with whom Justice Kennedy joins, concurring.
I join the Court’s opinion addressing two categories of facts excluded from consideration at capital sentencing proceedings by Booth v. Maryland, 482 U. S. 496 (1987), and South Carolina v. Gathers, 490 U. S. 805 (1989): information revealing the individuality of the victim and the impact of the crime on the victim’s survivors.1 As to these two categories, I believe Booth and Gathers were wrongly decided.
To my knowledge, our legal tradition has never included a general rule that evidence of a crime’s effects on the victim and others is, standing alone, irrelevant to a sentencing determination of the defendant’s culpability. Indeed, as the Court’s opinion today, see ante, at 819-821, and dissents in Booth, supra, at 519-520 (opinion of Scalia, J.) and Gathers, supra, at 817-820 (opinion of O’Connor, J.), make clear, criminal conduct has traditionally been categorized and penalized differently according to consequences not specifically *836intended, but determined in part by conditions unknown to a defendant when he acted. The majority opinion in Booth, supra, at 502-503, nonetheless characterized the consideration in a capital sentencing proceeding of a victim’s individuality and the consequences of his death on his survivors as “irrelevant” and productive of “arbitrary and capricious” results, insofar as that would allow the sentencing authority to take account of information not specifically contemplated by the defendant prior to his ultimate criminal decision. This condemnation comprehends two quite separate elements. As to one such element, the condemnation is merited but insufficient to justify the rule in Booth, and as to the other it is mistaken.
Evidence about the victim and survivors, and any jury argument predicated on it, can of course be so inflammatory as to risk a verdict impermissibly based on passion, not deliberation. Cf. Penry v. Lynaugh, 492 U. S. 302, 319-328 (1989) (capital sentence should be imposed as a “‘reasoned moral response’”) (quoting California v. Brown, 479 U. S. 538, 545 (1987) (O’Connor, J., concurring)); Gholson v. Estelle, 675 F. 2d 734, 738 (CA5 1982) (“If a person is to be executed, it should be as a result of a decision based on reason and reliable evidence”). But this is just as true when the defendant knew of the specific facts as when he was ignorant of their details, and in each case there is a traditional guard against the inflammatory risk, in the trial judge’s authority and responsibility to control the proceedings consistently with due process, on which ground defendants may object and, if necessary, appeal. See Darden v. Wainwright, 477 U. S. 168, 178-183 (1986) (due process standard of fundamental fairness governs argument of prosecutor at sentencing); United States v. Serhant, 740 F. 2d 548, 551-552 (CA7 1984) (applying due process to purportedly “inflammatory” victim impact statements); see also Lesko v. Lehman, 925 F. 2d 1527, 1545-1547 (CA3 1991); Coleman v. Saffle, 869 F. 2d 1377, 1394-1396 (CA10 1989), cert. denied, 494 U. S. 1090 *837(1990); Rushing v. Butler, 868 F. 2d 800, 806-807 (CA5 1989). With the command of due process before us, this Court and the other courts of the state and federal systems will perform the “duty to search for constitutional error with painstaking care,” an obligation “never more exacting than it is in a capital case.” Burger v. Kemp, 483 U. S. 776, 785 (1987).
Booth, supra,2 nonetheless goes further and imposes a blanket prohibition on consideration of evidence of the victim’s individuality and the consequential harm to survivors as irrelevant to the choice between imprisonment and execution, except when such evidence goes to the “circumstances of the crime,” id., at 502, and probably then only when the facts in question were known to the defendant and relevant to his decision to kill, id., at 505. This prohibition rests on the belief that consideration of such details about the victim and survivors as may have been outside the defendant’s knowledge is inconsistent with the sentencing jury’s Eighth Amendment duty “in the unique circumstance of a capital sentencing hearing ... to focus on the defendant as a ‘uniquely individual human bein[g].’” Id., at 504 (quoting Woodson v. North Carolina, 428 U. S. 280, 304 (1976) (plurality opinion of Stewart, Powell, and Stevens, JJ.)). The assumption made is that the obligation to consider the defendant’s uniqueness limits the data about a crime’s impact, on which a defendant’s moral guilt may be calculated, to the facts he specifically knew and presumably considered. His uniqueness, in other words, is defined by the specifics of his knowledge and the reasoning that is thought to follow from it.
To hold, however, that in setting the appropriate sentence a defendant must be considered in his uniqueness is not to require that only unique qualities be considered. While a defendant’s anticipation of specific consequences to the victims of his intended act is relevant to sentencing, such detailed *838foreknowledge does not exhaust the category of morally relevant fact. One such fact that is known to all murderers and relevant to the blameworthiness of each one was identified by the Booth majority itself when it barred the sentencing authority in capital cases from considering “the full range of foreseeable consequences of a defendant’s actions.” 482 U. S., at 504. Murder has foreseeable consequences. When it happens, it is always to distinct individuals, and, after it happens, other victims are left behind. Every defendant knows, if endowed with the mental competence for criminal responsibility, that the life he will take by his homicidal behavior is that of a unique person, like himself, and that the person to be killed probably has close associates, “survivors,” who will suffer harms and deprivations from the victim’s death. Just as defendants know that they are not faceless human ciphers, they know that their victims are not valueless fungibles; and just as defendants appreciate the web of relationships and dependencies in which they live, they know that their victims are not human islands, but individuals with parents or children, spouses or friends or dependents. Thus, when a defendant chooses to kill, or to raise the risk of a victim’s death, this choice necessarily relates to a whole human being and threatens an association of others, who may be distinctly hurt. The fact that the defendant may not know the details of a victim’s life and characteristics, or the exact identities and needs of those who may survive, should not in any way obscure the further facts that death is always to a “unique” individual, and harm to some group of survivors is a consequence of a successful homicidal act so foreseeable as to be virtually inevitable.
That foreseeability of the killing’s consequences imbues them with direct moral relevance, cf. Penry v. Lynaugh, supra, at 328 (death penalty should be “ ‘reasoned moral response’ ”), and evidence of the specific harm caused when a homicidal risk is realized is nothing more than evidence of the risk that the defendant originally chose to run despite the *839kinds of consequences that were obviously foreseeable. It is morally both defensible and appropriate to consider such evidence when penalizing a murderer, like other criminals, in light of common knowledge and the moral responsibility that such knowledge entails. Any failure to take account of a victim’s individuality and the effects of his death upon close survivors would thus more appropriately be called an act of lenity than their consideration an invitation to arbitrary sentencing. Indeed, given a defendant’s option to introduce relevant evidence in mitigation, see, e. g., Eddings v. Oklahoma, 455 U. S. 104, 113-114 (1982); Lockett v. Ohio, 438 U. S. 586, 604 (1978), sentencing without such evidence of victim impact may be seen as a significantly imbalanced process. See Mills v. Maryland, 486 U. S. 367, 397 (1988) (Rehnquist, C. J., dissenting).
I so view the relevance of the two categories of victim impact evidence at issue here, and I fully agree with the majority’s conclusion, and the opinions expressed by the dissenters in Booth and Gathers, that nothing in the Eighth Amendment’s condemnation of cruel and unusual punishment would require that evidence to be excluded. See ante, at 827 (“[I]f the State chooses to permit the admission of victim impact evidence and prosecutorial argument on that subject, the Eighth Amendment erects no per se bar”); Booth, supra, at 515-516 (White, J., dissenting) (nothing “'cruel or unusual’ or otherwise unconstitutional about the legislature’s decision to use victim impact statements in capital sentencing hearings”); Gathers, 490 U. S., at 816-821 (O’Connor, J., dissenting); id., at 823-825 (Scalia, J., dissenting).
I do not, however, rest my decision to overrule wholly on the constitutional error that I see in the cases in question. I must rely as well on my further view that Booth sets an unworkable standard of constitutional relevance that threatens, on its own terms, to produce such arbitrary consequences and uncertainty of application as virtually to guarantee a result far diminished from the case’s promise of appropriately *840individualized sentencing for capital defendants. 482 U. S., at 502. These conclusions will be seen to result from the interaction of three facts. First, although Booth was prompted by the introduction of a systematically prepared “victim impact statement” at the sentencing phase of the trial, Booth’s restriction of relevant facts to what the defendant knew and considered in deciding to kill applies to any evidence, however derived or presented. Second, details of which the defendant was unaware, about the victim and survivors, will customarily be disclosed by the evidence introduced at the guilt phase of the trial. Third, the jury that determines guilt will usually determine, or make recommendations about, the imposition of capital punishment.
A hypothetical case will illustrate these facts and raise what I view as the serious practical problems with application of the Booth standard. Assume that a minister, unidentified as such and wearing no clerical collar, walks down a street to his church office on a brief errand, while his wife and adolescent daughter wait for him in a parked car. He is robbed and killed by a stranger, and his survivors witness his death. What are the circumstances of the crime that can be considered at the sentencing phase under Booth? The defendant did not know his victim was a minister, or that he had a wife and child, let alone that they were watching. Under Booth, these facts were irrelevant to his decision to kill, and they should be barred from consideration at sentencing. Yet evidence of them will surely be admitted at the guilt phase of the trial. The widow will testify to what she saw, and, in so doing, she will not be asked to pretend that she was a mere bystander. She could not succeed at that if she tried. The daughter may well testify too. The jury will not be kept from knowing that the victim was a minister, with a wife and child, on an errand to his church. This is so not only because the widow will not try to deceive the jury about her relationship, but also because the usual standards of trial relevance afford factfinders enough information about *841surrounding circumstances to let them make sense of the narrowly material facts of the crime itself. No one claims that jurors in a capital case should be deprived of such common contextual evidence, even though the defendant knew nothing about the errand, the victim’s occupation, or his family. And yet, if these facts are not kept from the jury at the guilt stage, they will be in the jurors’ minds at the’ sentencing stage.
Booth thus raises a dilemma with very practical consequences. If we were to require the rules of guilt-phase evidence to be changed to guarantee the full effect of Booth’s promise to exclude consideration of specific facts unknown to the defendant and thus supposedly without significance in morally evaluating his decision to kill, we would seriously reduce the comprehensibility of most trials by depriving jurors of those details of context that allow them to understand what is being described. If, on the other hand, we are to leave the rules of trial evidence alone, Booth’s objective will not be attained without requiring a separate sentencing jury to be empaneled. This would be a major imposition on the States, however, and I suppose that no one would seriously consider adding such a further requirement.
But, even if Booth were extended one way or the other to exclude completely from the sentencing proceeding all facts about the crime’s victims not known by the defendant, the case would be vulnerable to the further charge that it would lead to arbitrary sentencing results. In the preceding hypothetical, Booth would require that all evidence about the victim’s family, including its very existence, be excluded from sentencing consideration because the defendant did not know of it when he killed the victim. Yet, if the victim’s daughter had screamed “Daddy, look out,” as the defendant approached the victim with drawn gun, then the evidence of at least the daughter’s survivorship would be admissible even under a strict reading of Booth, because the defendant, prior to killing, had been made aware of the daughter’s existence, *842which therefore became relevant in evaluating the defendant’s decision to kill. Resting a decision about the admission of impact evidence on such a fortuity is arbitrary.
Thus, the status quo is unsatisfactory, and the question is whether the case that has produced it should be overruled. In this instance, as in any other, overruling a precedent of this Court is a matter of no small import, for “the doctrine of stare decisis is of fundamental importance to the rule of law.” Welch v. Texas Dept. of Highways and Public Transportation, 483 U. S. 468, 494 (1987). To be sure, stare decisis is not an “inexorable command,” Burnet v. Coronado Oil & Gas Co., 285 U. S. 393, 405 (1932) (Brandeis, J., dissenting); and our “considered practice [has] not [been] to apply stare decisis as rigidly in constitutional [cases] as in nonconstitutional cases,” Glidden Co. v. Zdanok, 370 U. S. 530, 543 (1962). See Burnet, supra, at 405-407; Patterson v. McLean Credit Union, 491 U. S. 164, 172-173 (1989). But, even in constitutional cases, the doctrine carries such persuasive force that we have always required a departure from precedent to be supported by some “special justification.” Arizona v. Rumsey, 467 U. S. 203, 212 (1984).
The Court has a special justification in this case. Booth promises more than it can deliver, given the unresolved tension between common evidentiary standards at the guilt phase and Booth’s promise of a sentencing determination free from the consideration of facts unknown to the defendant and irrelevant to his decision to kill. An extension of the case to guarantee a sentencing authority free from the influence of information extraneous under Booth would be either an unworkable or a costly extension of an erroneous principle and would itself create a risk of arbitrary results. There is only one other course open to us. We can recede from the erroneous holding that created the tension and extended the false promise, and there is precedent in our stare decisis jurisprudence for doing just this. In prior cases, when this Court has confronted a wrongly decided, unworkable *843precedent calling for some further action by the Court, we have chosen not to compound the original error, but to overrule the precedent. See Swift & Co. v. Wickham, 382 U. S. 111 (1965);3 Continental T. V., Inc. v. GTE Sylvania Inc., 433 U. S. 36 (1977);4 see also Patterson v. McLean Credit *844Union, supra, at 173. Following this course here has itself the support not only of precedent but of practical sense as well. Therefore, I join the Court in its partial overruling of Booth and Gathers.
This case presents no challenge to the Court’s holding in Booth v. Maryland that a sentencing authority should not receive a third category of information concerning a victim’s family members’ characterization of and opinions about the crime, the defendant, and the appropriate sentence. See ante, at 830, n. 2.
Because this discussion goes only to the underlying substantive rule in question, for brevity I will confine most references to Booth alone.
In Swift & Co. v. Wickham, the Court overruled Kesler v. Department of Public Safety of Utah, 369 U. S. 163 (1962). The issue presented in both Swift and Kesler concerned the application of the three-judge district court statute, 28 U. S. C. § 2281 (1970 ed.), in cases of alleged state statutory pre-emption by federal law. The Court had held in Kesler that “§ 2281 comes into play only when the Supremacy Clause of the Federal Constitution is immediately drawn in question, but not when issues of federal or state statutory construction must first be decided even though the Supremacy Clause may ultimately be implicated.” 382 U. S., at 115.
Three years later in Swift & Co. v. Wickham, a majority of the Court disagreed with the Kesler analysis of the question, finding it inconsistent with the statute and earlier precedents of this Court. 382 U. S., at 122 (“The upshot of these decisions seems abundantly clear: Supremacy Clause cases are not within the purview of § 2281”). The Court concluded that there were
“[t]wo possible interpretations of § 2281 [that] would provide a more practical rule for three-judge court jurisdiction. The first is that Kesler might be extended to hold, as some of its language might be thought to indicate, that all suits to enjoin the enforcement of a state statute, whatever the federal ground, must be channeled through three-judge courts. The second is that no such suits resting solely on ‘supremacy’ grounds fall within the statute.” Id., at 125 (footnote omitted).
Rather than extend the incorrectly decided opinion in Kesler, the Court decided to overrule it. 382 U. S., at 126-127.
In Continental T. V., Inc. v. GTE Sylvania Inc., the Court overruled United States v. Arnold, Schwinn & Co., 388 U. S. 365 (1967), which had held that “[u]nder the Sherman Act, it is [per se] unreasonable ... for a manufacturer to seek to restrict and confine areas or persons with whom an article may be traded after the manufacturer has parted with dominion over it.” Id., at 379. The decision distinguished between restrictions on retailers based on whether the underlying transaction was a sale, in which case the Court applied a per se ban, or not a sale, in which case the arrangement would be subject to a “rule of reason” analysis. In Continental T. V., Inc., the Court reconsidered this per se rule in light of our traditional reliance on a “rule of reason” analysis for § 1 claims under the Sherman Act and the “continuing controversy and confusion, both in the *844scholarly journals and in the federal courts” caused by the sale/nonsale distinction drawn by the Court in Schwinn. 433 U. S., at 47-56. The Court proceeded to reexamination and concluded “that the distinction drawn in Schwinn between sale and nonsale transactions is not sufficient to justify the application of a per se rule in one situation and a rule of reason in the other. The question remains whether the per se rule stated in Schwinn should be expanded to include nonsale transactions or abandoned in favor of a return to the rule of reason.” Id., at 57. The Court found “no persuasive support for expanding the per se rule,” and Schwinn was overruled. 433 U. S., at 57.