decisions in the retroactivity field and proceeding to administer the doctrine on principle. See my dissenting opinion in Desist, supra. What is done today leads me again, see ibid., to urge that the time has come for us to take a fresh look at the whole problem of retroactivity.

I would reverse the judgment of the Supreme Court of Delaware. It would be less than frank were I not to say that I cast this vote with reluctance, feeling as I do about the unsoundness of Miranda.