with whom Mr. Justice Stevens joins, concurring in the judgment.

This Court announced its decision in Burch v. Louisiana, 441 U. S. 130 (1979), while the petitioner’s objection to the nonunanimous verdict was pending on direct appeal. Ante, at 326. Since I believe that new constitutional rules should apply retroactively “in cases still pending on direct review,” Hankerson v. North Carolina, 432 U. S. 233, 248 (1977) (Powell, J., concurring in judgment), I concur in the judgment reversing the petitioner’s conviction.

dissenting.

I am in agreement with the Court on the content of the applicable standards for gauging the need for retroactivity, but I cannot concur in the Court’s application of those standards in this case. The most important question here is whether it is probable that the Louisiana juries convicting on a vote of 5 to 1 convicted innocent persons. As the Court *338states, “only when an assessment of those probabilities indicates that the condemned practice casts doubt upon the reliability of the determinations of guilt in past criminal cases must the new procedural rule be applied retroactively.” Ante, at 329. Neither our precedents nor common experience supports the Court’s conclusion that the 5-to-l vote is inherently unreliable. Just as I think the Court has overstated the probabilities of jury error, I think it has unfairly understated the State’s reliance on our prior law and the burdens on the administration of the Louisiana justice system which will be associated with today’s ruling.

A

In Williams v. United States, 401 U. S. 646, 655, n. 7 (1971), we held that retroactivity is only appropriate where the former practice “presents substantial likelihood that the results of a number of those trials were factually incorrect.” In Hankerson v. North Carolina, 432 U. S. 233, 243 (1977), we similarly concluded that the “major purpose” of the new rule must be to correct a process which “substantially impairs its truth-finding function” raising “serious questions about the accuracy of guilty verdicts in past trials” before a rule should be retroactively imposed. Quite simply, when five-sixths of the deliberating jurors reach a finding of guilt, I do not think that there is a substantial probability that their decision was wrong.

The Court stresses the part of Mr. Justice Blackmun’s opinion in Ballew v. Georgia, 435 U. S. 223 (1978), suggesting that some studies had indicated that the reliability of the truth-finding process declines when the jury size is reduced. But I do not think that those citations can be used here to support the conclusion that the jury verdicts in issue were probably inaccurate. First, the opinion in Ballew relies heavily on the conclusions that a jury of only five is too small in number to ensure effective deliberation and to ensure that someone among the group will have memory abilities suffi*339cient to aid the jury in those deliberations. Id., at 241. These concerns are satisfied when the jury is composed of six members, even if one of those members is in the dissent. In fact, as indicated by Johnson v. Louisiana, 406 U. S. 356, 361 (1972), the presence of a dissenting juror strongly supports the inference that the jury has engaged in meaningful deliberation:

“We have no grounds for believing that majority jurors, aware of their responsibility and power over the liberty of the defendant, would simply refuse to listen to arguments presented to them in favor of acquittal, terminate discussion, and render a verdict. On the contrary it is far more likely that a juror presenting reasoned argument in favor of acquittal would either have his arguments answered or would carry enough other jurors with him to prevent conviction. A majority will cease discussion and outvote a minority only after reasoned discussion has ceased to have persuasive effect or to serve any other purpose — when a minority, that is, continues to insist upon acquittal without having persuasive reasons in support of its position. At that juncture there is no basis for denigrating the vote of so large a majority of the jury or for refusing to accept their decision as being, at least in their minds, beyond a reasonable doubt.”

Thus the jury that convicted petitioner satisfied the requirements of jury deliberation that the Court in Ballew found so critical. Further, our cases have indicated quite clearly that the degree of persuasion evidenced by a 5-to-l vote is sufficient to meet the requirement that guilt be proved beyond a reasonable doubt. In Johnson, supra, this Court held that a 9-to-3 verdict could satisfy due process, or in other words, satisfy the requirement that guilt be proved beyond a reasonable doubt. The degree of persuasion found acceptable there was far less impressive than that demonstrated by the jury which convicted petitioner. And yet we said that guilt was *340proved beyond a reasonable doubt in Johnson. I think here, too, we must then conclude that guilt was proved beyond a reasonable doubt. Since that is true, there has been no constitutionally unacceptable risk of erroneous convictions and Burch need not be applied retroactively.

There is a further weakness in the Court’s estimation of the probabilities. We simply have no way of knowing whether the person voting to acquit would have held firm with further pressure by his fellow jurors. The Court’s speculation about what would have happened had unanimity been required of Louisiana’s six-man juries amounts to just that: speculation. As long as this Court has approved “Allen charges” in federal cases over which it may exercise its supervisory authority, it is difficult to say that a holdout juror might not ultimately have been persuaded by the five-member majority.

The Court’s ruling is also at odds with our decisions in Gosa v. Mayden, 413 U. S. 665 (1973), and DeStefano v. Woods, 392 U. S. 631 (1968). In both of those cases, the Court declined to give retroactive effect to rulings that the right to jury trial had been totally denied under circumstances where our system of fairness required that it be afforded. Nevertheless, as we stated in DeStefano, the “values implemented by the right to jury trial would not measurably be served by requiring retrial of all persons convicted in the past by procedures not consistent with the Sixth Amendment right to jury trial.” Id., at 634. The deprivations addressed in those cases were no less based on procedural reliability than was the decision in Burch v. Louisiana, 441 U. S. 130 (1979).

B

I also think that the Court has unduly minimized Louisiana’s reliance on pre-Burch standards, and greatly underestimated the impact its ruling will have on the Louisiana judicial system. We have every reason to credit Louisiana with the presumption that its law was enacted in good faith. Prior to 1974, the Louisiana Constitution allowed for conviction by *341unanimous five-person juries for certain offenses. La. Const., Art. 7, §41 (1921). In 1974 this constitutional provision was replaced with the nonunanimous six-person jury provision. The coordinator of legal research for the Constitutional Convention explained in 1974 that he believed this provision satisfied the Federal Constitution, reasoning:

“A six-man jury was upheld in Williams v. Florida, 399 U. S. 78 (1970). If 75 per cent concurrence (%2) was enough for a verdict as determined in Johnson v. Louisiana, 406 U. S. 356 (1972), then requiring 83 per cent concurrence (%) ought to be within the permissible limits of Johnson.” Hargrave, The Declaration of Rights of the Louisiana Constitution of 1974, 35 La. L. Rev. 1, 56, n. 300 (1974).

The record similarly suggests that the administrative impact is substantial. In the first four months of 1979 in just Orleans Parish alone, 39 defendants were tried by six-person juries. Brief for Respondent 24, n. 43. The various courts in Louisiana apparently do not necessarily keep a record of the jury vote. Id., at 28, n. 49. With this large number of six-person jury trials, the potential for disruption is substantial. And although the Court states that the decision will only have an impact where the defendant was “in fact” convicted by less than six, how is it to be established what “in fact” occurred without clear records? Ante, at 336. As stated in the opinion of Mr. Justice Blackmun in Gosa, supra:

“Wholesale invalidation of convictions rendered years ago could well mean that convicted persons would be freed without retrial, for witnesses ... no longer may be readily available, memories may have faded, records may be incomplete or missing, and physical evidence may have disappeared. Society must not be made to tolerate a result of that kind when there is no significant question concerning the accuracy of the process by which judgment *342was rendered or, in other words, when essential justice is not involved.” 413 U. S., at 685.

Since Burch and Ballew held little more than that “lines must be drawn somewhere” 441 U. S., at 137; 435 U. S., at 239, Louisiana should not be required to retry defendants found guilty by reliable factfinders.

announced the judgment of the Court and delivered an opinion, in which Mr. Justice Stewart, Mr. Justice Marshall, and Mr. Justice Blackmun joined.

Burch v. Louisiana, 441 U. S. 130 (1979), held that conviction of a nonpetty criminal offense by a nonunanimous six-person jury violates the accused’s right to trial by jury guaranteed by the Sixth and Fourteenth Amendments. The issue *325in this case is whether the constitutional principle announced in Burch is to be given retroactive application.

I

On July 31, 1978, petitioner Darnell Brown was charged by bill of information in Orleans Parish with simple burglary, a felony punishable by confinement in the parish prison or state penitentiary for a maximum term of 12 years. La. Rev. Stat. Ann. § 14:62 (West Supp. 1979). At the time, the Louisiana Constitution and Code of Criminal Procedure provided that such crimes should be tried by a jury of six persons, five of whom must concur to render a verdict.1 Before trial, petitioner filed a motion to quash pursuant to Art. 532 (9) of the Louisiana Code of Criminal Procedure, arguing that his “due *326process rights under the Sixth and the Fourteenth Amendments to the United States Constitution as enunciated in Ballew v. Georgia, [435 U. S. 223 (1978),] will be violated by a less than unanimous vote by a six person jury.” App. 5. Petitioner therefore requested the trial judge to order a jury of 12 or, in the alternative, to require a unanimous verdict of the jury of 6.

Petitioner’s motion was denied, and on August 23 his trial commenced before a six-member jury. That same afternoon, after deliberating for approximately one hour, the jury returned a verdict of guilty. At petitioner’s request, the court polled the jurors and ascertained that their vote was 5 to 1 to convict. Sentencing was set for August 30, at which time petitioner renewed his objection to the nonunanimous six-person verdict by a motion for new trial. The trial judge again denied the motion and sentenced petitioner to a term of 22 years’ imprisonment at hard labor.2

Petitioner appealed his conviction to the Louisiana Supreme Court, assigning as principal error the trial judge’s refusal to grant the motion to quash. On April 17, 1979, while petitioner’s case was still pending on direct review in the Louisiana courts, Burch v. Louisiana, supra, was decided, holding unconstitutional those provisions of the Louisiana Constitution and Code of Criminal Procedure that sanctioned conviction of a nonpetty offense by a nonunanimous jury of six. Some five weeks later, on May 21, 1979, the Louisiana Supreme Court affirmed petitioner’s conviction. Although it implicitly acknowledged that Burch requires unanimous verdicts by six-person juries in all future prosecutions of simple burglary,3 the court nonetheless concluded, without elabora*327tion, that “the rule of Burch, supra, should not be applied retroactively to juries empaneled prior to the date of the Burch decision.” 371 So. 2d 746, 748 (1979) (emphasis in original). We granted certiorari. 444 U. S. 990 (1979). We reverse.

II

Linkletter v. Walker, 381 U. S. 618 (1965), was the first instance in which the Court declined to apply a new doctrine respecting one of the provisions of the Bill of Rights retroactively for the benefit of a previously convicted defendant. In the intervening 15 years, we have often considered the question of the retroactivity of decisions expounding new constitutional rules of criminal procedure, and have endeavored to elaborate appropriate standards for determining which rules are to be accorded retrospective and which only prospective effect. From the welter of case law that has developed in this area, several unequivocal principles emerge to guide our analysis in the present case.

It is by now uncontroverted that “the Constitution neither prohibits nor requires retrospective effect.” Id., at 629. Thus, although before Linkletter new constitutional rules had been applied to cases that had become final before promulgation of the rule, see id., at 628, and n. 13, that decision firmly settled that “in appropriate cases the Court may in the interest of justice make the rule prospective . . . where the exigencies of the situation require such an application.” Id., at 628; Johnson v. New Jersey, 384 U. S. 719, 726-727 (1966).

Similarly, it is clear that resolution of the question of retro-activity does not automatically turn on the particular provision of the Constitution on which the new prescription is based. “Each constitutional rule of criminal procedure has its own distinct functions, its own background of precedent, and its own impact on the administration of justice, and the way in which these factors combine must inevitably vary with the dictate involved.” Id., at 728. Accordingly, the test con*328sistently employed by the Court to decide whether a new constitutional doctrine should be applied retroactively contemplates the consideration of three criteria: “(a) the purpose to be served by the new standards, (b) the extent of the reliance by law enforcement authorities on the old standards, and (c) the effect on the administration of justice of a retroactive application of the new standards.” Stovall v. Denno, 388 U. S. 293, 297 (1967).

Moreover, our decisions establish that “[f foremost among these factors is the purpose to be served by the new constitutional rule,” Desist v. United States, 394 U. S. 244, 249 (1969), and that we will give controlling significance to the measure of reliance and the impact on the administration of justice “only when the purpose of the rule in question [does] not clearly favor either retroactivity or prospectivity.” Id., at 251; Michigan v. Payne, 412 U. S. 47, 55 (1973); see also Hankerson v. North Carolina, 432 U. S. 233, 242-244 (1977) ; Adams v. Illinois, 405 U. S. 278, 280 (1972) (plurality opinion of Brennan, J.). “Where the major purpose of new constitutional doctrine is to overcome an aspect of the criminal trial that substantially impairs its truth-finding function and so raises serious questions about the accuracy of guilty verdicts in past trials, the new rule has been given complete retroactive effect. Neither good-faith reliance by state or federal authorities on prior constitutional law or accepted practice, nor severe impact on the administration of justice has sufficed to require prospective application in these circumstances.” Williams v. United States, 401 U. S. 646, 653 (1971) (plurality opinion of White, J.). Accord, Hankerson v. North Carolina, supra, at 243; Gosa v. May den, 413 U. S. 665, 679 (1973) (plurality opinion of Blackmun, J.); Ivan V. v. City of New York, 407 U. S. 203, 204 (1972).

Finally, we have recognized that the extent to which the purpose of a new constitutional rule requires its retroactive application “is necessarily a matter of degree.” Johnson v. New Jersey, supra, at 729. Constitutional protections are *329frequently fashioned to serve multiple ends; while a new standard may marginally implicate the reliability and integrity of the factfinding process, it may have been designed primarily to foster other, equally fundamental values in our system of jurisprudence.4 Not every rule that “tends incidentally” to avoid unfairness at trial must be accorded retroactive effect. Gosa v. Mayden, supra, at 680 (plurality opinion of Blackmun, J.). So, too, additional safeguards may already exist that minimize the likelihood of past injustices.5 In short, “[t]he extent to which a condemned practice infects the integrity of the truth-determining process at trial is a ‘question of probabilities.’ ” Stovall v. Denno, supra, at 298 (quoting Johnson v. New Jersey, supra, at 729). And only when an assessment of those probabilities indicates that the condemned practice casts doubt upon the reliability of the determinations of guilt in past criminal cases must the new procedural rule be applied retroactively.6

*330Ill

With these principles in mind, then, we turn to consideration of the issue presented by this case: whether the rule of Burch v. Louisiana must be given retroactive effect. We conclude that it must.

A

The right to jury trial guaranteed by the Sixth and Fourteenth Amendments “is a fundamental right, essential for preventing miscarriages of justice and for assuring that fair trials are provided for all defendants.” Duncan v. Louisiana, 391 U. S. 145, 158 (1968). Trial by jury in serious criminal cases has long been regarded as an indispensable protection against the possibility of governmental oppression; the history of the jury’s development demonstrates “a long tradition attaching great importance to the concept of relying on a body of one’s peers to determine guilt or innocence as a safeguard against arbitrary law enforcement.” Williams v. Florida, 399 U. S. 78, 87 (1970). “Given this purpose, the essential feature of a jury obviously lies in the interposition between the accused and his accuser of the commonsense judgment of a group of laymen, and in the community participation and shared responsibility that results from that group’s determination of guilt or innocence.” Id., at 100.

Although we have held that the constitutional guarantee of trial by jury prescribes neither the precise number that can constitute a jury, Williams v. Florida, supra (six-person jury does not violate Sixth and Fourteenth Amendments), nor *331the exact proportion of the jury that must concur in the verdict, Apodaca v. Oregon, 406 U. S. 404 (1972) (10-to-2 vote in state trial does not violate the Constitution), we have also declared that there do exist size and unanimity limits that cannot be transgressed if the essence of the jury trial right is to be maintained. Thus Ballew v. Georgia, 435 U. S. 223 (1978), held that a reduction in the size of a jury to below six persons in nonpetty criminal cases raises such substantial doubts as to the fairness of the proceeding and the jury’s ability to represent the true sense of the community that it deprives the accused of his right to trial by j ury. For “much the same reasons,” we concluded in Burch that “conviction for a nonpetty offense by only five members of a six-person jury presents a similar threat to preservation of the substance of the jury trial guarantee” and hence violates the Sixth Amendment as applied to the States through the Fourteenth. 441 U. S., at 138. Though the line separating the permissible jury practice from the impermissible may not be the brightest, cf. Burch v. Louisiana, supra, at 137; Ballew v. Georgia, 435 U. S., at 231-232 (opinion of Blacknum, J.); id., at 245-246 (opinion of Powell, J.), a line must be drawn somewhere, and the constitutional inviolability of that border must be scrupulously respected lest the purpose and functioning of the jury be seriously impaired.

We think it apparent that the rationale behind the constitutional rule announced in Burch mandates its retroactive application. Mr. Justice Blackmun’s opinion in Ballew 7 *332cataloged the several considerations that led the Court to conclude that the operation of the jury was inhibited to a constitutionally significant degree by reducing its size to five members. Prominent among these concerns was the recognition, supported by a number of empirical studies,8 that a decline in jury size leads to less accurate factfinding and a greater risk of convicting an innocent person. Id., at 232-235.9 In addition, statistical and empirical data established that because of a concomitant decrease in the number of hung juries, a reduction in the size of the jury panel in criminal uases unfairly disadvantages one side — the defense. Id., at 236.10 Lastly, the opinion noted that the opportunity for meaningful and appropriate minority representation diminishes with the size of the jury. Id., at 236-237.11

*333Identical considerations underlay our decision in Burch. The threat which conviction by a 5-to-0 verdict poses to the fairness of the proceeding and the proper role of the jury is not significantly alleviated when conviction is instead obtained by the addition of a sixth, but dissenting, ballot. When the requirement of unanimity is abandoned, the vote of this “additional” juror is essentially superfluous. The prosecution’s demonstrated inability to convince all the jurors of the accused’s guilt certainly does nothing to allay our concern about the reliability and accuracy of the jury’s verdict. And while the addition of another juror to the five-person panel may statistically increase the- representativeness of that body, relinquishment of the unanimity requirement removes any guarantee that the minority voices will actually be heard.12

*334In sum, Burch established that the concurrence of six jurors was constitutionally required to preserve the substance of the jury trial right and assure the reliability of its verdict. It is difficult to envision a constitutional rule that more fundamentally implicates “the fairness of the trial — the very integrity of the fact-finding process.” Linkletter v. Walker, 381 U. S., at 639. “The basic purpose of a trial is the determination of truth,” Tehan v. United States ex rel. Shott, 382 U. S. 406, 416 (1966), and it is the jury to whom we have entrusted the responsibility for making this determination in serious criminal cases. Any practice that threatens the jury’s ability properly to perform that function poses a similar threat to the truth-determining process itself. The rule in Burch was directed toward elimination of just such a practice. Its purpose, therefore, clearly requires retroactive application.13

*335B

Due regard for countervailing considerations — the State's good-faith reliance on the old standards and the impact of retroactivity on the administration of justice — does not counsel a contrary result. The element of justifiable reliance on pre-Burch standards is minimal here. Unlike other cases that have been accorded prospective effect only, Burch did not overrule any prior decisions of this Court or invalidate a practice of heretofore unquestioned legitimacy. See, e. g., Desist v. United States, 394 U. S., at 250-251; Stovall v. Denno, 388 U. S., at 300; Tehan v. United States ex rel. Shott, supra, at 417. “Therefore, to build a ease for good-faith reliance the State must wring from our decision [s] the negative implication” that conviction by a nonunanimous six-person jury does not offend the Sixth Amendment's guarantee. See Adams v. Illinois, 405 U. S., at 293 (Douglas, J., dissenting). Yet if any implication is to be drawn from our opinions prior to Burch, it could only be that such a procedure was of doubtful constitutionality. Williams v. Florida, 399 U. S. 78 (1970), for example, highlighted the fact that the *336six-member jury approved in that case was required to render a unanimous verdict. Id., at 100, and n. 46. And Burch’s rule was distinctly foreshadowed by our decision in Ballew, which was handed down more than five months before petitioner’s trial and which was specifically cited to the trial court as mandating unanimity in the verdict of a six-member jury.' See supra, at 325-326. Cf. Berger v. California, 393 U. S. 314, 315 (1969).

Similarly, we are confident that retroactive application of the Burch rule will not have a devastating impact on the administration of the criminal law. It appears that by 1979 only two States — Louisiana and Oklahoma — permitted conviction of nonpetty offenses by a nonunanimous six-member jury, see Burch v. Louisiana, 441 U. S., at 138, and n. 12, and Louisiana, at least, did not institute its scheme until 1975.14 Furthermore, today’s decision will not affect the validity of all convictions obtained under Louisiana’s unconstitutional jury practice during that 4-year period, but only those in which it can be shown that the vote was in fact less than unanimous. Thus the number of persons who would have to *337be retried or released does not approach the magnitude involved in some of our previous cases. See, e. g., Linkletter v. Walker, 381 U. S., at 637 (retroactive application would “tax the administration of justice to the utmost”); Tehan v. United States ex rel. Shott, 382 U. S., at 419 (“an impact upon the administration of their criminal law so devastating as to need no elaboration”); DeStefano v. Woods, 392 U. S. 631, 634 (1968). What little disruption to the administration of justice results from retroactive application of Burch “must be considered part of the price we pay for former failures to provide fair procedures.” Adams v. Illinois, supra, at 297 (Douglas, J., dissenting).

Accordingly, the judgment of the Supreme Court of Louisiana is reversed. The case is remanded for further proceedings not inconsistent with this opinion.

It is so ordered.