dissenting.
In 1972, respondent was tried in a Louisiana state court for possessing and distributing cocaine and heroin. Pursuant to the state law applicable at that time, the jury consisted of five members. La. Code Crim. Proc. Ann., Art. 782 (West 1967) (amended 1975). Respondent raised no objection to its size. The jury unanimously voted to convict respondent, and he was sentenced to a prison term.
More than six years later, after exhausting state remedies, respondent sought habeas corpus in Federal District Court.1 *954Relying on this Court’s decision in Ballew v. Georgia, 435 U. S. 223 (1978), respondent argued that his trial violated the Sixth Amendment (applicable to the States through the Fourteenth), because a jury of five persons was constitutionally inadequate. The District Court granted the writ in reliance on Ballew, App. to Pet. for Cert. 11a, 22a-31a, and the Court of Appeals for the Fifth Circuit affirmed, 623 F. 2d 383 (1980). The Court of Appeals recognized that respondent’s conviction had become final long before Ballew held that five-member juries are unconstitutional. It nevertheless held that Ballew should be applied retroactively to invalidate all convictions rendered by juries of that size.2
1 believe that the Court of Appeals improperly applied Ballew to reverse respondent’s conviction. I therefore would grant the petition for certiorari and reverse the decision of the Court of Appeals.
Three recent cases govern respondent’s claim. In Ballew, supra, we held that juries in criminal cases must have at least six members to meet constitutional requirements. A smaller jury may be insufficient to “foster effective group deliberation,” 435 U. S., at 232, 234 (opinion of Blackmun, J.), and to provide a “fair cross-section” of the community, id., at 245 (White, J., concurring in judgment). Our decision in Ballew was reaffirmed one year later in Burch v. Louisiana, 441 U. S. 130, 137 (1979), when we identified similar constitutional flaws in a conviction reached by a nonunanimous *955six-member jury. Cf. Apodaca v. Oregon, 406 U. S. 404 (1972); Johnson v. Louisiana, 406 U. S. 356 (1972).
Soon thereafter, in Brown v. Louisiana, 447 U. S. 323 (1980), the Court applied Burch retroactively to reverse a conviction reached by a nonunanimous six-member jury. The Court divided three ways in Broum, and it is essential for our present purposes to identify the divergent views. Justice Brennan wrote for a plurality of four Justices and concluded that any conviction reached by a nonunanimous six-member jury should be reversed. Two other Justices concurred in the result, stating that Burch should be applied retroactively only to cases pending on direct review at the time Burch was decided. 447 U. S., at 337 (Powell, J., with whom Stevens, J., joined), quoting Hankerson v. North Carolina, 432 U. S. 233, 248 (1977) (Powell, J., concurring in judgment). Three Justices dissented, arguing against any retroactive application of the new rule, because there was no “ 'substantial likelihood’ ” that a 5-to-l jury reached a result that was “ Tactually incorrect.’ ” 447 U. S., at 338 (Rehnquist, J., with whom Burger, C. J., and White, J., joined), quoting Williams v. United States, 401 U. S. 646, 656, n. 7 (1971).
In sum, in Burch, as in Ballew, we identified constitutional defects in jury composition. Though the system challenged in each case differed somewhat, we invalidated each one for essentially the same reason: the Constitution requires that criminal juries be structured in a manner conducive to highly reliable adjudication. Ballew, supra, at 232, 234 (opinion of Blackmun, J.); Burch, supra, at 137. It does not follow from either case, however, that unanimous five-member juries and nonunanimous six-member juries frequently — or even occasionally — render incorrect decisions. A criminal defendant’s interest in a new trial, based on post hoc “speculation about what would have happened” with a jury of different size or structure, Brown v. Louisiana, supra, at 340 (Rehnquist, J., dissenting), need not always prevail over the *956public interest in assuring the finality of convictions. See Sumner v. Mata, 449 U. S. 539, 550-551, n. 3 (1981) (“both society and the individual criminal defendant have [an interest] ‘in insuring that there will at some point be the certainty that comes with an end to litigation’ ”), quoting Sanders v. United States, 373 U. S. 1, 24-25 (1963) (Harlan, J., dissenting); Mackey v. United States, 401 U. S. 667, 675-702 (1971) (Harlan, J., concurring in part and dissenting in part); Hankerson v. North Carolina, supra, at 247-248 (Powell, J., concurring in judgment). At some point, the costs of retroactive application of new constitutional principles of jury size or structure exceed the possibility of enhanced reliability that may be obtained upon retrial.3
The present case involves a conviction rendered by a unanimous five-member jury. If the case now were to be tried, it is plain in light of Ballew that such a jury is not of constitutionally adequate size. But this case was tried in 1972 — more than six years before Ballew — and it is now before us on collateral review. The retroactivity analysis of the plurality in Brown v. Louisiana thus is not controlling. Instead, the governing position is that represented by the combined views of the other five Justices in Broion. Because the Court of Appeals in this case improperly relied on the reasoning of the Brown plurality to apply Ballew retroactively, I would grant certiorari and reverse its judgment.
Respondent did not take a direct appeal from his conviction. He exhausted his state remedies by seeking state collateral relief, relying for the *954first time on his jury-composition claim. The state trial court did not address respondent’s failure to enter a contemporaneous objection to the five-member jury. Its denial of relief thus was on the merits. App. to Pet. for Cert. 34a. The Louisiana Supreme Court denied review. State ex rel. Thomas v. Blackburn, 361 So. 2d 1218 (1978).
Respondent failed to object at trial to the size of his jury. N. 1, supra. The Court of Appeals correctly disregarded this procedural default and reached the merits of respondent’s petition because the state trial court on collateral review had ruled upon the merits, see ibid. County Court of Ulster v. Allen, 442 U. S. 140, 152-153, 154 (1979); Franks v. Delaware, 438 U. S. 154, 161-162 (1978).
The State is greatly disadvantaged when a conviction, long thought to be final, is reversed on collateral review. The State’s opportunity to hold a retrial under these circumstances may be only theoretical. Witnesses disappear and memories fade with the passage of time. See Hankerson v. North Carolina, 432 U. S. 233, 247 (1977) (Powell, J., concurring in judgment).