It is a bedrock principle of our capital punishment jurisprudence that, in deciding whether to impose a sentence of death, a sentencer must consider not only the nature of the offense but also the “‘character and propensities of the of*387fender.”’ Woodson v. North Carolina, 428 U. S. 280, 304 (1976) (plurality opinion) (quoting Pennsylvania ex rel. Sullivan v. Ashe, 302 U. S. 51, 55 (1937)); see also ante, at 381-382. Without question, our commitment to individualized sentencing in capital proceedings provides some hope that we can avoid administering the death penalty “discriminatorily, wantonly and freakishly.” Gregg v. Georgia 428 U. S. 153, 220-221 (1976) (White, J., concurring in judgment) (footnotes omitted). The insistence in our law that the sentencer know and consider the defendant as a human being before deciding whether to impose the ultimate sanction operates as a shield against arbitrary execution and enforces our abiding judgment that an offender’s circumstances, apart from his crime, are relevant to his appropriate punishment.
The Court holds today that Richard Boyde’s death sentence must be affirmed even if his sentencing jury reasonably could have believed that it could not consider mitigating evidence regarding his character and background. Eschewing the fundamental principle that “the risk that the death penalty will be imposed in spite of factors which may call for a less severe penalty ... is unacceptable and incompatible with the commands of the Eighth and Fourteenth Amendments,” Lockett v. Ohio, 438 U. S. 586, 605 (1978), the Court adopts an unduly stringent standard for reviewing a challenge to a sentencing instruction alleged to be constitutionally deficient. Under the majority’s approach, a capital sentence will stand unless “there is a reasonable likelihood that the jury has applied the challenged instruction” unconstitutionally. Ante, at 380. Because the majority’s “reasonable likelihood” standard is not met where a “‘reasonable’ juror could or might have interpreted” a challenged instruction unconstitutionally, ibid, that standard is inconsistent with our longstanding focus, in reviewing challenged instructions in all criminal contexts, on whether a juror could reasonably interpret the instructions in an unconstitutional manner. See. *388e. g., Sandstrom v. Montana, 442 U. S. 510 (1979). Even more striking, the majority first adopts this standard in its review of a capital sentencing instruction. I have long shared this Court’s assessment that death is qualitatively different from all other punishments, see Spaziano v. Florida, 468 U. S. 447, 468, and n. 2 (1984) (Stevens, J., concurring in part and dissenting in part) (collecting cases), but I have never understood this principle to mean that we should review death verdicts with less solicitude than other criminal judgments. By adopting its unprecedented standard, the majority places too much of the risk of error in capital sentencing on the defendant.
Further, the majority’s conclusion that “there is not a reasonable likelihood that the jurors in petitioner’s case understood the challenged instructions to preclude consideration of relevant mitigating evidence,” ante, at 386, is belied by both the plain meaning of the instructions and the context in which they were given. Because the instructions given to Boyde’s jury were constitutionally inadequate under any standard, including the one adopted by the Court today, I dissent.
I
At the penalty phase of his trial, Richard Boyde presented extensive mitigating evidence regarding his background and character. He presented testimony regarding his impoverished background, his borderline intelligence, his inability to get counseling, and his efforts to reform. Friends and family testified that, notwithstanding his criminal conduct, Boyde possesses redeeming qualities, including an ability to work well with children.
In accordance with California’s then-operative capital jury instructions, the trial court instructed the jury that it should “consider, take into account and be guided by” 11 sentencing factors in deciding whether to return a verdict of death. 1 California Jury Instructions, Criminal 8.84.1 (4th ed. 1979) (CALJIC). Because none of these factors explicitly in*389formed the jury that it could consider evidence of a defendant’s background and character, see People v. Easley, 34 Cal. 3d 858, 878, 671 P. 2d 813, 825 (1983), Boyde argues that the trial court’s instructions were constitutionally inadequate.. The State responds that the instructions fully informed the jury of its responsibility to consider character and background evidence through factor (k), which provided that a jury could consider “[a]ny other circumstance which extenuates the gravity of the crime even though it is not a legal excuse for the crime. ” Boyde replies that a reasonable juror could have understood factor (k) as permitting consideration only of evidence related to the circumstances of the crime.
II
It is an essential corollary of our reasonable-doubt standard in criminal proceedings that a conviction, capital or otherwise, cannot stand if the jury’s verdict could have rested on unconstitutional grounds. See, e. g., Stromberg v. California, 283 U. S. 359, 367-368 (1931); Williams v. North Carolina, 317 U. S. 287, 291-292 (1942); Cramer v. United States, 325 U. S. 1, 36, n. 45 (1945); Yates v. United States, 354 U. S. 298, 312 (1957); Leary v. United States, 395 U. S. 6, 31-32 (1969); Bachellar v. Maryland, 397 U. S. 564, 571 (1970); see also Chapman v. California, 386 U. S. 18, 24 (1967) (“[Bjefore a federal constitutional error can be held harmless, the court must be able to declare a belief that it was harmless beyond a reasonable doubt”). In a society that values the presumption of innocence and demands resolution of all reasonable doubt before stripping its members of liberty or life, the decision to leave undisturbed a sentence of death that could be constitutionally infirm is intolerable.
Contrary to the majority’s intimation that the legal standard is “less than clear from our cases,” see ante, at 378, we have firmly adhered to a strict standard in our review of challenged jury instructions. In Sandstrom v. Montana, supra, the petitioner claimed that the trial court’s instructions un*390constitutionally shifted to him the burden of proof regarding his intent at the time of the crime. Rejecting the State’s claim that the jury might not have understood the instruction in an unconstitutional manner, we declared that “whether a defendant has been accorded his constitutional rights depends upon the way in which a reasonable juror could have interpreted the instruction.” Id., at 514 (emphasis added). Because we had “no way of knowing that Sandstrom was not convicted on the basis of the unconstitutional instruction,” id., at 526, we held that his conviction must be set aside. Likewise, in Francis v. Franklin, 471 U. S. 307, 319 (1985), we applied Sandstrom to invalidate a conviction where “a reasonable juror could . . . have understood” that the instructions placed the burden of proof on the defendant. We emphasized that the “[t]he question ... is not what the State Supreme Court declares the meaning of the charge to be, but rather what a reasonable juror could have understood the charge as meaning.” 471 U. S., at 315-316 (citing Sandstrom, supra, at 516-517) (emphasis added).
Sandstrom is equally applicable to claims challenging the constitutionality of capital sentencing instructions. See, e. g., California v. Brown, 479 U. S. 538, 541 (1987) (in deciding whether a “mere sympathy” instruction impermissibly excludes consideration of mitigating evidence, “‘[t]he question . . . [is] what a reasonable juror could have understood the charge as meaning’ ”) (quoting Francis, supra, at 315-316). As recently as Mills v. Maryland, 486 U. S. 367 (1988), this Court unequivocally confirmed that, in reviewing sentencing instructions alleged to preclude full consideration of mitigating circumstances, “[t]he critical question ... is whether petitioner’s interpretation of the sentencing process is one a reasonable jury could have drawn from the instructions given by the trial judge.” Id., at 375-376 (citing Francis, supra, at 315-316; Sandstrom, 442 U. S., at 516-517; and Brown, supra, at 541) (emphasis added).
*391These cases leave no doubt as to the appropriate standard of review.1 To be sure, the dissent in Francis disagreed with what it acknowledged to be “the Court’s legal standard, which finds constitutional error where a reasonable juror could have understood the charge in a particular manner.” 471 U. S., at 332 (Rehnquist, J., dissenting). But the Francis majority squarely and unqualifiedly rejected the dis*392sent’s proposal that, for constitutional error to be found, there must be something more than a “reasonable possibility of an unconstitutional understanding” of the challenged instruction. Id., at 323, n. 8. As the Francis Court stated, “it has been settled law since Stromberg v. California, 283 U. S. 359 (1931), that when there exists a reasonable possibility that the jury relied on an unconstitutional understanding of the law in reaching a guilty verdict, that verdict must be set aside.” Ibid.
The majority defends the adoption of its “reasonable likelihood” standard on the ground that it “better accommodates the concerns of finality and accuracy than does a standard which makes the inquiry dependent on how a single hypothetical ‘reasonable’ juror could or might have interpreted the instruction.” Ante, at 380. The majority fails, however, to explain how the new standard differs from Sandstrom’s “could have” standard other than to suggest that the new standard, unlike Sandstrom’s, requires more than “speculation” to overturn a capital sentence. Ibid. It is difficult to conceive how a reasonable juror could interpret an instruction unconstitutionally where there is no “reasonable likelihood” that a juror would do so. Indeed, if the majority did not explicitly allow for such a possibility, lower courts would have good reason to doubt that the two standards were different at all; the majority’s more stringent version of the “reasonable likelihood” standard is inconsistent with the cases from which the majority appropriates that standard.
The “reasonable likelihood” language first appeared in Napue v. Illinois, 360 U. S. 264 (1959), in which the Court reversed a state-court determination that a prosecutor’s failure to correct perjured testimony did not affect the verdict. The Court rejected the claim that it was “bound by [the state court’s] determination that the false testimony could not in any reasonable likelihood have affected the judgment of the jury.” Id., at 271 (emphasis added). Based on its own re*393view of the record, the Court overturned the defendant’s conviction because the false testimony “may have had an effect on the outcome of the trial.” Id., at 272 (emphasis added). The language in Napue thereafter provided the governing standard for determining whether a prosecutor’s knowing use of perjured testimony mandates reversal of a sentence. See United States v. Bagley, 473 U. S. 667, 679, n. 9 (1985) (opinion of Blackmun, J.).
As Justice Blackmun explained in Bagley, the “reasonable likelihood” standard should be understood to be an equivalent of the “harmless error” standard adopted in Chapman v. California:
“The rule that a conviction be obtained by the knowing use of perjured testimony must be set aside if there is any reasonable likelihood that the false testimony could have affected the jury’s verdict derives from Napue v. Illinois. Napue antedated Chapman v. California, 386 U. S. 18 (1967), where the ‘harmless beyond a reasonable doubt’ standard was established. The Court in Chapman noted that there was little, if any, difference between a rule formulated, as in Napue, in terms of ‘whether there is a reasonable possibility that the evidence complained of might have contributed to the conviction,’ and a rule ‘requiring the beneficiary of a constitutional error to prove beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained.’ 386 U. S., at 24. It is therefore clear . . . that this Court’s precedents indicate that the standard of review applicable to the knowing use of perjured testimony is equivalent to the Chapman harmless-error standard.” 473 U. S., at 679-680, n. 9 (citations and internal quotation marks omitted).
The history of the “reasonable likelihood” standard thus confirms that the majority’s version of the standard has no precedential support; where the Court has used “reasonable like*394lihood” language in the past, it has regarded such language as focusing, no less than the standards in Chapman and Sandstrom, on whether an error could have affected the outcome of a trial. See supra, at 389-393.2
To the extent the Court’s new standard does require a defendant to make a greater showing than Sandstrom, the malleability of the standard encourages ad hoc review of challenged instructions by lower courts. Although the standard, as the majority adopts it, requires a defendant challenging the constitutionality of an instruction to demonstrate more than a reasonable “possibility” that his jury was “impermissibly inhibited by the instruction,” a defendant “need not establish that the jury . . . more likely than not” was misled. Ante, at 380. Beyond this suggestion that error must be more than possible but less than probable, the Court is silent. Thus, appellate courts, familiar with applying the Sandstrom standard to ambiguous instructions, are now required to speculate whether an instruction that could have been misunderstood creates a “reasonable likelihood” that it was in fact misunderstood. Ante, at 380. I cannot discern how principled review of alleged constitutional errors is advanced by *395this standard.3 That this Court has regarded the two standards as identical in prior cases, see supra, at 393, will no doubt contribute to confusion in the lower courts.
More fundamentally, the majority offers no persuasive basis for altering our standard of review regarding capital instructions alleged to be constitutionally infirm. Despite the majority’s declaration to the contrary, our “strong policy in favor of accurate determination of the appropriate sentence in a capital case” is not equaled by our “strong policy against retrials” based on alleged deficiencies in jury instructions. Ante, at 380. We have long embraced a commitment to resolving doubts about the accuracy of a death verdict in favor of a capital defendant. See, e. g., Beck v. Alabama, 447 U. S. 625, 637 (1980) (“[T]he risk of an unwarranted conviction . . . cannot be tolerated in a case in which the defendant’s life is at stake”). Indeed, to characterize our commitment to accurate capital verdicts as a “policy” is inappropriately dismissive of our heightened dedication to fairness and accuracy in capital proceedings. See, e. g., Bullington v. Missouri, 451 U. S. 430, 445-446 (1981); Woodson, 428 U. S., at 304 (plurality opinion).
Moreover, the finality concerns to which the majority alludes are far less compelling in this context than the majority suggests. In addressing certain post-trial challenges to presumptively valid convictions, this Court has identified specific justifications for requiring a heightened showing by a defendant. Thus, the Court demands a showing greater than the “possibility” of error in reviewing a defendant’s request *396for a new trial based on newly discovered evidence, INS v. Abudu, 485 U. S. 94, 107, n. 12 (1988), because the “finality concerns are somewhat weaker” in the context of such claims. Strickland v. Washington, 466 U. S. 668, 694 (1984). Our adoption of this “high standard for newly discovered evidence claims presuppose[d] that all the essential elements of a presumptively accurate and fair proceeding were present in the proceeding whose result [was] challenged.” Ibid.
Likewise, in Strickland, the Court held that a defendant “must show that there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Ibid. In adopting this more demanding standard, the Court relied heavily on the special circumstances which give rise to ineffective-assistance claims. In particular, the Court emphasized the government’s inability to assure a defendant effective counsel in a given case and the difficulties reviewing courts face in discerning the precise effects of various representation-related errors:
“Conflict of interest claims aside, actual ineffectiveness claims alleging a deficiency in attorney performance are subject to a general requirement that the defendant affirmatively prove prejudice. The government is not responsible for, and hence not able to prevent, attorney errors that will result in reversal of a conviction or sentence. Attorney errors come in an infinite variety and are as likely to be utterly harmless in a particular case as they are to be prejudicial. They cannot be classified according to likelihood of causing prejudice. Nor can they be defined with sufficient precision to inform defense attorneys correctly just what conduct to avoid. Representation is an art, and an act or omission that is unprofessional in one case may be sound or even brilliant in another.” Id., at 693.
For these reasons, the Court in Strickland refused to overturn a verdict whenever a defendant shows that the errors of *397his attorney “had some conceivable effect on the outcome of the proceeding.” Ibid. Instead, the Court determined that the “reasonable probability” test more appropriately addresses the risk of error that attaches to ineffective-assistance claims in light of the fact that all trial decisions have some effect on the course of a trial.
In contrast, this case does not require the Court to relitigate facts or to speculate about the possible effects of alternative representation strategies that Boyde’s counsel might have pursued at trial. Quite simply, the issue herd is whether the trial court properly instructed the jury regarding its capital sentencing role. Such a challenge goes to the core of the accuracy of the verdict; it asks whether the defendant was sentenced by the jury according to the law. Bollenbach v. United States, 326 U. S. 607, 613 (1946) (“A conviction ought not to rest on an equivocal direction to the jury on a basic issue”). In such a circumstance, a capital defendant’s interest in an exacting review of the alleged error is unquestionably at its height, because there is no “presumptive validity” regarding the jury’s sentence. The State, on the other hand, retains no strong reliance interest in sustaining a capital verdict that may have been obtained based on a misunderstanding of the law.
Our refusal to apply a standard less protective than “reasonable doubt” to alleged errors in criminal trials in part guarantees the reliability of the jury’s determination. But it also reflects our belief that appellate courts should not “invad[e] [the] factfinding function which in a criminal case the law assigns solely to the jury.” Carella v. California, 491 U. S. 263, 268 (1989) (Scalia, J., concurring in judgment) (internal quotation marks omitted; citations omitted). Thus, where jury instructions are unclear, an appellate court may not choose the preferred construction because “[t]o do so would transfer to the jury the judge’s function in giving the law and transfer to the appellate court the jury’s function of *398measuring the evidence by appropriate legal yardsticks.” Bollenbach, supra, at 614.
This reasoning is no less applicable to California’s capital sentencing proceedings, in which the factfinding function is assigned to the jury. See Hicks v. Oklahoma, 447 U. S. 343, 346 (1980) (where defendant “has a substantial and legitimate expectation that he will be deprived of his liberty only to the extent determined by the jury in the exercise of its statutory discretion,” it violates due process to affirm his sentence “simply on the frail conjecture that a jury might have imposed a sentence equally as harsh” had they been properly instructed). To ignore a reasonable possibility that jurors were misled about the range of mitigating evidence that they could consider is to undermine confidence that the jury actually decided that Boyde should be sentenced to death in accordance with the law. It overrides California’s “fundamental decision about the exercise of official power — a reluctance to entrust plenary powers over the life and liberty of the citizen to one judge or to a group of judges.” Duncan v. Louisiana, 391 U. S. 145, 156 (1968).
Accordingly, I would review the challenged instructions in this case to determine whether a reasonable juror could have understood them to preclude consideration of mitigating evidence regarding Boyde’s character and background.
Ill
Under any standard, though, the instructions are inadequate to ensure that the jury considered all mitigating evidence. The majority’s conclusion that factor (k) would be understood by reasonable jurors to permit consideration of mitigating factors unrelated to the crime does not accord with the plain meaning of the factor’s language.4 A “circum*399stance which extenuates the gravity of the crime” unambiguously refers to circumstances related to the crime. Jurors, relying on ordinary language and experience, would not view the seriousness of a crime as dependent upon the background or character of the offender. A typical juror would not, for example, describe a particular murder as “a less serious crime” because of the redeeming qualities of the murderer; surely Boyde’s murder of Gibson could not be considered less grave, as the majority suggests, because Boyde demonstrated that his “criminal conduct was an aberration from otherwise good character,” ante, at 382-383, n. 5.5 Rather, an offender’s background and character unrelated to his crime should be considered by the sentencer because of society’s deeply felt view that punishment should reflect both the seriousness of a crime and the nature of the offender. See, e. g., Penry v. Lynaugh, 492 U. S. 302, 319 (1989) (a sentence should “ ‘reflect a reasoned moral response to the defendant’s *400background, character, and crime’” (quoting California v. Brown, 479 U. S., at 545 (O’Connor, J., concurring)).
A
The majority resists the natural understanding of the instruction by focusing on language in Penry that describes “ ‘the belief, long held by this society, that defendants who commit criminal acts that are attributable to a disadvantaged background, or to emotional and mental problems, may be less culpable than defendants who have no such excuse.’” Ante, at 382 (quoting Penry, supra, at 319) (emphasis added by majority). According to the majority, this statement reveals that jurors could understand background and character evidence as extenuating the seriousness of a crime. But this language does not prove what the majority would have it prove. The language tells us, as is clear from several of our cases, that a criminal defendant may be considered less culpable and thus less deserving, of severe punishment if he encountered unusual difficulties in his background, suffers from limited intellectual or emotional resources, or possesses redeeming qualities. See, e. g., Woodson, 428 U. S., at 304 (plurality opinion). The language in Penry does not, however, suggest that because an offender’s culpability is lessened his crime, too, is less serious. Rather than answering the central question of this case — whether character and-background evidence can be regarded as “extenuat[ing] the gravity of the crime” as opposed to lessening the offender’s moral culpability — Penry simply confirms that an offender’s background and character, apart from his crime, must be considered in fixing punishment.6
*401The majority appears to rest its position on the assumption that it would be nonsensical, given society’s “long held” belief that character and background evidence is relevant to a sentencing determination, to conclude that the jury might have thought that it could not consider such evidence. Ante, at 381-382. If the value of giving effect to such mitigating evidence is so deeply held, the assumption holds, surely the jury could not have been misled by the trial court’s instructions. The sad irony of the majority’s position is that, under its reasoning, the more fundamentally rooted a legal principle is in our constitutional values, the less scrutiny we would apply to jury instructions that run counter to that principle. For example, because “the presumption of innocence [is] that bedrock ‘axiomatic and elementary’ principle whose ‘enforcement lies at the foundation of the administration of our criminal law,”’ In re Winship, 397 U. S. 358, 363 (1970) (quoting Coffin v. United States, 156 U. S. 432, 453 (1895)), the majority apparently would resolve doubts about the adequacy of a reasonable-doubt instruction against the accused on the assumption that jurors share our “long held” belief in the presumption of innocence. The majority’s position would therefore encourage trial courts to be exacting in their instructions regarding legal minutiae but leave in barest form instructions regarding those principles “indispensable to command the respect and confidence of the community in applications of the criminal law.” 397 U. S., at 364. Because this argument inverts the degree of concern we should exhibit toward fundamental errors in criminal proceedings, it is unacceptable.
B
As the majority maintains, the adequacy of an instruction must be judged “‘in the context of the overall charge.’” *402Ante, at 378 (citations omitted).- Nothing in the charge here, however, overcame the constitutional inadequacy of factor (k) in failing to instruct the jury to consider all mitigating evidence.
The majority suggests that factor (k), by referring to “‘[a]ny other circumstance which extenuates the gravity of the crime’” (emphasis added), signaled that character and background evidence could be considered because “[ojther factors listed in CALJIC 8.84.1 allow for consideration of mitigating evidence not associated with the crime itself.” Ante, at 378, 383. The majority thus believes that the jury would be unlikely to read a limitation into factor (k) that was not shared by all of the “other” factors to which the prefatory language in factor (k) refers. But the “any other” language in factor (k) need not refer to all of the preceding factors; it could well refer solely to those factors that permit consideration of mitigating evidence related to the offense.7 The understanding of the instruction must turn on the meaning of “circumstance which extenuates the gravity of the crime,” not on factor (k)’s prefatory language. Because that phrase unambiguously refers to circumstances related to the crime, one cannot reasonably conclude on the basis of the scope of the other factors that the jury understood factor (k) to encompass mitigating evidence regarding Boyde’s character and background.
Equally unpersuasive is the majority’s claim that Boyde’s presentation of extensive background and character evidence itself suggests that the jurors were aware of their responsibility to consider and give effect to that evidence. This argument is foreclosed by Penry, where we stated that “it is not enough simply to allow the defendant to present mitigating evidence to the sentencer. The sentencer must also be *403able to consider and give effect to that evidence in imposing sentence.” 492 U. S., at 319. Thus, mere presentation of mitigating evidence, in the absence of a mechanism for giving effect to such evidence, does not satisfy constitutional requirements.
The majority attempts to avoid this conclusion by characterizing this case as unlike those in which the instructions “clearly directed the sentencer to disregard evidence.” Ante, at 384. Implicit in this claim is the view that the Constitution is satisfied when the sentencing instructions do not explicitly preclude the jury from considering all mitigating evidence. In other words, the Constitution provides no affirmative guarantee that the jury will be informed of its proper sentencing role. This view is unsupportable.
The Court in Lockett, faced with statutory restrictions on the consideration of mitigating evidence, framed the relevant question in that case to be whether the instructions “prevent[ed] the sentencer . . . from giving independent mitigating weight to aspects of the defendant’s character.” 438 U. S., at 605. We have understood this principle affirmatively to require the sentencing court to alert the jury to its constitutional role in capital sentencing. Thus, in Penry, we overturned a death sentence because the jury was not informed that it could consider mitigating evidence regarding Penry’s mental retardation and childhood abuse. It was “the absence of instructions informing the jury that it could consider and give effect to the mitigating evidence” that was dispositive. 492 U. S., at 328 (emphasis added); see also Brown, 479 U. S., at 545 (O’Connor, J., concurring) (“[T]he jury instructions — taken as a whole — must clearly inform the jury that they are to consider any relevant mitigating evidence about a defendant’s background and character” (emphasis added)); cf. Sumner v. Shuman, 483 U. S. 66, 76 (1987) (“Not only [does] the Eighth Amendment require that capital-sentencing schemes permit the defendant to present any relevant mitigating evidence, but ‘Lockett requires the *404sentencer to listen’ to that evidence”) (quoting Eddings v. Oklahoma, 455 U. S. 104, 115, n. 10 (1982)). The Court cannot fairly conclude, then, that the mere presentation of evidence satisfied Boyde’s right to a constitutionally adequate sentencing determination.
Finally, in examining the context of the sentencing instructions, the majority finds “no objectionable prosecutorial argument” that would reinforce an impermissible interpretation of factor (k). Although the prosecutor “‘never suggested that the background and character evidence could not be considered,”’ ante, at 385 (quoting 46 Cal. 3d 212, 251, 758 P. 2d 25, 47 (1988)), he did not need to. Factor (k) accomplished that purpose on its own, and the prosecutor, to make his point, needed only to repeat that language to the jury.
In his opening penalty phase argument to the jury, the prosecutor described some of the background and character evidence that Boyde had offered and asked rhetorically: “[D]oes this in any way relieve him or... in any way suggest that this crime is less serious or that the gravity of the crime is any less; I don’t think so.” App. 24. The majority suggests that this argument merely went to the weight the jury should assign to Boyde’s character and background evidence. Ante, at 385-386. But the argument directly tracks the language of factor (k) specifying what evidence may be considered, not what weight should be attached to such evidence. The argument does not suggest that Boyde’s background and character evidence was untrue or insubstantial, but rather emphasizes that the evidence did not, indeed could not in any way, lessen the seriousness or the gravity of the crime itself.
The prosecutor’s closing statement likewise reinforced the message that evidence unrelated to the crime did not fall within the scope of factor (k): “If you look and you read what it says about extenuation, it says, ‘To lessen the seriousness of a crime as by giving an excuse.’ Nothing I have heard lessens the seriousness of this crime.” App. 29. Again, the prosecutor designed his argument to bring home to the jury *405the plain meaning of the sentencing instructions. That the argument focuses more on the language of factor (k) than on the substance of Boyde’s mitigating evidence confirms that the prosecutor sought to prevent the jury from considering non-crime-related evidence.
Nor is this a case in which potentially misleading prosecutorial argument can be discounted because the trial court’s instructions satisfactorily informed the jury of its proper sentencing role. Rather, the prosecutor exploited the constitutional inadequacy of factor (k) and sought to ensure that the limited scope of factor (k) did not escape the attention of the jury. Thus, both the prosecutor’s comments and the trial court’s charge failed to communicate to the jury that it could give effect to mitigating character and background evidence. At the very least, a reasonable juror could have understood the charge and the prosecutor’s arguments as so limited. Accordingly, neither the words of the charge nor the context in which they were given provide sufficient assurance that the jury considered all mitigating evidence.
IV
“When the State brings a criminal action to deny a defendant liberty or life, . . . ‘the interests of the defendant are of such magnitude that historically and without any explicit constitutional requirement they have been protected by standards of proof designed to exclude as nearly as possible the likelihood of an erroneous judgment.’” Santosky v. Kramer, 455 U. S. 745, 755 (1982) (quoting Addington v. Texas, 441 U. S. 418, 423 (1979)). I cannot conclude with any confidence that Boyde’s jury understood that it could consider, as mitigating factors, evidence of Boyde’s difficult background and limited personal resources.8 That the majority regards *406confidence in such a conclusion as unnecessary to its affirmance of Boyde’s death sentence reflects the Court’s growing and unjustified hostility to claims of constitutional violation by capital defendants. When we tolerate the possibility of error in capital proceedings, and “leav[e] people in doubt,” In re Winship, 397 U. S., at 364, whether defendants undeserving of that fate are put to their death, we hasten our return to the discriminatory, wanton, and freakish administration of the death penalty that we found intolerable in Furman v. Georgia, 408 U. S. 238 (1972).
V
Adhering to my view that the death penalty is in all circumstances cruel and unusual punishment prohibited by the Eighth and Fourteenth Amendments, Gregg v. Georgia, 428 U. S., at 231-241 (Marshall, J., dissenting), I would in any case vacate the decision below affirming Boyde’s death sentence.
The majority attributes some of the uncertainty regarding the proper standard to this Court’s decision in Andres v. United States, 333 U. S. 740, 752 (1948), quoting the Court as follows: “ ‘That reasonable men might derive a meaning from the instructions given other than the proper meaning ... is probable.” Ante, at 379 (ellipsis and emphasis added by majority). The majority fails to quote the Court’s following sentence, in which the Court declared that “[i]n death cases doubts such as those presented here should be resolved in favor of the accused.” 333 U. S., at 752. Read in context, the passage suggests only that in a case where an instruction was probably misinterpreted, any doubt must be resolved in favor of the accused; it does not suggest, as the majority implies, that it must be probable that an instruction could be misinterpreted before a conviction will be overturned.
The majority likewise mischaraeterizes this Court’s holding in Bachellar v. Maryland, 397 U. S. 564, 571 (1970). The majority suggests that Bachellar turned on the fact that it was “ ‘equally likely that the verdict . . . rested on an unconstitutional ground,’” ante, at 380 (quoting 397 U. S., at 571) (ellipsis added by majority), and that Bachellar thus reflects only our refusal “to choose between two such likely possibilities,” ante, at 380. The majority’s misrepresentation of the Bachellar holding becomes apparent when the ellipsis inserted by the majority is removed:
“[S]o far as we can tell, it is equally likely that the verdict resulted ‘merely because [petitioners’ views about Vietnam were] themselves offensive to their hearers.’ Street v. New York, [394 U. S. 576, 592 (1969)]. Thus, since petitioners’ convictions could' have rested on an unconstitutional ground, they must be set aside.” 397 U. S., at 571 (emphasis added).
As the complete quotation makes clear, the holding in Bachellar is that a conviction cannot stand if it “could have rested on an unconstitutional ground.” The Court’s observation that, in the case before it, the verdict was “equally likely” to be unconstitutional was just that — an observation. See also id., at 569 (“[I]n light of the instructions given by the trial judge, the jury could have rested its verdict on a number of grounds”) (emphasis added).
That the majority perceives little difference between our longstanding approach to challenged jury instructions and its reformulated “reasonable likelihood” standard suggests an alarming insensitivity to the premises underlying our criminal justice system. Just as the “reasonable doubt” standard at trial reflects our awareness of the meaning and serious consequences that our society attaches to a criminal conviction, the insistence on reasonable certainty in the correctness of capital sentencing instructions is commensurate with our heightened concern for accuracy in capital proceedings. Thus, the majority’s assertion that “there may not be great differences among these various phrasings,” ante, at 379, is unfounded. To the contrary, in reviewing criminal judgments we have described the difference between a standard that demands reasonable certainty on the one hand, ahd one that tolerates significant doubt on the other, as the difference that sets apart “a society that values the good name and freedom of every individual." In re Winship, 397 U. S. 358, 363-364 (1970).
Our repudiation of such a malleable standard in Francis v. Franklin, 471 U. S. 307 (1985), where we rejected a proposed “more likely than not” standard, is no less applicable here:
“This proposed alternative standard provides no sound basis for appellate review of jury instructions. Its malleability will certainly generate inconsistent appellate results and thereby compound the confusion that has plagued this area of the law. Perhaps more importantly, the suggested approach provides no incentive for trial courts to weed out potentially infirm language from jury instructions . . . .” Id., at 322-323, n. 8.
As the majority concedes, see ante, at 374, n. 2, several years after Boyde’s trial, the California Supreme Court recognized the “potential misunderstanding” generated by the instructions challenged in his case and thereafter required lower courts to supplement the unadorned factor (k) instruction with language that would explicitly inform the jury that it could *399consider any “ ‘aspect of [the] defendant’s character or record . . . that the defendant proffers as a basis for a sentence less than death.’” People v. Easley, 34 Cal. 3d 858, 878, n. 10, 671 P. 2d 813, 826, n. 10 (1983) (quoting Lockett v. Ohio, 438 U. S. 586, 604 (1978)).
Thus, it is unsurprising that a criminal treatise, in describing the evolution of offense classification in our criminal system, reports that “serious offenses” such as murder, manslaughter, rape, and arson came to be called felonies, whereas other, presumably “less serious” offenses, came to be called misdemeanors. 1 C. Torcía, Wharton’s Criminal Law § 17, p. 81 (14th ed. 1978); see also Argersinger v. Hamlin, 407 U. S. 25, 34 (1972) (“[E]ven in prosecutions for offenses less serious than felonies, a fair trial may require the presence of a lawyer”). The characterization of felonies, which are defined by certain offense-related elements, as serious crimes regardless of the nature of the offender captures our intuitive sense that a crime is not made less serious by factors extrinsic to it, but only by circumstances surrounding the offense itself. For similar reasons, the doctrine of justification and excuse in our criminal law focuses solely on factors related to the commission of the crime, such as duress, necessity, entrapment, and ignorance or mistake. See, e. g.,1 W. LaFave & A. Scott, Substantive Criminal Law, Ch. 5 (1986).
To the extent it has spoken to the issue, this Court supports the view "that circumstances that extenuate the gravity of a crime are analytically . distinct from evidence regarding an offender’s character and background. The commitment to considering background and character evidence in our capital punishment jurisprudence is traceable, in part, through Woodson, to the following passage in Pennsylvania ex rel. Sullivan v. Ashe, 302 U. S. 51, 54-55 (1937) (emphasis added): “[Pjunishment of like crimes may *401be made more severe if committed by ex-convicts. . . . For the determination of sentences, justice generally requires consideration of more than the particular acts by which the crime was committed and that there be taken into account the circumstances of the offense together with the character and propensities of the offender.”
Indeed, at least seven of the ten factors preceding factor (k) — factors (a), (d), (e), if), (g), (h), and (j) — relate solely to circumstances surrounding the commission of the offense. See ante, at 373-374, n. 1 (quoting complete instruction).
For the reasons canvassed in Justice Brennan’s dissent in Blystone v. Pennsylvania, ante, p. 299, I also believe that the mandatory language of California’s sentencing scheme deprives a capital defendant of an independent judgment by the sentencer that death is the appropriate punishment. Like the instruction in Blystone, Boyde’s instruction required the *406sentence!’ to deliver a verdict of death if the aggravating circumstance or circumstances, no matter how insubstantial, outweighed the mitigating circumstances. Channeling sentencing discretion is indeed an essentia! aspect of a constitutional capital punishment scheme, but it should not be understood to deprive the sentencer of the choice to reject the ultimate sanction where the aggravating circumstances do not warrant it.