with whom Justice Brennan joins, and Justice Stevens joins as to Part I, dissenting.
Adhering to my view that the death penalty is in all circumstances cruel and unusual punishment prohibited by the *247Eighth and Fourteenth Amendments, Gregg v. Georgia, 428 U. S. 153, 231-241 (1976) (Marshall, J., dissenting), I would vacate the decision below insofar as it left undisturbed the death sentence imposed in this case.
Even if I did not hold this view, I would vacate petitioner’s sentence of death for two independent reasons. First, the jury that sentenced Leslie Lowenfield was subjected during the penalty phase of the trial to a combination of practices that courts have viewed as coercive in far less sensitive situations. The use of these practices in this case presents an unacceptable risk that the jury returned a sentence of death for reasons having nothing to do with proper constitutional considerations. Second, even in the absence of coercion, the jury’s sentence of death could not stand because it was based on a single statutory aggravating circumstance that duplicated an element of petitioner’s underlying offense. This duplication prevented Louisiana’s sentencing scheme from adequately guiding the discretion of the sentencing jury in this case and relieved the jury of the requisite sense of responsibility for its sentencing decision. As we have recognized frequently in the past, such failings may have the effect of impermissibly biasing the sentencing process in favor of death in violation of the Eighth and the Fourteenth Amendments.
I
After many hours of deliberations, petitioner’s sentencing jury informed the court that it was “having great distress” and unable to reach a verdict. App. 17. Had the jury remained deadlocked, petitioner would have received a sentence of life imprisonment by the operation of Louisiana law. But the presiding judge intervened to aid the jury in reaching a verdict, and petitioner now charges that the judge’s intervention was coercive.
Two principles should guide our evaluation of petitioner’s claim. First, recognizing that “impartiality” is a state difficult to define and “coercion” an event difficult to discern in concrete situations, we must be careful to focus on the par*248ticular facts of this case in order to assess “all the circumstances” surrounding the jury’s progress from deadlock to unanimity. Jenkins v. United States, 380 U. S. 445, 446 (1965) (per curiam). Second, we often have acknowledged that the unique nature of the death penalty demands a greater degree of reliability in capital sentencings than in other criminal proceedings. See, e. g., Lockett v. Ohio, 438 U. S. 586, 604 (1978) (opinion of Burger, C. J., joined by Stewart, Powell, and Stevens, JJ.); Woodson v. North Carolina, 428 U. S. 280, 305 (1976) (opinion of Stewart, Powell, and Stevens, JJ.).
The Court in this case pays mere lipservice to these concerns, citing the relevant portions of Jenkins and Lockett but then proceeding to ignore their teachings. The Court offers a sanitized rendition of the facts, ignoring or glossing over evidence of coercion in its examination of “all the circumstances” of the sentencing proceeding. The Court then performs a mechanical and cramped application of our precedents regarding jury coercion, essentially restricting these cases to their facts. Moreover, the Court focuses on the impact of each challenged practice in isolation, never addressing their cumulative effect. Finally, the Court neglects to consider how the capital sentencing context of this case affects the application of principles forged in other contexts. In sum, the Court’s approach fails to take seriously petitioner’s challenge and consequently fails to recognize its force. The Court’s decision to condone the coercive practices at issue here renders hollow our pronouncement that “the decision whether a man deserves to live or die must be made on scales that are not deliberately tipped toward death.” Witherspoon v. Illinois, 391 U. S. 510, 521-522, n. 20 (1968).
The starting point for any determination of jury coercion is the facts of a given case. The opinion of the Court, however, does not keep its promise to examine “all the circumstances,” failing to mention several significant events in evaluating whether the trial court’s conduct improperly influenced the *249jury’s decision. First, the Court treats as irrelevant all events preceding the polling of the jury and the administration of the Allen charge. Allen v. United States, 164 U. S. 492 (1896). The Court fails to recognize that the guilt phase of petitioner’s trial, which immediately preceded the sentencing phase, shaped the jury’s collective state of mind and relationship to the court. The jury’s deliberations on the issue of guilt or innocence lasted 13 hours over 2 days. After 11 hours of deliberation, the jury informed the court that it was experiencing “much distress” and requested to view some physical evidence. App. 16. The court refused the request and stated: “7 order you to go back to the jury room and to deliberate and arrive at a verdict.” Id., at 24 (emphasis added). Defense counsel objected to this instruction and raised that objection again during the penalty phase on the ground that it might have a “residual effect” on the sentencing process. Id., at 51. After the jury returned its guilt phase verdict at 3 p.m., the court gave the jury only an hour’s break before conducting the sentencing hearing and sending the jury to deliberate at 8:17 p.m. These events might well have suggested to the jury that the court was anxious for a verdict to be reached and reached quickly. That impression might have received further support when, at 11:55 p.m. and after a total of more than 13 hours that day in court, the jury requested permission to retire for the night, and the court inquired: “Is there any way you could continue deliberating tonight and arrive at a verdict?” Id., at 48. This background, which the Court ignores, is important to an understanding and evaluation of the jury’s reaction to the polling procedures and the modified Allen charge of the following day.
Second, although the Court notes that the jury was instructed at the commencement of the sentencing phase that its failure to reach a verdict would result in a life sentence rather than a second sentencing hearing, the Court fails to observe that this instruction was repeated three more times *250during the proceeding. First, the jury sent a note specifically asking whether a life sentence would result automatically from a failure to reach a verdict, to which the court .responded affirmatively. Second, the court repeated this instruction before polling the jury as to whether further deliberations would be helpful. Finally, the court repeated it again immediately after twice polling the jury and immediately before giving the modified Allen charge. The court’s reiteration of the consequences of failing to reach a verdict before the jury poll and again before the verdict-urging charge may well have been understood by the jury as an expression by the court of reluctance to impose only a life sentence and an admonition to reach a verdict.
Third, the Court does not mention the refusal of defense counsel’s specific request that the jury be instructed that it was not required by law to return a verdict. In Hyde v. United States, 225 U. S. 347 (1912), we rejected a claim of jury coercion by relying on the court’s statement to the jurors that if “ ‘they could not conscientiously and freely agree upon a verdict they would be discharged.’ ” Id., at 382. We concluded that “[i]t is hard to believe that . . . with that promise expressly made to them, they were coerced by a threat of confinement to acquit those who they were convinced were guilty or convict those who they were convinced were innocent.” Id., at 383. In the decided absence of such an instruction, the possibility of coercion runs much stronger.
None of the above circumstances is by itself a reason to conclude that the jury was coerced in this case. But these circumstances, as well as those recounted by the Court, make up the “totality” in light of which we must judge the practices challenged in this case. Both the polling procedures and the Allen charge administered by the court must be examined against this background, first individually and then as a cumulative whole.
The Court makes quick work of petitioner’s challenge to the court’s polling of the jury in this case. Observing that *251the court’s inquiries regarding the helpfulness of further deliberations were “clearly independent” of an inquiry regarding the jury’s stance on the merits, the Court concludes that the pollings of the jury did not “ ‘reveal the nature or extent of its division.’” Ante, at 240 (quoting Brasfield v. United States, 272 U. S. 448, 450 (1926)). Such a conclusion might be accurate in an ordinary trial setting, where a hung jury leads only to retrial. But the Court ignores the fact that the jury in this case had been instructed repeatedly that failure to reach a verdict would automatically result in a life sentence. With this background knowledge, a juror’s vote against further deliberations indicated acceptance of the life sentence that would necessarily follow. The jury’s response to the polling questions in this case thus very probably revealed the nature and extent of its substantive division.
Not only does the Court refuse to acknowledge that the polling in this case is similar in nature to the polling we condemned in Brasfield, it also ignores the ways in which this case is worse than Brasfield. The court here polled the jury not once, but twice, increasing whatever coercive effect a single poll would have had. Moreover, the second poll whittled down the minority jurors from 4 to 1, creating enormous pressure on the lone holdout, as compared to the minority of three jurors in Brasfield. Finally, the jurors in this case were asked to identify themselves by name in both polls, exposing the identities of the minority jurors to the court. Under such circumstances, the polling procedures used here posed an even greater risk of “improper influence upon the jury” than the poll we examined and rejected in Brasfield. Id., at 450.
The Court’s treatment of petitioner’s challenge to the Allen charge is similarly dismissive. The Court begins by suggesting that the validity of such charges is “beyond dispute,” citing cases from all of the Circuits in which some form of an Allen charge has been upheld. Ante, at 237-238, and n. 1. This sweeping statement denigrates the serious res*252ervations expressed by many federal1 and state2 courts concerning the coercive nature of the traditional Allen charge. These reservations, voiced in the context of ordinary criminal trials, have particular significance for the instant case. The usual justifications for a verdict-urging charge are the time, expense, and possible loss of evidence that a new trial would entail. None of these justifications was present here, where a hung jury would have resulted in a life sentence. Moreover, in an ordinary criminal trial, an Allen charge will not steer the jury one way or the other on the merits, because it is as likely that the minority jurors are for conviction as for acquittal. Here, the charge inevitably made a verdict of death more likely, because a continued deadlock would have achieved a substantive outcome of a life sentence rather than simply another sentencing hearing. These considerations indicate that the State’s interest in a verdict in this case was relatively weak, whereas the defendant’s interest in preserving the integrity of a dissenting vote was correspondingly strong. The general reservations voiced by other courts *253about the coerciveness of verdict-urging charges should be given special attention under these circumstances.
The opinion of the Court, however, persistently refuses to recognize the unique posture of this case. Instead, it blandly notes that this case is factually distinguishable from our other significant jury coercion cases, Jenkins v. United States, 380 U. S. 445 (1965), and United States v. United States Gypsum Co., 438 U. S. 422 (1978). This analysis, if such it may be called, fails to recognize the animating principle of both Jenkins and Gypsum: If the jury might believe from the trial court’s statements or actions that the court is insisting upon a verdict “‘one way or the other,’” 438 U. S., at 460, that message poses an impermissible risk of jury coercion. Just such a risk was posed here, when the court gave a verdict-urging charge to a jury that knew, and indeed had just been instructed, that its failure to reach unanimity would result in the substantive outcome of a life sentence.
The Court’s most significant analytical failure, however, lies in its refusal to consider petitioner’s charge of coercion in other than a piecemeal fashion. Content that the polling procedures did not contravene Brasfield and that the verdict-urging charge satisfied Allen, the Court never considers the two practices in tandem. Other federal courts have recognized that an Allen charge given on the heels of a jury poll poses special risks of coercion. See United States v. Sae-Chua, 725 F. 2d 530, 532 (CA9 1984); Cornell v. Iowa, 628 F. 2d 1044, 1048, n. 2 (CA8 1980), cert. denied, 449 U. S. 1126 (1981); Williams v. United States, 119 U. S. App. D. C. 190, 193, 338 F. 2d 530, 533 (1964). These courts have noted that when a jury poll is followed by an Allen charge, “the impression is inherently conveyed to the jury that the revelation of their division prompted the giving of the subsequent verdict-urging instruction and that it is, therefore, directed toward the minority jurors.” Cornell v. Iowa, supra, at 1048, n. 2. In this case, the charge was given after the polling had pared down the minority to a single juror, identified *254to the court by name. That juror could not help feeling that the verdict-urging charge was directed at him and him alone. The polling and the charge in this case together created an atmosphere far more charged with coercion than either practice alone possibly could have engendered. Such coercion is strongly evidenced by the fact that the jury returned a verdict of death a mere 30 minutes after the court gave the verdict-urging charge.3
It is an open and a far closer question whether the practices challenged in this case should be deemed coercive in an ordinary criminal context. We have recognized often and reiterated last Term that practices entirely appropriate in other contexts may be improper in capital sentencing proceedings. See Booth v. Maryland, 482 U. S. 496, 509, n. 12 (1987). The Court in this case, however, fails to recognize this principle and makes no attempt to assess how the capital sentencing context affects the legitimacy of the challenged practices. This failure is troubling not only because we require greater reliability in capital sentencings, but also because the nature of the capital sentencing process makes the practices challenged here more dangerous. The capital sentencing jury is asked to make a moral decision about whether a particular individual should live or die. Despite the objective factors that are introduced in an attempt to guide the exercise of the jurors’ discretion, theirs is largely a sub*255jective judgment. Given the amorphous and volatile nature of their inquiry, capital sentencing juries that have reached an impasse in their deliberations may be particularly prone to coercion from the court. This concern leads me to conclude that the jury polling and Allen charge used in this case created an unacceptable risk of jury coercion and thus were “inconsistent with the reasoned decisionmaking we require in capital cases.” Booth v. Maryland, supra, at 509 (footnote omitted).
II
Even had the jury reached its verdict free from any improper influence by the court, that verdict still could not stand. The principles established by our prior cases preclude the imposition of the death penalty when it is based on a single statutory aggravating circumstance that merely duplicates an element of the underlying offense. We have insisted repeatedly that the discretion of the sentencer be guided by a narrowing of the class of people eligible for the death penalty and that the sentencer be fully cognizant of its responsibility for the imposition of a sentence of life or death. Both of these principles have been violated by the operation of the Louisiana sentencing scheme in this case.
Since our decision in Furman v. Georgia, 408 U. S. 238 (1972), we have required that there be a “meaningful basis for distinguishing the few cases in which [the death sentence] is imposed from the many cases in which it is not.” Id., at 313 (White, J., concurring). We have held consistently that statutory aggravating circumstances considered during the sentencing process provide one of the means by which the jury’s discretion is guided in making such constitutionally mandated distinctions. See, e. g., McCleskey v. Kemp, 481 U. S. 279, 305 (1987) (describing “the role of the aggravating circumstance in guiding the sentencing jury’s discretion”); Zant v. Stephens, 462 U. S. 862, 877 (1983) (holding that “an aggravating circumstance must genuinely narrow the class of persons eligible for the death penalty”); Gregg *256v. Georgia, 428 U. S., at 197 (opinion of Stewart, Powell, and Stevens, JJ.) (explaining that the finding of statutory-aggravating circumstances helps the jury “to consider the circumstances of the crime and the criminal before it recommends sentence”).
The Court today suggests that our emphasis on aggravating circumstances has been mere happenstance and holds that the critical narrowing function may be performed prior to and distinct from the sentencing process.4 This holding misunderstands the significance of the narrowing requirement. The Court treats the narrowing function as a merely technical requirement that the number of those eligible for *257the death penalty be made smaller than the number of those convicted of murder. But narrowing the class of death eligible offenders is not “an end in itself” any more than aggravating circumstances are. See ante, at 244. Rather, as our cases have emphasized consistently, the narrowing requirement is meant to channel the discretion of the sentencer. It forces the capital sentencing jury to approach its task in a structured, step-by-step way, first determining whether a defendant is eligible for the death penalty and then determining whether all of the circumstances justify its imposition. The only conceivable reason for making narrowing a constitutional requirement is its function in structuring sentencing deliberations. By permitting the removal of the narrowing function from the sentencing process altogether, the Court reduces it to a mechanical formality entirely unrelated to the choice between life and death.
The Court’s relegation of the narrowing function to the guilt phase of a capital trial implicates the concerns we expressed in another context in Caldwell v. Mississippi, 472 U. S. 320 (1985). In Caldwell, we vacated petitioner’s sentence of death when the prosecutor had argued to the jury that the appellate court would review the imposition of the death sentence for correctness, concluding that “it is constitutionally impermissible to rest a death sentence on a determination made by a sentencer who has been led to believe that the responsiblity for determining the appropriateness of the defendant’s death rests elsewhere.” Id., at 328-329. Here, the sentencing jurors were led to believe that petitioner’s eligibility for the death sentence was already established by their findings during the guilt phase — findings arrived at without any contemplation of their implication for petitioner’s sentence. Indeed, the court specifically instructed the jury at the start of its guilt phase deliberations: “You are not to discuss, in any way, the possibility of any penalties whatsoever.” Record 2283. Then, during the penalty hearing, the prosecutor twice reminded the jury that *258it had already found during the guilt phase one of the aggravating circumstances that the State urged was applicable to petitioner’s sentence. Id., at 2311, 2319. The prosecutor’s argument might well have convinced the jury that it had no choice about and hence no responsibility for the defendant’s eligibilty for the death penalty. This situation cannot be squared with our promise to ensure that “a capital sentencing jury recognizes the gravity of its task and proceeds with the appropriate awareness of its Truly awesome responsibility.’” Caldwell v. Mississippi, supra, at 341.
In sum, the application of the Louisiana sentencing scheme in cases like this one, where there is a complete overlap between aggravating circumstances found at the sentencing phase and elements of the offense previously found at the guilt phase, violates constitutional principles in ways that will inevitably tilt the sentencing scales toward the imposition of the death penalty. The State will have an easier time convincing a jury beyond a reasonable doubt to find a necessary element of a capital offense at the guilt phase of a trial if the jury is unaware that such a finding will make the defendant eligible for the death penalty at the sentencing phase. Then the State will have an even easier time arguing for the imposition of the death penalty, because it can remind the jury at the sentencing phase, as it did in this case, that the necessary aggravating circumstances already have been established beyond a reasonable doubt. The State thus enters the sentencing hearing with the jury already across the threshold of death eligibility, without any awareness on the jury’s part that it had crossed that line. By permitting such proceedings in a capital case, the Court ignores our early pronouncement that “a State may not entrust the determination of whether a man should live or die to a tribunal organized to return a verdict of death.” Witherspoon v. Illinois, 391 U. S., at 521 (footnote omitted).
*259III
After a total of 22 hours of almost continuous deliberations regarding petitioner’s guilt and appropriate sentence, the jury in this case informed the court that it was in “great distress.” Had the jury not broken its deadlock, petitioner would have been entitled to a life sentence without retrial. Thus, at 3 p.m. on May 16, 1984, Leslie Lowenfield’s life hung delicately in the balance. It is impossible to know what finally prompted the jury to return its sentence of death, but the coercive practices engaged in by the trial court, or the prosecutor’s argument that a key aggravating circumstance already had been established at the guilt phase, may well have tipped the balance. Neither of these factors has any place in capital sentencing deliberations, and their presence in this case convinces me that petitioner was denied the individualized and reasoned consideration of his penalty that the Constitution promises him. I dissent.
See, e. g., United States v. Rey, 811 F. 2d 1458, 1458, 1460 (CA11) (“The modern judicial trend ... is against the Allen charge. ... As we see it, the Allen charge interferes with the jurors when they are performing their most important role: determining guilt or innocence in a close case. It unjustifiably increases the risk that an innocent person will be convicted as a result of the juror abandoning his honestly-held beliefs”), cert. denied, 484 U. S. 830 (1987); United States v. Blandin, 784 F. 2d 1048, 1050 (CA10 1986) (“We have approved the Allen instruction as permissible in the Tenth Circuit, but urge caution in its use. ... It should not be given during the course of deliberations”); United States v. Seawell, 550 F. 2d 1159, 1162, 1163 (CA9 1977) (“Problems arising from the inherently coercive effect of the Allen charge have caused other courts of appeals and state courts to prohibit or to restrict severely its use.... A single Allen charge, without more, stands at the brink of impermissible coercion”) (footnotes omitted).
See, e. g., People v. Gainer, 19 Cal. 3d 835, 566 P. 2d 997 (1977) (banning use of traditional Allen charge in all criminal cases); State v. Czachor, 82 N. J. 392, 413 A. 2d 593 (1980) (same); Kersey v. State, 525 S. W. 2d 139 (Tenn. 1975) (same).
The Court argues that the failure of petitioner’s counsel to object to the polling or the Allen charge suggests that their coercive potential was not “apparent.” Ante, at 240. The Court fails to acknowledge, however, that at the time of the polling and charge, defense counsel had already moved three separate times for a mistrial during the sentencing phase: once when the jury had been out for more than five hours, again when the jury sent a note indicating its deadlock, and again after the first polling revealed an 8-to-4 division as to whether further deliberations would be helpful. Defense counsel may well have reasoned that renewing his motions during the second polling or the Allen charge would be unavailing. In any case, counsel’s repeated mistrial motions clearly demonstrate his awareness of the jury’s confusion and distress.
The Court argues that our opinion in Jurek v. Texas, 428 U. S. 262 (1976), establishes that the narrowing requirement may constitutionally be met at the guilt phase rather than the sentencing phase. It focuses on dicta in the opinion announcing the judgment to the effect that the classes of capital murder established by the Texas Legislature serve “ ‘much the same purpose’ ” as a list of statutory aggravating circumstances. Ante, at 245 (quoting Jurek v. Texas, supra, at 270 (opinion of Stewart, Powell, and Stevens, JJ.)). The Court ignores our later recognition that the three questions posed to the jury during the sentencing phase under the scheme approved in Jurek establish additional aggravating circumstances not already determined during the guilt phase. See, e. g., Skipper v. South Carolina, 476 U. S. 1, 5 (1986) (“[Ejvidence that a defendant would in the future pose a danger to the community if he were not executed may be treated as establishing an ‘aggravating factor’ for purposes of capital sentencing”); Adams v. Texas, 448 U. S. 38, 46 (1980) (In answering the three sentencing questions affirmatively, jurors in Texas “must consider both aggravating and mitigating circumstances, whether appearing in the evidence presented at the trial on guilt or innocence or during the sentencing proceedings”). Hence, Jurek cannot establish our approval of a divorce of the narrowing requirement from the sentencing proceedings.
Moreover, even if Jurek did stand for the proposition advanced by the Court, it would still be distinguishable from the instant case. Under the Texas capital sentencing statute evaluated in Jurek, jurors are explicitly instructed at the guilt phase that their findings would make the defendant eligible for the death penalty. See Jurek v. State, 522 S. W. 2d 934, 938 (Tex. Crim. App. 1975). In the instant case, the jurors were specifically instructed not to consider the penalty that might result from their findings during the guilt phase. See Record 2283.