with whom Justice Powell joins, concurring in the judgment.
To remedy the jeopardy-barred prosecution and conviction of respondent James Michael Mathews for aggravated mur*249der, the Ohio appellate court modified the conviction to one for the lesser included offense of murder, which was not jeopardy barred. The United States Court of Appeals for the Sixth Circuit held this remedy constitutionally insufficient because there was a “reasonable possibility” that the presence of the aggravated-murder charge prejudiced respondent’s defense against the charge of murder. I think the Court of Appeals applied the right standard but reached the wrong result. Accordingly, I concur in today’s judgment but do not join the Court’s opinion.
H-
Respondent concedes that after he pleaded guilty to armed robbery Ohio could have tried him for murder without violating the Double Jeopardy Clause. Disagreeing with the Court of Appeals, however, he contends that the presence of the jeopardy-barred charge of aggravated murder in his subsequent trial automatically rendered unconstitutional any conviction resulting from that trial. The Court correctly points out that this position cannot be reconciled with the terms of the judgment in Benton v. Maryland, 395 U. S. 784 (1969). See ante, at 246. Moreover, respondent’s primary argument for his position is unconvincing. He asserts that modifying his conviction from aggravated murder down to simple murder cannot possibly cure the violation, because the trial itself constituted the violation. The Double Jeopardy Clause, however, did not prohibit Ohio from holding a trial, only from seeking a conviction for aggravated murder. (It also barred the State from obtaining a conviction for aggravated murder, but the violation of that prohibition was remedied by the Ohio appellate court’s modification of the judgment.) The error, then, was not that the trial was held, but that it was improperly broad. The prosecution did something it was not allowed to do: it tried respondent for aggravated murder in addition to simple murder. It is true that the prosecutorial conduct at issue is prohibited not simply for *250its potential effect on the trial’s outcome, but also for the ordeal through which it put the defendant. As a consequence, it is also true that reducing respondent’s sentence does not make him “whole” for the violation: it does not compensate him, for example, for any mental anguish inflicted upon him by the prosecution for the aggravated offense. See Price v. Georgia, 398 U. S. 323, 331, and n. 10 (1970). But it hardly follows from these considerations that the appropriate remedy must always be to set aside the entire conviction and have yet another trial, particularly since one of the purposes of the Double Jeopardy Clause is to promote finality by avoiding multiple trials for the same offense.
Accordingly, the Court of Appeals held that respondent was entitled to a new trial on the murder charge only if he demonstrated a “‘reasonable possibility that he was prejudiced’” by the violation. Mathews v. Marshall, 754 F. 2d 158, 162 (CA6 1985), quoting Graham v. Smith, 602 F. 2d 1078, 1083 (CA2), cert. denied, 444 U. S. 995 (1979). The majority here now rejects that standard and holds that a defendant must demonstrate “a reliable inference of prejudice.” Ante, at 246. This means, the majority explains, that respondent must show that without the error “there is a reasonable probability” that he would not have been convicted of murder. Ibid, (emphasis added). In yet a third formulation of its standard, the Court announces: “To prevail in a case like this, the defendant must show that, but for the improper inclusion of the jeopardy-barred charge, the result of the proceeding probably would have been different.” Ante, at 247. To the extent that these standards differ from the “reasonable possibility” test applied by the Court of Appeals, they are, in my view, unprecedented and inappropriate.
I — I I — I
The Court starts out on the wrong foot by asserting that “this is not a ‘harmless error’ case.” Ante, at 244-245. Fundamentally, this is a “harmless error” case. Ohio concedes that it violated the Double Jeopardy Clause. To say *251that the remedy imposed by the state courts was constitutionally adequate is simply to say that the State’s acknowledged transgression may be deemed harmless with respect to respondent’s conviction for the lesser included offense. In Chapman v. California, 386 U. S. 18 (1967), this Court rejected the argument that no constitutional violation can ever be harmless. Some constitutional rights, of course, are “so basic to a fair trial” that their denial automatically requires reversal, id., at 23, but I agree with the Court that this category does not include double jeopardy violations of the sort involved here when the ultimate conviction is not for a jeopardy-barred offense. Under Chapman, therefore, respondent’s conviction for simple murder may be sustained if the State shows “beyond a reasonable doubt” that its error did not contribute to the conviction. Id., at 24; see also, e. g., United States v. Hasting, 461 U. S. 499, 511 (1983). As was noted in Chapman, the “harmless . . . beyond a reasonable doubt” standard is essentially the same as a requirement of reversal whenever there is a “reasonable possibility” that the error contributed to the conviction. 386 U. S., at 23-24.
The “reasonable possibility” standard originated in Fahy v. Connecticut, 375 U. S. 85 (1963), where it was applied to the improper introduction of illegally seized evidence. Less than two years prior to this Court’s decision in Chapman, the Court of Appeals for the Second Circuit persuasively demonstrated that the Fahy standard was equally applicable to situations of the kind involved here, i. e., jeopardy-barred prosecutions that ultimately result in convictions on lesser included charges that are not barred. See United States ex rel. Hetenyi v. Wilkins, 348 F. 2d 844 (1965), cert. denied sub nom. Mancusi v. Hetenyi, 383 U. S. 913 (1966). Hetenyi was charged by the State with first-degree murder but convicted only of second-degree murder, a lesser included offense. After his conviction was overturned on appeal, he again was prosecuted for first-degree murder, and ultimately *252convicted once more only of second-degree murder.1 Writing for the Court of Appeals, then-judge Marshall noted that the Constitution forbade the reprosecution of Hetenyi for an offense of which he had been impliedly acquitted in the first trial, but that the State constitutionally could have prosecuted Hetenyi again for second-degree murder. Nonetheless, the Court of Appeals invalidated Hetenyi’s reconviction for the lesser offense, because there was a “reasonable possibility that he was prejudiced” by the fact that he was charged with first-degree murder. 348 F. 2d, at 864 (emphasis in original). “For example,” Judge Marshall explained, “it is entirely possible that without the inclusion of the first degree murder charge, the jury, reflecting a not unfamiliar desire to compromise might have returned a guilty verdict on the first degree manslaughter charge,” a lesser included offense of second-degree murder. Id., at 866. The court refused to apply a more lenient test for harmless error, noting:
“The ends of justice would not be served by requiring a factual determination that the accused was actually prejudiced ... by being prosecuted for and charged with first degree murder, nor would the ends of justice be served by insisting upon a quantitative measurement of that prejudice. The energies and resources consumed by such inquiry would be staggering and the attainable level of certainty most unsatisfactory.” Id., at 864.
This Court relied on Hetenyi in Price v. Georgia, 398 U. S. 323 (1970), a case with similar facts. Price was tried for murder and found guilty of manslaughter. His conviction *253was overturned on appeal, he was retried for murder, and he again was found guilty of manslaughter. This Court held that the reprosecution for murder was barred on double jeopardy grounds, and rejected the State’s argument that the error was rendered harmless by the fact that the second jury convicted Price only of the unbarred offense. Citing Hetenyi, the Court noted that “we cannot determine whether or not the murder charge against petitioner induced the jury to find him guilty of the less serious offense of voluntary manslaughter rather than to continue to debate his innocence.” 398 U. S., at 331. The Court did not explicitly employ the “reasonable possibility” standard, but it observed: “The Double Jeopardy Clause ... is cast in terms of the risk or hazard of trial and conviction, not of the ultimate legal consequence of the verdict. To be charged and to be subjected to a second trial for first-degree murder is an ordeal not to be viewed lightly.” Ibid. The Court certainly gave no indication that it would consider the error harmless unless Price could show that “but for the improper inclusion of the jeopardy-barred charge, the result of the proceeding probably would have been different.” See ante, at 247.2
*254The Court today offers virtually no explanation for departing from Chapman and Fahy in favor of a more lenient approach. It cites no support at all for the “reliable inference” and “probably would have been different” formulations of the new test it announces. In support of the “reasonable probability” formulation, the Court refers to Strickland v. Washington, 466 U. S. 668 (1984), which used the same words but did not concern the adequacy of a proffered remedy for an acknowledged constitutional violation. The question in Strickland was whether there had been a constitutional violation in the first place. Id., at 691-692. We held that a professionally unreasonable mistake by defense counsel constitutes ineffective assistance of counsel under the Sixth Amendment only if in retrospect there is a “reasonable probability” that the mistake altered the verdict — that is, “a probability sufficient to undermine confidence in the outcome.” Id., at 694; cf. United States v. Bagley, 473 U. S. 667 (1985) (prosecutor’s failure to disclose evidence favorable to the accused violates due process where there is a “reasonable probability” that disclosure would have affected the outcome).3 In this case, however, it is common ground that *255respondent’s rights under the Double Jeopardy Clause were violated when Ohio tried him for aggravated murder. The question is not whether Ohio has also violated the Sixth Amendment or the Due Process Clause. The question is whether the State has sufficiently contained the damage from its acknowledged violation of the Double Jeopardy Clause, or whether that transgression taints even the conviction for simple murder. At issue is the extent to which the law will tolerate a conviction that may have been obtained through abridgment of a defendant’s constitutional rights. Once it is established that the State has violated the Constitution in the course of a prosecution, the proceedings lose whatever presumption of regularity they formerly enjoyed, and the State properly bears a heavy burden in arguing that the result should nonetheless be treated as valid.
By ruling, despite Chapman, that a defendant in a case such as this must show more than a reasonable possibility of prejudice to invalidate the conviction, the Court makes double jeopardy violations more readily excusable than any other kind of constitutional error. For me, that makes little sense. Violations of the Double Jeopardy Clause are no less serious than violations of other constitutional protections. Their excusability should be judged by the same standard. The Court offers no real explanation for the special leniency it announces today, and there is none.
I — I I — I
The proper question in this case is thus whether Ohio has shown beyond a reasonable doubt that the aggravated-murder charge did not contribute to respondent’s conviction for simple murder. Under Chapman and Fahy this means *256that the State must convincingly disprove any reasonable possibility that the greater charge prejudiced respondent’s defense against the lesser. With most errors, the prejudice to be feared is so obvious that there is no need for the defendant to spell it out. When illegally obtained evidence is introduced, or when the prosecutor makes forbidden remarks, it is understood that the danger is that the illegal evidence or the remarks will influence the jury. In this case, however, there is no obvious way in which the defendant could have been prejudiced. The conviction for the jeopardy-barred offense was reduced to one for an unbarred, lesser included offense. There is no possibility of a compromise verdict, as in Hetenyi or Price, because here the jury convicted for the greater offense charged. Nor is there any reason to believe that evidence, regarding the robbery, admitted to prove the aggravating factor in respondent’s trial, would not have been admitted in a trial for simple murder. Like the Federal Rules of Evidence, Ohio’s evidence code allows the introduction of evidence of other crimes to show “proof of motive, opportunity, intent, preparation, plan, knowledge, identity, or absence of mistake or accident.” Ohio Rule Evid. 404(B); see also Fed. Rule Evid. 404(b). The murder in this case was committed while in flight from the bank robbery. The State’s theory is that respondent murdered his accomplice in order to silence him so that respondent could escape prosecution for the robbery and, he hoped, return later to pick up the stolen money he had hidden. There consequently is no apparent reason why evidence of the robbery would not have been admissible to show motive and opportunity for the murder.
Given all this, respondent is obligated to spell out with some specificity how the trial might have gone better for him had the State charged only simple murder. He has not done so; instead, he has simply speculated that all sorts of things might have been different. That is not enough to prevent this Court from “declaring] a belief that [the error] was *257harmless beyond a reasonable doubt.” Chapman, 386 U. S., at 24. If it were, the remand in Benton v. Maryland, 395 U. S. 784 (1969), would have been inappropriate: the Court there simply would have vacated the burglary conviction, because there was no telling what would have happened had the defendant not been forced to defend himself against the larceny charge. Perhaps different trial tactics would have been tried; perhaps defense counsel would have prepared more fully on the burglary charge. Indéed, if abstract speculations of this sort sufficed to create a “reasonable doubt” that an error was harmless, it is difficult to see how any constitutional error ever would qualify.
I therefore concur in the Court’s judgment, although I see no justification for departing from the traditional and established standards for deciding questions of this kind.
Hetenyi was actually tried three times, each time for first-degree murder. In the second trial he was convicted of that offense, but the second conviction, like the first, was reversed on appeal on grounds other than double jeopardy. The third trial resulted in a conviction for second-degree murder.
Since Price, three Courts of Appeals have reviewed whether a double jeopardy violation was adequately remedied by modifying the jeopardy-barred conviction to one for a lesser included offense that was not jeopardy-barred; like the Court of Appeals for the Sixth Circuit in this case, each of the other two courts has followed Hetenyi. In Graham v. Smith, 602 F. 2d 1078 (CA2), cert. denied, 444 U. S. 995 (1979), the court addressed a situation similar to the one involved here. Graham’s conviction for second-degree murder had been reduced on appeal to manslaughter. The modified conviction was then set aside on federal habeas, and the State successfully retried Graham for second-degree murder. Recognizing the double jeopardy violation, a state appellate court reduced the new conviction to manslaughter. This remedy was judged insufficient by the Court of Appeals because Graham’s testimony at the second trial resulted in the introduction of extremely damaging impeachment evidence, and he claimed that he might not have testified had the State charged only manslaughter. The Court of Appeals found this claim doubtful but sufficiently plausible to *254create a “‘reasonable possibility that [Graham] was prejudiced.’” 602 F. 2d, at 1083, quoting Hetenyi, 348 F. 2d, at 864.
The Court of Appeals for the Fifth Circuit took a similar approach in Tapp v. Lucas, 658 F. 2d 383 (1981), cert. denied, 456 U. S. 972 (1982), which also involved facts resembling those here. Tapp was prosecuted for murder and convicted of manslaughter. The conviction was set aside, and he was retried for murder, this time with success. To remedy the double jeopardy violation, the State Supreme Court reduced the second conviction to manslaughter. Citing Graham v. Smith, the Court of Appeals upheld this remedy as “a common sense solution to the problem of avoiding yet another trial.” 658 F. 2d, at 386. The court distinguished Price on the ground that “the possibility of prejudicial jury compromise is simply absent from this case.” Ibid, (emphasis added).
As in Strickland, the Court today defines a “reasonable probability” as “a probability sufficient to undermine confidence in the outcome.” Ante, at 247; Strickland, 466 U. S., at 694. In Strickland, however, we specifically refused to require a showing “that counsel’s deficient conduct more *255likely than not altered the outcome in the ease.” Id., at 693 (emphasis added). The Court’s reliance on some of the language of Strickland therefore renders particularly puzzling its wholly unprecedented demand one paragraph later that the respondent here demonstrate that without the jeopardy-barred charge “the result of the proceeding probably would have been different.” Ante, at 247.