Opinion

JUSTICE O'CONNOR, with whom JUSTICE WHITE joins, concurring in the judgment.

Kevin Taylor admitted that he had killed Scott Siniscalchi.

He contended, however, that he had "act[ed] under a sudden and intense passion resulting from serious provocation by [Siniscalchi]." Ill. Rev. Stat., ch. 38,, 9-2 (1985). If Taylor's account is to be believed, then, under the law of the State of Illinois, he is not guilty of murder but rather of manslaughter. Ibid. At trial, Taylor took the stand and admitted to the two elements of murder. He asked only that the jury consider his state of mind when he acted and convict him of voluntary manslaughter, acquitting him of murder. Illinois law is clear that this put the jury to a choice: Taylor could be convicted only of manslaughter or murder-not of both. Indeed, because Taylor produced sufficient evidence to raise the defense of sudden passion, Illinois law required the State to negate Taylor's defense beyond a reasonable doubt. People v. Reddick, 123 Ill. 2d 184, 197, 526 N. E. 2d 141, 146 (1988). As a result, the jury should not have been permitted to convict Taylor of murder if there was so much as a reasonable possibility that Taylor's manslaughter defense had merit. Ibid.

In Falconer v. Lane, 905 F.2d 1129 (1990), the Court of Appeals for the Seventh Circuit held that instructions similar to those given at Taylor's trial did not comport with Illinois law and were ambiguous at best. In Taylor's case, according to the Court of Appeals, this ambiguity resulted in a reasonable likelihood that the jury misunderstood those instructions, and that once it found Taylor guilty of the two elements of murder (to which Taylor had admitted), the jury simply stopped deliberating without considering the possibility that Taylor was guilty only of manslaughter. 954 F. 2d 441, 442 (1992). In other words, the court concluded that there was a reasonable likelihood that the jury never considered Taylor's defense of sudden and provoked passion, even though the trial court thought there was sufficient evidence of the defense for the issue to reach the jury and even though the State bore the burden of proving its absence beyond a reasonable doubt. This, the court held, violated due process. Id., at 450.

The Court of Appeals, however, understood that our decision in Teague v. Lane, 489 U. S. 288 (1989) (plurality opinion), bars the announcement of new rules on habeas corpus. 954 F. 2d, at 451. Accordingly, it examined our precedents to determine whether its decision was "dictated" by our prior decisions. In so doing, the court construed our cases in Boyde v. California, 494 U. S. 370 (1990), and Connecticut v. Johnson, 460 U. S. 73 (1983) (plurality opinion), as compelling its conclusion that the instructions used in Taylor's case violated due process. 954 F. 2d, at 452-453. It therefore held that its rule was not "new" and ordered that a writ of habeas corpus issue unless Taylor was retried within 120 days. Id., at 453.

I agree with the majority today that the rule the Court of Appeals announced was at least susceptible to debate among reasonable jurists. See Butler v. McKellar, 494 U. S. 407, 415 (1990). For that reason, I agree that under Teague a federal court cannot issue a writ of habeas corpus based on the ambiguous instructions in dispute here. In so deciding, however, I would not reach out to decide the merits of the rule, nor would I construe our cases so narrowly as the Court does. For that reason, I write separately.

Prior to Boyde, we phrased the standard for reviewing jury instructions in a variety of ways, not all of which were consistent. Compare Mills v. Maryland, 486 U. S. 367, 384 (1988) (constitutional error occurs when there is a "substantial probability" the instructions precluded consideration of constitutionally relevant evidence), with Sandstrom v. Mon O'CONNOR, J., concurring in judgment

tana, 442 U. S. 510, 523 (1979) (constitutional error occurs when jurors "could reasonably have concluded" that the instructions created a presumption of guilt on an element of the crime). In Boyde, we clarified that when the claim is that a single jury "instruction is ambiguous and therefore subject to an erroneous interpretation," the proper inquiry is "whether there is a reasonable likelihood that the jury has applied the challenged instruction in a way that prevents the consideration of constitutionally relevant evidence." 494 U. S., at 380. As the Court notes, we chose the more restrictive standard in that case, and, as a result, Boyde itself did not state a new rule. The Court, however, finds Boyde inapplicable because it was a capital case. Ante, at 342.

It is true that we clarified the standard for reviewing jury instructions in a capital case, but Boyde did not purport to limit application of that standard to capital cases, nor have we so limited it. In Estelle v. McGuire, 502 U. S. 62 (1991), for example, the Court reviewed an ambiguous state-law instruction in a noncapital case. Although I disagreed with the Court's conclusion regarding the effect of that ambiguous instruction, see id., at 76-80 (O'CONNOR, J., concurring in part and dissenting in part), I agreed with the standard it used in reaching its conclusion: "'whether there is a reasonable likelihood that the jury has applied the challenged instruction in a way' that violates the Constitution." Id., at 72 (quoting Boyde v. California, supra). It is clear that the "reasonable likelihood" standard of Boyde applies to noncapital cases.

Although the Court's opinion today might be read as implying that erroneous jury instructions may never give rise to constitutional error outside of capital cases, ante, at 342, such an implication would misconstrue our precedent. When the Court states that "instructions that contain errors of state law may not form the basis for federal habeas relief," ibid. (citing Estelle v. McGuire, supra), it must mean that a mere error of state law, one that does not rise to the level of a constitutional violation, may not be corrected on federal habeas. Some erroneous state-law instructions, however, may violate due process and hence form the basis for relief, even in a noncapital case. In McGuire, a majority of the Court found that the particular erroneous instruction at issue did not give rise to a constitutional violation, but the very fact that the Court scrutinized the instruction belies any assertion that erroneous instructions can violate due process only in capital cases.

We have not held that the Eighth Amendment's requirement that the jury be allowed to consider and give effect to all relevant mitigating evidence in capital cases, see, e. g., Boyde, supra, applies to noncapital cases. Nevertheless, we have held that other constitutional amendments create "constitutionally relevant evidence" that the jury must be able to consider. See, e. g., Rock v. Arkansas, 483 U. S. 44, 51 (1987) ("The right to testify on one's own behalf at a criminal trial has sources in several provisions of the Constitution"); Delaware v. Van Arsdall, 475 U. S. 673, 678-679 (1986) (REHNQUIST, J.) ("[T]he Confrontation Clause guarantees an opportunity for effective cross-examination" (internal quotation marks omitted)). The category of "constitutionally relevant evidence" is not limited to capital cases.

In this case, the question is not whether application of the "reasonable likelihood" standard of Boyde is a new rule. It is not. See ante, at 341-342; supra, at 348. Nor is the question whether jury instructions may be so erroneous under state law as to rise to the level of a constitutional violation. It is clear to me that they may. See, e. g., McGuire, 502 U. S., at 72; id., at 78 (O'CONNOR, J., concurring in part and dissenting in part). The question is whether reasonable jurists could disagree over whether the particular erroneous instruction at issue here-which we assume created a reasonable likelihood that the jury did not consider Taylor's affirmative defense once it determined the two elements of murder were established-violated the Constitution. O'CONNOR, J., concurring in judgment

Our cases do not provide a clear answer to that question.

Due process, of course, requires that the State prove every element of a criminal offense beyond a reasonable doubt. In re Winship, 397 U. S. 358 (1970). This straightforward proposition has spawned a number of corollary rules, among them the rule that the State may not "us[e] evidentiary presumptions in a jury charge that have the effect of relieving the State of its burden of persuasion beyond a reasonable doubt of every essential element of a crime." Francis v. Franklin, 471 U. S. 307, 313 (1985). Accord, Rose v. Clark, 478 U. S. 570, 580 (1986); Connecticut v. Johnson, 460 U. S., at 84-85 (plurality opinion); Sandstrom, supra, at 521-523. The Court of Appeals extended these cases-which themselves are the "logical extension" of Winship, see Rose, supra, at 580-one step further. It read them as standing for the proposition that any instruction that leads "the jury to ignore exculpatory evidence in finding the defendant guilty of murder beyond a reasonable doubt" violates due process; it disregarded as meaningless the distinction between elements of the offense and affirmative defenses. 954 F. 2d, at 453.

Our opinions in Martin v. Ohio, 480 U. S. 228 (1987), and Patterson v. New York, 432 U. S. 197 (1977), however, make clear that at least in some circumstances the distinction is not meaningless. In Patterson, we held that the Due Process Clause did not require the State to prove the absence of the affirmative defense of extreme emotional disturbance beyond a reasonable doubt; the State instead could place the burden of proving the defense on the defendant. Id., at 210. We reaffirmed this holding in Martin, supra, and rejected petitioner's claim that requiring her to prove self-defense by a preponderance of the evidence shifted to petitioner the burden of disproving the elements of the crime. Id., at 233234. (Although Martin was decided after Taylor's conviction became final, its holding, like Boyde's, was not a new rule.) This case differs from Martin and Patterson in at least two ways. First, Taylor had only the burden of production and not the burden of persuasion; once he produced sufficient evidence for the issue to go to the jury, the State was required to prove the absence of his defense beyond a reasonable doubt. See Reddick, 123 Ill. 2d, at 197, 526 N. E. 2d, at 146. Second, Taylor's contention does not concern the allocation of burdens of proof; he argues that the jury did not consider his defense at all. Nevertheless, I cannot say that our prior cases compel the rule articulated by the Court of Appeals. At the very least, Martin and Patterson confirm that the rule the Court of Appeals promulgated here goes beyond what we hitherto have said the Constitution requires.

The purpose of Teague is to promote the finality of statecourt judgments. When a state court makes a "reasonable, good-faith interpretatio[n]" of our precedents as they exist at the time of decision, that decision should not be overturned on federal habeas review. Butler, 494 U. S., at 413414. Whatever the merits of the Court of Appeals' constitutional holding, an issue that is not before us, the Illinois courts were not unreasonable in concluding that the error in Taylor's instructions was not constitutional error. The State is not required to allow the defense of sudden and provoked passion at all, and the State is free to allow it while requiring the defendant to prove it. Martin, supra; Patterson, supra. It is not a begrudging or unreasonable application of these principles to hold that jury instructions that create a reasonable likelihood the jury will not consider the defense do not violate the Constitution.

Because our cases do not resolve conclusively the question whether it violates due process to give an instruction that is reasonably likely to prevent the jury from considering an affirmative defense, or a hybrid defense such as the State of Illinois permits, resolution of the issue on habeas would require us to promulgate a new rule. Like the Court, I be lieve that this rule does not fall within either of Teague's exceptions to nonretroactive application of new rules on habeas. The rule does not place any conduct, much less" 'primary, private individual conduct[,] beyond the power of the criminal law-making authority to proscribe.''' Teague, 489 U. S., at 311 (quoting Mackey v. United States, 401 U. S. 667, 675 (1971) (Harlan, J., concurring in judgments in part and dissenting in part)). Nor does the rule embody a "procedur[e] without which the likelihood of an accurate conviction is seriously diminished." 489 U. S., at 313. As noted above, the Constitution does not require the State to provide an affirmative defense to murder; a rule that, once such a defense is provided, the instructions must not prevent the jury from considering it is "a far cry from the kind of absolute prerequisite to fundamental fairness that is implicit in the concept of ordered liberty." Id., at 314 (internal quotation marks omitted).

The rule the Court of Appeals promulgated is not compelled by precedent, nor does it fall within one of the two Teague exceptions. I therefore agree with the Court that the Court of Appeals erred in applying that rule in this case. I do not join the Court's opinion, however, because it could be read (wrongly, in my view) as suggesting that the Court of Appeals' decision in this case applied not only a new rule, but also an incorrect one. I would reserve that question until we address it on direct review.