dissenting.
Having joined Justice Blackmun’s dissent from the plurality’s opinion in Walton v. Arizona, 497 U. S. 639, 649-652 (1990), I necessarily also subscribe to the views expressed by Justice Souter today. I write separately for two reasons: to explain why agreement with Justice Blackmun’s dissent is fully consistent with refusing to read Walton as “controlling],” but see ante, at 169 (opinion of the Court), and to explain why the grant of certiorari in this case was a misuse of our discretion.
Under Justice Blackmun’s understanding of Arizona law, Walton did present exactly the same issue before us today. The Arizona statute at issue required the judge to impose death upon finding aggravating factors if “‘there are no mitigating circumstances sufficiently substantial to call for leniency.’” 497 U. S., at 644 (quoting Ariz. Rev. Stat. Ann. *200§ 13-703(E) (West 1989)). In Justice Blackmun’s view, Arizona case law indicated “that a defendant’s mitigating evidence will be deemed ‘sufficiently substantial to call for leniency’ only if the mitigating factors ‘outweigh’ those in aggravation.” 497 U. S., at 687. Accordingly, Justice Blackmun believed that we confronted the constitutionality of a statute that mandated death when the scales were evenly balanced. Ibid.
But Justice Blackmun never concluded that the plurality similarly read Arizona case law as “requir[ing] a capital sentence in a case where aggravating and mitigating circumstances are evenly balanced.” Id., at 688. To the contrary, he observed that “the plurality does not even acknowledge that this is the dispositive question.” Ibid. Because Justice Blackmun did not read the plurality opinion as confronting the problem of equipoise that he believed Arizona law to present, my join of his dissent is consistent with my conclusion that stare decisis does not bind us today. As Justice Souter explains, post, at 203-204, n. 1 (dissenting opinion), the Walton plurality painstakingly avoided an express endorsement of a rule that allows a prosecutor to argue, and allows a judge to instruct the jury, that if the scales are evenly balanced when the choice is between life and death, the law requires the more severe penalty.
There is a further difference between this case and Walton—one that should have kept us from granting certiorari in the first place. In Walton, the defendant petitioned for certiorari, and our grant enabled us to consider whether the Arizona Supreme Court had adequately protected his rights under the Federal Constitution. In this case, by contrast, the State of Kansas petitioned us to review a ruling of its own Supreme Court on the grounds that the Kansas court had granted more protection to a Kansas litigant than the Federal Constitution required. A policy of judicial restraint would allow the highest court of the State to be the final *201decisionmaker in a case of this kind. See Brigham City v. Stuart, 547 U. S. 398, 409 (2006) (Stevens, J., concurring).
There is a remarkable similarity between the decision to grant certiorari in this case and our comparable decision in California v. Ramos, 463 U. S. 992 (1983). In Ramos, we reviewed a decision of the California Supreme Court that had invalidated a standard jury instruction concerning the Governor’s power to commute life without parole sentences—an instruction that was unique to California. By a vote of 5 to 4, the Court reversed the judgment of the state court, concluding—somewhat ironically—that “the wisdom of the decision to permit juror consideration of possible commutation is best left to the States.” Id., at 1014.
In response I asked, as I do again today, “what harm would have been done to the administration of justice by state courts if the [Kansas] court had been left undisturbed in its determination^]” Id., at 1030. “If it were true that this instruction may make the difference between life and death in a ease in which the scales are otherwise evenly balanced, that is a reason why the instruction should not be given— not a reason for giving it.” Ibid. “No matter how trivial the impact of the instruction may be, it is fundamentally wrong for the presiding judge at the trial—who should personify the evenhanded administration of justice—to tell the jury, indirectly to be sure, that doubt concerning the proper penalty should be resolved in favor of [death].” Ibid.
As in Ramos, in this case “no rule of law commanded the Court to grant certiorari.” Id., at 1031. Furthermore, “[n]o other State would have been required to follow the [Kansas] precedent if it had been permitted to stand. Nothing more than an interest in facilitating the imposition of the death penalty in [Kansas] justified this Court’s exercise of its discretion to review the judgment of the [Kansas] Supreme Court.” Ibid. And “[t]hat interest, in my opinion, is not sufficient to warrant this Court’s review of the validity of a *202jury instruction when the wisdom of giving that instruction is plainly a matter that is best left to the States.” Ibid*
We decided Ramos on the same day as Michigan v. Long, 463 U. S. 1032 (1983). Prior to that time, “we had virtually no interest” in criminal cases where States sought to set aside the rulings of their own courts. Id., at 1069 (Stevens, *203J., dissenting). Although in recent years the trend has been otherwise, I continue to hope “that a future Court will recognize the error of this allocation of resources,” id., at 1070, and return to our older and better practice of restraint.
Justice Souter, with whom Justice Stevens, Justice Ginsburg, and Justice Breyer join, dissenting.
I
Kansas’s capital sentencing statute provides that a defendant “shall be sentenced to death” if, by unanimous vote, “the jury finds beyond a reasonable doubt that one or more of the aggravating circumstances . . . exist and . . . that the existence of such aggravating circumstances is not outweighed by any mitigating circumstances which are found to exist.” Kan. Stat. Ann. §21-4624(e) (1995). The Supreme Court of Kansas has read this provision to require imposition of the death penalty “[i]n the event of equipoise, [that is,] the jury’s determination that the balance of any aggravating circumstances and any mitigating circumstances weighed equal.” 278 Kan. 520, 534, 102 P. 3d 445, 457 (2004) (case below); see also State v. Kleypas, 272 Kan. 894, 1016, 40 P. 3d 139, 232 (2001) (per curiam) (stating that the language of § 21-4624(e) “provides that in doubtful cases the jury must return a sentence of death”). Given this construction, the state court held the law unconstitutional on the ground that the Eighth Amendment requires that a “ Tie g[o] to the defendant’ when life or death is at issue.” Ibid. Because I agree with the Kansas justices that the Constitution forbids a mandatory death penalty in what they describe as “doubtful cases,” when aggravating and mitigating factors are of equal weight, I respectfully dissent.1
*204II
More than 30 years ago, this Court explained that the Eighth Amendment’s guarantee against cruel and unusual punishment barred imposition of the death penalty under statutory schemes so inarticulate that sentencing discretion produced wanton and freakish results. See Furman v. Georgia, 408 U. S. 238, 309-310 (1972) (Stewart, J., concurring) (“[T]he Eighth and Fourteenth Amendments cannot tolerate the infliction of a sentence of death under legal systems that permit this unique penalty to be ... wantonly and . . . freakishly imposed” on a “capriciously selected random handful” of individuals). The Constitution was held to require, instead, a system structured to produce reliable, Woodson v. North Carolina, 428 U. S. 280, 305 (1976) (plurality opinion), rational, Jurek v. Texas, 428 U. S. 262, 276 (1976) (joint opinion of Stewart, Powell, and Stevens, JJ.), and rationally reviewable, Woodson, supra, at 303, determinations of sentence.
Decades of back-and-forth between legislative experiment and judicial review have made it plain that the constitutional demand for rationality goes beyond the minimal requirement to replace unbounded discretion with a sentencing structure; a State has much leeway in devising such a structure and in selecting the terms for measuring relative culpability, but a system must meet an ultimate test of constitutional reliability in producing “ ‘a reasoned moral response to the defendant’s background, character, and crime,’” Penry v. Lynaugh, 492 U. S. 302, 319 (1989) (quoting California v. Brown, 479 U. S. 538, 545 (1987) (O’Connor, J., concurring); emphasis deleted); cf. Gregg v. Georgia, 428 U. S. 153, 206 (1976) (joint opinion of Stewart, Powell, and Stevens, JJ.) (sanctioning *205sentencing procedures that “focus the jury’s attention on the particularized nature of the crime and the particularized characteristics of the individual defendant”). The Eighth Amendment, that is, demands both form and substance, both a system for decision and one geared to produce morally justifiable results.
The State thinks its scheme is beyond questioning, whether as to form or substance, for it sees the tie-breaker law as equivalent to the provisions examined in Blystone v. Pennsylvania, 494 U. S. 299 (1990), and Boyde v. California, 494 U. S. 370 (1990), where we approved statutes that required a death sentence upon a jury finding that aggravating circumstances outweighed mitigating ones. But the crucial fact in those systems was the predominance of the aggravators, and our recognition of the moral rationality of a mandatory capital sentence based on that finding is no authority for giving States free rein to select a different conclusion that will dictate death.
Instead, the constitutional demand for a reasoned moral response requires the state statute to satisfy two criteria that speak to the issue before us now, one governing the character of sentencing evidence, and one going to the substantive justification needed for a death sentence. As to the first, there is an obligation in each case to inform the jury’s choice of sentence with evidence about the crime as actually committed and about the specific individual who committed it. See Spaziano v. Florida, 468 U. S. 447, 460, and n. 7 (1984). Since the sentencing choice is, by definition, the attribution of particular culpability to a criminal act and defendant, as distinct from the general culpability necessarily implicated by committing a given offense, see Penry, supra, at 327-328; Spaziano, supra, at 460; Zant v. Stephens, 462 U. S. 862, 879 (1983), the sentencing decision must turn on the uniqueness of the individual defendant and on the details of the crime, to which any resulting choice of death must be “directly” related, Penry, supra, at 319.
*206Second, there is the point to which the particulars of crime and criminal are relevant: within the category of capital crimes, the death penalty must be reserved for “the worst of the worst.” See, e. g., Roper v. Simmons, 543 U. S. 551, 568 (2005) (“Capital punishment must be limited to those offenders who commit ‘a narrow category of the most serious crimes’ and whose extreme culpability makes them ‘the most deserving of execution’” (quoting Atkins v. Virginia, 536 U. S. 304, 319 (2002))). One object of the structured sentencing proceeding required in the aftermath of Furman is to eliminate the risk that a death sentence will be imposed in spite of facts calling for a lesser penalty, Penry, supra, at 328-329, and the essence of the sentencing authority’s responsibility is to determine whether the response to the crime and defendant “must be death,” Spaziano, supra, at 461; cf. Gregg, supra, at 184 (joint opinion of Stewart, Powell, and Stevens, JJ.). Of course, in the moral world of those who reject capital punishment in principle, a death sentence can never be a moral imperative. The point, however, is that within our legal and moral system, which allows a place for the death penalty, “must be death” does not mean “may be death.”
Since a valid capital sentence thus requires a choice based upon unique particulars identifying the crime and its perpetrator as heinous to the point of demanding death even within the class of potentially capital offenses, the State’s provision for a tiebreaker in favor of death fails on both counts. The dispositive fact under the tiebreaker is not the details of the crime or the unique identity of the individual defendant. The determining fact is not directly linked to a particular crime or particular criminal at all; the law operates merely on a jury’s finding of equipoise in the State’s own selected considerations for and against death. Nor does the tiebreaker identify the worst of the worst, or even purport to reflect any evidentiary showing that death must be the reasoned moral response; it does the opposite. The statute *207produces a death sentence exactly when a sentencing impasse demonstrates as a matter of law that the jury does not see the evidence as showing the worst sort of crime committed by the worst sort of criminal, in a combination heinous enough to demand death. It operates, that is, when a jury has applied the State’s chosen standards of culpability and mitigation and reached nothing more than what the Supreme Court of Kansas calls a “tie,” Kleypas, 272 Kan., at 1016, 40 P. 3d, at 232 (internal quotation marks omitted). It mandates death in what that court identifies as “doubtful cases,” ibid. The statute thus addresses the risk of a morally unjustifiable death sentence, not by minimizing it as precedent unmistakably requires, but by guaranteeing that in equipoise cases the risk will be realized, by “placing a ‘thumb [on] death’s side of the scale,’” Sochor v. Florida, 504 U. S. 527, 532 (1992) (quoting Stringer v. Black, 503 U. S. 222, 232 (1992); alteration in original).
In Kansas, when a jury applies the State’s own standards of relative culpability and cannot decide that a defendant is among the most culpable, the state law says that equivocal evidence is good enough and the defendant must die. A law that requires execution when the case for aggravation has failed to convince the sentencing jury is morally absurd, and the Court’s holding that the Constitution tolerates this moral irrationality defies decades of precedent aimed at eliminating freakish capital sentencing in the United States.
Ill
That precedent, demanding reasoned moral judgment, developed in response to facts that could not be ignored, the kaleidoscope of life and death verdicts that made no sense in fact or morality in the random sentencing before Furman was decided in 1972. See 408 U. S., at 309-310 (Stewart, J., concurring). Today, a new body of fact must be accounted for in deciding what, in practical terms, the Eighth Amendment guarantees should tolerate, for the period starting in *2081989 has seen repeated exonerations of convicts under death sentences, in numbers never imagined before the development of DNA tests. We cannot face up to these facts and still hold that the guarantee of morally justifiable sentencing is hollow enough to allow maximizing death sentences, by requiring them when juries fail to find the worst degree of culpability: when, by a State’s own standards and a State’s own characterization, the case for death is “doubtful.”
A few numbers from a growing literature will give a sense of the reality that must be addressed. When the Governor of Illinois imposed a moratorium on executions in 2000, 13 prisoners under death sentences had been released since 1977 after a number of them were shown to be innocent, as described in a report which used their examples to illustrate a theme common to all 13, of “relatively little solid evidence connecting the charged defendants to the crimes.” State of Illinois, G. Ryan, Governor, Report of the Governor’s Commission on Capital Punishment: Recommendations Only 7 (Apr. 2002) (hereinafter Report); see also id., at 5-6, 7-9. During the same period, 12 condemned convicts had been executed. Subsequently the Governor determined that four more death row inmates were innocent. See id., at 5-6; Warden, Illinois Death Penalty Reform, 95 J. Crim. L. & C. 381, 382, and n. 6 (2005).2 Illinois had thus wrongly con*209victed and condemned even more capital defendants than it had executed, but it may well not have been otherwise unique; one recent study reports that between 1989 and 2003, 74 American prisoners condemned to death were exonerated, Gross, Jacoby, Matheson, Montgomery, & Patil, Exonerations in the United States 1989 Through 2003, 95 J. Crim. L. & C. 523, 531 (2006) (hereinafter Gross), many of them cleared by DNA evidence, ibid.3 Another report states that “more *210than 110” death row prisoners have been released since 1973 upon findings that they were innocent of the crimes charged, and “[hjundreds of additional wrongful convictions in potentially capital cases have been documented over the past century.” Lanier & Acker, Capital Punishment, the Moratorium Movement, and Empirical Questions, 10 Psychology, Public Policy & Law 577, 593 (2004). Most of these wrongful convictions and sentences resulted from eyewitness misidentification, false confession, and (most frequently) perjury, Gross 544, 551-552, and the total shows that among all prosecutions homicide cases suffer an unusually high incidence of false conviction, id., at 532, 552, probably owing to the combined difficulty of investigating without help from the victim, intense pressure to get convictions in homicide cases, and the corresponding incentive for the guilty to frame the innocent, id., at 532.
We are thus in a period of new empirical argument about how “death is different,” Gregg, 428 U. S., at 188 (joint opinion of Stewart, Powell, and Stevens, JJ.): not only would these false verdicts defy correction after the fatal moment, the Illinois experience shows them to be remarkable in number, and they are probably disproportionately high in capital cases. While it is far too soon for any generalization about the soundness of capital sentencing across the country, the cautionary lesson of recent experience addresses the tie-breaking potential of the Kansas statute: the same risks of falsity that infect proof of guilt raise questions about sen*211tences, when the circumstances of the crime are aggravating factors and bear on predictions of future dangerousness.
In the face of evidence of the hazards of capital prosecution, maintaining a sentencing system mandating death when the sentencer finds the evidence pro and con to be in equipoise is obtuse by any moral or social measure. And unless application of the Eighth Amendment no longer calls for reasoned moral judgment in substance as well as form, the Kansas law is unconstitutional.
Justice Scalia takes issue with my approach, suggesting that the federal interests vindicated by our review are equally weighty whether the state court found for the defendant or for the State. Ante, at 182-185 (concurring opinion). In so doing, he overlooks the separate federal interest in ensuring that no person be convicted or sentenced in violation of the Federal Constitution—an interest entirely absent when the State is the petitioner. It is appropriate—and certainly impartial, but see ante, at 185—to take this difference in federal interests into account in considering whether to grant a petition for writ of certiorari.
Justice Scalia also fails to explain why there is such an urgent need “to ensure the integrity and uniformity of federal law.” Ante, at 183. If this perceived need is a “primary basis for the Constitution’s allowing us to be accorded jurisdiction to review state-court decisions,” ibid, (citing Art. III, § 2, els. 1 and 2), then one would think that the First Judiciary Act would have given us jurisdiction to review all decisions based on the Federal Constitution coming out of state courts. But it did not. Unconcerned about Justice Scalia’s “crazy quilt,” ante, at 185, the First Congress only provided us with jurisdiction over such cases “where [there] is drawn in question the validity of a statute of, or an authority exercised under any State, on the ground of their being repugnant to the constitution, treaties or laws of the United States, and the decision is in favour of such their validity.” Act of Sept. 24, 1789, §25, 1 Stat. 85 (emphasis added). Not until 1914 did we have jurisdiction over decisions from state courts which arguably overprotected federal constitutional rights at the expense of state laws. Act of Dec. 23, 1914, ch. 2, 38 Stat. 790; see also Delaware v. Van Arsdall, 475 U. S. 673, 694-697 (1986) (Stevens, J., dissenting). Even then, our review was only by writ of certiorari, whereas until 1988 defendants had a right to appeal to us in eases in which state courts had upheld the validity of state statutes challenged on federal constitutional grounds. See 28 U. S. C. § 1257 (1982 ed.). In other words, during the entire period between 1789 and 1988, the laws enacted by Congress placed greater weight on the vindication of federal rights than on the interest in the uniformity of federal law.
The majority views Walton v. Arizona, 497 U. S. 639 (1990), as having decided this issue. But Walton is ambiguous on this point; while the Court there approved Arizona’s practice of placing the burden on capital defendants to prove, “by a preponderance of the evidence, the existence *204of mitigating circumstances sufficiently substantial to call for leniency,” id., at 649 (plurality opinion), it did not quantify the phrase “sufficiently substantial.” Justice Blackmun clearly thought otherwise, see id., at 687 (dissenting opinion), but he cried a greater foul than one can get from the majority opinion. Stare decisis does not control this ease.
The Illinois Report emphasizes the difference between exoneration of a convict because of actual innocence, and reversal of a judgment because of legal error affecting conviction or sentence but not inconsistent with guilt in fact. See Report 9 (noting that, apart from the 13 released men, a “broader review” discloses that more than half of the State’s death penalty eases “were reversed at some point in the process”). More' importantly, it takes only a cursory reading of the Report to recognize that it describes men released who were demonstrably innocent or convicted on grossly unreliable evidence. Of one, the Report notes “two other persons were subsequently convicted in Wisconsin of” the murders. Id., at 8. Of two others, the Report states that they were released after “DNA tests revealed that none of them were the source of the semen found in the victim. That same year, two other men confessed to the crime, pleaded *209guilty and were sentenced to life in prison, and a third was tried and convicted for the crime.” Ibid. Of yet another, the Report says that “another man subsequently confessed to the crime for which [the released man] was convicted. He entered a plea of guilty and is currently serving a prison term for that crime.” Id., at 9.
A number were subject to judgments as close to innocence as any judgments courts normally render. In the case of one of the released men, the Supreme Court of Illinois found the evidence insufficient to support his conviction. See People v. Smith, 185 Ill. 2d 532, 708 N. E. 2d 365 (1999). Several others obtained acquittals, and still more simply had the charges against them dropped, after receiving orders for new trials.
At least 2 of the 13 were released at the initiative of the executive. We can reasonably assume that a State under no obligation to do so would not release into the public a person against whom it had a valid conviction and sentence unless it were certain beyond all doubt that the person in custody was not the perpetrator of the crime. The reason that the State would forgo even a judicial forum in which defendants would demonstrate grounds for vacating their convictions is a matter of common sense: evidence going to innocence was conclusive.
The authors state the criteria for their study: “As we use the term, ‘exoneration’ is an official act declaring a defendant not guilty of a crime for which he or she had previously been convicted. The exonerations we have studied occurred in four ways: (1) In forty-two cases governors (or other appropriate executive officers) issued pardons based on evidence of the defendants’ innocence. (2) In 263 cases criminal charges were dismissed by courts after new evidence of innocence emerged, such as DNA. (3) In thirty-one cases the defendants were acquitted at a retrial on the basis of evidence that they had no role in the crimes for which they were originally convicted. (4) In four cases, states posthumously acknowledged the innocence of defendants who had already died in prison . .. .” Gross 524 (footnote omitted).- The authors exclude from their list of exonerations “any ease in which a dismissal or an acquittal appears to have been *210based on a decision that while the defendant was not guilty of the charges in the original conviction, he did play a role in the crime and may be guilty of some lesser crime that is based on the same conduct. For our purposes, a defendant who is acquitted of murder on retrial, but convicted of involuntary manslaughter, has not been exonerated. We have also excluded any case in which a dismissal was entered in the absence of strong evidence of factual innocence, or in which—despite such evidence—there was unexplained physical evidence of the defendant’s guilt.” Id., at 524, n. 4.