Supreme Court of the United States.
*229 Catherine Hancock argued the cause for petitioner. With her on the briefs was Elizabeth W. Cole.
Dorothy A. Pendergast argued the cause for respondent. With her on the brief were John M. Mamoulides and Terry M. Boudreaux.[*]
JUSTICE KENNEDY delivered the opinion of the Court.
We must decide in this case whether a prisoner whose murder conviction became final before our decision in Caldwell v. Mississippi, 472 U.S. 320 (1985), is entitled to use that decision to challenge his capital sentence in a federal habeas corpus action. We hold that he cannot, for Caldwell announced a new rule as defined by Teague v. Lane, 489 U.S. 288 (1989), and the new rule does not come within Teague's exception for watershed rules fundamental to the integrity of the criminal proceeding.
Over 10 years ago, petitioner Robert Sawyer murdered Frances Arwood, a visitor in the New Orleans, Louisiana, residence petitioner shared with his girlfriend, Cynthia *230 Shano. On September 29, 1979, petitioner and his accomplice Charles Lane arrived at the residence after a night of drinking. They argued with Arwood and accused her of giving drugs to Shano's children. For reasons that are not clear, petitioner and Lane struck Arwood repeatedly with their fists and dragged her by the hair into the bathroom. There they stripped the victim naked, literally kicked her into the bathtub, and subjected her to scalding, dunkings, and additional beatings. Petitioner left Lane to guard the victim, and apparently to rape her, while petitioner went to the kitchen to boil water to scald her. Petitioner kicked Arwood in the chest, causing her head to strike the tub or a windowsill and rendering her unconscious. The pair then dragged Arwood into the living room, where they continued to beat and kick her. Petitioner poured lighter fluid on the unconscious victim, particularly her torso and genital area, and set the lighter fluid afire. He told Lane that he had done this to show "just how cruel he could be." There were further brutalities we do not recount. Arwood later died of her injuries.
Petitioner was convicted and sentenced to death for the crime by a Louisiana jury in September 1980. At issue in this case are remarks made by the prosecutor in his closing argument during the sentencing phase of the trial. The prosecutor first stated, after discussing the proof of aggravating circumstances under Louisiana law:
"The law provides that if you find one of those circumstances then what you are doing as a juror, you yourself will not be sentencing Robert Sawyer to the electric chair. What you are saying to this Court, to the people of this Parish, to any appellate court, the Supreme Court of this State, the Supreme Court possibly of the United States, that you the people as a fact finding body from all the facts and evidence you have heard in relationship to this man's conduct are of the opinion that there are aggravating circumstances as defined by the statute, by *231 the State Legislature that this is a type of crime that deserves that penalty. It is merely a recommendation so try as he may, if Mr. Weidner tells you that each and every one of you I hope can live with your conscience and try and play upon your emotions, you cannot deny, it is a difficult decision. No one likes to make those type of decisions but you have to realize if but for this man's actions, but for the type of life that he has decided to live, if of his own free choosing, I wouldn't be here presenting evidence and making argument to you. You wouldn't have to make the decision." Tr. 982.
After emphasizing the brutal nature of the crime for which they had convicted petitioner, the prosecutor told the jury:
"There is really not a whole lot that can be said at this point in time that hasn't already been said and done. The decision is in your hands. You are the people that are going to take the initial step and only the initial step and all you are saying to this court, to the people of this Parish, to this man, to all the Judges that are going to review this case after this day, is that you the people do not agree and will not tolerate an individual to commit such a heinous and atrocious crime to degrade such a fellow human being without the authority and the impact, the full authority and impact of the law of Louisiana. All you are saying is that this man from his actions could be prosecuted to the fullest extent of the law. No more and no less." Id., at 984.
Finally, the prosecutor emphasized again that the jury's decision would be reviewed by later decisionmakers:
"It's all [you're] doing. Don't feel otherwise. Don't feel like you are the one, because it is very easy for defense lawyers to try and make each and every one of you feel like you are pulling the switch. That is not so. It is not so and if you are wrong in your decision believe me, believe me there will be others who will be behind *232 you to either agree with you or to say you are wrong so I ask that you do have the courage of your convictions." Id., at 985.
The Louisiana Supreme Court affirmed petitioner's conviction and sentence. State v. Sawyer, 422 So. 2d 95 (1982). This Court granted certiorari and remanded the case with instructions to the Louisiana Supreme Court to reconsider its decision in light of Zant v. Stephens, 462 U.S. 862 (1983). Sawyer v. Louisiana, 463 U.S. 1223 (1983). The Louisiana Supreme Court reaffirmed the capital sentence on remand, Sawyer v. Louisiana, 442 So. 2d 1136 (1983). His conviction and sentence became final on April 2, 1984, when we denied certiorari, Sawyer v. Louisiana, 466 U.S. 931. Petitioner sought state collateral relief, which was denied. Sawyer v. Maggio, 479 So. 2d 360 (La. 1985); Sawyer v. Maggio, 480 So. 2d 313 (La. 1985).
Petitioner then filed the federal habeas corpus petition now before us, raising a host of constitutional claims. Relevant here is petitioner's claim that the prosecutor's closing argument violated the Eighth Amendment of the United States Constitution by diminishing the jury's sense of responsibility for the capital sentencing decision, in violation of our decision in Caldwell v. Mississippi, 472 U.S. 320 (1985). Caldwell was decided over one year after petitioner's conviction became final.
The District Court denied relief, concluding that the prosecutor's remarks were of a different character from those in Caldwell, and that there was no reasonable probability that the sentence would have been different in the absence of the comments. A divided panel of the Court of Appeals for the Fifth Circuit affirmed. 848 F.2d 582 (1988). The panel held that the facts in this case were "a far cry from those in Caldwell," in large part due to the absence of any judicial approval of the prosecutor's comments. Id., at 596. Following the panel decision, the Fifth Circuit granted rehearing en banc. Id., at 606.
*233 After the en banc court heard oral argument, but while the case was pending, a plurality held in Teague v. Lane, 489 U.S. 288 (1989), that a rule of constitutional law established after a petitioner's conviction has become final may not be used to attack the conviction on federal habeas corpus unless the rule falls within one of two narrow exceptions. The Fifth Circuit requested supplemental briefing from the parties on the question whether Teague barred petitioner's claim for relief under Caldwell. The en banc court held that Caldwell announced a new rule within the meaning of Teague, a rule not within Teague's second exception for watershed rules of criminal procedure that guarantee the accuracy of a criminal proceeding. Accordingly, the Court of Appeals affirmed the denial of habeas corpus relief. 881 F.2d 1273 (1989).
We granted certiorari, 493 U.S. 1042 (1990), to resolve a conflict among the Courts of Appeals, see Hopkinson v. Shillinger, 888 F.2d 1286 (CA10 1989), and now affirm.
We must address first whether, in relying on Caldwell, petitioner claims the benefit of a new rule, as defined by our decision in Teague. In Caldwell, we held that the Eighth Amendment prohibits the imposition of a death sentence by a sentencer that has been led to the false belief that the responsibility for determining the appropriateness of the defendant's capital sentence rests elsewhere. See 472 U.S., at 328-329; id., at 342 (opinion of O'CONNOR, J.). We determined that false information of this type might produce "substantial unreliability as well as bias in favor of death sentences." Id., at 330.
At the outset we note that the parties dispute whether Caldwell, even if its rule applies, could support any claim for relief in petitioner's case. The State emphasizes that the judge in this case, unlike Caldwell, see id., at 339, did not approve the prosecutor's argument, and that the remarks *234 in this case were less likely to mislead. Petitioner, on the other hand, contends that the prosecutor's remarks were similar to those in Caldwell, and were not cured by the judge's instructions to the jury. We need not address the significant questions concerning the merits of petitioner's Caldwell claim on these facts, or the question whether application of Caldwell to the facts presented here would itself involve a new rule of law. Rather, we address only whether Caldwell is available to petitioner as a ground upon which he may seek relief. Cf. Dugger v. Adams, 489 U.S. 401, 408, n. 4 (1989) (merit of Caldwell claim immaterial to disposition of case on procedural bar grounds).
Our review of the relevant precedents that preceded Caldwell convinces us that it is a new rule for purposes of Teague. On this point we are in accord with the Court of Appeals, as well as the other two Courts of Appeals that have addressed the question. See Clark v. Dugger, 901 F.2d 908 (CA11 1990); Hopkinson v. Shillinger, supra. The rule of Teague serves to "validat[e] reasonable, good-faith interpretations of existing precedents made by state courts even though they are shown to be contrary to later decisions." Butler v. McKellar, 494 U.S. 407, 414 (1990). Thus, we have defined new rules as those that were not "dictated by precedent existing at the time the defendant's conviction became final." Teague, supra, at 301 (plurality opinion). The principle announced in Teague serves to ensure that gradual developments in the law over which reasonable jurists may disagree are not later used to upset the finality of state convictions valid when entered. This is but a recognition that the purpose of federal habeas corpus is to ensure that state convictions comply with the federal law in existence at the time the conviction became final, and not to provide a mechanism for the continuing reexamination of final judgments based upon later emerging legal doctrine.
Caldwell, of course, was not decided upon a clean slate. As the Court in Caldwell recognized, we had earlier addressed *235 the question of improper prosecutorial comment in Donnelly v. DeChristoforo, 416 U.S. 637 (1974). We stated in Donnelly that improper remarks by a prosecutor could at some point "so infec[t] the trial with unfairness as to make the resulting conviction a denial of due process." Id., at 643. No such pervasive error was established in that case, and we took the occasion to warn against "holding every improper and unfair argument of a state prosecutor to be a federal due process violation." Caldwell, supra, at 338. Caldwell, unlike Donnelly, was a capital case; and while noting the principle set forth in Donnelly, the Court in Caldwell determined to rely not on the Due Process Clause but on more particular guarantees of sentencing reliability based on the Eighth Amendment. In Donnelly we had reversed a Court of Appeals opinion vacating a conviction because prosecutorial comments were "potentially" misleading, 416 U.S., at 641, but in Caldwell we found that the need for reliable sentencing in capital cases required a new sentencing proceeding because false prosecutorial comment created an "unacceptable risk that `the death penalty [may have been] meted out arbitrarily or capriciously,'" 472 U.S., at 343 (opinion of O'CONNOR, J.).
Examination of our Eighth Amendment authorities that preceded Caldwell shows that it was not dictated by prior precedent existing at the time the defendant's conviction became final. In Caldwell itself we relied on Eddings v. Oklahoma, 455 U.S. 104 (1982); Lockett v. Ohio, 438 U.S. 586 (1978) (plurality opinion); Gardner v. Florida, 430 U.S. 349 (1977) (plurality opinion); and Woodson v. North Carolina, 428 U.S. 280 (1976) (plurality opinion), in support of the result. We cited these decisions for the general proposition that capital sentencing must have guarantees of reliability, and must be carried out by jurors who would view all of the relevant characteristics of the crime and the criminal, and take their task as a serious one. Petitioner, too, cites these and other cases in support of the argument that Caldwell was *236 "rooted" in the Eighth Amendment command of reliable sentencing, and that application of these cases to misleading prosecutorial comment "[b]y analogy" would lead to the predictable Caldwell result. Brief for Petitioner 16.
We do not doubt that our earlier Eighth Amendment cases lent general support to the conclusion reached in Caldwell. But neither this fact, nor petitioner's contention that state courts "would have found Caldwell to be a predictable development in Eighth Amendment law," Brief for Petitioner 8, suffices to show that Caldwell was not a new rule. In petitioner's view, Caldwell was dictated by the principle of reliability in capital sentencing. But the test would be meaningless if applied at this level of generality. Cf. Anderson v. Creighton, 483 U.S. 635, 639 (1987) ("[I]f the test of `clearly established law' were to be applied at this level of generality,... [p]laintiffs would be able to convert the rule of qualified immunity that our cases plainly establish into a rule of virtually unqualified liability simply by alleging violation of extremely abstract rights").
It is beyond question that no case prior to Caldwell invalidated a prosecutorial argument as impermissible under the Eighth Amendment. Eddings and Lockett invalidated statutory schemes that imposed an absolute prohibition against consideration of certain mitigating evidence by the sentencer. Woodson invalidated a capital sentencing statute providing for mandatory capital sentencing. Gardner invalidated a capital sentence based on information of which the defendant had no notice or opportunity to respond. These cases do not speak to the issue we decided in Caldwell. What we said in Saffle v. Parks, 494 U.S. 484, 491 (1990), applies here: "Even were we to agree with [petitioner's] assertion that our decisions in Lockett and Eddings inform, or even control or govern, the analysis of his claim, it does not follow that they compel the rule that [petitioner] seeks." Certainly Caldwell was not seen as compelled by the three Justices of this Court who found a "lack of authority" in our *237 Eighth Amendment precedents for the approach taken there. See 472 U.S., at 350 (REHNQUIST, J., dissenting).
From the point of view of a state court considering petitioner's claim at the time his conviction became final, Saffle, supra, at 488, there were in fact indications in our decisions that the Caldwell rule was not a requirement of the Eighth Amendment. In a previous case raising an Eighth Amendment challenge to prosecutorial comment, we had rejected the petitioner's claim. California v. Ramos, 463 U.S. 992 (1983). Indeed, the Mississippi Supreme Court had held without dissent in Caldwell that Ramos stood for the proposition that "states may decide whether it is error to mention to jurors the matter of appellate review." See Caldwell v. State, 443 So. 2d 806, 813 (1983). The Mississippi court's characterization of Ramos, of course, later proved to be incorrect. But this nonetheless suggests that prior to Caldwell our cases did not put other courts on notice that the Eighth Amendment compelled the Caldwell result.
Our opinion in Maggio v. Williams, 464 U.S. 46 (1983), provides more direct evidence that the rule of Caldwell cannot be described as dictated by existing law at the time petitioner's claim became final. In Williams we vacated a stay of execution in a case presenting a claim very similar to that in Caldwell. JUSTICE STEVENS' opinion concurring in the judgment described at length the prosecutor's argument in that case, 464 U.S., at 53-54, one similar to the argument made in Caldwell. The Court, however, found that the prisoner's challenge to the prosecutor's statements "warrant[ed] little discussion." 464 U.S., at 49. Although we stated that the failure to raise the claim of improper prosecutorial argument in an earlier habeas petition was "inexcusable," we noted that the District Court in the second petition had given the claim "full consideration" under the "standard established in Donnelly v. DeChristoforo, 416 U.S. 637 (1974)," and had found that the prosecutor's closing argument "did not render Williams' trial fundamentally unfair." Id., at 49-50. Our *238 opinion concluded by describing this and other claims raised by Williams as "insubstantial." Id., at 52. Williams, of course, did not represent a rejection on the merits of the rule announced in Caldwell. But given our statements concerning so similar a claim in Williams, we do not think a state court viewing petitioner's case at the time his conviction became final could have concluded that our Eighth Amendment precedents compelled such a rule.
We note also that, when petitioner's conviction became final, there was some reason for doubt as to this Court's view concerning what became a major premise of Caldwell, that misleading prosecutorial comment might cause a "bias in favor of death sentences." 472 U.S., at 330. At the time of petitioner's trial and appeal there was at least "some suggestion," see Dugger v. Adams, 489 U. S., at 409, that comments tending to diminish the jury's sense of sentencing responsibility would skew the result toward leniency rather than a death sentence. See Dobbert v. Florida, 432 U.S. 282, 294, and n. 7 (1977) (Florida's change to a system in which jury's verdict was advisory might benefit defendants, as the jury "may have chosen leniency when they knew [the sentencing] decision rested ultimately on the shoulders of the trial judge, but might not have followed the same course if their vote were final").
Petitioner places primary reliance on numerous state cases, decided prior to the finality of his conviction, that prohibited prosecutorial statements of the type later held to violate the Eighth Amendment in Caldwell. See, e. g., Ward v. Commonwealth, 695 S.W.2d 404, 408 (Ky. 1985); Ice v. Commonwealth, 667 S.W.2d 671, 676 (Ky.), cert. denied, 469 U.S. 860 (1984); Wiley v. State, 449 So. 2d 756, 762 (Miss. 1984), cert. denied, 479 U.S. 906 (1986); Williams v. State, 445 So. 2d 798, 811-812 (Miss. 1984), cert. denied, 469 U.S. 1117 (1985); State v. Robinson, 421 So. 2d 299, 233-234 (La. 1982); State v. Willie, 410 So. 2d 1019, 1033-1035 (La. 1982), cert. denied, 465 U.S. 1051 (1984); State v. Jones, 296 *239 N. C. 495, 501-502, 251 S.E.2d 425, 427-429 (1979); State v. Gilbert, 273 S. C. 690, 696-698, 258 S.E.2d 890, 894 (1979); State v. Tyner, 273 S. C. 646, 659-660, 258 S.E.2d 559, 566 (1979); Hawes v. State, 240 Ga. 327, 334-335, 240 S.E.2d 833, 839 (1977); Fleming v. State, 240 Ga. 142, 145-146, 240 S.E.2d 37, 40 (1977), cert. denied, 444 U.S. 885 (1979); State v. White, 286 N. C. 395, 403-404, 211 S.E.2d 445, 450 (1975); Prevatte v. State, 233 Ga. 929, 932-933, 214 S.E.2d 365, 367-368 (1975); State v. Hines, 286 N. C. 377, 381-386, 211 S.E.2d 201, 204-207 (1975). Petitioner argues that these authorities show that state courts anticipated the rule of Caldwell, and that no state reliance interest could be upset by retroactive application of the federal rule to overturn a state conviction that became final before Caldwell was decided.
The flaw in this argument is that "the availability of a claim under state law does not of itself establish that a claim was available under the United States Constitution." Dugger v. Adams, supra, at 409. All of the cases cited by petitioner, with one arguable exception, are decisions of state law, and do not purport to construe the Eighth Amendment. These cases, moreover, apply state common-law rules prohibiting any mention of appellate review; they do not condemn false prosecutorial statements under the Eighth Amendment analysis employed in Caldwell. Reliance on state-law cases for the proposition that the rule adopted in Caldwell was an old one misapprehends the function of federal habeas corpus. As we have said, the "`relevant frame of reference'" for the new rule inquiry "`is not the purpose of the new rule whose benefit the [defendant] seeks, but instead the purposes for which the writ of habeas corpus is made available.'" Teague, 489 U. S., at 306 (plurality opinion) (quoting Mackey v. United States, 401 U.S. 667, 682 (1971)). Federal habeas corpus serves to ensure that state convictions comport with the federal law that was established at the time petitioner's conviction became final.
*240 Petitioner points out, to support his argument that Caldwell applied an old rule, that our opinion there was based in part on the adoption by many state courts of rules that prohibited prosecutorial comments that could diminish the jury's sense of sentencing responsibility. Brief for Petitioner 11; see 472 U.S., at 333-334, and n. 4. It is true that our cases have looked to the decisions of state courts and legislatures to inform Eighth Amendment analysis. But petitioner's attempt to use this fact to show that Caldwell is an old rule is untenable. Under this view, state-court decisions would both inform this Court's decisions on the substantive content of the Eighth Amendment and, by simultaneous effect, impose those standards back upon the States themselves with retroactive effect. This view is also inconsistent with our citation in Penry v. Lynaugh, 492 U.S. 302, 329-330 (1989), of Ford v. Wainwright, 477 U.S. 399 (1986), which relied for its Eighth Amendment analysis on the statutory or common law of a majority of the States, see id., at 408-409, as an example of a new rule.
One Louisiana case cited by petitioner disapproving prosecutorial comment on appellate review does discuss Eighth Amendment principles rather than relying solely on state law. Even in this case, however, the court cited Eighth Amendment cases only in its discussion of prosecutorial reference to the possibility of pardon. Its discussion of prosecutorial comment on appellate review, the issue before us here, referred to state-law rules. See State v. Willie, supra, at 1033 (La. 1982), cert. denied, 465 U.S. 1051. Petitioner also cites post-Caldwell Louisiana cases, which cite Caldwell and state cases interchangeably, and state that Caldwell did not change prior law in the State. See State v. Smith, 554 So. 2d 676, 685 (La. 1989); State v. Clark, 492 So. 2d 862, 870-871 (La. 1986); State ex rel. Busby v. Butler, 538 So. 2d 164, 173 (La. 1988). To the extent these cases reflect state-court recognition that general Eighth Amendment principles pointed toward adoption of a Caldwell rule, or that Caldwell *241 is congruent with pre-existing state law, they cannot serve to show that Caldwell was dictated by our Eighth Amendment precedent. State courts as well as federal can be expected to engage in application of the principles announced in prior Eighth Amendment decisions that are "susceptible to debate among reasonable minds." Butler, 494 U. S., at 415.
Petitioner appears to contend that state courts will recognize federal constitutional protections only if they are compelled to do so by federal precedent and the threat of federal habeas review. Since some state courts had recognized a principle similar to Caldwell's, this argument goes, the result in Caldwell must have been compelled by Eighth Amendment precedent. This argument is premised on a skepticism of state courts that we decline to endorse. State courts are coequal parts of our national judicial system and give serious attention to their responsibilities for enforcing the commands of the Constitution. It is not surprising that state courts, whether applying federal constitutional protections or seeking fair administration of their own state capital punishment law, would have taken care to exclude misleading prosecutorial comment. But this conscientious exercise of their powers of supervision and review could not dictate Caldwell as a principle of federal law under the Eighth Amendment.
Under Teague, new rules may be applied in habeas corpus proceedings only if they come within "one of two narrow exceptions." Saffle, 494 U. S., at 486. The first of these applies to new rules that place an entire category of primary conduct beyond the reach of the criminal law, Teague, supra, at 311 (plurality opinion), or new rules that prohibit imposition of a certain type of punishment for a class of defendants because of their status or offense, Penry, supra, at 330. This exception has no application here. The second Teague exception applies to new "watershed rules of criminal procedure" that are necessary to the fundamental fairness of the *242 criminal proceeding. Saffle, supra, at 495; Teague, 489 U. S., at 311-313 (plurality opinion). Petitioner here challenges the Court of Appeals' conclusion that Caldwell does not come within this exception.
Petitioner contends that the second Teague exception should be read to include new rules of capital sentencing that "preserve the accuracy and fairness of capital sentencing judgments." Brief for Petitioner 30. But this test looks only to half of our definition of the second exception. Acceptance of petitioner's argument would return the second exception to the broad definition that Justice Harlan first proposed in Desist, but later abandoned in Mackey, under which new rules that "significantly improve the pre-existing fact-finding procedures are to be retroactively applied on habeas." Desist v. United States, 394 U.S. 244, 262 (1969). In Teague, we modified Justice Harlan's test to combine the accuracy element of the Desist test with the Mackey limitation of the exception to watershed rules of fundamental fairness. It is thus not enough under Teague to say that a new rule is aimed at improving the accuracy of trial. More is required. A rule that qualifies under this exception must not only improve accuracy, but also "`alter our understanding of the bedrock procedural elements'" essential to the fairness of a proceeding. Teague, supra, at 311 (plurality opinion) (quoting Mackey, 401 U. S., at 693).
The scope of the Teague exceptions must be consistent with the recognition that "[a]pplication of constitutional rules not in existence at the time a conviction became final seriously undermines the principle of finality which is essential to the operation of our criminal justice system." Teague, supra, at 309 (plurality opinion) (citing Friendly, Is Innocence Irrelevant? Collateral Attacks on Criminal Judgments, 38 U. Chi. L. Rev. 142, 150 (1970)). The "costs imposed upon the State[s] by retroactive application of new rules of constitutional law on habeas corpus thus generally far outweigh the benefits of this application." Solem v. *243 Stumes, 465 U.S. 638, 654 (1984) (opinion of Powell, J.). As we stated in Teague, because the second exception is directed only at new rules essential to the accuracy and fairness of the criminal process, it is "unlikely that many such components of basic due process have yet to emerge." 489 U.S., at 313 (plurality opinion).
It is difficult to see any limit to the definition of the second exception if cast as proposed by petitioner. All of our Eighth Amendment jurisprudence concerning capital sentencing is directed toward the enhancement of reliability and accuracy in some sense. Indeed, petitioner has not suggested any Eighth Amendment rule that would not be sufficiently "fundamental" to qualify for the proposed definition of the exception, and at oral argument in this case counsel was unable to provide a single example. Tr. of Oral Arg. 17. In practical effect, petitioner asks us to overrule our decision in Penry that Teague applies to new rules of capital sentencing. This we decline to do.
At the time of petitioner's trial and appeal, the rule of Donnelly was in place to protect any defendant who could show that a prosecutor's remarks had in fact made a proceeding fundamentally unfair. It was always open to this petitioner to challenge the prosecutor's remarks at his sentencing proceeding, by making the showing required by Donnelly. See Dugger v. Adams, 489 U. S., at 410 (defendant whose trial and appeal occurred prior to Caldwell "could have challenged the improper remarks by the trial judge at the time of his trial as a violation of due process. See Donnelly v. De-Christoforo, 416 U.S. 637 (1974)"); Maggio v. Williams, 464 U. S., at 49-50 (discussing application of Donnelly to improper remarks at sentencing). Petitioner has not contested the Court of Appeals' finding that he has no claim for relief under the Donnelly standard. And as the Court of Appeals stated: "[T]he only defendants who need to rely on Caldwell rather than Donnelly are those who must concede that the prosecutorial argument in their case was not so harmful as *244 to render their sentencing trial `fundamentally unfair.'" 881 F.2d, at 1293.
Rather than focusing on the prejudice to the defendant that must be shown to establish a Donnelly violation, our concern in Caldwell was with the "unacceptable risk" that misleading remarks could affect the reliability of the sentence. See 472 U.S., at 343 (opinion of O'CONNOR, J.). Caldwell must therefore be read as providing an additional measure of protection against error, beyond that afforded by Donnelly, in the special context of capital sentencing. See Darden v. Wainwright, 477 U.S. 168, 183-184, n. 14 (1986). The Caldwell rule was designed as an enhancement of the accuracy of capital sentencing, a protection of systemic value for state and federal courts charged with reviewing capital proceedings. But given that it was added to an existing guarantee of due process protection against fundamental unfairness, we cannot say this systemic rule enhancing reliability is an "absolute prerequisite to fundamental fairness," 489 U.S., at 314, of the type that may come within Teague's second exception.
Discussions of the nature of Caldwell error from other contexts also support our conclusion. In Dugger v. Adams, supra, we held that failure to consider a Caldwell claim would not come within a "fundamental miscarriage of justice" exception to the doctrine of procedural default. Id., at 412, n. 6; see Murray v. Carrier, 477 U.S. 478 (1986). We rejected the dissent's contention that a fundamental miscarriage of justice had been shown in that "the very essence of a Caldwell claim is that the accuracy of the sentencing determination has been unconstitutionally undermined." Dugger, supra, at 412, n. 6. Similarly, in Williams, supra, JUSTICE STEVENS concluded his discussion of a Caldwell-type claim by stating: "I question whether it can be said that this trial was fundamentally unfair. See Rose v. Lundy, [455 U.S. 509,] 543, and n. 8 [(1982)] (STEVENS, J., dissenting)." 464 U.S., at 56. These cases, of course, involved different *245 rules and contexts. Yet we think their rationale reflects a rejection of the argument that Caldwell represents a rule fundamental to the criminal proceeding.
Because petitioner seeks the benefit of a new rule that does not come within either of the Teague exceptions, his claim for habeas corpus relief is without merit. The judgment of the Court of Appeals is therefore
Affirmed.
JUSTICE MARSHALL, with whom JUSTICE BRENNAN joins, JUSTICE BLACKMUN joins as to Parts I-IV, and JUSTICE STEVENS joins as to Parts I-III, dissenting.
In his closing argument in the sentencing phase of Robert Sawyer's trial, the prosecutor emphatically argued to the jury that a sentence of death would be "merely a recommendation" and that "others" would be able to correct the decision if it turned out to be "wrong." This argument misrepresented the scope of appellate review of capital sentences under Louisiana law. La. Code Crim. Proc. Ann., Art. 905.9 (West 1984) (review by State Supreme Court is limited to question whether sentence of death is "excessive"). The prosecutor's effort to minimize the jury's sense of responsibility is precisely the type of misleading argument that we condemned in Caldwell v. Mississippi, 472 U.S. 320 (1985), and is therefore "fundamentally incompatible with the Eighth Amendment's heightened `need for reliability in the determination that death is the appropriate punishment.'" Id., at 340 (quoting Woodson v. North Carolina, 428 U.S. 280, 305 (1976) (plurality opinion)).
The Court refuses to address Sawyer's Caldwell claim on the merits. Instead, it holds that Caldwell created a "new" rule within the meaning of Teague v. Lane, 489 U.S. 288 (1989), ante, at 234, and that. Caldwell's protection against misleading prosecutorial argument is not a "`watershed rul[e] of criminal procedure'" essential to the fundamental fairness of a capital proceeding, ante, at 241 (quoting Saffle v. Parks, 494 U.S. 484, 495 (1990)). To reach this result, the majority *246 misrepresents the source and function of Caldwell's prohibitions, thereby applying its newly crafted retroactivity bar to a case in which the State has no legitimate interest in the finality of the death sentence it obtained through intentional misconduct. I dissent.
In Teague, the plurality declared that a case announces a new rule "if the result was not dictated by precedent existing at the time the defendant's conviction became final." 489 U.S., at 301. This Term, the Court held that the "`new rule' principle ... validates reasonable, good-faith interpretations of existing precedents made by state courts even though they are shown to be contrary to later decisions." Butler v. McKellar, 494 U.S. 407, 414 (1990). Accord, Parks, supra, at 488 (quoting Butler, supra, at 414). I continue to regard the Court's effort to curtail the scope of federal habeas as inconsistent with Congress' intent to provide state prisoners with an opportunity to redress "unlawful state deprivations of their liberty interests through a fresh and full review of their claims by an Article III court." Butler, supra, at 427 (BRENNAN, J., joined by MARSHALL, BLACKMUN, and STEVENS, JJ., dissenting). Even under the majority's standard, though, if the answer to a legal question is not "susceptible to debate among reasonable minds," Butler, supra, at 415, or if existing precedent would have "compelled" state courts to provide relief at the time the defendant's conviction became final, Parks, supra, at 488, then the decision does not announce a "new" legal rule within the meaning of Teague. In such circumstances, a defendant is entitled to the retroactive benefit of the decision he seeks to invoke.
The "new rule" inquiry spelled out in Teague, Butler, and Parks confirms that Caldwell did not create a new rule. The roots of the Caldwell rule can be traced directly to this Court's Eighth Amendment decisions demanding heightened *247 reliability in capital sentencing. Woodson v. North Carolina, supra (plurality opinion); Lockett v. Ohio, 438 U.S. 586 (1978) (plurality opinion); Eddings v. Oklahoma, 455 U.S. 104, 118-119 (1982) (O'CONNOR, J., concurring). In Woodson, Lockett, and Eddings, the Court considered and rejected States' efforts after Furman v. Georgia, 408 U.S. 238 (1972), to eliminate arbitrariness in the administration of the death penalty by limiting or withdrawing the sentencer's discretion. These decisions, as well as the post-Furman decisions in which the Court upheld capital sentencing schemes, see, e. g., Gregg v. Georgia, 428 U.S. 153 (1976); Proffitt v. Florida, 428 U.S. 242 (1976); Jurek v. Texas, 428 U.S. 262 (1976), emphasized that sentencers must confront their "truly awesome responsibility of decreeing death for a fellow human ... with due regard for the consequences of their decision." Lockett, supra, at 598 (plurality opinion) (internal quotation marks and citation omitted). By the time of Caldwell, "this Court's Eighth Amendment jurisprudence ha[d] taken as a given that capital sentencers would view their task as the serious one of determining whether a specific human being should die at the hands of the State." 472 U.S., at 329.
The majority nonetheless insists that the "principle of reliability in capital sentencing" is framed at such a high "level of generality" that treating it as the relevant principle for determining whether Caldwell is new law would render Teague "meaningless." Ante, at 236. This argument ignores the centrality of the Caldwell rule to reliability in capital sentencing. Caldwell error affects not just the consideration of some relevant sentencing factors, but the entire decisionmaking process itself. When a prosecutor misleadingly tells the jury that its verdict may be corrected on appeal, the prosecutor invites the jury to shirk its sentencing responsibility. The prosecutor essentially informs the jury that its verdict is less important because no execution will occur without the independent approval of higher authorities. To the extent *248 the prosecutor's comments are "focused, unambiguous, and strong," Caldwell, supra, at 340, such misconduct casts irredeemable doubt on the resulting verdict.
Some rules in capital proceedings do not contribute fundamentally to reliability; as to such rules, the majority's rejection of the reliability principle as too general may be apt. For example, the rule of Batson v. Kentucky, 476 U.S. 79 (1986), prohibiting the state from exercising peremptory challenges in a racially discriminatory manner does not have a fundamental impact on the accuracyas opposed to the integrity of the criminal process. See Allen v. Hardy, 478 U.S. 255, 259 (1986). The Caldwell rule, though, is a prerequisite to reliability in capital sentencing. Not unlike the right to counsel, the right to a jury that understands the gravity of its task is essential to the vindication of the other sentencing guarantees. Meticulous presentation of evidence and careful instruction on the law are of minimal value to a defendant whose jury has been led to believe that its verdict is of little or no consequence. The majority's observation that Caldwell's prohibition against misleading prosecutorial argument is specific thus does not undermine Sawyer's assertion that it was dictated by the Eighth Amendment's general insistence on reliability in capital sentencing.
The majority's assertion that "there were in fact indications in our decisions that the Caldwell rule was not a requirement of the Eighth Amendment," ante, at 237, is unsupported by the cases on which the majority relies. In California v. Ramos, 463 U.S. 992 (1983), the defendant had challenged California's requirement that trial courts instruct capital juries about the Governor's power to commute life sentences. In rejecting the Eighth Amendment challenge, the Court emphasized that the challenged instruction was accurate. The Court distinguished Gardner v. Florida, 430 U.S. 349 (1977), in which the Court had struck down a death *249 sentence based in part on information contained in a presentence report that had not been disclosed to defense counsel. Unlike Gardner, where there was a "risk that some of the information [relied on in sentencing] . . . may [have] be[en] erroneous," id., at 359 (plurality opinion), the sentencing decision in Ramos did not rest "in part on erroneous or inaccurate information." 463 U.S., at 1004. See also ibid. (the "need for reliability in capital sentencing" did not require reversal because the challenged instruction gave the jury "accurate information"). Cf. Caldwell, 472 U. S., at 342 (O'CONNOR, J., concurring in part and concurring in judgment) ("In my view, the prosecutor's remarks were impermissible because they were inaccurate and misleading in a manner that diminished the jury's sense of responsibility. I agree that there can be no `valid state penological interest' in imparting inaccurate or misleading information that minimizes the importance of the jury's deliberations in a capital sentencing case") (quoting id., at 336 (majority opinion)). The Ramos Court's approval of California's decision to provide capital juries with accurate information respecting commutation cannot reasonably be read as an approval of misleading or inaccurate prosecutorial argument concerning the scope of appellate review.
That the Mississippi Supreme Court in Caldwell erroneously read Ramos so broadly does not, as the majority argues, "sugges[t] that prior to Caldwell our cases did not put other courts on notice that the Eighth Amendment compelled the Caldwell result." Ante, at 237. Some courts will misconstrue our precedents notwithstanding their clarity, see, e. g., McKoy v. North Carolina, 494 U.S. 433, 439-441 (1990) (state court failed to adhere to clear direction of Mills v. Maryland, 486 U.S. 367 (1988)), and the mere fact that a single court adopts a position contrary to the one dictated by our precedents does not confirm that the case law was unclear. Indeed, if that were the standard, almost every Supreme Court decision would announce a new rule, as we *250 seldom take cases to resolve issues as to which the lower courts are in universal agreement. Moreover, under the majority's view, state-court decisions, by misconstruing the scope of this Court's Eighth Amendment decisions, would simultaneously limit the reach of those decisions as a matter of federal law. Cf. ante, at 240.
Ironically, the majority regards one errant decision by the Mississippi Supreme Court as evidence of uncertainty and yet dismisses as irrelevant to its "new rule" inquiry the States' near-unanimous rejection of Caldwell-type prosecutorial arguments prior to Caldwell, supra, at 333-334, and n. 4 (collecting cases). Even the Mississippi Supreme Court declared that "[a]ny argument by the state which distorts or minimizes the solemn obligation and responsibility of the jury is serious error." Hill v. State, 432 So. 2d 427, 439 (1983) (refusing to rule on defendant's Caldwell-type claim, however, because of the absence of a contemporaneous objection). State decisions, even if they are not premised on federal law, play a part in determining the status of constitutional protections under the Eighth Amendment. That Amendment "draw[s] its meaning from the evolving standards of decency that mark the progress of a maturing society," Trop v. Dulles, 356 U.S. 86, 101 (1958) (plurality opinion), and this Court has often looked to the laws of the States as a barometer of contemporary values, see, e. g., Penry v. Lynaugh, 492 U.S. 302, 330-331 (1989); Ford v. Wainwright, 477 U.S. 399, 408-409 (1986). Of course, the recognition of a right under state law does not translate automatically into the existence of federal constitutional protection. But a consensus among States regarding an essential ingredient to "a fair trial in the sentencing phase," State v. Berry, 391 So. 2d 406, 418 (La. 1980) (opinion on rehearing), is evidence that the right is cognizable under the Federal Constitution. The States' strong pre-Caldwell condemnation of misleading prosecutorial arguments regarding the scope of appellate review *251 is thus additional evidence that our Eighth Amendment decisions compelled the result in Caldwell.
Moreover, the majority's contention that the state courts based their decisions solely on "state common law," ante, at 239, assumes that States' capital punishment jurisprudence has evolved independently of our Eighth Amendment decisions. But state decisions regarding capital sentencing procedures even those that do not explicitly mention federal law are surely informed by federal principles and should thus be accorded some weight in discerning the scope of federal protections. Only an especially condescending federalism would protect States from retroactive application of federal law by dismissing state decisions concerning capital sentencing as irrelevant to the lineage of the federal law.[1]
This Court's approach to improper prosecutorial comments in Donnelly v. DeChristoforo, 416 U.S. 637 (1974), also supports a finding that Caldwell did not establish a new rule. In Donnelly, the prosecutor hinted that the defendant might have been willing to accept a lesser penalty for his crime, implicitly suggesting that the defendant had acknowledged his guilt. The Court held that this comment did not violate the Due Process Clause because it was ambiguous, corrected by the trial court, and too fleeting to have influenced the jury. Id., at 643-645. The Donnelly Court specifically *252 confined its decision to prosecutorial comments that did not implicate "specific guarantees of the Bill of Rights." Id., at 643. Had the claim implicated such rights, the Court acknowledged that "special care" would be required "to assure that prosecutorial conduct in no way impermissibly infringe[d] them." Ibid.
Donnelly was decided prior to the Court's explicit recognition in the cases following Gregg that the Eighth Amendment affords special protections to defendants facing the death penalty. The Court's decisions in the decade after Donnelly but before Caldwell made unmistakably clear that the death penalty's qualitatively different character from all other punishments necessitates "a corresponding difference in the need for reliability in the determination that death is the appropriate punishment in a specific case." Woodson, 428 U. S., at 305 (plurality opinion). See also Beck v. Alabama, 447 U.S. 625, 637-638 (1980) (quoting Gardner, 430 U. S., at 357-358 (plurality opinion)). Moreover, our jurisprudence by the time of Caldwell indicated unambiguously that the Eighth Amendment protects against the risk that the death penalty would be imposed in an arbitrary or capricious manner. Gregg, 428 U. S., at 188 (joint opinion of Stewart, Powell, and STEVENS, JJ.); see also Lockett, 438 U. S., at 605 (plurality opinion). In light of the Court's repeated emphasis on indispensable safeguards guaranteed in capital sentencing by a provision of the Bill of Rights, a court faced with misleading prosecutorial comments about the jury's sentencing role just prior to Caldwell could not reasonably have concluded on the basis of Donnelly that such comments would survive this Court's scrutiny.
The majority's contrary conclusion rests on a misunderstanding of the relationship between Caldwell and Donnelly. The majority endorses the Fifth Circuit's view that "`[t]he only defendants who need to rely on Caldwell rather than Donnelly are those who must concede that the prosecutorial argument in their case was not so harmful as to render their *253 sentencing trial "fundamentally unfair."'" Ante, at 243-244 (quoting 881 F.2d 1273, 1293 (1989)). But Caldwell is not, as the majority argues, "an additional measure of protection against error, beyond that afforded by Donnelly, in the special context of capital sentencing." Ante, at 244. This analysis erroneously presumes precisely what Caldwell denies, that "focused, unambiguous, and strong," prosecutorial arguments that mislead a jury about its sentencing role in the capital context can ever be deemed harmless. Caldwell rests on the view that any strong, uncorrected, and unequivocal prosecutorial argument minimizing the jury's sense of responsibility for its capital sentencing decision "presents an intolerable danger that the jury will in fact choose to minimize the importance of its role." 472 U.S., at 333. Caldwell thus tells us that a capital trial in which the jury has been misled about its sentencing role is fundamentally unfair and therefore violates Donnelly as well.
The majority's claim that Maggio v. Williams, 464 U.S. 46 (1983), provides more "direct evidence" that the rule of Caldwell was not clear at the time petitioner's conviction became final, ante, at 237, is likewise unconvincing. In Williams, the Court vacated the Fifth Circuit's entry of a stay in a capital case because Williams' contentions were "insubstantial." 464 U.S., at 52. Williams alleged, inter alia, that the prosecutor's closing argument had "elicited a decision based on passion rather than reason." Id., at 49. Some, but not all, of the prosecutor's argument referred to the scope of appellate review. See id., at 53-54 (STEVENS, J., concurring in judgment). When the motion to vacate the stay came to this Court, the sole issue was whether there was "a reasonable probability" that four Members of the Court would vote to grant certiorari. Id., at 48 (internal quotation marks omitted). In view of Williams' prior unsuccessful efforts to secure relief on similar claims, the Court applied "a strict standard of review" to Williams' application. Id., at 55 (STEVENS, J., concurring in judgment). The Court did not discuss the *254 merits of Williams' claim regarding the prosecutorial argument other than to note that the District Court had given it "full consideration," id., at 49, and had found "that it did not render Williams' trial fundamentally unfair," id., at 50. The Court's vacation of the stay in these circumstances thus reflects only the Court's view that Williams' claims, in such a posture, did not "warrant certiorari and plenary consideration." Id., at 48.[2] In sum, because the cases that dictated the result in Caldwell were decided before Sawyer's conviction became final in 1984, he is entitled to careful review of the merits of his Caldwell claim.
Even if Caldwell established a "new rule," that rule nonetheless is available on federal habeas because it is a rule "without which the likelihood of an accurate [verdict] is seriously diminished," Teague, 489 U. S., at 313 (plurality opinion). The devastating impact of prosecutorial argument that diminishes jurors' sense of responsibility is revealed in the state-court decisions condemning such argument. See, e. g., Fleming v. State, 240 Ga. 142, 146, 240 S.E.2d 37, 40 (1977) (holding that "this type of remark has an unusual potential for corrupting the death sentencing process"); State v. Berry, 391 So. 2d, at 418 ("If the reference conveys the message that the jurors' awesome responsibility is lessened by the fact that their decision is not the final one, or if the reference contains inaccurate or misleading information, then the defendant has not had a fair trial in the sentencing phase, and the penalty should be vacated"); Ward v. Commonwealth, 695 S. W. 2d *255 404, 408 (Ky. 1985) (holding that "the prosecutor clearly sought to divert from the minds of the jurors their true responsibility in this case by implying that the ultimate responsibility would fall to the trial judge, this court, [or] other appellate courts . . . . This is clearly an error of reversible magnitude"); Hill v. State, 432 So. 2d, at 439 ("Any argument by the state which distorts or minimizes this solemn obligation and responsibility of the jury is serious error. . . . [I]n a death penalty case a jury should never be given false comfort that any decision they make will, or can be, corrected"); Wiley v. State, 449 So. 2d 756, 762 (Miss. 1984) ("While a jury is not literally `the hangman,' only they [sic] may supply the hangman's victims. All notions of justice require that the jurors as individuals, and as a body, recognize and appreciate the gravity of their role").
The majority's underestimation of Caldwell's importance rests on the defect discussed above, supra, at 252-253, namely, the view that a Caldwell error will not render a trial fundamentally unfair.[3] The majority's vague suggestion that Caldwell serves as "a protection of systemic value for state and federal courts charged with reviewing capital proceedings," ante, at 244, does not disguise its inability to identify, in concrete terms, a situation in which Caldwell error occurs and yet the capital proceeding can be described as fundamentally *256 fair. See Caldwell, 472 U. S., at 341 (holding that if improper prosecutorial comment occurs the sentencing decision "does not meet the standard of reliability that the Eighth Amendment requires").
Nor does Dugger v. Adams, 489 U.S. 401 (1989), undermine Caldwell's status as a fundamental rule. The issue there was whether a particular defendant who had failed to object to misleading prosecutorial argument at sentencing had suffered sufficient prejudice to justify overlooking a state procedural bar. 489 U.S., at 406. The Court's denial of relief rested largely on the importance of the State's "interest in having the defendant challenge a faulty instruction in a timely manner so that it can correct the misstatement." Id., at 409; see also Wainwright v. Sykes, 433 U.S. 72, 87 (1977). The stringent standard for excusing procedural defaults against a particular defendant is premised on "the dual notion that, absent exceptional circumstances, a defendant is bound by the tactical decisions of competent counsel, and that defense counsel may not flout state procedures and then turn around and seek refuge in federal court from the consequences of such conduct." Reed v. Ross, 468 U.S. 1, 13 (1984) (citations omitted).
No such concern with enforcing state procedural rules against a particular defendant is at stake when we decide whether to apply new constitutional principles retroactively to all federal habeas cases. Our inquiry instead focuses on the importance of the new principle generally to the fairness and accuracy of the proceedings in which that principle went unobserved. Whereas the Dugger inquiry focuses on the general necessity of a rule to ensure an accurate verdict in all cases, the Court will overlook a clear procedural default only if the error has "probably resulted in the conviction of one who is actually innocent," 489 U.S., at 412, n. 6 (internal quotation marks omitted). The strict procedural default rule is designed in part to protect the State's interest unique in *257 the context of procedural default in correcting error in the first instance. Sykes, supra, at 88-90.
Finally, the fundamental importance of Caldwell cannot be denied on the ground that "it is `unlikely that many [new rules] of basic due process [essential to accuracy and fairness] have yet to emerge.'" Ante, at 243 (quoting Teague, 489 U. S., at 313 (plurality opinion)). The majority cannot bind the future to present constitutional understandings of what is essential for due process. See, e. g., Hurtado v. California, 110 U.S. 516, 530-531 (1884). We would rightly regard such a statement as an expression of hubris were we to discover it in a volume of the United States Reports from 100, 50, or even 20 years ago, at which time, incidentally, this Court, "[i]n light of history, experience, and the present limitations of human knowledge," rejected the argument "that committing to the untrammeled discretion of the jury the power to pronounce life or death in capital cases is offensive to anything in the Constitution." McGautha v. California, 402 U.S. 183, 207 (1971) (footnote omitted); cf. Gregg, 428 U. S., at 189 (joint opinion of Stewart, Powell, and STEVENS, JJ.) ("Furman mandates that where discretion is afforded a sentencing body on a matter so grave as the determination of whether a human life should be taken or spared, that discretion must be suitably directed and limited so as to minimize the risk of wholly arbitrary and capricious action"). Moreover, the notion that we have already discovered all those procedures central to fundamental fairness is squarely inconsistent with our Eighth Amendment methodology, under which "bedrock" Eighth Amendment principles emerge in light of new societal understandings and experience. See, e. g., Coker v. Georgia, 433 U.S. 584, 593-597 (1977) (plurality opinion).
The Court's refusal to allow Sawyer the benefit of Caldwell reveals the extent to which Teague and its progeny unjustifiably *258 limit the retroactive application of accuracy-enhancing criminal rules. Prior to Teague, our retroactivity jurisprudence always recognized a difference between rules aimed primarily at deterring police conduct and those designed to promote the accuracy of criminal proceedings. Although the former generally were not applied retroactively, see, e. g., Linkletter v. Walker, 381 U.S. 618, 636-637 (1965), the Court routinely afforded defendants the benefit of "new constitutional doctrine [whose purpose] 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." Williams v. United States, 401 U.S. 646, 653 (1971) (plurality opinion). We departed from the general rule favoring retroactive application of accuracy-enhancing rules only in special cases in which retroactivity would have undermined substantial reliance interests of law enforcement officials and prosecutors who acted in good faith prior to the change in the law. See, e. g., Stovall v. Denno, 388 U.S. 293 (1967).
The state prosecutor in this case surely could not claim a good-faith belief in the legitimacy of the conduct proscribed in Caldwell misleading and inaccurate argument designed to minimize the jury's sentencing responsibility. Indeed, respondent seems to concede as much, framing the State's reliance interest, beyond its general interest in the finality of its convictions, as the right to have "misleading prosecutorial remarks. . . reviewed under the fundamental fairness standard of due process" rather than the Eighth Amendment. Brief for Respondent 16. This purported reliance interest depends on the erroneous view that Caldwell error could survive review under the Due Process Clause. See, supra, at 257. But even granting a distinction in the degree of scrutiny applied by Donnelly and Caldwell, the State's claimed interest in having its intentional misconduct reviewed under a less demanding standard is hardly worth crediting.
*259 The State is thus left to rely solely on its general interest in the finality in its criminal proceedings. Before today, such an interest was never alone sufficient to preclude vindication of constitutional rights on federal habeas. See Reed v. Ross, 468 U. S., at 15. Teague itself, of course, stated that it was departing from our traditional approach. But that case, as well as Butler and Parks, involved rules that the Court did not recognize as contributing meaningfully to the accuracy of criminal proceedings. See Teague, 489 U. S., at 315 (plurality opinion) (failure to apply rule does not "seriously diminish the likelihood of obtaining an accurate conviction"); Butler, 494 U. S., at 416 (failure to apply rule "would not seriously diminish the likelihood of obtaining an accurate determinationindeed, it may increase that likelihood"); Parks, 494 U. S., at 495 ("The objectives of fairness and accuracy are more likely to be threatened than promoted" by the rule); cf. ante, at 244 (acknowledging that Caldwell's central purpose is to enhance "the accuracy of capital sentencing"). Those cases thus could have been decided in the same way under our prior retroactivity doctrine, which weighed the State's finality and reliance interests against the defendant's interests protected by the new rule.
No such balancing of the competing concerns occurs today. The Court instead simply elevates its preference for finality in state proceedings over Congress' commitment "to provide a federal forum for state prisoners . . . by extending the habeas corpus powers of the federal courts to their constitutional maximum," Fay v. Noia, 372 U.S. 391, 426 (1963). This raw preference for finality is unjustified. Although a State undoubtedly possesses a legitimate interest in the finality of its convictions, when the State itself undermines the accuracy of a capital proceeding, that general interest must give way to the demands of justice.
The jury that sentenced Sawyer to death was deliberately misled about the significance of its verdict. That Sawyer *260 was thus denied a fundamentally fair trial was as apparent when Sawyer's conviction became final as it is today. The Court's refusal to allow a federal habeas court to correct this error is yet another indication that the Court is less concerned with safeguarding constitutional rights than with speeding defendants, deserving or not, to the executioner. I dissent.
Even if I did not believe that Sawyer was entitled to federal habeas review of his Caldwell claim, I would nonetheless vacate his death sentence. I adhere to my view that the death penalty is in all circumstances cruel and unusual punishment. Gregg v. Georgia, 428 U. S., at 231 (MARSHALL, J., dissenting).
[*] Julius L. Chambers filed a brief for the NAACP Legal Defense and Educational Fund, Inc., as amicus curiae urging reversal.
Kent S. Scheidegger filed a brief for the Criminal Justice Legal Foundation as amicus curiae urging affirmance.
Briefs of amici curiae were filed for the American Bar Association by Stanley Chauvin, Jr., Jay Topkis, Ronald J. Tabak, and Eric M. Freedman; and for Stephen H. Sachs et al. by Randy Hertz and Michael Millemann.
[1] That Penry v. Lynaugh, 492 U.S. 302, 329 (1989), and Teague v. Lane, 489 U.S. 288, 301 (1989), cite Ford v. Wainwright, 477 U.S. 399 (1986), as crafting a "new" rule does not establish that state decisions are irrelevant in assessing the status of a right under the Federal Constitution. Cf. ante, at 240. Neither of these opinions discussed the citation to Ford, and the force of their conclusions is undermined by this Court's subsequent reliance on state decisions in Saffle v. Parks, 494 U.S. 484 (1990), to determine whether the rule invoked in that case was compelled by our Eighth Amendment decisions, see id., at 490-491 (citing state decisions). State decisions cannot be deemed relevant to the Teague inquiry only to the extent that they disprove the rootedness of a constitutional right.
[2] The majority nonetheless views Williams as casting some doubt on the ultimate disposition of Caldwell v. Mississippi, 472 U.S. 320 (1985) because the prosecutor's argument in Williams was "very similar to [the argument] in Caldwell." Ante, at 237. That position, though, is overbroad. The District Court's finding that Williams' trial was not fundamentally unfair under Donnelly v. DeChristoforo, 416 U.S. 637 (1974), was tantamount to a finding that Williams' jury was not misled about its sentencing role. See Williams v. King, 573 F. Supp. 525, 530-531 (MD La. 1983).
[3] The majority's rejection of the States' view that Caldwell's prohibitions are vital to the fairness of a capital proceeding reveals a tension in the Court's retroactivity doctrine. At the same time that the majority insists that Caldwell was not dictated by our Eighth Amendment decisions, the majority also argues that Caldwell is not a fundamental rule because it affected only an incremental change in capital sentencing. See ante, at 244 (stating that Caldwell provides merely an "additional measure of protection against error, beyond that afforded by Donnelly"). A rule may be "new" even if it is desigued to serve interests substantially similar to an "old" rule. The majority's extensive effort in its "new rule" analysis to demonstrate that Caldwell's "additional" protections marked a departure in our Eighth Amendment jurisprudence, however, seems disingenuous in light of its conclusion that the departure did not amount to much.
