with whom
Justice Brennan and Justice Blackmun join, dissenting.Under Arkansas law, a defendant who is convicted of a class B felony and “who has previously been convicted of. . . [or] found guilty of four [4] or more felonies” may be sentenced to an enhanced term of imprisonment ranging from 20 years to 40 years. Ark. Stat. Ann. §41-1001(2)(b) (1977) (current version at Ark. Code Ann. § 5 — 4—501(b)(3) (1987)). At the March 1982 sentencing trial held after Johnny Lee Nelson pleaded guilty to the class B felony of burglary,1 the State of Arkansas introduced evidence indicating that Nelson *43had four prior felony convictions. Nelson protested that he had received a gubernatorial pardon for one of the convictions. The prosecutor and the trial judge disbelieved Nelson’s claim, however, and the jury sentenced him to 20 years in prison. Three and a half years later — during which time Nelson, from jail, persistently implored Arkansas courts to investigate his pardon claim — a Federal District Court finally ordered the State to check its records. Lo and behold, it turned out that Nelson had been pardoned — and Arkansas soon announced its intention to try Nelson, once again, as a habitual offender.2
The majority holds today that, although Arkansas attempted once and failed to prove that Nelson had the four prior convictions required for habitual offender status, it does not violate the Double Jeopardy Clause for Arkansas to attempt again. I believe, however, that Nelson’s retrial is squarely foreclosed by Burks v. United States, 437 U. S. 1 *44(1978), where we held that a State may not retry a defendant where it failed initially to present sufficient evidence of guilt. The majority rushes headlong past those facets of Nelson’s case and of Arkansas law that reveal the prosecution’s failure to present sufficient evidence of guilt in this case, in order to answer the open and narrow question of double jeopardy law on which the Court granted certiorari. By virtue of the majority’s haste, Nelson now faces a new sentencing trial, and Arkansas will be able to augment the evidence it presented at Nelson’s initial trial with evidence of prior convictions it opted not to introduce in the first place. Because this result embodies the classic double jeopardy evil of a State “honing its trial strategies and perfecting its evidence through successive attempts at conviction,” Tibbs v. Florida, 457 U. S. 31, 41 (1982), I dissent.
I
The Double Jeopardy Clause is “designed to protect an individual from being subjected to the hazards of trial and possible conviction more than once for an alleged offense.” Green v. United States, 355 U. S. 184, 187 (1957). Reflecting this principle, we held in Burks that the prohibition against double jeopardy prevents retrial where a State’s evidence at trial is found insufficient. See also Hudson v. Louisiana, 450 U. S. 40 (1981); Greene v. Massey, 437 U. S. 19 (1978). The Burks rule is based on the time-honored notion that the State should be given only “one fair opportunity to offer whatever proof it [can] assemble.” Burks, supra, at 16. Unlike a finding of reversible trial error, which traditionally has not barred retrial, see United States v. Tateo, 377 U. S. 463 (1964); United States v. Ball, 163 U. S. 662 (1896), reversal for evidentiary insufficiency “constitute^] a decision to the effect that the government has failed to prove its case.” Burks, supra, at 15.
This case is troubling in a number of respects, not the least of which is that no one in the Arkansas criminal justice system seems to have taken Nelson’s pardon claim at all seri*45ously. At bottom, however, this case is controlled by the Burks insufficiency principle. For under Arkansas’ law of pardons, the State’s evidence against Nelson in his sentencing trial was at all times insufficient to prove four valid prior convictions. The majority errs in treating this as a case of mere trial error, and in reaching the unsettled issue whether, after a trial error reversal based on the improper admission of evidence, a reviewing court should evaluate the sufficiency of the evidence by including, or excluding, the tainted evidence. See Greene v. Massey, supra, at 26, n. 9 (expressly reserving this question). This case has nothing to do with inadmissible evidence and everything to do with Arkansas’ defective proof.
As the District Court noted in ruling for Nelson, Arkansas decisional law holds that pardoned convictions have no probative value in sentence enhancement proceedings. See 641 F. Supp. 174, 183 (ED Ark. 1986) (under Arkansas law: “[A] pardon renders the conviction a nullity. . . . [F]or purposes of the enhancement statute, a conviction which has been pardonned [sic] is not a conviction”). The District Court cited a 1973 decision of the Arkansas Supreme Court, Duncan v. State, 254 Ark. 449, 494 S. W. 2d 127 (1973), which held that a pardoned conviction cannot be counted toward the four prior convictions required under the State’s sentence enhancement statute. The Duncan court, id., at 451, 494 S. W. 2d, at 129, quoted with approval this Court’s decision in Ex parte Garland, 4 Wall. 333, 380 (1867), where we stated: “A pardon reaches both the punishment prescribed for the offense and the guilt of the offender; and when the pardon is full, it releases the punishment and blots out of existence the guilt, so that in the eye of the law the offender is as innocent as if he had never committed the offense.” Drawing upon that state-court holding, the District Court in this case concluded: “The truth is that the state could not *46have provided any evidence to rebut the petitioner’s contention because it did not exist.” 641 F. Supp., at 184.3
That Arkansas was not roused to investigate Nelson’s pardon claim until long after his trial does not transform the State’s failure of proof — fatal for double jeopardy purposes under Burks — into a mere failure of admissibility. As the District Court noted, Arkansas law establishes “that the prosecutor must carry the significant burden of ferreting out information regarding the validity of prior convictions whenever he seeks enhancement.” 641 F. Supp., at 184 (citing Roach v. State, 255 Ark. 773, 503 S. W. 2d 467 (1973)). The delay in the discovery of Nelson’s pardon does not change the essential fact that, as a matter of state law, the paper evidence of the disputed conviction presented by the prosecutor was devoid of probative value from the moment the conviction was expunged by the pardon. A pardon simply “blots out of existence” the conviction as if it had never happened. Duncan v. State, supra, at 451, 494 S. W. 2d, at 129. If, in seeking to prove Nelson’s four prior convictions, the State had offered documented evidence to prove three valid prior convictions and a blank piece of paper to prove a fourth, no one would doubt that Arkansas had produced insufficient evidence and that the Double Jeopardy Clause barred retrial. There is no constitutionally significant difference between that hypothetical and this case.4
*47In sum, Arkansas had “one fair opportunity to offer whatever proof it could assemble” that Nelson had four prior convictions, Burks, 437 U. S., at 16, but it “failed to prove its case.” Id., at 15. In reversing both the District Court and the Court of Appeals to give Arkansas a second chance to sentence Nelson as a habitual offender, the majority pays no more than lipservice to the Burks insufficiency principle. I would therefore hold that the Double Jeopardy Clause prohibits Arkansas from subjecting Nelson to a new sentencing trial at which it can “supply evidence” of a fourth conviction “which it failed to muster in the first proceeding.” Id., at 11.
II
Even if I did not regard this as a case of insufficient evidence controlled by Burks, I could not join my colleagues in the majority. The question whether a reviewing court, in evaluating insufficiency for double jeopardy purposes, should look to all the admitted evidence, or just the properly admitted evidence, is a complex one. It is worthy of the thoughtful consideration typically attending this Court’s decisions concerning the Double Jeopardy Clause.
The majority instead resolves this issue as if it had already been decided. Ante, at 40-41. In the majority’s view: “It is quite clear from our opinion in Burks that a reviewing court must consider all of the evidence admitted by the trial court in deciding whether retrial is permissible under the Double Jeopardy Clause.” Ibid. Burks decided no such thing. At issue in Burks was whether a finding of initial insufficiency bars a defendant’s retrial; we held that it did. *48Burks did not presume to decide the completely distinct issue, raised by this case, of by what measure a reviewing court evaluates insufficiency in cases where a piece of evidence which went to the jury is later ruled inadmissible. Indeed, had Burks settled or even logically foreclosed this issue, there would have been no reason for us specifically to reserve its resolution in Greene v. Massey, 437 U. S., at 26, n. 9 — a case decided the very same day as Burks.5
It seems to me that the Court’s analysis of this issue should begin with the recognition that, in deciding when the double jeopardy bar should apply, we are balancing two weighty interests: the defendant’s interest in repose and society’s interest in the orderly administration of justice. See, e. g., United States v. Tateo, 377 U. S., at 466. The defendant’s interest in avoiding successive trials on the same charge reflects the idea that the State
“should not be allowed to make repeated attempts to convict an individual for an alleged offense, thereby subjecting him to embarrassment, expense and ordeal and compelling him to live in a continuing state of anxiety and insecurity, as well as enhancing the possibility that even though innocent he may be found guilty.” Green v. United States, 355 U. S., at 187-188.
*49See also Burks, supra, at 11. Society’s corresponding interest in the sound administration of justice reflects the fact that “[i]t would be a high price indeed for society to pay were every accused granted immunity from punishment because of any defect sufficient to constitute reversible error in the proceedings leading to conviction.” United States v. Tateo, supra, at 466.
I do not intend in this dissenting opinion to settle what rule best accommodates these competing interests in cases where a reviewing court has determined that a portion of a State’s proof was inadmissible. At first blush, it would seem that the defendant’s interest is every bit as great in this situation as in the Burks situation. Society’s interest, however, would appear to turn on a number of variables. The chief one is the likelihood that retrying the defendant will lead to conviction. See United States v. Tateo, supra, at 466 (noting society’s interest “in punishing one whose guilt is clear”). In appraising this likelihood, one might inquire into whether prosecutors tend in close cases to hold back probative evidence of a defendant’s guilt; if they do not, there would be scant societal interest in permitting retrial given that the State’s remaining evidence is, by definition, insufficient.6 Alternatively, one might inquire as to why the evidence at issue was deemed inadmissible. Where evidence was stricken for reasons having to do with its unreliability, it would seem curious to include it in the sufficiency calculus. Inadmissible hearsay evidence, for example, or evidence deemed defective or nonprobative as a matter of law thus might not be included. By contrast, evidence stricken in compliance with evidentiary rules grounded in other public policies — the policy of encouraging subsequent remedial measures embodied in Federal Rule of Evidence 407, for ex*50ample, or the policy of deterring unconstitutional searches and seizures embodied in the exclusionary rule — might more justifiably be included in a double jeopardy sufficiency analysis.7
The Court today should have enunciated rules of this type, rules calibrated to accommodate, as best as possible, the defendant’s interest in repose with society’s interest in punishing the guilty. Regrettably, the majority avoids such subtlety in its terse opinion. Instead, it opts for a declaration that our decision in Burks — although no one knew it at the time — was settling the issue on which we granted certiorari here. This is ipse dixit jurisprudence of the worst kind. I dissent.
Nelson pleaded guilty to having taken $45.00 from a vending machine in 1979. See 641 F. Supp. 174, 175 (ED Ark. 1986).
The conviction for which Nelson was pardoned was a 1960 conviction for assault with intent to rape. He was pardoned in 1964 by Arkansas Governor Orval E. Faubus. App. 6 (text of pardon).
The record in this case shows that Nelson attempted unsuccessfully both during and after his trial to alert state authorities to this pardon. During the trial, Nelson stated that after serving three years in jail, he “had the case investigated and the governor at the time Faubus which [sic] gave me a pardon for my sentence.” Id., at 8 (abridged transcript of sentencing trial). He added: “[A]t my home I have documents of that pardon on that [sic].” Id., at 9. The prosecutor did not question Nelson about this claim. Instead, the prosecutor moved to strike Nelson’s testimony on the ground that Nelson was “confused as to the meaning of the pardon and a commutation.” Id., at 11. The prosecutor further stated: “I think the records are clear that are in the court. . . .” Id., at 11-12. Ultimately, the trial judge, and Nelson’s own defense counsel — who like the prosecutor had never investigated Nelson’s claim of pardon — accepted this account. Id., at 12.
After receiving the enhanced sentence, Nelson sought both on direct appeal and in state postconviction actions to have his claim investigated. Only after a Federal District Court ordered Arkansas to investigate Nelson’s claim did Nelson’s pardon finally come to light — in August 1985. Id., at 1-4.
The Court of Appeals did not disturb this determination of the District Court. Rather, it focused upon, and rejected, Arkansas’ separate contention that double jeopardy does not attach to sentence enhancement trials. See 828 F. 2d 446, 449 (CA8 1987). That issue is not before this Court, Arkansas having conceded the validity of this aspect of the Court of Appeals’ ruling. See ante, at 36-37, n. 4. The Court of Appeals also rejected as incorrect Arkansas’ claim that, in cases of trial error, reviewing courts should not engage in any subsequent review for insufficiency, however measured. 828 F. 2d, at 450.
The majority offers its own analogy: the discovery of Nelson’s pardon, it states, is like “newly discovered evidence.” Ante, at 41, n. 7. The majority overlooks a critical distinction. The emergence of new evidence in *47no way strips the old evidence of all probative value; while new evidence may cast doubt on the persuasiveness of the old evidence, its emergence does not render once sufficient evidence “insufficient.” Arkansas’ law of pardons, by contrast, robs evidence of a pardoned conviction of all probative value. It was thus not the discovery of Nelson’s pardon that stripped his prior conviction of evidentiary weight, but rather the fact of the pardon itself. The discovery of Nelson’s pardon merely called the parties’ attention to this critical fact.
None of the numerous appellate court cases cited by the majority in support of its resolution of this issue, ante, at 41, n. 8, interpreted Burks as disposing of the sufficiency question before us. Rather, with varying degrees of analysis, these courts evaluated the ramifications of including or excluding tainted evidence in a sufficiency analysis upon the interests of the defendant and of society — precisely the analytic approach I urge in the succeeding paragraphs. See, e. g., United States v. Tranowski, 702 F. 2d 668, 671 (CA7 1983) (concluding that policy arguments favor including tainted evidence in insufficiency analysis), cert. denied, 468 U. S. 1217 (1984); Bullard v. Estelle, 665 F. 2d 1347, 1358-1361 (CA5 1982) (using similar interest analysis in case involving retrial for sentence enhancement and concluding that inadmissible evidence should not be included in insufficiency analysis).
It is no answer to say that prosecutors who initially lacked sufficient admissible evidence may gather more before a retrial. Such conduct is-precisely what the Double Jeopardy Clause was designed to guard against. See Tibbs v. Florida, 457 U. S. 31, 41 (1982).
Arkansas suggests a “clear trial court ruling” test as a means of accommodating defense and societal interests. Under this test, where a trial court has affirmatively ruled that a piece of evidence is admissible, a State is entitled to rely on that ruling by counting this evidence in a subsequent insufficiency analysis — even if a reviewing court had ruled the evidence inadmissible. Brief for Petitioner 12. This test furthers a societal interest of which this Court took note in United States v. Tateo, 377 U. S. 463, 466 (1964): the interest in not deterring appellate courts from safeguarding defendants’ rights. It is not at all clear, however, that Arkansas’ test would authorize retrial in this case. Far from having refrained from introducing evidence of additional convictions in reliance on a trial court’s determination that Nelson had not received a pardon, the prosecutor in this case seems to have done all he could to lead the trial court to believe that Nelson’s pardon claim was meritless. See n. 2, supra.