(delivered the opinion of the Court)
Petitioner Jalil Abdul-Kabir, formerly known as Ted Calvin Cole, contends that there is a reasonable likelihood that the trial judge’s instructions to the Texas jury that sentenced him to death prevented jurors from giving meaningful consideration to constitutionally relevant mitigating evidence. He further contends that the judgment of the Texas Court of Criminal Appeals (CCA) denying his application for post-conviction relief on November 24, 1999, misapplied the law as clearly established by earlier decisions of this Court, thereby warranting relief under the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA),28 U. S. C. §2254. We agree with both contentions. Although the relevant state-court judgment for purposes of our review under AEDPA is that adjudicating the merits of Cole’s state habeas application, in which these claims were properly raised, we are persuaded that the same result would be dictated by those cases decided before the state trial court entered its judgment affirming Cole’s death sentence on September 26, 1990. Accordingly, we reverse the judgment of the Court of Appeals and remand for further proceedings consistent with this opinion.
In December 1987, Cole, his stepbrother Michael Hickey, and Michael’s wife, Kelly, decided to rob and kill Kelly’s grandfather, Raymond Richardson, to obtain some cash. Two days later they did so. Cole strangled Richardson with a dog leash; the group then searched the house and found $20 that they used to purchase beer and food. The next day, Michael and Kelly surrendered to the police and confessed. The police then arrested Cole who also confessed.
Cole was tried by a jury and convicted of capital murder. After a sentencing hearing, the jury was asked to answer two special issues (App. 127–128):
Was the conduct of the defendant, TED CALVIN COLE, that caused the death of the deceased, RAYMOND C. RICHARDSON, committed deliberately and with the reasonable expectation that the death of the deceased or another would result?
. . . .
Is there a probability that the defendant, TED CALVIN COLE, would commit criminal acts of violence that would constitute a continuing threat to society?
The trial judge instructed the jury to take into consideration evidence presented at the guilt phase as well as the sentencing phase of the trial but made no reference to mitigating evidence. Under the provisions of the Texas criminal code, the jury’s affirmative answers to these two special issues required the judge to impose a death sentence. See Tex. Code Crim. Proc. Ann., Art. 37.071 (Vernon 2006).
At the sentencing hearing, the State introduced evidence that Cole pleaded guilty to an earlier murder when he was only 16. Shortly after being released on parole, Cole pleaded guilty to charges of aggravated sexual assault on two boys and was sentenced to 15 more years in prison. As evidence of Cole’s propensity for future dangerousness, the State introduced Cole’s diary which, according to the State’s expert psychiatrist, Dr. Richard Coons, revealed a compulsive attraction to young boys and an obsession with criminal activity. Dr. Coons described Cole as a sociopath who lacked remorse and would not profit or learn from his experiences.
In response, Cole presented two categories of mitigating evidence. The first consisted of testimony from his mother and his aunt, who described his unhappy childhood. Cole’s parents lived together “off and on” for 10 years, over the course of which they had two children, Cole, and his younger sister, Carla. App. 35. Shortly after Cole was born, his father was arrested for robbing a liquor store. Cole’s father deserted the family several times, abandoning the family completely before Cole was five years old. On the last occasion that Cole saw his father, he dropped Cole off a block from where he thought Cole’s mother lived, told Cole to “go find her,” and drove off. Id., at 42. Cole had no contact with his father during the next 10 years. Ibid. After Cole’s father left, his mother found herself unable to care for Cole and his sister and took the children to live with her parents in Oklahoma. Cole’s grandparents were both alcoholics – Cole’s mother was herself a self-described “drunk” – and lived miles away from other children. Eventually, because Cole’s grandparents did not want their daughter or her children living with them, Cole’s mother placed him in a church-run children’s home, although she kept her daughter with her. Over the next five years Cole’s mother visited him only twice. Cole’s aunt, who visited him on holidays, testified that Cole seemed incapable of expressing any emotion and that his father never visited him at all.
The second category of mitigating evidence came from two expert witnesses – a psychologist and the former chief mental health officer for the Texas Department of Corrections – who discussed the consequences of Cole’s childhood neglect and abandonment. Dr. Jarvis Wright, the psychologist, spent 8 to 10 hours interviewing Cole and administering an “extensive battery of psychological tests.” Id., at 63. He testified that Cole had “real problems with impulse control” apparently resulting from “central nervous damage” combined with “all the other factors of [his] background.” Id., at 69. He also testified that Cole had likely been depressed for much of his life, that he had a “painful” background, and that he had “never felt loved and worthwhile in his life.” Id., at 73, 86. Providing an analogy for Cole’s early development, Dr. Wright stated that “the manufacturing process [had] botched the raw material horribly.” Id., at 73.
When specifically asked about future dangerousness, Dr. Wright acknowledged that “if Ted were released today on the street, there’s a much greater probability of dangerous behaviour than with the rest of us.” Id., at 74. Although he acknowledged the possibility of change or “burn out,” he admitted that Cole would likely pose a threat of future dangerousness until “years from now.” Ibid. Except for his prediction that Cole would change as he grew older, Dr. Wright’s testimony did not contradict the State’s claim that Cole was a dangerous person, but instead sought to provide an explanation for his behaviour that might reduce his moral culpability.
Dr. Wendell Dickerson, a psychologist who had not previously examined Cole, observed that it was difficult to predict future dangerousness, but that “violent conduct is predominantly, overwhelmingly the province of the young” with the risk of violence becoming rare as people grow older. Id., at 95. On cross-examination, in response to a hypothetical question about a person with Cole’s character and history, Dr. Dickerson acknowledged that he would be “alarmed” about the future conduct of such a person because “yes, there absolutely is a probability that they would commit .... future acts of violence.” Id., at 113. In sum, the strength of Cole’s mitigating evidence was not its potential to contest his immediate dangerousness, to which end the experts’ testimony was at least as harmful as it was helpful. Instead, its strength was its tendency to prove that his violent propensities were caused by factors beyond his control – namely, neurological damage and childhood neglect and abandonment.
It was these latter considerations, however, that the prosecutor discouraged jurors from taking into account when formulating their answers to the special issues. During the voir dire, the prosecutor advised the jurors that they had a duty to answer the special issues based on the facts, and the extent to which such facts objectively supported findings of deliberateness and future dangerousness, rather than their views about what might be an appropriate punishment for this particular defendant. For example, juror Beeson was asked:
[I]f a person had a bad upbringing, but looking at those special issues, you felt that they [sic] met the standards regarding deliberateness and being a continuing threat to society, could you still vote ‘yes,’ even though you felt like maybe they’d [sic] had a rough time as a kid? If you felt that the facts brought to you by the prosecution warranted a ‘yes’ answer, could you put that out of your mind and just go by the facts?
. . . .
[T]hat would not keep you from answering ‘yes,’ just because a person had a poor upbringing, would it?
XI Voir Dire Statement of Facts filed in No. CR88–0043–A (Dist. Ct. Tom Green Cty., Tex., 51st Jud. Dist.), p. 1588.
The prosecutor began his final closing argument with a reminder to the jury that during the voir dire they had “promised the State that, if it met its burden of proof,” they would answer “yes” to both special issues. App. 145. The trial judge refused to give any of several instructions requested by Cole that would have authorized a negative answer to either of the special issues on the basis of “any evidence which, in [the jury’s] opinion, mitigate[d] against the imposition of the Death Penalty, including any aspect of the Defendant’s character or record.” Id., at 115; see also id., at 117–124. Ultimately, the jurors answered both issues in the affirmative and Cole was sentenced to death.
On direct appeal, the sole issue raised by Cole was that the evidence was insufficient to support the jury’s verdict. The CCA rejected Cole’s claim and affirmed the judgment of the trial court on September 26, 1990.
On March 2, 1992, the lawyer who then represented Cole filed an application for a writ of habeas corpus in the Texas trial court, alleging 21 claims of error. Counsel later withdrew, and after delays caused in part by a letter from Cole to the trial judge stating that he wished to withdraw his “appeal,” the judge ultimately “had petitioner bench warranted” to a hearing on September 4, 1998. Id., at 152–153. During that hearing, Cole advised the court that he wished to proceed with his habeas proceedings and to have the CCA appoint counsel to represent him. Without counsel having been appointed to represent Cole, and without conducting an evidentiary hearing, the trial court entered its findings and conclusions recommending denial of the application.
Three of Cole’s 21 claims related to the jury’s inability to consider mitigating evidence. The trial judge rejected the first – “that his mitigating evidence was not able to be properly considered and given effect by the jury under the special issues,” id., at 157 – because he concluded that the record, and “especially” the testimony of the two expert witnesses, “provide[d] a basis for the jury to sufficiently consider the mitigating evidence offered by petitioner,” id., at 161. With respect to Cole’s second claim, the judge agreed that appellate counsel had been ineffective for failing to assign error based on “the trial court’s failure to instruct the jury on mitigating evidence as contemplated by the Pendry [sic] decision.” Id., at 166. He nevertheless found that the result on appeal would have been the same had the point been raised. Ibid. On the third claim relating to mitigating evidence, the judge rejected Cole’s argument that the trial court’s failure to specifically instruct the jury to consider mitigating evidence and offer a definition of “mitigating” was error. Id., at 173.
Over the dissent of two members of the court, and after adopting the trial court’s findings of fact and conclusions of law with only minor changes, the CCA denied Cole’s application for state collateral relief. Ex parte Cole, No. 41,673–01 (Nov. 24, 1999) (per curiam), App. 178–179. We consolidated this case with Brewer v Quarterman, post, p. ___, and granted certiorari, 549 U. S. ___ (2006).
After the Federal District Court granted Cole’s motion for the appointment of counsel, he filed a timely petition for a federal writ of habeas corpus pursuant to 28 U. S. C. §2254. His principal claim then, as it is now, was that the sentencing jury “was unable to consider and give effect to the mitigating evidence in his case,” in violation of the Constitution. Cole v Johnson, Civ. Action No. 6:00–CV–014–C (ND Tex., Mar. 6, 2001), p. 5, App. 184.
In its opinion denying relief, the District Court began by summarizing Cole’s mitigating evidence, highlighting his “destructive family background.” Ibid. The court then correctly described our decision in Penry I, 492 U. S. 302 , in these words:
In [Penry] the Supreme Court found that when the defendant places mitigating evidence before the jury, Texas juries must be given instructions which allow the jury to give effect to that mitigating evidence and to express its reasoned moral response to that evidence in determining whether to impose the death penalty.
Civ. Action No. 6:00–CV–014–C, at 8–9, App. 188.
The court next noted that the Fifth Circuit had formulated its own analysis for evaluating Penry claims. Under that analysis, for mitigating evidence to be constitutionally relevant, it “must show (1) a uniquely severe permanent handicap with which the defendant is burdened through no fault of his own, .... and (2) that the criminal act was attributable to this severe permanent condition.” Civ. Action No. 6:00–CV–014–C, at 9, App. 189 (quoting Davis v Scott, 51 F. 3d 457, 460–461 (CA5 1995) (internal quotation marks omitted; emphasis added)). Ultimately, Cole’s inability to show a “nexus” between his troubled family background and his commission of capital murder doomed his Penry claim. Civ. Action No. 6:00–CV–014–C, at 13, App. 193.
The Court of Appeals denied Cole’s application for a certificate of appealability (COA), holding that “reasonable jurists would not debate the district court’s conclusion that Cole’s evidence was not constitutionally relevant mitigating evidence.” Cole v Dretke, 418 F. 3d 494, 498 (CA5 2005). Shortly thereafter, however, we held that the Fifth Circuit’s “screening test” for determining the “ ‘constitutional relevance’ ” of mitigating evidence had “no foundation in the decisions of this Court.” Tennard v Dretke,542 U. S. 274, 284 (2004) . Accordingly, we vacated its order denying a COA in this case and remanded for further proceedings. On remand, the Court of Appeals reviewed Cole’s Penry claim on the merits and affirmed the District Court’s judgment denying the writ.
Focusing primarily on the testimony of petitioner’s two experts rather than that of his mother and his aunt, the Court of Appeals reviewed our recent decisions and concluded “that the Texas special issues allowed the jury to give ‘full consideration and full effect’ to the mitigating evidence that Cole presented at the punishment phase of his trial.” 418 F. 3d, at 511. With two judges dissenting, the court denied the petition for rehearing en banc.
Because Cole filed his federal habeas petition after the effective date of AEDPA, the provisions of that Act govern the scope of our review. We must therefore ask whether the CCA’s adjudication of Cole’s claim on the merits “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U. S. C. §2254(d)(1). We conclude that it did.
A careful review of our jurisprudence in this area makes clear that well before our decision in Penry I, our cases had firmly established that sentencing juries must be able to give meaningful consideration and effect to all mitigating evidence that might provide a basis for refusing to impose the death penalty on a particular individual, notwithstanding the severity of his crime or his potential to commit similar offenses in the future. Three of the five cases decided on the same day in 1976 – Woodson v North Carolina,428 U. S. 280 (1976), Proffitt v Florida,428 U. S. 242 (1976), and Jurek v Texas,428 U. S. 262 (1976) – identified the background principles we would apply in later cases to evaluate specific rules inhibiting the jury’s ability to give meaningful effect to such mitigating evidence.
In Woodson v North Carolina, we invalidated a statute that made death the mandatory sentence for all persons convicted of first-degree murder. One of the statute’s constitutional shortcomings was its “failure to allow the particularized consideration of relevant aspects of the character and record of each convicted defendant before the imposition upon him of a sentence of death.” 428 U. S., at 303 (plurality opinion). In Proffitt v Florida and Jurek v Texas, the joint opinions rejected facial challenges to the sentencing statutes enacted in Florida and Texas, assuming in both cases that provisions allowing for the unrestricted admissibility of mitigating evidence would ensure that a sentencing jury had adequate guidance in performing its sentencing function. As a majority of the Court later acknowledged, our holding in Jurek did not preclude the possibility that the Texas sentencing statute might be found unconstitutional as applied in a particular case. See n. 15, infra.
Two years later, in Lockett v Ohio,438 U. S. 586 (1978) , a plurality concluded “that the Eighth and Fourteenth Amendment s require that the sentencer, in all but the rarest kind of capital case, not be precluded from considering, as a mitigating factor, any aspect of a defendant’s character or record and any of the circumstances of the offense that the defendant proffers as a basis for a sentence less than death.” Id., at 604 (footnote omitted). Because Ohio’s death penalty statute was inconsistent with this principle, it was declared unconstitutional. The plurality noted the possible tension between a holding that the Ohio statute was invalid and our decisions in Proffitt and Jurek upholding the Florida and Texas statutes, but distinguished those cases because neither statute “clearly operated at that time to prevent the sentencer from considering any aspect of the defendant’s character and record or any circumstances of his offense as an independently mitigating factor.” 438 U. S., at 607.
While Chief Justice Burger’s opinion in Lockett was joined by only three other Justices, the rule it announced was endorsed and broadened in our subsequent decisions in Eddings v Oklahoma,455 U. S. 104 (1982) , and Skipper v South Carolina, 476 U. S. 1 (1986) . In those cases, we emphasized the severity of imposing a death sentence and that “the sentencer in capital cases must be permitted to consider any relevant mitigating factor.” Eddings, 455 U. S., at 112 (emphasis added).
In the wake of our decision in Lockett, Ohio amended its capital sentencing statute to give effect to Lockett’s holding. Neither Florida nor Texas did so, however, until after our unanimous decision in Hitchcock v Dugger, 481 U. S. 393 (1987) , unequivocally confirmed the settled quality of the Lockett rule. As Justice Scalia’s opinion for the Court explained, the defendant had introduced some rather atypical mitigating evidence that was not expressly authorized by the Florida statute, 481 U. S., at 397:
In the sentencing phase of this case, petitioner’s counsel introduced before the advisory jury evidence that as a child petitioner had the habit of inhaling gasoline fumes from automobile gas tanks; that he had once passed out after doing so; that thereafter his mind tended to wander; that petitioner had been one of seven children in a poor family that earned its living by picking cotton; that his father had died of cancer; and that petitioner had been a fond and affectionate uncle to the children of one of his brothers.
As the opinion further explained, the Florida courts had construed the state statute to preclude consideration of mitigating factors unmentioned in the statute. Accordingly, despite our earlier decision in Proffitt upholding the statute against a facial challenge, it was necessary to set aside Hitchcock’s death sentence. We explained, 481 U. S., at 398–399:
|We think it could not be clearer that the advisory jury was instructed not to consider, and the sentencing judge refused to consider, evidence of non-statutory mitigating circumstances, and that the proceedings therefore did not comport with the requirements of Skipper v South Carolina,476 U. S. 1 (1986) , Eddings v Oklahoma,455 U. S. 104 (1982) , and Lockett v Ohio,438 U. S. 586 (1978) (plurality opinion). Respondent has made no attempt to argue that this error was harmless, or that it had no effect on the jury or the sentencing judge. In the absence of such a showing our cases hold that the exclusion of mitigating evidence of the sort at issue here renders the death sentence invalid. See Skipper, supra (evidence that defendant had adapted well to prison life); Eddings, supra (evidence of 16-year-old defendant’s troubled family history and emotional disturbance).|
Of course, our reference to “exclusion” of the evidence did not refer to its admissibility, but rather to its exclusion from meaningful consideration by the jury. Had Jurek and Proffitt truly stood for the proposition that the mere availability of relevant mitigating evidence was sufficient to satisfy the Constitution’s requirements, Hitchcock could never have been decided as it was.
In the year following our decision in Hitchcock, we made clear that sentencing under the Texas statute, like that under the Florida statute, must accord with the Lockett rule. In Franklin v Lynaugh,487 U. S. 164, 172, 177, 183 (1988) , the plurality rejected the claim that the judge’s instructions did not allow the jury to give adequate weight to whatever “ ‘residual doubts’ ” it may have had concerning the defendant’s guilt, or to evidence of the petitioner’s good behaviour while in prison. That particular holding is unremarkable because we have never held that capital defendants have an Eighth Amendment right to present “residual doubt” evidence at sentencing, see Oregon v Guzek, 546 U. S. 517, 523–527 (2006) , and in most cases evidence of good behaviour in prison is primarily, if not exclusively, relevant to the issue of future dangerousness. What makes Franklin significant, however, is the separate opinion of Justice O’Connor, and particularly those portions of her opinion expressing the views of five Justices, see infra, at 18, and n. 15. After summarizing the cases that clarified Jurek’s holding, she wrote, 487 U. S., at 183, 184–185 (opinion concurring in judgment) :
In my view, the principle underlying Lockett, Eddings, and Hitchcock is that punishment should be directly related to the personal culpability of the criminal defendant.
In light of this principle it is clear that a State may not constitutionally prevent the sentencing body from giving effect to evidence relevant to the defendant’s background or character or the circumstances of the offense that mitigates against the death penalty. Indeed, the right to have the sentencer consider and weigh relevant mitigating evidence would be meaningless unless the sentencer was also permitted to give effect to its consideration.
Under the sentencing procedure followed in this case the jury could express its views about the appropriate punishment only by answering the special verdict questions regarding the deliberateness of the murder and the defendant’s future dangerousness. To the extent that the mitigating evidence introduced by petitioner was relevant to one of the special verdict questions, the jury was free to give effect to that evidence by returning a negative answer to that question. If, however, petitioner had introduced mitigating evidence about his background or character or the circumstances of the crime that was not relevant to the special verdict questions, or that had relevance to the defendant’s moral culpability beyond the scope of the special verdict questions, the jury instructions would have provided the jury with no vehicle for expressing its ‘reasoned moral response’ to that evidence.
Justice O’Connor’s opinion for the Court in Penry I endorsed the views she had expressed in Franklin and unquestionably governs the facts of this case. Penry contended that his mitigating evidence of mental retardation and an abusive childhood provided a basis for a sentence of life imprisonment rather than death and that the jury should have been instructed that it could consider that evidence when making its sentencing decision. In response to that contention, our opinion first held that Penry was not asking us to make new law because he was relying on a rule that was “dictated” by earlier cases, see n. 10, supra, and explained why Justice O’Connor’s separate opinion in Franklin correctly defined the relevant rule of law. In Franklin, we noted, “both the concurrence and the dissent stressed that ‘the right to have the sentencer consider and weigh relevant mitigating evidence would be meaningless unless the sentencer was also permitted to give effect to its consideration’ in imposing sentence.” 492 U. S., at 321 (citing Franklin, 487 U. S., at 185 (O’Connor, J., concurring in judgment); id., at 199 (Stevens, J., dissenting)).
Applying that standard, we held that neither the “deliberateness” nor the “future dangerousness” special issue provided the jury with a meaningful opportunity to give effect to Penry’s mitigating evidence. With respect to the former, we explained, 492 U. S., at 323:
In the absence of jury instructions defining ‘deliberately’ in a way that would clearly direct the jury to consider fully Penry’s mitigating evidence as it bears on his personal culpability, we cannot be sure that the jury was able to give effect to the mitigating evidence of Penry’s mental retardation and history of abuse in answering the first special issue. Without such a special instruction, a juror who believed that Penry’s retardation and background diminished his moral culpability and made imposition of the death penalty unwarranted would be unable to give effect to that conclusion if the juror also believed that Penry committed the crime ‘deliberately.’ Thus, we cannot be sure that the jury’s answer to the first special issue reflected a ‘reasoned moral response’ to Penry’s mitigating evidence.
With respect to the future dangerousness issue, we emphasized the fact that Penry’s evidence of mental retardation was relevant only as an aggravating factor. Id., at 323–324. More broadly, we noted that the evidence of Penry’s mental retardation and childhood abuse functioned as a “two-edged sword,” because it “may diminish his blameworthiness for his crime even as it indicates that there is a probability that he will be dangerous in the future.” Id., at 324. We therefore held that, in the absence of an appropriate instruction directing the “jury to consider fully” mitigating evidence as it bears on the extent to which a defendant is undeserving of a death sentence, “we cannot be sure” that it did so. Id., at 323. As our discussion of the deliberateness issue demonstrates, we did not limit our holding in Penry I to mitigating evidence that can only be viewed as aggravating. When the evidence proffered is double edged, or is as likely to be viewed as aggravating as it is as mitigating, the statute most obviously fails to provide for adequate consideration of such evidence.
The former special issues (as composed at the time of both Penry’s and Cole’s sentencing proceedings) provided an adequate vehicle for the evaluation of mitigating evidence offered to disprove deliberateness or future dangerousness. As Judge Reavley noted in his opinion for the Court of Appeals in Penry I, however, they did not tell the jury as to what “to do if it decided that Penry, because of retardation, arrested emotional development and a troubled youth, should not be executed.” Id., at 324 (internal quotation marks omitted).
In recommending denial of Cole’s application for collateral relief, the Texas trial judge did not analyze Penry I itself. Under the framework set forth in Penry I, the testimony of Cole’s mother and aunt, as well as the portions of the expert testimony suggesting that his dangerous character may have been the result of his rough childhood and possible neurological damage, were not relevant to either of the special verdict questions, except, possibly, as evidence supporting the State’s argument that Cole would be dangerous in the future. This would not satisfy the requirement of Penry I, however, that the evidence be permitted its mitigating force beyond the scope of the special issues. Therefore, it would have followed that those questions failed to provide the jury with a vehicle for expressing its “reasoned moral response” to that evidence.
Instead of relying on Penry I, the trial judge relied on three later Texas cases and on our opinion in Graham v Collins,506 U. S. 461 (1993) , as having held that nine different categories of mitigating evidence – including a troubled family background, bipolar disorder, low IQ, substance abuse, paranoid personality disorder, and child abuse – were sufficiently considered under the Texas special issues. App. 159–160. Applying those cases, the judge defined the legal issue “whether the mitigating evidence can be sufficiently considered” as one that “must be determined on a case by case basis, depending on the nature of the mitigating evidence offered and whether there exists other testimony in the record that would allow consideration to be given.” Id., at 160. As we have noted, in endorsing this formulation of the issue, neither the trial judge nor the CCA had the benefit of any input from counsel representing petitioner. See Part II, supra. In our view, denying relief on the basis of that formulation of the issue, while ignoring the fundamental principles established by our most relevant precedents, resulted in a decision that was both “contrary to” and “involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U. S. C. §2254(d).
The state court’s primary reliance on Graham, to the exclusion of our other cases in this line of jurisprudence, was misguided. In Graham, we held that granting collateral relief to a defendant who had been sentenced to death in 1984 would require the announcement of a new rule of constitutional law in contravention of Teague v Lane,489 U. S. 288 (1989) . In reaching that conclusion we relied heavily on the fact that in 1984 it was reasonable for judges to rely on the interpretation of Jurek that the plurality had espoused in Franklin. See 506 U. S., at 468–472; see also n. 15, supra. But as we have explained, in both Franklin and Penry I, a majority of the Court ultimately rejected the plurality’s interpretation of Jurek. Neither Franklin nor Penry I was inconsistent with Graham’s narrow holding, but they do suggest that our later decisions – including Johnson v Texas, 509 U. S. 350 (1993) , in which we refused to adopt the rule that Graham sought – are of more relevance to Cole’s case than Graham. The relevance of those cases lies not in their results – in several instances, we concluded, after applying the relevant law, that the special issues provided for adequate consideration of the defendant’s mitigating evidence – but in their failure to disturb the basic legal principle that continues to govern such cases: The jury must have a “meaningful basis to consider the relevant mitigating qualities” of the defendant’s proffered evidence. Johnson, 509 U. S., at 369; see also Graham, 506 U. S., at 474 (explaining that Penry was entitled to additional instructions “[b]ecause it was impossible [for the jury] to give meaningful mitigating effect to Penry’s evidence by way of answering the special issues”).
Before turning to those more recent cases, it is appropriate to identify the reasons why the CCA’s ruling was not a reasonable application of Penry I itself. First, the ruling ignored the fact that even though Cole’s mitigating evidence may not have been as persuasive as Penry’s, it was relevant to the question of Cole’s moral culpability for precisely the same reason as Penry’s. Like Penry’s evidence, Cole’s evidence of childhood deprivation and lack of self-control did not rebut either deliberateness or future dangerousness but was intended to provide the jury with an entirely different reason for not imposing a death sentence. Second, the judge’s assumption that it would be appropriate to look at “other testimony in the record” to determine whether the jury could give mitigating effect to the testimony of Cole’s mother and aunt is neither reasonable nor supported by the Penry opinion. App. 160. Third, the fact that the jury could give mitigating effect to some of the experts’ testimony, namely, their predictions that Cole could be expected to become less dangerous as he aged, provides no support for the conclusion that the jury understood it could give such effect to other portions of the experts’ testimony or that of other witnesses. In sum, the judge ignored our entire line of cases establishing the importance of allowing juries to give meaningful effect to any mitigating evidence providing a basis for a sentence of life rather than death. His recommendation to the CCA was therefore unsupported by either the text or the reasoning in Penry I.
The same principles originally set forth in earlier cases such as Lockett and Eddings have been articulated explicitly by our later cases, which explained that the jury must be permitted to “consider fully” such mitigating evidence and that such consideration “would be meaningless” unless the jury not only had such evidence available to it, but also was permitted to give that evidence meaningful, mitigating effect in imposing the ultimate sentence. Penry I, 492 U. S., at 321, 323 (internal quotation marks omitted); Graham, 506 U. S., at 475 (acknowledging that a “constitutional defect” has occurred not only when a jury is “precluded from even considering certain types of mitigating evidence,” but also when “the defendant’s evidence [i]s placed before the sentencer but the sentencer ha[s] no reliable means of giving mitigating effect to that evidence”).
Four of our more recent cases lend support to the conclusion that the CCA’s decision was unsupported by either the text or the reasoning of Penry I. In Johnson v Texas, we held that the Texas special issues allowed adequate consideration of petitioner’s youth as a mitigating circumstance. Indeed, we thought it “strain[ed] credulity to suppose that the jury would have viewed the evidence of petitioner’s youth as outside its effective reach” because its relevance was so obvious. 509 U. S., at 368. There is of course a vast difference between youth – a universally applicable mitigating circumstance that every juror has experienced and which necessarily is transient – and the particularized childhood experiences of abuse and neglect that Penry I and Cole described – which presumably most jurors have never experienced and which affect each individual in a distinct manner.
Evidence of youth, moreover, has special relevance to the question of future dangerousness. A critical assumption motivating the Court’s decision in Johnson was that juries would in fact be able to give mitigating effect to the evidence, albeit within the confines of the special issues. See 509 U. S., at 370 (“If any jurors believed that the transient qualities of petitioner’s youth made him less culpable for the murder, there is no reasonable likelihood that those jurors would have deemed themselves foreclosed from considering that in evaluating petitioner’s future dangerousness”). Prosecutors in some subsequent cases, however, have undermined this assumption, taking pains to convince jurors that the law compels them to disregard the force of evidence offered in mitigation. Cole’s prosecution is illustrative: the State made jurors “promise” they would look only at the questions posed by the special issues, which, according to the prosecutor, required a juror to “put .... out of [his] mind” Cole’s mitigating evidence and “just go by the facts.” Supra, at 6. Arguments like these are at odds with the Court’s understanding in Johnson that juries could and would reach mitigating evidence proffered by a defendant. Nothing in Johnson forecloses relief in these circumstances. See 509 U. S., at 369 (“Penry remains the law and must be given a fair reading”).
This conclusion derives further support from the fact that, in Johnson, the Court understood that the defendant’s evidence of youth – including testimony from his father that “his son’s actions were due in large part to his youth,” id., at 368, and counsel’s corresponding arguments that the defendant could change as he grew older – was “readily comprehended as a mitigating factor,” id., at 369, in the context of the special issues. The evidence offered in this case, however, as well as that offered by the petitioner in Brewer, post, at 2–3, and n.1, is closer in nature to that offered by the defendant in Penry I than that at issue in Johnson. While the consideration of the defendant’s mitigating evidence of youth in Johnson could easily have directed jurors towards a “no” answer with regard to the question of future dangerousness, a juror considering Cole’s evidence of childhood neglect and abandonment and possible neurological damage or Brewer’s evidence of mental illness, substance abuse, and a troubled childhood could feel compelled to provide a “yes” answer to the same question, finding himself without a means for giving meaningful effect to the mitigating qualities of such evidence. In such a case, there is a reasonable likelihood that the special issues would preclude that juror from giving meaningful consideration to such mitigating evidence, as required by Penry I. See Johnson, 509 U. S., at 367 (explaining that in Boyde v California,494 U. S. 370, 380 (1990) , “we held that a reviewing court must determine ‘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’ ”).
In three later cases, we gave Penry I the “fair reading” required by Johnson and repudiated several Fifth Circuit precedents providing the basis for its narrow reading of that case. First, in our review of Penry’s resentencing, at which the judge had supplemented the special issues with a nullification instruction, we again concluded that the jury had not been provided with an adequate “vehicle for expressing its reasoned moral response” to his mitigating evidence. Penry v Johnson,532 U. S. 782, 797 (2001) (Penry II). Indeed, given that the resentencing occurred after the enactment of AEDPA, we concluded (contrary to the views of the Fifth Circuit, which had denied Penry a COA) that the CCA’s judgment affirming the death sentence was objectively unreasonable. Id., at 803–804. Second, and as we have already noted, in Tennard we held that the Fifth Circuit’s test for identifying relevant mitigating evidence was incorrect. 542 U. S., at 284. Most recently, in Smith v Texas,543 U. S. 37 (2004) (per curiam), and again contrary to the views of the Fifth Circuit, we held that a nullification instruction that was different from the one used in Penry’s second sentencing hearing did not foreclose the defendant’s claim that the special issues had precluded the jury from “expressing a ‘reasoned moral response’ to all of the evidence relevant to the defendant’s culpability.” Id., at 46.
Our line of cases in this area has long recognized that before a jury can undertake the grave task of imposing a death sentence, it must be allowed to consider a defendant’s moral culpability and decide whether death is an appropriate punishment for that individual in light of his personal history and characteristics and the circumstances of the offense. As Chief Justice Burger wrote in Lockett, 438 U. S., at 605:
There is no perfect procedure for deciding in which cases governmental authority should be used to impose death. But a statute that prevents the sentencer in all capital cases from giving independent mitigating weight to aspects of the defendant’s character and record and to circumstances of the offense proffered in mitigation creates the risk that the death penalty will be imposed in spite of factors which may call for a less severe penalty. When the choice is between life and death, that risk is unacceptable and incompatible with the commands of the Eighth and Fourteenth Amendments.
Our cases following Lockett have made clear that when the jury is not permitted to give meaningful effect or a “reasoned moral response” to a defendant’s mitigating evidence – because it is forbidden from doing so by statute or a judicial interpretation of a statute – the sentencing process is fatally flawed. For that reason, our post-Penry cases are fully consistent with our conclusion that the judgment of the Court of Appeals in this case must be reversed. The case is remanded for further proceedings consistent with this opinion.
It is so ordered.
Chief Justice Roberts
(with whom Justice Scalia, Justice Thomas, and Justice Alito join, dissenting)
A jury imposed a sentence of death in each of these cases, despite hearing mitigating evidence from the defendants about their troubled backgrounds. The convictions and sentences were upheld on direct review. On state collateral review, each defendant claimed that the jury instructions did not allow sufficient consideration of the mitigating evidence. This Court had considered similar challenges to the same instructions no fewer than five times in the years before the state habeas courts considered the challenges at issue here. See Jurek v Texas,428 U. S. 262 (1976); Franklin v Lynaugh,487 U. S. 164 (1988); Penry v Lynaugh,492 U. S. 302 (1989) (Penry I); Graham v Collins, 506 U. S. 461 (1993); Johnson v Texas,509 U. S. 350 (1993) . Four of the cases rejected the defendant’s challenge. Only one – Penry I – upheld it. The guidance the Court gave in these five cases on whether the jury instructions at issue allowed sufficient consideration of mitigating evidence amounted to – it depends. It depends on the particular characteristics of the evidence in a specific case. The state courts here rejected the claim as applied to the particular mitigating evidence in these cases, and the defendants sought federal habeas review.
Under the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), however, a state-court decision can be set aside on federal habeas review only if it is “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U. S. C. §2254(d)(1). When this Court considers similar challenges to the same jury instructions five separate times, it usually is not because the applicable legal rules are “clearly established.” The Court today nonetheless picks from the five precedents the one that ruled in favour of the defendant – Penry I – and anoints that case as the one embodying “clearly established Federal law.” In doing so the Court fails to give any meaningful weight to the two pertinent precedents subsequent to Penry I – Graham and Johnson – even though those cases adopted a more “limited view” of Penry I than the Court embraces today. Johnson, supra, at 365. Indeed, the reading of Penry I in Graham and Johnson prompted every one of the remaining Justices who had been in the majority in Penry I on the pertinent question to dissent in Graham and Johnson, on the ground that the Court was failing to adhere to Penry I.
I suppose the Court today is free to ignore the import of Graham and Johnson on the question of what Penry I means, but in 1999 or 2001, respectively – when petitioners were denied collateral relief – the state courts did not have that luxury. They should not be faulted today for concluding – exactly as the Graham and Johnson dissenters did – that the Court had cut back significantly on Penry I.
We give ourselves far too much credit in claiming that our sharply divided, ebbing and flowing decisions in this area gave rise to “clearly established” federal law. If the law were indeed clearly established by our decisions “as of the time of the relevant state-court decision,” Williams v Taylor,529 U. S. 362, 412 (2000) , it should not take the Court more than a dozen pages of close analysis of plurality, concurring, and even dissenting opinions to explain what that “clearly established” law was. Ante, at 10–24. When the state courts considered these cases, our precedents did not provide them with “clearly established” law, but instead a dog’s breakfast of divided, conflicting, and ever-changing analyses. That is how the Justices on this Court viewed the matter, as they shifted from being in the majority, plurality, concurrence, or dissent from case to case, repeatedly lamenting the failure of their colleagues to follow a consistent path. Whatever the law may be today, the Court’s ruling that ’twas always so – and that state courts were “objectively unreasonable” not to know it, Williams, supra, at 409 – is utterly revisionist.
In 1987, Jalil Abdul-Kabir – referred to by his given name, Ted Calvin Cole, throughout this opinion, ante, at 1, n. 1 – was convicted of capital murder after he confessed to strangling 66-year-old Raymond Richardson with a dog leash to steal $20 from him. Among the 21 claims Cole raised on state collateral review was a challenge under Penry I, 492 U. S. 302 , to the application of Texas’s special issue jury instructions. In evaluating Cole’s challenge, the state habeas trial court stated, App. in No. 05–11284, pp. 159–160:
The issue is whether the sentencing jury had been unable to give effect to [Cole’s] mitigating evidence within the confines of the statutory ‘special issues.’ While [Penry I] held that evidence of a defendant’s mental retardation and abused childhood could not be given mitigating effect by a jury within the framework of the special issues, the cases that followed such as Graham v Collins, [506 U. S. 461 ] (1993), Garcia v State, 919 S. W. 2d 370 (1996), Mines v State, 888 S. W. 2d 816 (1994), and Zimmerman v State, 881 S. W. 2d 360 (1994) held that the mitigating evidence of alcoholism, drug abuse, bad family background, bipolar disorder, low I.Q., substance abuse, head injury, paranoid personality disorder and child abuse were sufficiently considered under the special issues. The issue of whether the mitigating evidence can be sufficiently considered must be determined on a case by case basis, depending on the nature of the mitigating evidence offered and whether there exists other testimony in the record that would allow consideration to be given.
Applying that standard, the state court concluded that “[t]he evidence presented at the punishment stage of the trial, especially evidence from [Cole’s] expert witnesses, provide[d] a basis for the jury to sufficiently consider the mitigating evidence.” Id., at 161. The Texas Court of Criminal Appeals adopted the trial court’s findings without substantive comment, and denied Cole’s application for habeas corpus relief on November 24, 1999. Id., at 178–179.
In finding that the state court’s decision was objectively unreasonable, the Court begins by stating that the principle the state court violated was “firmly established,” based on “[a] careful review of our jurisprudence in this area.” Ante, at 10. The only thing clear about our jurisprudence on the pertinent question in 1999, however, is that it was unsettled and confused.
In Jurek, the Court upheld Texas’s use of the special issues as facially constitutional, with the controlling opinion noting that “the constitutionality of the Texas procedures turns on whether the enumerated questions allow consideration of particularized mitigating factors.” 428 U. S., at 272 (joint opinion of Stewart, Powell, and Stevens, JJ.). In so doing, Jurek left open the possibility that some mitigating evidence might not be within the reach of the jury under the special issues; other types of mitigating evidence, of course, would be. Cf. id., at 272–273 (suggesting that the future dangerousness special issue allowed the jury to consider prior criminal conduct, age, duress, and whether the defendant was under extreme mental pressure).
The next occasion the Court had to consider mitigating evidence under the Texas special issues arose in Franklin, in which the Court concluded that the defendant’s mitigating evidence of good behaviour in prison was taken into account under the future dangerousness special issue. 487 U. S., at 178–179 (plurality opinion); id., at 186–187 (O’Connor, J., concurring in judgment). A plurality of the Court also rejected the argument that a jury must be permitted to give “independent” effect to mitigating evidence – beyond the special issues – concluding that “this submission is foreclosed by Jurek” and rejecting the dissent’s argument to the contrary. Id., at 179–180, and n. 10; see also id., at 199–200 (Stevens, J., dissenting).
The Court today places great weight on the opinion by Justice O’Connor concurring in the judgment in Franklin, an opinion joined only by Justice Blackmun. Ante, at 15–18. That separate opinion expressed “doubts” about the plurality’s view that mitigating evidence need not be given effect beyond the special issues, noting that if the petitioner in Franklin had introduced evidence not covered by the special issues, “we would have to decide whether the jury’s inability to give effect to that evidence amounted to an Eighth Amendment violation.” 487 U. S., at 183, 185. The separate opinion concluded, however, that “this is not such a case.” Id., at 185. According to the Court today, a discerning state judge should have seen that federal law was “clearly established” on the point by the concurring and dissenting opinions, not the plurality. Ante, at 15–18.
Penry I, decided the following Term, concluded that in that case the Texas instructions did not allow the jury to give mitigating effect to evidence of Penry’s mental retardation and abusive childhood. 492 U. S., at 328, 315 (“Penry does not .... dispute that some types of mitigating evidence can be fully considered by the sentencer in the absence of special jury instructions. Instead, Penry argues that, on the facts of this case, the jury was unable to fully consider and give effect to the mitigating evidence .... in answering the three special issues” (emphasis added; citations omitted)). In granting relief, the Court, quoting the Franklin concurrence, noted that Penry’s evidence “ ‘had relevance to [his] moral culpability beyond the scope of the special verdict questions,’ ” 492 U. S., at 322 (quoting 487 U. S., at 185 (O’Connor, J., concurring in judgment); some alterations deleted), and that it was relevant to the special issues “only as an aggravating factor.” 492 U. S., at 323 (emphasis in original). According to the Court today, the views of the Franklin concurrence and dissent were thus elevated to the opinion of the Court in Penry I, again clearly establishing federal law. Ante, at 17–18, and n. 15. The four dissenters in Penry I complained that the Court’s holding “flatly contradic[ted]” Jurek, and that in finding a constitutional violation, the Court was “throwing away Jurek in the process.” 492 U. S., at 355, 354 (Scalia, J., concurring in part and dissenting in part).
A state court looking at our pertinent precedents on the Texas special issue instructions would next have to consider the significance of Saffle v Parks,494 U. S. 484 (1990) . That case – issued less than nine months after Penry I – considered Oklahoma instructions, but extensively analyzed Penry I in doing so. See 494 U. S., at 491–492. The Court concluded that the mitigating evidence in that case could be adequately considered by the jury under the instructions given. The four dissenters in Saffle – including the author of today’s opinion – complained that the majority’s discussion of Penry I was “strangely reminiscent” of the position of the Penry I dissenters. 494 U. S., at 504 (opinion of Brennan, J.). The Saffle dissenters asserted that the majority’s failure to reject the position of the Penry I dissenters “creates considerable ambiguity about which Lockett [v Ohio,438 U. S. 586 (1978) ] claims a federal court may hereafter consider on habeas corpus review.” 494 U. S., at 504–505.
In Graham, 506 U. S., at 476–477, decided three years later, the Court sought to clarify the interplay between Jurek, Franklin, and Penry I:
It seems to us, however, that reading Penry as petitioner urges – and thereby holding that a defendant is entitled to special instructions whenever he can offer mitigating evidence that has some arguable relevance beyond the special issues – would be to require in all cases that a fourth ‘special issue’ be put to the jury: ‘ “Does any mitigating evidence before you, whether or not relevant to the above [three] questions, lead you to believe that the death penalty should not be imposed?” ’ The Franklin plurality rejected precisely this contention, finding it irreconcilable with the Court’s holding in Jurek, and we affirm that conclusion today.
[citation omitted; second emphasis added]
Thus, in Graham the Court rejected the reading of Franklin and Penry I that the Court today endorses, reasoning that it would require a new sentencing in every case, and would be impossible to square with Jurek.
Although the Court today tells us it was clear that the applicable federal law was established by the Franklin concurrence and dissent, and that Penry I had to be read in that light, ante, at 17–18, the Court majority in Graham specifically relied instead upon the Franklin plurality in rejecting the same broad reading of Penry I the Court resuscitates today, nunc pro tunc. Graham, supra, at 476–477. The dissenters in Graham – including every remaining Member of the Penry I majority – were adamant that Penry I should have been controlling in Graham. See, e.g., 506 U. S., at 507 (opinion of Souter, J., joined by Blackmun, Stevens, and O’Connor, JJ.) (“Our description of Penry’s claim applies .... almost precisely to Graham’s claim”); id., at 508 (“[Graham’s] position is identical to that of Penry”); id., at 512 (“Penry controls in this respect, and we should adhere to it”); id., at 520 (“[T]he case is controlled by Penry”). The issue is not whether the majority or the dissenters in Graham were right about how to read Penry I, but whether it was reasonable for a state court in 1999 to read it the way the majority in Graham plainly did.
Later the same Term, in Johnson, the Court reaffirmed the “limited view of Penry” it had adopted in Graham. 509 U. S., at 365. Once again the Court majority specifically relied on the Franklin plurality – not the concurrence and dissent. See 509 U. S., at 370–371. And once again the dissenters – including every remaining Member of the Penry I majority – lamented the Court’s asserted failure to adhere to Penry I. 509 U. S., at 385–386 (opinion of O’Connor, J., joined by Blackmun, Stevens, and Souter, JJ.). The dissent – by the Penry I author – made precisely the same point made by the Court today about how to read the Franklin concurrence and dissent. 509 U. S., at 385–386. The difference, of course, was that in Johnson the point was made in dissent. It cannot have been “objectively unreasonable” for a state court, in 1999, to have been guided by the Johnson majority on this question, rather than by the dissent.
In short, a state court reading our opinions would see an ongoing debate over the meaning and significance of Penry I. That state court would see four dissenters in Graham and Johnson – including every remaining Member of the Penry I majority – arguing that the Court was failing to follow or sharply limiting Penry I in those cases. On the flip side, the state court would see four dissenters in Penry I – every one later joining the majorities in Graham and Johnson – suggesting that the Penry I majority departed from Jurek. It is in that context that the Court today tells us that the state courts should have regarded Penry I as “clearly established Federal law, as determined by the Supreme Court of the United States.” §2254(d)(1).
The Court asserts that Graham and Johnson did not “disturb the basic legal principle” at issue, ante, at 23, and that we cite no post-Penry I cases inconsistent with its reading of that case, ante, at 17, n. 14. I do not understand how the author of today’s opinion can say that Graham did not disturb the principle of Penry I, however, when he joined a dissent in Graham stating that “[Graham’s] position is identical to that of Penry” and that Graham’s case “is controlled by Penry.” 506 U. S., at 508, 520 (opinion of Souter, J.) (emphasis added). That would seem to suggest that Graham was inconsistent with Penry I. I do not understand how the author of today’s opinion can say that Johnson had no effect on Penry I, when he joined a dissent in Johnson stating that the majority opinion “upset our settled Eighth Amendment jurisprudence.” 509 U. S., at 382 (opinion of O’Connor, J.). Now Johnson is dismissed as just an application of “basic legal principle[s],” over which Justices can disagree, ante, at 23; back then it “upset our settled Eighth Amendment jurisprudence.” And what of Saffle? There the author of today’s opinion joined a dissent claiming that the majority was adopting the rule rejected in Penry I. 494 U. S., at 504 (opinion of Brennan, J.). Again, that would seem to suggest inconsistency with Penry I.
In fact, Penry I is not even consistent with the reading the Court ascribes to it – in that case the Court concluded that a jury could only view Penry’s mitigating evidence as aggravating, and thus could not give the evidence any mitigating effect. 492 U. S., at 323 (Penry’s evidence was “relevant only as an aggravating factor” (emphasis in original)); see also Graham, supra, at 473 (“Although Penry’s evidence of mental impairment and childhood abuse indeed had relevance to the ‘future dangerousness’ inquiry, its relevance was aggravating only” (emphasis in original)). The Court concedes that Cole’s evidence in the present case was not purely aggravating, see ante, at 24 (“[T]he jury could give mitigating effect to some of the experts’ testimony”), thus drawing into even starker contrast the rule that was established by a fair reading of Penry I in 1999 versus the rule the Court today reads Penry I to have “clearly established.”
As might be expected in light of the foregoing, judges called upon to apply these precedents were confused by the ambiguity of this Court’s pronouncements. See, e.g., Mines v Texas, 888 S. W. 2d 816, 820 (Tex. Crim. App. 1994) (Baird, J., concurring) (“The Supreme Court’s holdings in Penry, Graham and Johnson do not provide an analytical framework to determine when our capital sentencing scheme fails to allow the jury to consider and give effect to mitigating evidence ....”); see also Brewer v Dretke, 442 F. 3d 273, 279, n. 16 (CA5 2006) (per curiam) (remarking, in applying Graham and Penry I, that “[t]here is no easy way to locate [the defendant] at either pole”). Commentators at the time likewise concluded that Graham and Johnson “put a cap on Penry’s principles.” Denno, Testing Penry and Its Progeny, 22Am. J. Crim. L. 1, 10 (1994) (“In Graham, the Court made clear that it did not interpret Penry ‘as effecting a sea change’ in its evaluation of the constitutionality of the former Texas death penalty statute ....”). See also Twenty-Eighth Annual Review of Criminal Procedure, 87 Geo. L. J. 1756, 1770 (1999) (“The possible reach of Penry has been circumscribed by [Graham] and [Johnson]”).
It is a familiar adage that history is written by the victors, but it goes too far to claim that the meaning and scope of Penry I was “clearly established” in 1999, especially in the wake of Graham and Johnson. In applying AEDPA, we have recognized that “[a] federal court may not overrule a state court for simply holding a view different from its own, when the precedent from this Court is, at best, ambiguous.” Mitchell v Esparza, 540 U. S. 12, 17 (2003) (per curiam); see also Lockyer v Andrade, 538 U. S. 63, 72–73 (2003) (declining to find federal law “clearly established” when “our precedents in [the] area have not been a model of clarity”).
When the state court rejected Cole’s claim, it knew that mitigating evidence of mental retardation and severe childhood abuse could not be given effect under the special issues, Penry I, 492 U. S., at 328, but that evidence of youth and a transient upbringing could be, Graham, 506 U. S., at 476; Johnson, 509 U. S., at 368. The court concluded that Cole’s mitigating evidence – a troubled childhood and “impulse control” disorder – was more like that considered in Johnson and Graham than in Penry I. And because Cole’s mitigating evidence was not as troubling as that at issue in Penry I, the state court did not act unreasonably in concluding that the collateral damage of his upbringing and impulse control disorder would, like youth in Johnson, dissipate over time, so that Cole would be less of a danger in the future. It is irrelevant that the ill effects of Cole’s upbringing and impulse control disorder might not wear off for some time – there was no suggestion in Johnson that the petitioner in that case would become less dangerous any time soon.
In other words, our precedents – which confirmed that the permanence of a mitigating feature was highly relevant, and that the correct answer was a case-specific matter turning on the particular facts – did not provide a clear answer, because the particular evidence before the court fell somewhere between the guideposts established by those precedents. As we have recognized, “the range of reasonable judgment can depend in part on the nature of the relevant rule .... [Some] rules are more general, and their meaning must emerge in application over the course of time.” Yarborough v Alvarado,541 U. S. 652, 664 (2004) . See also Brown v Payton, 544 U. S. 133, 143 (2005) (reviewing state-court application of Supreme Court precedent “to similar but not identical facts” and concluding that “[e]ven on the assumption that its conclusion was incorrect, it was not unreasonable, and is therefore just the type of decision that AEDPA shields on habeas review”).
The state court’s approach to the question was plainly correct; indeed, we engaged in a similar comparison in Graham, 506 U. S., at 476 itself in determining that the evidence presented in that case was cognizable under the special issues:
Jurek is reasonably read as holding that the circumstance of youth is given constitutionally adequate consideration in deciding the special issues. We see no reason to regard the circumstances of Graham’s family background and positive character traits in a different light. Graham’s evidence of transient upbringing and otherwise nonviolent character more closely resembles Jurek’s evidence of age, employment history, and familial ties than it does Penry’s evidence of mental retardation and harsh physical abuse.
The state court thought that Cole’s evidence “more closely resemble[d]” Johnson and Graham than Penry I. That cannot be said to be “contrary to, or .... an unreasonable application of, clearly established Federal law.” §2254(d)(1). See Brown, supra, at 143, 147; Williams, 529 U. S., at 411.
The Court further holds that the jury instructions did not permit Cole’s evidence to have “mitigating force beyond the scope of the special issues,” ante, at 21, as it now reads Penry I to require. At the time the state court ruled, however, Graham and Johnson, decided after Penry I, had expressly rejected the notion that a jury must “be able to give effect to mitigating evidence in every conceivable manner in which the evidence might be relevant,” so long as the jury could consider “in some manner all of a defendant’s relevant mitigating evidence.” Johnson, supra, at 372–373. The state court found that Cole’s mitigating evidence could be “sufficiently consider[ed]” by the jury “within the confines of the statutory ‘special issues,’ ” App. in No. 05–11284, at 161, 159, a holding consistent with this Court’s precedents as of 1999 – and certainly not contrary to clearly established federal law.
In reaching today’s result, the Court also takes advantage of eight years of hindsight and relies on three cases that postdate the state court’s ruling. Ante, at 28 (citing Penry v Johnson,532 U. S. 782 (2001) (Penry II), Tennard v Dretke, 542 U. S. 274 (2004), and Smith v Texas,543 U. S. 37 (2004) (per curiam)). What is pertinent under AEDPA, however, is whether federal law was clearly established by our decisions when the state court acted. Williams, supra, at 412. AEDPA requires state courts to reasonably apply clearly established federal law. It does not require them to have a crystal ball.
In 1991, petitioner Brent Ray Brewer was convicted of murder committed during the course of a robbery. Like Cole, Brewer claims that the Texas special issues prevented the jury from giving effect to mitigating evidence that he suffered from depression and had been abused as a teenager. The Texas courts rejected these claims on both direct and collateral review.
In evaluating Brewer’s claim, the Court focuses on the so-called “two-edged sword” nature of the evidence found to be beyond the jury’s reach in Penry I, and concludes that Brewer’s mitigating evidence is similarly double edged. The state court distinguished Penry I, however, stating that “a stay in a mental hospital does not evidence a long term mental illness which would affect appellant’s ability to conform to the requirements of society,” App. in No. 05–11287, p. 141 (internal quotation marks omitted), in contrast to Penry’s “organic brain disorder .... which made it impossible for him to appreciate the wrongfulness of his conduct or to conform his conduct to the law,” Penry I, 492 U. S., at 309. The state court determined that the nature of Brewer’s evidence allowed the jury to find that he would not be a future danger, whereas Penry’s did not.
The Court rejects this distinction, noting that while Brewer’s mitigating evidence may have been less compelling than Penry’s, “that difference does not provide an acceptable justification for refusing to apply the reasoning in Penry I to this case.” Ante, at 6, and n. 5. This misses the point. The state court’s distinction goes not to the relative strength of the mitigating evidence, but rather its character – an episodic rather than permanent mental disorder. As discussed in the context of Cole, see supra, at 12, the distinction was not a “refus[al] to apply the reasoning in Penry I,” ante, at 6, but rather an application of Penry I that can hardly be said to be “objectively unreasonable” based on this Court’s decisions as of 2001. Indeed, in considering future dangerousness, it is difficult to imagine a more pertinent distinction than whether a mental condition is or is not permanent.
The Court concedes that “[t]he transient quality of [Brewer’s] mitigating evidence may make it more likely to fall in part within the ambit of the special issues,” and yet still finds the state court’s decision unreasonable because the evidence may have had relevance beyond the special issues. Ante, at 7. As in Cole’s case, this conclusion squarely conflicts with the Court’s rejection in Graham of the proposition that “a defendant is entitled to special instructions whenever he can offer mitigating evidence that has some arguable relevance beyond the special issues.” 506 U. S., at 476 (emphasis in original). That rejection was confirmed in Johnson, see 509 U. S., at 372–373 (rejecting a rule that “would require that a jury be able to give effect to mitigating evidence in every conceivable manner in which the evidence might be relevant” in favour of the rule “that a jury be able to consider in some manner all of a defendant’s relevant mitigating evidence”). Once again, the Court rejects the state court’s reasonable reading of existing cases in favour of its own revisionist reading of this Court’s doctrine, heavily informed by subsequent decisions that the state court had no means to predict.
In AEDPA, Congress “work[ed] substantial changes” to the power of federal courts to grant habeas corpus relief. Felker v Turpin, 518 U. S. 651, 654 (1996) . In today’s decisions, the Court trivializes AEDPA’s requirements and overturns decades-old sentences on the ground that they were contrary to clearly established federal law at the time – even though the same Justices who form the majority today were complaining at that time that this Court was changing that “clearly established” law.
Still, perhaps there is no reason to be unduly glum. After all, today the author of a dissent issued in 1988 writes two majority opinions concluding that the views expressed in that dissent actually represented “clearly established” federal law at that time. So there is hope yet for the views expressed in this dissent, not simply down the road, but tunc pro nunc. Encouraged by the majority’s determination that the future can change the past, I respectfully dissent.
(with whom Justice Thomas joins, and with whom Justice Alito joins as to Part I, dissenting)
I remain of the view “that limiting a jury’s discretion to consider all mitigating evidence does not violate the Eighth Amendment.” Ayers v Belmontes, 549 U. S. ___, ___ (2006) (slip op., at 1) (Scalia, J., concurring) (citing Walton v Arizona,497 U. S. 639, 673 (1990) (Scalia, J., concurring in part and concurring in judgment)).
But even under this Court’s precedents to the contrary, the state-court decisions in these two cases were hardly objectively unreasonable under the Antiterrorism and Effective Death Penalty Act of 1996, as The Chief Justice’s dissenting opinion demonstrates. That is all which is needful to demonstrate the error of today’s judgments. The full truth is worse than that, however. There was in fact clearly established law that governed these cases, and it favoured the State. When the state courts rendered their decisions, Johnson v Texas, 509 U. S. 350 (1993) , was this Court’s most recent pronouncement on the Texas special issues. And in that case, the Court unambiguously drew back from the broader implications of its prior decision in Penry v Lynaugh, 492 U. S. 302 (1989) (Penry I). Reiterating what it had recently said in Graham v Collins,506 U. S. 461, 475 (1993) , the Court made clear that “ ‘[i]n Penry, the defendant’s evidence was placed before the sentencer but the sentencer had no reliable means of giving mitigating effect to that evidence.’ ” Johnson, supra, at 366 (emphasis added). Penry I, said Johnson, stood for the proposition that habeas relief was appropriate where jurors had been unable to give any mitigating effect to the evidence at issue. 509 U. S., at 369; see also Graham, supra, at 475. Penry I in no way meant to imply, Johnson warned, “that a jury [must] be able to give effect to mitigating evidence in every conceivable manner in which the evidence might be relevant.” 509 U. S., at 372 (emphasis added). Johnson thus established, in no uncertain terms, that jurors need only “be able to consider in some manner all of a defendant’s relevant mitigating evidence.” Ibid. (emphasis added); see generally id., at 372–373.
The dissenters in Johnson very much disagreed with that analysis. They read Penry I for the more expansive proposition that “the Texas special issues violated the Eighth Amendment to the extent they prevented the jury from giving full consideration and effect to a defendant’s relevant mitigating evidence.” 509 U. S., at 385 (opinion of O’Connor, J.) (citing Penry I, supra; emphasis added and deleted). “[H]aving some relevance to [a special] issue,” the dissent said, “was not sufficient.” 509 U. S., at 385. And because youth (the mitigating feature in Johnson) had obvious relevance beyond the special issues, an additional instruction was needed. Id., at 375. The differences between the Johnson majority and dissenters could not have been more pronounced.
Today the Court overrules Johnson sub silentio, and reinstates the “full effect” interpretation of Penry I. For as The Chief Justice explains, ante, at 12, 15 (dissenting opinion), it was not objectively unreasonable for the state courts to conclude that the ill effects of petitioners’ mental illnesses and difficult childhoods would wear off in due time, allowing the jury to give that mitigating evidence some effect through the future dangerousness instruction – just as could be done for the mitigating factor of youth in Johnson. The Court nonetheless reverses these sentences because the juries were unable to give effect to “any independent concern” (independent, that is, of the Texas special issues) that the defendants “may not be deserving of a death sentence,” Brewer, ante, at 6, or to consider the evidence’s “relevance to the defendant’s moral culpability beyond the scope of the special verdict questions,” id., at 7 (internal quotation marks omitted). The Court does not acknowledge that it is overruling Johnson, but makes the Court of Appeals the scapegoat for its change of heart.
The Fifth Circuit in both of these cases relied heavily on Johnson when denying relief. See Cole v Dretke, 418 F. 3d 494, 505 (2005); Brewer v Dretke, 442 F. 3d 273, 278, 281 (2006) (relying on Cole). How does the Court manage to distinguish it? The Court tries two main lines of argument. First, the Court explains, Abdul-Kabir, ante, at 26:
A critical assumption motivating the Court’s decision in Johnson was that juries would in fact be able to give mitigating effect to the evidence, albeit within the confines of the special issues .... Prosecutors in some subsequent cases, however, have undermined this assumption, taking pains to convince jurors that the law compels them to disregard the force of evidence offered in mitigation.
Because Johnson’s “critical assumption” has now been “undermined,” the Court says, Johnson cannot be said to “foreclos[e] relief in these circumstances.” Abdul-Kabir, ante, at 26.
This attempt to “distinguish” Johnson wilts under even the mildest scrutiny. Since when does this Court craft constitutional rules that depend on the beneficence of the prosecutor? (Never mind that this “critical assumption” of Johnson was not so critical as to be mentioned in the case.) And more importantly, how can prosecutorial style have any bearing on whether the Eighth Amendment requires a jury to be able to give “some effect,” as opposed to “full effect,” to a defendant’s mitigating evidence? It is of course true that a prosecutor’s arguments may be relevant evidence in the final analysis of whether a capital trial has met the “some effect” test. But it has absolutely no relevance to which test is selected in the first place.
Second, the Court explains that “the consideration of the defendant’s mitigating evidence of youth in Johnson could easily have directed jurors towards a ‘no’ answer with regard to the question of future dangerousness,” whereas a juror considering petitioners’ mitigating evidence “could feel compelled to provide a ‘yes’ answer to the same question.” Abdul-Kabir, ante, at 27. But it is quite apparent that jurors considering youth in Johnson could also have “fe[lt] compelled to provide a ‘yes’ answer” to the future dangerousness question. While one can believe that “the impetuousness and recklessness that may dominate in younger years can subside,” Johnson, 509 U. S., at 368, one can also believe that a person who kills even in his younger years is fundamentally depraved, and more prone to a life of violent crime. Johnson itself explicitly recognized this point, denying relief despite “the fact that a juror might view the evidence of youth as aggravating, as opposed to mitigating.” Ibid.
As the Court’s opinion effectively admits, nothing of a legal nature has changed since Johnson. What has changed are the moral sensibilities of the majority of the Court. For those in Texas who have already received the ultimate punishment, this judicial moral awakening comes too late. Johnson was the law, until today. And in the almost 15 years in-between, the Court today tells us, state and lower federal courts in countless appeals, and this Court in numerous denials of petitions for writ of certiorari, have erroneously relied on Johnson to allow the condemned to be taken to the death chamber. See, e.g., Robison v Johnson, 151 F. 3d 256, 269 (CA5 1998) (denying petition for rehearing), cert. denied,526 U. S. 1100 (1999) (petitioner executed Jan. 21, 2000); Motley v Collins, 18 F. 3d 1223, 1233–1235 (CA5), cert. denied sub nom. Motley v Scott,513 U. S. 960 (1994) (petitioner executed Feb. 7, 1995).
The individuals duly tried and executed between Johnson and today’s decisions were not, in my view (my view at the time of Johnson, and my view now), entitled to federal judicial invalidation of their state-imposed sentences. That is because in my view the meaning of the Eighth Amendment is to be determined not by the moral perceptions of the Justices du jour, but by the understanding of the American people who adopted it – which understanding did not remotely include any requirement that a capital jury be permitted to consider all mitigating factors. If, however, a majority of the Justices are going to govern us by their moral perceptions, in this area at least they ought to get their moral perceptions right the first time. Whether one regards improvised death-is-different jurisprudence with disdain or with approval, no one can be at ease with the stark reality that this Court’s vacillating pronouncements have produced grossly inequitable treatment of those on death row. Relief from sentence of death because of the jury’s inability to give “full effect” to all mitigating factors has been made available only to those who have managed to drag out their habeas proceedings until today. This is not justice. It is caprice.
 For purposes of consistency with testimony given by witnesses at trial and sentencing, we refer to petitioner throughout the opinion by his given name, Ted Cole.
 These were the two standard Texas special issues in place at the time of Cole’s sentencing. In 1991, the Texas Legislature amended the special issues in response to this Court’s decision in Penry v Lynaugh, 492 U. S. 302 (1989) (Penry I), to include language instructing the jury to decide “[w]hether, taking into consideration all of the evidence, including the circumstances of the offense, the defendant’s character and background, and the personal moral culpability of the defendant, there is a sufficient mitigating circumstance or circumstances to warrant that a sentence of life imprisonment without parole rather than a death sentence be imposed.” Tex. Code Crim. Proc. Ann., Art. 37.071, §2(e)(1) (Vernon 2006).
 Although Cole had not raised any of the 21 claims presented in his state habeas application on direct appeal – including his claim that the jury heard significant mitigating evidence which it could neither consider nor give effect to under the Texas sentencing statute, in violation of Penry I – under state law, his Penry claim remained cognizable on state habeas review. See Ex parte Kunkle, 852 S. W. 2d 499, 502, n. 3 (Tex. Crim. App. 1993) (en banc) (holding that “we have held that [allegations of Penry error occurring in cases tried before Penry] are cognizable via habeas corpus despite an applicant’s failure to raise them on direct appeal”). Nor did Cole’s failure to raise this claim on direct appeal affect its later review under AEDPA by the United States Court of Appeals for the Fifth Circuit. See Jackson v Johnson, 150 F. 3d 520, 523 (CA5 1998) (holding that Texas’ post-conviction procedures provide petitioners “adjudication on the merits” sufficient to satisfy 28 U. S. C. §2254(d)).
 The trial judge also noted that there were “no controverted, previously unresolved factual issues regarding petitioner’s Pendry [sic] claim.” App. 161.
 The contrast between the District Court’s succinct statement of Penry I’s holding and the prosecutor’s explanation at voir dire of the jurors’ duty to answer the special issues on the basis of the facts presented and not their views about Cole’s moral culpability, see Part I, supra, could not be more stark.
 The Court of Appeals distinguished Penry I on the ground that Penry’s evidence of mental retardation could only have been considered as aggravating, whereas this “record does not suggest that the jury viewed Cole’s mitigating evidence as an aggravating factor only .... [T]his evidence fits well within the broad scope of the future dangerousness special issue ....” 418 F. 3d, at 506–507, and n. 54.
 In his dissent, Judge Dennis argued that the panel had improperly “used another Fifth Circuit gloss upon a Supreme Court decision, i.e., the double edged evidence limitation of Penry I, that has no basis in the Supreme Court decisions, to avoid confronting the real issue.” Cole v Dretke, 443 F. 3d 441, 442 (CA5 2006) (per curiam).
 The opinion also referred to a proposition that “cannot fairly be denied – that death is a punishment different from all other sanctions in kind rather than degree,” and continued on to conclude that “[a] process that accords no significance to relevant facets of the character and record of the individual offender or the circumstances of the particular offense excludes from consideration in fixing the ultimate punishment of death the possibility of compassionate or mitigating factors stemming from the diverse frailties of humankind.” Woodson, 428 U. S., at 303–304.
 “By authorizing the defence to bring before the jury at the separate sentencing hearing whatever mitigating circumstances relating to the individual defendant can be adduced, Texas has ensured that the sentencing jury will have adequate guidance to enable it to perform its sentencing function.” Jurek, 428 U. S., at 276 (joint opinion of Stewart, Powell, and Stevens, JJ.); see also Proffitt, 428 U. S., at 257–258 (same).
 In Penry I itself, the Court noted that the rule sought by Penry – “that when such mitigating evidence is presented, Texas juries must, upon request, be given jury instructions that make it possible for them to give effect to that mitigating evidence in determining whether the death penalty should be imposed – is not a ‘new rule’ under Teague [v Lane, 489 U. S. 288 (1989),] because it is dictated by Eddings and Lockett.” 492 U. S., at 318–319.
 See Ohio Rev. Code Ann. §2929.04(B)(7) (Anderson 1982) (amended 1981) (adding, as a mitigating circumstance, “[a]ny other factors that are relevant to the issue of whether the offender should be sentenced to death”).
 To the extent that Jurek implied at the time it was decided that all that was required by the Constitution was that the defence be authorized to introduce all relevant mitigating circumstances, and that such information merely be before the jury, it has become clear from our later cases that the mere ability to present evidence is not sufficient. The only mitigating evidence presented in Jurek – offered to rebut the State’s witnesses’ testimony about Jurek’s bad reputation in the community – appears to have consisted of Jurek’s father’s testimony that Jurek had “always been steadily employed since he had left school and that he contributed to his family’s support.” 428 U. S., at 267. Therefore, the question presented in our later cases – namely, whether the jury was precluded from giving meaningful effect to mitigating evidence, particularly that which may go to a defendant’s lack of moral culpability – was not at issue in that case. When we deemed the Texas sentencing scheme constitutionally adequate in Jurek, we clearly failed to anticipate that when faced with various other types of mitigating evidence, the Texas special issues would not provide the sentencing jury with the requisite “adequate guidance.”
 “In Jurek v Texas, 428 U. S. 262 (1976), this Court held that the Texas capital sentencing procedures satisfied the Eighth Amendment requirement that the sentencer be allowed to consider circumstances mitigating against capital punishment. It was observed that even though the statute did not explicitly mention mitigating circumstances, the Texas Court of Criminal Appeals had construed the special verdict question regarding the defendant’s future dangerousness to permit jury consideration of the defendant’s prior criminal record, age, mental state, and the circumstances of the crime in mitigation. Id., at 271–273. Since the decision in Jurek, we have emphasized that the Constitution guarantees a defendant facing a possible death sentence not only the right to introduce evidence mitigating against the death penalty but also the right to consideration of that evidence by the sentencing authority. Lockett v Ohio, 438 U. S. 586 (1978), established that a State may not prevent the capital sentencing authority ‘from giving independent mitigating weight to aspects of the defendant’s character and record and to circumstances of the offense proffered in mitigation.’ Id., at 605 (plurality opinion). We reaffirmed this conclusion in Eddings v Oklahoma, 455 U. S. 104 (1982) , and in Hitchcock v Dugger, 481 U. S. 393 (1987) .” Franklin, 487 U. S., at 183–184 (emphasis added).
 The Chief Justice’s dissent incorrectly assumes that our holding today adopts the rule advocated by the petitioner in Graham v Collins, 506 U. S. 461 (1993), namely, that “ ‘a defendant is entitled to special instructions whenever he can offer mitigating evidence that has some arguable relevance beyond the special issues.’ ” Post, at 7 (quoting Graham, 506 U. S., at 476; emphasis in Graham). The rule that we reaffirm today – a rule that has been clearly established since our decision in Penry I – is this: Special instructions are necessary when the jury could not otherwise give meaningful effect to a defendant’s mitigating evidence. The rule is narrower than the standard urged by Graham because special instruction is not required when mitigating evidence has only a tenuous connection – “some arguable relevance” – to the defendant’s moral culpability. But special instruction is necessary when the defendant’s evidence may have meaningful relevance to the defendant’s moral culpability “beyond the scope of the special issues.” Penry I, 492 U. S., at 322–323. Despite the dissent's colourful rhetoric, it cites no post-Penry I cases inconsistent with this reading of its holding.
 “In Franklin, however, the five concurring and dissenting Justices did not share the plurality’s categorical reading of Jurek. In the plurality’s view, Jurek had expressly and unconditionally upheld the manner in which mitigating evidence is considered under the special issues. Id., at 179–180, and n. 10. In contrast, five Members of the Court read Jurek as not precluding a claim that, in a particular case, the jury was unable to fully consider the mitigating evidence introduced by a defendant in answering the special issues. 487 U. S., at 183 (O’Connor, J., concurring in judgment); id., at 199–200 (Stevens, J., dissenting). Indeed, both the concurrence and the dissent understood Jurek as resting fundamentally on the express assurance that the special issues would permit the jury to fully consider all the mitigating evidence a defendant introduced that was relevant to the defendant’s background and character and to the circumstances of the offense.” Penry I, 492 U. S., at 320–321; see also id., at 318 (“[T]he facial validity of the Texas death penalty statute had been upheld in Jurek on the basis of assurances that the special issues would be interpreted broadly enough to enable sentencing juries to consider all of the relevant mitigating evidence a defendant might present”).
 It is also clear that Penry I applies in cases involving evidence that is neither double edged nor purely aggravating, because in some cases a defendant’s evidence may have mitigating effect beyond its ability to negate the special issues. See, e.g., Tennard v Dretke,542 U. S. 274, 288–289 (2004) (holding that petitioner was entitled to a COA on his Penry claim where his evidence of low IQ and impaired intellectual functioning had “mitigating dimension beyond the impact it has on the individual’s ability to act deliberately”). In Tennard, the majority declined to accept the dissent’s argument that the petitioner’s evidence of low intelligence did “not necessarily create the Penry I ‘two-edged sword,’ ” and therefore could be given adequate mitigating effect within the context of the future dangerousness special issue. 542 U. S., at 293 (Rehnquist, C. J., dissenting). Cf. Johnson v Texas, 509 U. S. 350 (O’Connor, J., dissenting), in turn citing Penry I, 492 U. S., at 355 (Scalia, J., concurring in part and dissenting in part) (“The Court today holds that ‘the constitutionality turns on whether the [special] questions allow mitigating factors not only to be considered ...., but also to be given effect in all possible ways, including ways that the questions do not permit’ ” (emphasis in original)); cf. also Smith v Texas, 543 U. S. 37, 41, 46–48 (2004) (per curiam) (reversing the CCA’s denial of post-conviction relief because the special issues did not provide an adequate vehicle for expressing a “ ‘reasoned moral response’ ” to petitioner’s evidence of low IQ and a troubled upbringing).
 The lynchpin of The Chief Justice’s dissent is his assumption that Justice O’Connor’s opinions in Franklin and Penry I merely described two ad hoc judgments – see post, at 2, 5–6 – rather than her understanding of the governing rule of law announced in Lockett, Eddings, and Hitchcock v Dugger, 481 U. S. 393 (1987) . In his view, our line of cases in this area has flip-flopped, depending on the composition of the majority, rather than slowly defining core principles by eliminating those interpretations of the rule that are unsupportable. The fact that Justice O’Connor’s understanding of the law was confirmed by the Court in Penry I in 1989 – well before AEDPA was enacted – is a sufficient response to most of the rhetoric in the dissent. Neither Justice O’Connor’s opinion for the Court in Penry I, nor any other opinion she joined, ever endorsed the “ ‘some arguable relevance’ ” position described by The Chief Justice, see post, at 7, 16, which mistakenly interprets our opinion as adopting the rule that the dissenters in Franklin and Saffle would have chosen, see post, at 7, 16. The fact that the Court never endorsed that broader standard is fully consistent with our conclusion that the narrower rule applied in Penry I itself is “clearly established.” Arguments advanced in later dissenting opinions do not affect that conclusion.
 The Texas cases relied upon by the court were Garcia v State, 919 S. W. 2d 370, 398–399 (Tex. Crim. App. 1996) (en banc) (holding that, in light of the fact that Garcia received a “Penry” instruction (included in the amended Texas special issues), which instructed the jury to consider the defendant’s character and background in determining whether to impose life rather than death, he was not entitled to any special instructions requiring the jury to consider his drug use, alcoholism, and family background as mitigating evidence); Mines v State, 888 S. W. 2d 816, 818 (Tex. Crim. App. 1994) (en banc) (holding, on remand after Johnson, that Mines’ mitigating evidence of bipolar disorder was “well within the effective reach of the jury”); and Zimmerman v State, 881 S. W. 2d 360, 362 (Tex. Crim. App. 1994) (en banc) (holding, also on remand after Johnson, that Zimmerman’s “mitigating” evidence of low IQ, past substance abuse, a diagnosis of paranoid personality disorder, and a disruptive family environment did not warrant an additional instruction under Johnson or Penry I).
 Graham claimed that the Texas system had not “allowed for adequate consideration of mitigating evidence concerning his youth, family background, and positive character traits”; in Johnson, we declined to adopt such a rule, even without the Teague bar that prevented us from doing so in Graham. 509 U. S., at 365–366.
 This fact should be reassuring to those who fear that the rule we endorse today – and which we have endorsed since Penry I – “would require a new sentencing in every case.” Post, at 8 (Roberts, C. J., dissenting).
 A jury may be precluded from doing so not only as a result of the instructions it is given, but also as a result of prosecutorial argument dictating that such consideration is forbidden. See Part VI, infra.
 Because The Chief Justice’s only concern is with the proper application of AEDPA, he finds it unnecessary to define the rule that he thinks post-Penry I cases either did or should have applied. What is most relevant under AEDPA, however, is the holdings set forth in majority opinions, rather than the views of dissenters who supported a different understanding of the law at the time those opinions were written.
 We came to the same conclusion in Graham, after distinguishing the defendant’s mitigating evidence in that case from that offered by the defendant in Penry I: “The jury was not forbidden to accept the suggestion of Graham’s lawyers that his brief spasm of criminal activity in May 1981 was properly viewed, in light of his youth, his background, and his character, as an aberration that was not likely to be repeated. Even if Graham’s evidence, like Penry’s, had significance beyond the scope of the first special issue, it is apparent that Graham’s evidence – unlike Penry’s – had mitigating relevance to the second special issue concerning his likely future dangerousness. Whereas Penry’s evidence compelled an affirmative answer to that inquiry, despite its mitigating significance, Graham’s evidence quite readily could have supported a negative answer.” 506 U. S., at 475–476.
 In Graham, we acknowledged that Penry I did not “effec[t] a sea change in this Court’s view of the constitutionality of the former Texas death penalty statute.” Graham, 506 U. S., at 474. The reason, of course, that this was not the case is because the rule set forth in Penry I was merely an application of the settled Lockett-Eddings-Hitchcock rule described by Justice O’Connor in her opinions.
 Without making any attempt to explain how the jury in either this case or in Brewer v Quarterman, post, p. __, could have given “meaningful effect” or a “reasoned moral response” to either defendant’s mitigating evidence, The Chief Justice concludes his dissent by lamenting the fact that the views shared by Justice O’Connor’s concurrence and the dissenters in Franklin in 1988 – and later endorsed in Penry I – “actually represented ‘clearly established’ federal law at that time. ” Post, at 16. To his credit, his concluding sentence does not go so far as to state that he favours a “tunc pro nunc” rejection of those views, an endorsement of the views expressed by the four dissenters in Penry I, or even agreement with the Fifth Circuit’s recently rejected test for identifying relevant mitigating evidence. See Nelson v Quarterman, 472 F. 3d 287, 291–293 (2006) (en banc) (recognizing the “now-defunct” nature of the Fifth Circuit’s “ ‘constitutional-relevance’ test” post-Tennard and that a “ ‘full-effect’ ” standard – meaning that “a juror be able to express his reasoned moral response to evidence that has mitigating relevance beyond the scope of the special issues” – was “clearly established” for purposes of AEDPA in 1994, when Nelson’s conviction became final).
 In evaluating the state court’s analysis, the Court criticizes its reliance on Graham because Graham primarily addressed retroactivity under Teague v Lane, 489 U. S. 288 (1989). Ante, at 23. But in considering whether the rule requested was dictated by precedent, Graham of course had to evaluate the scope of that precedent – including Penry I – and did so extensively. See 506 U. S., at 467–477. Moreover, as explained below, the Court in Johnson v Texas, 509 U. S. 350, 370–372 (1993), adopted the same reading of Penry I adopted in Graham, without considering the issue under Teague.
 The Court is correct that “[w]hat is most relevant under AEDPA .... is the holdings set forth in majority opinions, rather than the views of dissenters .... at the time those opinions were written.” Ante, at 25, n. 22. But that must include the majority opinions in all the pertinent cases, not just the lone one of the bunch that ruled in favour of the defendant. Here it must include the subsequent majority opinions in Saffle, Graham, and Johnson, as well as in Penry I, and it was not objectively unreasonable for a state court to view Saffle, Graham, and Johnson the same way today’s author did at the time – or at least to conclude that the Court’s current view of Penry I was not as clearly established as the Court would have it today.
 The Court criticizes this dissent for failing “to define the rule” that our post-Penry I cases either did or should have applied. Ante, at 25, n. 22. But the whole point is that “the rule,” far from being “clearly established” by our decisions, was – at the very least – unsettled and confused. Under AEDPA, those defending the finality of a state-court judgment challenged on federal habeas review do not have to show that the state-court judgment was consistent with some version of “clearly established Federal law” other than that offered by the challenger; AEDPA obviously contemplates that there may not be “clearly established Federal law.” The Court’s criticism only underscores how far the reasoning employed today strays from AEDPA’s mandate.
 Relatedly, the Court thinks Johnson distinguishable because jurors have “experienced” youth but “have never experienced” the “particularized childhood experiences of abuse and neglect” at issue here. Abdul-Kabir, ante, at 25–26. It is again quite impossible to understand, however, how that can have any bearing upon whether “some effect” or “full effect” is the required test.
all rights reserved