with whom Justice Thomas joins, dissenting.
The Court concludes, ante, at 951-956, 177 L. Ed. 2d, at 1032-1034, that the Superior Court of Butts County, Georgia, made errors of law in applying the prejudice inquiry for ineffective-assistance-of-counsel claims under Strickland v. Washington, 466 U.S. 668, 104 S. Ct. 2052, 80 L. Ed. 2d 674 (1984). In my view there was no error of law, and the Court today remands for the state court to do what it has already done: find no reasonable likelihood that the mitigation evidence the Court details in its opinion would have persuaded a jury to change its mind about the death sentence for this brutal rape-murder.
The state habeas court responsibly executed the first step in the Strickland. analysis, finding that the investigation of mitigation evidence by Sears’ trial counsel was deficient performance. The issue here is the second step: whether Sears was prejudiced by that deficiency. As the Court acknowledges, ante, at 952, 177 L. Ed. 2d, at 1032, the state habeas court correctly stated the prejudice standard under Strickland: The defendant has the burden to establish “a reasonable probability that, but for counsel’s deficient performance, the result of the proceeding would have been different.” App. to Pet. for Cert. 24B-25B (citing 466 U.S., at 688, 694, 104 S. Ct. 2052, 80 L. Ed. 2d 674). “When applied to the sentencing phase of death penalty trials,” that means “a reasonable probability that, absent [counsel’s] errors, the sentencer would have concluded that the balance of the aggravating and mitigating circumstances did not warrant death.” App. to Pet. for Cert. 25B-26B.
The Court today concludes that there were two errors in the application of that proper standard. First, it reasons that the court erroneously “curtailed a more probing prejudice inquiry because it placed undue reliance on the assumed reasonableness of counsel’s mitigation theory” at trial. Ante, at 953, 177 L. Ed. 2d, at 1032. That argument is flawed on several levels. To begin with, the state habeas court did not assume trial counsel’s mitigation theory was reasonable; it found, that it was.
[561 U.S. 958]
It said: “[Although counsel failed to investigate thoroughly, they did develop a *1036reasonable mitigation theory with evidence to support it.” App. to Pet. for Cert. 30B. After interviews of roughly a dozen potential mitigation witnesses, who, with the exception of Sears’ father, gave positive accounts of Sears and his family, see 7 Record 2025, 2051-2052; 8 id., at 2129, 2291-2344, Sears’ trial counsel developed a mitigation theory that Sears came from a good family and had a solid middle-class upbringing; that his offense was completely out of character; that he cooperated with police; and that sentencing Sears to death would devastate his family and friends, see id., at 2124-2125; 19 id., at 4861-4862, 4916-4917, 4954-4955; 20 id., at 5181. To support that approach his attorneys called seven witnesses, including Sears’ mother, four family friends, and his high school guidance counselor. See Pet. for Cert. 6-7 (citing trial transcript pages between 2375 and 2451). The state habeas court did not declare that this mitigation theory “might be reasonable, in the abstract,” as the Court puts it, ante, at 953, 177 L. Ed. 2d, at 1033. Rather, it concluded that counsel “put forth a reasonable theory with supporting evidence.” App. to Pet. for Cert. 30B.
The Court’s argument is also flawed because the habeas court’s reasonableness finding did not cause it to “curtai[l]” its prejudice inquiry, or lead to the conclusion that it could “obviate the need to analyze” whether pursuing a different mitigation theory would have made a difference. Ante, at 953, 177 L. Ed. 2d, at 1033. The reasonableness finding merely meant that the prejudice determination had to be made by asking, not whether the jury’s mind would probably have been changed by hearing Sears’ new mitigation theory instead of hearing no mitigation theory at all; but rather whether it would probably have been changed by substituting Sears’ new mitigation theory for the reasonable mitigation theory that was presented and rejected.1 After hearing all the witnesses
[561 U.S. 959]
and other evidence Sears presented before it, the state court concluded that “it is just not possible to know what effect a different mitigation theory would have had.” App. to Pet. for Cert. 30B (emphasis added).2
The second, “and more fundamental!],” legal error the Court alleges, ante, at 954, 177 L. Ed. 2d, at 1033, is really encased within the first. The Court claims that the state habeas court “limited the prejudice inquiry under Strickland, to cases in which there was only ‘little or no mitigation evidence’ presented.” Ibid, (quoting App. to Pet. for Cert. 30B). The court erred, we are told, by determining that “presentation of] some mitigation evidence should foreclose an inquiry into whether” Sears was prejudiced. Ante, at 955, 177 L. Ed. 2d, at 1034. That is not a fair reading of the opinion. The state court did not hold that a defendant could never suffer prejudice whenever his counsel provided any mitigation evidence. Rather, it stated that “|t]his *1037case cannot be fairly compared with those where little or no mitigation evidence is presented and where a reasonable prediction of outcome can be made.” App. to Pet. for Cert. 30B (emphasis added). That is absolutely correct. This case is not like the prejudice cases on which the Court relies, where it could readily be said that the overlooked mitigation theory would have made a much deeper impression on the jury than the utterly unsupported theory (or absence of any theory) offered at trial. See Porter v. McCollum, 558 U.S. 30, 41, 130 S. Ct. 447, 175 L. Ed. 2d 398 (2009) (per curiam); Rompilla v. Beard, 545 U.S. 374, 378, 393, 125 S. Ct. 2456, 162 L. Ed. 2d 360 (2005); Wiggins v. Smith, 539 U.S. 510,
[561 U.S. 960]
515, 537, 123 S. Ct. 2527, 156 L. Ed. 2d 471 (2003); Williams v. Taylor, 529 U.S. 362, 369, 120 S. Ct. 1495, 146 L. Ed. 2d 389 (2000). Sears’ trial counsel presented a reasonable mitigation theory and offered evidence sufficient to support it, so the prejudice inquiry was more difficult—so difficult that Sears could not make the requisite showing. Clearly referring to the evidence in this particular case, the court said:
“Although here, the Petitioner can argue that a prior appeal shows the difficulty one juror was having reaching the same verdict as the others, it is just not possible to know what effect a different mitigation theory would have had on her, just as it is impossible to know what effect it would have had on other jurors.” App. to Pet. for Cert. 30B.
Since the habeas court made no legal error en route to its Strickland conclusion, the only basis for reversing the judgment here would be disagreement with the conclusion itself: that Sears had not established that his new mitigation theory would probably have caused the jury to impose a life sentence instead of death.
The Court makes no attempt to contradict that conclusion. Doing so would require a fact-intensive inquiry into the 22-volume record to measure the persuasiveness of the evidence supporting Sears’ new mitigation theory—an inquiry the Court purports to disavow, ante, at 956, 177 L. Ed. 2d, at 1034, but nonetheless tendentiously undertakes, ante, at 948-950, 177 L. Ed. 2d, at 1029-1031. The reader might think the state habeas court’s conclusion highly questionable from the Court’s account, which recites as solid all the evidence supporting Sears’ new mitigation theory, see ante, at 948-551, 177 L. Ed. 2d, at 1029-1032. It is far from solid. Some is likely inadmissible as unreliable hearsay under Georgia law, see Gissendaner v. State, 272 Ga. 704, 714, 532 S.E.2d 677, 688-689 (2000); Gulley v. State, 271 Ga. 337, 347, 519 S.E.2d 655, 664 (1999)—such as much of the evidence for the uncorroborated second-hand claim that Sears “suffered sexual abuse at the hands of
[561 U.S. 961]
an adolescent male cousin,” ante, at 948, 177 L. Ed. 2d, at 1029.3 Other evidence a competent attorney would likely not have placed before the jury—such as all the *1038testimony about Sears’ childhood from his brother Demetrius, an admitted drug dealer and drug user, 6 Record 1682-1684, 1695, 1752, and a convicted felon (for bank fraud, wire fraud, identity theft, and cocaine trafficking), id., at 1687. No juror would have been impressed by such a character witness.
Some of the evidence is incredible, such as the psychiatrist’s assertion that Sears had “substantial deficits in mental cognition and reasoning ... as a result of several serious head injuries he suffered as a child,” ante, at 949, 177 L. Ed. 2d, at 1030. The serious head injuries consisted of Sears’ hitting his head at a rollerskating rink sometime early in elementary school, 1 Record 76; 2 id., at 225, running into an end table as a child, 6 id., at 1651, and getting hit with a golf club sometime later in elementary school, 1 id., at 79; 2 id., at 225.4 (The last of
[561 U.S. 962]
these major injuries might not have been introduced anyway, since that would have provided the prosecution an opportunity to refute both the extent of the injury and the mercy-worthiness of Sears, by introducing into evidence Sears’ boast that when he was 11 or 12 he “beat the s*** out of’ someone after he was hit on the head with a golf club, 8 id., at 2195.) Likewise incredible was the assertion that Demetrius “introduced Sears to a life of crime,” ante, at 950, 177 L. Ed. 2d, at 1031. According to testimony on which the Court relies, Demetrius would “never let [Sears] hang around” with him and his drug-dealing friends. 6 Record 1685-1686.
A jury also would have discredited the psychiatric testimony of Dr. Strickland that “[f]rom an etiological standpoint . . . Sears’ ‘history is replete with multiple head trauma, substance abuse and traumatic experiences of the type expected’ to lead to these significant [mental] impairments,” ante, at 949, 177 L. Ed. 2d, at 1030 (quoting 2 Record 150). As already noted, the evidence of brain-damaging trauma is nonexistent. The psychiatric testimony of Dr. Dudley relied upon the self-interested reporting of Sears himself and the testimony of his less-than-trustworthy brother, Demetrius, see, e.g., 1 id., at 122, 133. And then there are the unfavorable parts of Dr. Dudley’s testimony: Sears is a “narcissis [t],” id., at 135, with a “grandiose” opinion of himself, id., at 98-99; 2 id., at 246. Dr. Dudley’s affidavit portrays Sears as arrogant and self-centered, id., at 246, 247, and notes what he termed Sears’ “fantastical” boasting of his first sexual experience with a woman at the age of six and his other “innumerable sexual experiences,” 1 id., at 98-99, 100; 2 id., at 246-247. It is hard to see how it could be thought probable that Sears’ so-called “magical thinking,” 1 id., at 84, would have helped his plea for leniency, see ante, at 951, 177 L. Ed. 2d, at 1031. It seems to me more likely the *1039jury would conclude that Sears’ “pro-foun[d] personality
[561 U.S. 963]
disorder,” 1 Record 104, made him exactly the kind of person who would commit heinous crimes in the future.
And some of the evidence the Court recounts is so utterly unlikely to affect a jury’s determination that this brutal murder deserved death that its recitation is just plain hilarious. For example, the claim that Sears’ father “was ‘verbally abusive,’ ” ante, at 948, 177 L. Ed. 2d, at 1029, resting on nothing more than an art teacher’s recollection that Sears’ father “severely criticized” him—“and meant it”!—at a conference with the principal concerning his son’s poor academic performance, 6 Record 1747; the claim that his father “disciplined Sears with age-inappropriate military-style drills,” ante, at 948, 177 L. Ed. 2d, at 1030, which consisted of positively VonSteubenesque acts such as dousing the kid with cold water when he refused to get up for school, and making him run extra laps after sports practices, 6 Record 1622; and the claim that his mother’s “ ‘favorite word’ ”—actually three words—to refer to her sons was scatological, ante, at 948, 177 L. Ed. 2d, at 1029 (quoting 2 Record 265).
While the Court takes pains to describe all the elements of Sears’ new mitigation theory, down to the silliest, it does not trouble to describe the brutal circumstances of the crime— which are at least just as relevant to assessing whether the different mitigation theory would probably have altered the sentence. But the jury heard all about them. See Sears v. State, 268 Ga. 759, 759-760, 493 S.E.2d 180, 182 (1997). They heard Sears’ confession that he kidnaped, raped, and murdered Gloria Wilbur, a 59-year-old wife and mother. Sears, carrying a briefcase containing various instruments of mayhem—brass knuckles, knives, and handcuffs—and his accomplice, Phillip Williams, were surveying a supermarket parking lot on a Sunday evening in October 1990, looking for a car to steal to drive back home to Ohio from Georgia. As the victim was putting her groceries in the trunk of her car, Sears approached, punched her in the face with his brass knuckles, shoved her into the car, and drove to pick up Williams. Sears then handcuffed her and pulled her into the
[561 U.S. 964]
backseat as Williams drove. After they passed into Tennessee, Sears raped her. Later in the evening, after they had crossed into Kentucky, Sears told Williams to stop the car. Sears forced her, still handcuffed, into the woods by the side of the highway as she begged for her life. After throwing her on the ground, he stabbed her in the neck. In his confession he showed no regret or remorse for his heinous crimes.5
I do not know how anyone could disagree with the habeas court’s conclusion that it is impossible to say that substituting the “deprived-childhood-cum-brain-damage” defense for the “good-middle-class-kid-who-made-a-mistake” defense would probably have produced a different verdict. I respectfully dissent.
. The Court contends, ante, at 953, 177 L. Ed. 2d, at 1033, that there was a “tension” between the state court’s conclusion that the investigation was deficient and its conclusion that the mitigation theory presented to the jury was reasonable. This terribly misreads the state court’s opinion. It did not say (as the Court’s point assumes) that counsel’s using the mitigation theory they did was reasonable; it said that the theory itself was reasonable, making it hard to say whether a different theory would have persuaded the jury. This presents no conceivable “tension.”
. On the fair reading we owe the state court, its opinion provides no basis for inferring that it failed to “engag[e] with the evidence” and “did not even conduct any real analysis.” Ante, at 953, n. 9, 177 L. Ed. 2d, at 1032.
. The Court’s reliance on Green v. Georgia, 442 U.S. 95, 97, 99 S. Ct. 2150, 60 L. Ed. 2d 738 (1979) (per curiam), ante, at 950-951, n. 6, 177 L. Ed. 2d, at 1031, to suggest that this unreliable hearsay would be admissible for sentencing purposes is entirely misplaced. In Green, we held it violated constitutional due process to exclude testimony regarding a co-conspirator’s confession that he alone committed the capital murder with which the defendant was charged. Our holding depended on “th[e] unique circumstances” of the case: The testimony to be used at sentencing was “highly relevant” and “substantial[ly]” reliable as a statement against penal interest made to a close friend; it was corroborated by “ample” evidence and was used by the State to obtain a *1038conviction in a separate trial against the co-conspirator. 442 U.S., at 97, 99 S. Ct. 2150, 60 L. Ed. 2d 738. Here there are no such circumstances. The testimony is uncorroborated secondhand reporting from self-interested witnesses that is unreliable and therefore likely inadmissible.
. There is an unsubstantiated claim from Sears himself, 8 Record 2195, that when he was a teenager he was hit with a “hatchet” above his right eye. Of course, that is the same place where he collided with an end table, 6 id., at 1651, leaving the “lesion”—better known as a scar—on his head that Dr. Strickland noted, ante, at 950, n. 5, 177 L. Ed. 2d, at 1031 (quoting 1 Record 78). There is no corroborating evidence for this event: no medical records, 1 id., at 77, no other apparent scars, 2 id., at 245; 6 id., at 1651, and, tellingly, no family or friends to confirm what surely would have been memorable had it happened.
. The jury also heard from several corrections officers who testified that while Sears was incarcerated awaiting trial and sentencing, he racked up dozens of disciplinary infractions, including assaults on other inmates. “ ‘Predatory’ ” “ ‘[i]ncorrigible,’ ” and incapable of reform was how they described him. 10 id., at 2951-2957; 19 id., at 4868.