concurring in part and dissenting in part:
I agree with the majority that Mann is not entitled to relief on his claims of guilt-phase ineffective assistance of counsel. I respectfully disagree that he is not entitled to relief on his claim of ineffective assistance of counsel at sentencing. Therefore, I concur in part and dissent in part.
I
Before venturing into the dense thicket of AEDPA nuances, we must recognize at the onset that, although the law may not provide a remedy, Mann was sentenced to death under an unconstitutional sentencing scheme.
First, his death sentence was imposed by a judge, not a jury, under a system that the Supreme Court later declared unconstitutional as a violation of the Sixth Amendment right to trial by jury. Ring v. Arizona, 536 U.S. 584, 609, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002). Mann, of course, has no remedy for that constitutional infirmity because Ring does not apply retroactively. Schriro v. Summerlin, 542 U.S. 348, 358, 124 S.Ct. 2519, 159 L.Ed.2d 442 (2004).
Second, he was sentenced during a period in which Arizona unconstitutionally precluded the consideration of mitigating evidence that lacked a “causal nexus” to the crime, in irreconcilable conflict with controlling Supreme Court Eighth Amendment jurisprudence as decided in Eddings v. Oklahoma, 455 U.S. 104, 102 S.Ct. 869, 71 L.Ed.2d 1 (1982). McKinney v. Ryan, 813 F.3d 798, 802 (9th Cir. 2015) (en banc). However, because neither Mann’s trial nor Mann’s post-conviction counsel raised this Eighth Amendment claim before the state court, we are precluded from granting Mann direct relief on a meritorious constitutional claim.
Yet one more constitutional error infected Mann’s sentencing: ineffective assistance of counsel. Mann’s attorney did not investigate or present available and significant mitigating evidence to the sentencing judge. The majority does not seriously contest this constitutional error, but concludes that AEDPA deference prevents us from reaching the merits of that claim ourselves. I respectfully disagree that AEDPA precludes meaningful habeas review. I would reach the merits and grant relief.
II
The state post-conviction court was required to assess Mann’s ineffective assistance of counsel claim under Strickland v. Washington, which held that a criminal defendant’s Sixth Amendment right to counsel is violated if the defendant’s trial attorney performs deficiently and thereby prejudices the defendant. 466 U.S. 668, 687, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). However, the post-conviction court committed constitutional error in its consideration of the ineffective assistance of counsel claim in two respects: (1) it unconstitutionally applied a “causal nexus” test for consideration of mitigating evidence in violation of Eddings, and (2) it applied a “more likely than not” standard rather than the constitutionally required “reasonable probability” test in its assessment of prejudice in violation of Strickland.1 The post-conviction court’s analysis *1163of the Strickland claim in its decision was thus contrary to clearly established federal law as determined by the Supreme Court. Therefore, its decision is not entitled to AEDPA deference, and we should analyze the ineffective assistance claim de novo on the merits.
A
Strickland requires a state post-conviction court to “evaluate the totality of the available mitigation evidence — both that adduced at trial, and the evidence adduced in the habeas proceeding” — and reweigh it against the aggravating evidence to determine whether counsel’s deficiency prejudiced the defendant. Williams v. Taylor, 529 U.S. 362, 397-99, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (emphasis added).
Sentencers and reviewing courts may determine what weight to accord mitigating evidence, but “may not give [mitigating evidence] no weight by excluding such evidence from their consideration.” Eddings, 455 U.S. at 114-15, 102 S.Ct. 869. Rather, it is “[h]ighly relevant — if not essential— [to the] selection of an appropriate sentence” that the sentencer consider “the fullest information possible concerning the defendant’s life and characteristics.” Lockett v. Ohio, 438 U.S. 586, 602-03, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978) (alterations in original) (quoting Williams v. New York, 337 U.S. 241, 247, 69 S.Ct. 1079, 93 L.Ed. 1337 (1949)). Accordingly, courts cannot require a capital defendant to “establish a nexus between [mitigating evidence] and [the] crime.” Tennard v. Dretke, 542 U.S. 274, 287, 124 S.Ct. 2562, 159 L.Ed.2d 384 (2004).
Yet, we know that throughout Mann’s trial, sentencing, direct appeal, and post-conviction proceeding, Arizona courts “articulated and applied a ‘causal nexus’ test,” contrary to Eddings, “that forbade as a matter of law giving weight to mitigating evidence, such as ... mental condition, unless the ... mental condition was causally connected to the crime.” McKinney, 813 F.3d at 802-03 (concluding that Arizona courts applied a “causal nexus” test contrary to clearly established federal law from 1989 until approximately 2005). It was not until the Supreme Court held in Tennard that a similar Fifth Circuit test had “no basis in [Supreme Court] precedents and ... [was] inconsistent with the standard [the Court] adopted for relevance in the capital sentencing context” that Arizona’s practice changed. 542 U.S. at 287, 124 S.Ct. 2562.
Shortly after Tennard, the Arizona Supreme Court recognized and applied a new rule eliminating the causal nexus requirement in State v. Anderson, 210 Ariz. 327, 111 P.3d 369, 391 (2005). But during the period relevant to this case, Arizona post-conviction courts were applying the unconstitutional causal nexus test, and were thus unable to properly “reweigh” the totality of available mitigating and aggravating evidence as required by Strickland.
In this case, the post-conviction court applied a causal nexus test in violation of Eddings. First, we know that the post-*1164conviction court was required to do so by state law, as the Arizona Supreme Court consistently articulated and applied a causal nexus test both before and after Mann’s post-conviction proceeding. McKinney, 813 F.3d at 803. Just as federal courts must presume that state courts “know and follow” federal law, Woodford v. Visciotti, 537 U.S. 19, 24, 123 S.Ct. 357, 154 L.Ed.2d 279 (2002) (per curiam), perhaps even more so, federal courts cannot “presume ... lightly that a state court failed to apply its own law,” here, an unconstitutional causal nexus test, Bell v. Cone, 543 U.S. 447, 455, 125 S.Ct. 847, 160 L.Ed.2d 881 (2005) (per curiam) (emphasis added). I cannot presume, nor do I believe, that the post-conviction judge would act contrary to controlling Arizona precedent.
Second, the record shows that the post-conviction court actually applied the causal nexus test to Mann’s mitigating evidence. Indeed, the Court said so, stating “[t]he Court finds that Defendant has not proven the existence of a causal connection between the accident and its effects and the murders.” Moreover, the post-conviction court and the State emphasized the causal nexus test throughout Mann’s post-conviction proceeding. When clinical neuropsy-chologist Dr. James F. Comer testified about the results of Mann’s neuropsycho-logical evaluation, the judge asked Dr. Comer to clarify whether Mann’s head injury could have affected the crime. Dr. Comer responded by stating that he was “making no direct, causal connection” between Mann’s organic brain damage and the murders because that was “something ... the court need[ed] to determine” — a response that suggests he understood the court’s question to be legal, not factual. The State emphasized that colloquy in its oral argument at the close of evidence, noting that the “court specifically asked Dr. Comer the critical legal issue, whether there’s any nexus between any possible injury to the brain from that accident and to the crime.” Indeed, when discussing Dr. Comer’s findings, the State’s post-conviction hearing memo cited State v. Hoskins, 199 Ariz. 127, 14 P.3d 997 (2000), a case— as recognized by the majority — in which the Arizona Supreme Court “articulated and insisted on its unconstitutional causal nexus test.” McKinney, 813 F.3d at 814-15.
The post-conviction court’s written decision then directly responded to its own questioning of Dr. Comer and the State’s invocation of the causal nexus test. The decision first addressed Mann’s state law claim for resentencing, which could only be granted if “[njewly discovered material facts ... probably would have changed the verdict or sentence.” Ariz. R. Crim. Pro. 32.1(e). Then, as previously noted, the post-conviction court concluded that Mann had “not proven the existence of a causal connection between the accident and its effects and the murders.” Rather, Mann had “committed the[ ] murders for pecuniary gain, not for reasons traceable to his 1985 accident.” The post-conviction court thus concluded that Mann could not establish that any new evidence probably would have changed his sentence, as required for relief pursuant to Arizona Rule of Criminal Procedure 32.1(e). The post-conviction court’s cursory ineffective assistance of counsel analysis then imported the reasoning from that state law resentencing analysis. The court concluded that Mann was not prejudiced by his counsel’s failure to introduce evidence of the accident and Mann’s attendant organic brain damage because “[additional evidence that pertain[ed] to the 1985 accident and its effects [was] discussed ... above, where th[e] Court found that it would not have changed the sentence imposed.” That “discuss[ion] ... above” turned almost exclusively on the lack of “a causal connection between the accident and its effects and the murders.”
*1165The post-conviction court did not use the phrase “causal connection” to respond to “the core” of Mann’s claims, as the majority suggests. Mann did not argue that evidence relating to the 1985 accident was relevant solely as a potential contributing factor to the crime. Rather, Mann explicitly argued at the post-conviction hearing that such evidence “would have provided some free-standing mitigation of itself.” Once the post-conviction court concluded that there was no causal connection, however, it did not consider and respond to this additional mitigation argument. The court failed to address how Mann’s organic bráin damage and attendant personality change — apart from a causal nexus to the crime — could have otherwise reduced his moral culpability or altered his sentencing profile.
Further, I cannot presume that the post-conviction court’s invocation of the phrase “causal connection” was divorced from the legal meaning given to that phrase by the Arizona Supreme Court.2 At the time of Mann’s post-conviction proceeding, the term “causal connection” was a legal term of art, inextricably intertwined with the Arizona Supreme Court’s unconstitutional causal nexus test. See e.g., State v. Martinez, 196 Ariz. 451, 999 P.2d 795, 808 (2000) (“[T]here was simply no causal connection between [the defendant’s] personality disorder and his actions on [the day of the murder].” (emphasis added)); State v. Sharp, 193 Ariz. 414, 973 P.2d 1171, 1182 (1999) (“Appellant failed to show a causal connection between [the mitigating evidence] and his actions on the night of the murder. We have previously explained that we require a causal connection to justify considering evidence of a defendant’s background as a mitigating circumstance.” (emphasis added)); State v. Rienhardt, 190 Ariz. 579, 951 P.2d 454, 467 (1997) (“Since [the defendant] declined to present any evidence of a causal connection at his aggravation-mitigation hearing, we reject this mitigating factor.” (emphasis added)). The post-conviction court referenced the lack of “causal connection” because it was required to do so by state law, not to respond to “the core” of Mann’s claim. To suggest otherwise is to say that the court did not mean what it said.
In sum, the post-conviction court excluded from its prejudice analysis as a matter of law mitigating evidence that, irrespective of its relationship to the crime, could have shed light on Mann’s moral culpability, contrary to clearly established federal law. See Eddings, 455 U.S. at 113, 102 S.Ct. 869. Therefore, its decision is not entitled to deference under AEDPA.
B
The post-conviction court’s prejudice analysis was also “contrary to” clearly established federal law because it applied the “more-likely-than-not” standard that Strickland rejected. 466 U.S. at 693-94, 104 S.Ct. 2052. No one disagrees that it is constitutional error to apply a “more-*1166likely-than-not” standard in analyzing prejudice because Strickland rejected that approach and instead established a “reasonable probability” standard. Specifically, Strickland held that to establish prejudice at the penalty phase of a capital proceeding, the petitioner must show that “there is a reasonable probability that, absent [counsel’s] errors, the senteneer ... would have concluded that the balance of aggravating and mitigating circumstances did not warrant death.” Id. at 695, 104 S.Ct. 2052. The Court explicitly held that a “reasonable probability” is “a probability sufficient to undermine confidence in the outcome,” not a probability that it is “more likely than not” the result would be different. Id. at 693-94, 104 S.Ct. 2052.
Here, the post-conviction court denied Mann’s ineffective assistance of counsel claim for the reasons “discussed” in its state law resentencing analysis, “where [the] [c]ourt found that [the new mitigating evidence] would not have changed the sentence imposed.” At the time of the post-conviction court’s decision, Arizona law limited resentencing to cases where “[n]ewly discovered material facts probably exist and such facts probably would have changed the verdict or sentence.” Ariz. R. Crim. P. 32.1(e) (2000). “Probably” effectively meant “more likely than not.” See State v. Orantez, 183 Ariz. 218, 902 P.2d 824, 829 (1995) (concluding that new evidence “would have likely resulted in a different verdict” and thus “probably change[d] the verdict”). The post-conviction court’s ineffective assistance of counsel analysis did not invoke a different test for assessing prejudice. The most fair reading of the decision is therefore that the court denied constitutional relief for the same reason — and under the same procedural rules — that it denied state law relief: it probably would not “have changed the verdict or sentence imposed.” Although the majority may be correct that the difference between the “more-likely-than-not” and “reasonable probability” standards is outcome determinative “in the rarest case,” it is a difference that permits us to make that assessment de novo, rather than through AEDPA’s deferential lens.
The fact that the post-conviction judge cited State v. Nash, 143 Ariz. 392, 694 P.2d 222 (1985), an Arizona case that adopted Strickland’s “reasonable probability” standard, does not alter this conclusion. The post-conviction court’s decision cites Nash only for the general idea that Strickland sets out the performance and prejudice test for Sixth Amendment claims, not to provide a specific prejudice standard distinct from the more-likely-than-not standard otherwise applicable under Arizona law and previously applied in the decision.
And even if the citation to Nash meant that the post-conviction court implicitly adopted the correct prejudice standard, it is not enough to cite Strickland — a court’s analysis must reflect it too. The Supreme Court has held that even if a state court identifies the proper Strickland standard, by “failing to apply Strickland to assess the ineffective-assistance-of-counsel claim ... raised, the state court’s adjudication [is] contrary to clearly established federal law.” Lafler v. Cooper, — U.S. —, 132 S.Ct. 1376, 1390, 182 L.Ed.2d 398 (2012). Here, it is apparent that the post-conviction court did not actually “apply Strickland to assess” Mann’s claim and reweigh the aggravating and mitigating evidence.
The only potential prejudice analysis is contained in the post-conviction court’s state law resentencing discussion, which primarily notes Mann’s pecuniary motive for the crime and the lack of a causal connection. However, both of those observations are irrelevant to a proper reweighing of aggravating and mitigating factors under Strickland — the post-conviction court merely restated an established aggravating factor and otherwise failed to *1167respond to Mann’s contentions that the new evidence undermined the trial court’s rejection of the non-statutory mitigating factor of remorse. See id. (concluding that Strickland analysis was “contrary to” clearly established law in part because the court’s analysis consisted of “irrelevant observation[s]”). The only potential prejudice analysis in the post-conviction court’s opinion therefore suggests that the court did not actually place the powerful new mitigating evidence in the context of evidence established at trial and reweigh the totality of available mitigating evidence against the evidence in aggravation. Williams, 529 U.S. at 397-98, 120 S.Ct. 1495.
As a final note, the post-conviction court was not excused from its obligation to apply Strickland because the same judge presided over both Mann’s trial and post-conviction proceeding, and that judge concluded that the newly introduced evidence would not have changed his mind. Strickland’s prejudice analysis turns on a reasonable decisionmaker, not on the “idio-syncracies of [a] particular decisionmaker.” 466 U.S. at 695, 104 S.Ct. 2052. Indeed, it would be odd to give weight to a single decisionmaker in this context, given that judge sentencing in capital cases is unconstitutional. Ring, 536 U.S. at 609, 122 S.Ct. 2428.
Additionally and significantly, in states like Arizona, where the state supreme court “conducts an independent review of the aggravating and mitigating factors, reweighing them afresh,” Correll v. Ryan, 539 F.3d 938, 951 (9th Cir. 2008), the post-conviction court must assess whether there is a reasonable probability that “the sen-tencer — including an appellate court, to the extent it independently reweighs the evidence — would have concluded that the balance of aggravating and mitigating factors did not warrant death.” Strickland, 466 U.S. at 695, 104 S.Ct. 2052 (emphasis added). The fact that the post-conviction court spoke in terms of absolutes rather than degrees of probability thus provides additional evidence that the court failed to “apply Strickland to assess” prejudice by reweighing the evidence from the perspective of a neutral and reasonable sentencer.
C
In sum, the record demonstrates that the post-conviction court’s prejudice analysis was contrary to clearly established federal law in two respects. Because the state court’s decision is not entitled to AEDPA deference, I would review Mann’s Strickland claim de novo.
Ill
Reviewing the Strickland claim de novo, one can only conclude that Mann was denied his Sixth Amendment right to effective assistance of counsel and is entitled to habeas relief. As we know, to prevail on a claim for ineffective assistance of counsel, a defendant must first establish that “counsel’s performance was deficient” and, second, that the “deficient performance prejudiced the defense.” Strickland, 466 U.S. at 687, 104 S.Ct. 2052. Mann satisfies both of Strickland’s prongs.
A
Because the post-conviction court only reached Strickland’s prejudice prong, we must review Strickland’s deficiency prong de novo. Porter v. McCollum, 558 U.S. 30, 39, 130 S.Ct. 447, 175 L.Ed.2d 398 (2009) (per curiam).3 Reviewing that prong de *1168novo, one can only conclude that counsel’s performance at sentencing was constitutionally deficient.
Counsel’s performance is measured by “an objective standard of reasonableness,” as determined by “prevailing professional norms.” Strickland, 466 U.S. at 687-88, 104 S.Ct. 2062. Although there are no “specific guidelines” for measuring counsel’s performance, Cullen v. Pinholster, 563 U.S. 170, 195, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011) (quoting Strickland, 466 U.S. at 688, 104 S.Ct. 2052), “general principles have emerged ... that inform our view as to the ‘objective standard of reasonableness’ by which we assess attorney performance, particularly with respect to the duty to investigate,” Summerlin v. Schriro, 427 F.3d 623, 629 (9th Cir. 2005) (en banc).
Generally, for the penalty phase of capital proceedings, prevailing professional norms require counsel “to conduct a thorough investigation of the defendant’s background.” Williams, 529 U.S. at 396, 120 S.Ct. 1495. As recognized in Strickland, the American Bar Association attorney standards serve as “guides” to determining what constitutes the requisite “thorough investigation.” 466 U.S. at 688-90, 104 S.Ct. 2052. At the time of Mann’s sentencing, those standards provided that counsel’s penalty phase investigation should “comprise efforts to discover all reasonably available mitigating evidence.” ABA Guidelines for the Appointment and Performance of Counsel in Death Penalty Cases 11.4.1(C) (1989).
Specifically, because evidence of mental impairment may place a defendant’s character in context, thereby reducing a defendant’s moral culpability, capital defense attorneys have a “duty to investigate and present mitigating evidence of mental impairment.” Bean v. Calderon, 163 F.3d 1073, 1080 (9th Cir. 1998). This standard practice was likewise reflected in the ABA guidelines at the time of Mann’s trial, which provided that capital defense attorneys should investigate the defendant’s “medical history.” ABA Guidelines for the Appointment and Performance of Counsel in Death Penalty Cases 11.4.1(D)(2)(C) (1989). Therefore, the “[fjailure to investigate a defendant’s organic brain damage or other mental impairments may constitute ineffective assistance of counsel.” Caro v. Calderon (Caro I), 165 F.3d 1223, 1226 (9th Cir. 1998).
As measured against these prevailing professional norms, counsel’s penalty phase investigation fell below an objective standard of reasonableness. The majority does not suggest otherwise.
Mann’s counsel never investigated the circumstances or consequences of the 1985 accident, and made no effort to investigate whether Mann sustained organic brain damage. The failure to do so is unsurprising given that Mann’s counsel admitted that he “was not focused on mitigation.” Indeed, counsel never even requested that a background investigation or any long term research be performed with regard to Mann. Counsel declined to pursue such avenues of mitigation relief despite the Capital Representation Project’s death *1169penalty expert’s recommendation to investigate whether Mann had “organic brain damage” and had been “self-medicating with drugs.”
Mann’s counsel also never attempted to obtain Mann’s medical records, whether relating to the 1985 accident or otherwise. Yet, obtaining “readily available documentary evidence such as ... medical records” is “fundamental to preparing for virtually every capital sentencing proceeding.” Robinson v. Schriro, 595 F.3d 1086, 1108-09 (9th Cir. 2010). Mann’s counsel recognized as much, and requested.a continuance to obtain “all the records” counsel could “from prior schooling, hospitalizations, [Mann’s] prior criminal record, and prior presentence reports.”
Finally, Mann’s counsel did not speak with Karen Miller, again, contrary to the recommendation of the Capital Representation Project. Counsel was aware, however, that Miller lived with Mann for ten years, including both before and after the 1985 accident; that she was someone who “care[d] about” Mann; and that she observed a change in Mann’s behavior during their relationship. Counsel recognized the importance of their relationship and invoked it during his sentencing argument, noting that Mann “has tried to maintain some kind of a relationship with [Miller]” and that she “even went to the jail a couple times to visit” Mann. Counsel was thus aware that Miller “could have provided important background information” about Mann, “providing leads for further investigation,” likely regarding the 1985 accident and its impact on Mann’s character. Robinson, 595 F.3d at 1110. Like other instances where counsel fails to “even take the first step of interviewing witnesses or requesting records,” counsel’s decision not to investigate the circumstances or consequences of the 1985 accident “did not reflect reasonable professional judgment” subject to this Court’s deference. Porter, 558 U.S. at 39-40, 130 S.Ct. 447.
If there could be any doubt that counsel’s actions fell below an objective standard of reasonableness, it is obviated by counsel’s concession at trial that such an investigation was the prevailing professional norm in the state of Arizona. See Wiggins v. Smith, 539 U.S. 510, 524, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003) (concluding that counsel performed deficiently where counsel acknowledged the relevant professional norm); see also Pinholster, 563 U.S. at 196, 131 S.Ct. 1388 (distinguishing Wiggins as a case in which “the defendant’s trial counsel specifically acknowledged a standard practice for capital cases ... that was inconsistent with what he had done”). When counsel requested a continuance, he told the court that a continuance was necessary so that counsel could fulfill his constitutional and state-law obligations to conduct a “comprehensive psychological profile and background investigation,” and obtain “all the records” available, including those from hospitalizations.4 Indeed, counsel conceded that the failure to conduct such an investigation *1170would constitute “reversible error.” Yet, whether due to time constraints or “inattention,” counsel never conducted the very investigation he described to the court as required by law. See Wiggins, 539 U.S. at 526, 123 S.Ct. 2527 (concluding that counsel’s “failure to investigate thoroughly resulted from inattention, not reasoned strategic judgment”).
In the post-conviction proceeding, counsel did not attempt to insulate his failure to investigate Mann’s medical history as a strategic decision. But, even if he had done so, cloaking a decision as “strategic” does not eliminate counsel’s “duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.” Strickland, 466 U.S. at 691, 104 S.Ct. 2052. It is often only after a preliminary investigation that “reasonably diligent counsel may draw a line [because] they have good reason to think further investigation would be a waste.” Rompilla v. Beard, 545 U.S. 374, 382-83, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005). After all, an “uninformed strategy is not a reasoned strategy. It is, in fact, no strategy at all.” Correll, 539 F.3d at 949.
Here, counsel did conduct a narrow mitigation investigation. He spoke with members of Mann’s family to develop evidence of Mann’s “unstable and abusive family background” and “positive relationships with his daughters and mother.” Counsel also requested a “neutral” psychological evaluation, provided by a person who was neither “defense [n]or State oriented.” However, “counsel’s duty to investigate all potentially mitigating evidence related to a defendant’s mental health [and] family background ... is not discharged merely by conducting a limited investigation of [those] issues.” Lambright v. Schriro, 490 F.3d 1103, 1120 (9th Cir. 2007) (en banc). Particularly where, like here, counsel “uncovered no evidence in [his] investigation to suggest that a mitigation case, in its own right, would have been counterproductive, or that further investigation would have been fruitless.” Wiggins, 539 U.S. at 525, 123 S.Ct. 2527.
Indeed, the failure to investigate Mann’s medical history could not be considered “strategic” in light of the mitigating evidence available to counsel. Mann’s biographical reference to the 1985 car accident and attendant concussion put counsel on notice of potential powerful mitigating evidence of organic brain damage, especially when coupled with the Capital Representation Project’s specific recommendation to investigate whether Mann suffered from brain damage. “In assessing the reasonableness of an attorney’s investigation ... a court must consider not only the quantum of evidence already known to counsel, but also whether the known evidence would lead a reasonable attorney to investigate further.” Id. at 527, 123 S.Ct. 2527. Any reasonable attorney would further investigate whether a car accident that left two out of three passengers dead resulted in any permanent physical or psychological damage to Mann, the sole survivor.
Counsel’s failure to investigate the 1985 accident and whether Mann suffered any organic brain damage was neither reasoned nor strategic; it was objectively unreasonable. Therefore, Mann’s counsel provided constitutionally deficient representation during the penalty phase of Mann’s trial.
B
Counsel’s failure to investigate and present mitigating evidence relating to the 1985 accident prejudiced Mann. If Mann’s counsel had conducted a reasonable investigation, the sentencing judge would have learned about Mann’s organic brain damage and ability to feel remorse, mitigators that could have affected the sentencer’s *1171view of Mann’s moral culpability. Indeed, “[m]ore than any other singular factor, mental defects have been respected as a reason for leniency in our criminal justice system.” Caro v. Woodford (Caro II), 280 F.3d 1247, 1258 (9th Cir. 2002) (citing 4 William Blackstone, Commentaries, *24-25). Given the impact that classic mitigating evidence would have had on Mann’s sentencing profile, the newly introduced evidence is more than “sufficient to undermine confidence” in Mann’s death sentence. Strickland, 466 U.S. at 694, 104 S.Ct. 2052.
To establish prejudice from counsel’s errors during the sentencing phase of a capital trial, the petitioner must show “a reasonable probability that, absent the errors, the sentencer ... would have concluded that the balance of aggravating and mitigating circumstances did not warrant death.” Id. at 695, 104 S.Ct. 2052. This requires the court to “couple the omitted evidence with the mitigating evidence presented at trial,” and assess whether the omitted evidence might have influenced the sentencer’s assessment of the defendant’s moral culpability. Caro II, 280 F.3d at 1256-57. Evidence of brain injury is a classic form of mitigating evidence and thus often alters the balance of aggravating and mitigating circumstances. Correll, 539 F.3d at 954.
Here, there was substantial evidence of organic brain damage that was not presented to the sentencing judge. According to post-conviction testimony, Mann lost consciousness for a few hours after the accident. Once in the hospital, Mann’s doctors, which may have included up to five neurologists, expressed concern about Mann’s head injuries. After the accident, Mann experienced symptoms associated with organic brain damage, including as-tro-projection and severe lingering headaches.
Mann’s post-conviction neurological expert, Dr. Comer, also concluded that Mann had “cognitive deficits consistent with head injury.” Dr. Comer effectively ruled out substance abuse as the cause of Mann’s cognitive deficits because the “cluster of weak test performances that [Mann] demonstrated tended to occur almost entirely on tests that had been well-documented to be sensitive to the subtle effects of head injury.” To further support that conclusion, both Mann and Miller presented evidence that it was only after the 1985 accident that Mann began using and selling cocaine, a conclusion consistent with Mann’s presentence report. This compelling evidence suggests that Mann suffered organic brain damage in the accident, from which, according to Dr. Comer, he did not “successfully] recover! ] emotionally and behaviorally.” The newly adduced mitigating evidence thus would have permitted the sentencer to more accurately assess Mann’s “character and record” — an “indispensable part of the process of inflicting the penalty of death.” Woodson v. North Carolina, 428 U.S. 280, 304, 96 S.Ct. 2978, 49 L.Ed.2d 944 (1976) (plurality opinion). Indeed, it “is precisely the type [of mitigating evidence] most likely to affect a [sentencer’s] evaluation of the punishment” Mann should have received. Caro I, 165 F.3d at 1227.
The mitigating evidence relating to the 1985 accident would have been of particular import in Mann’s case because Arizona courts “place significant weight on brain injuries as mitigating evidence.” Correll, 539 F.3d at 950 n. 3. Further, “a major personality change can amount to a mitigating factor in Arizona.” Smith v. Stewart, 140 F.3d 1263, 1271 (9th Cir. 1998); see also State v. Rockwell, 161 Ariz. 5, 775 P.2d 1069, 1079 (1989) (vacating death sentence in case where defendant’s “violent and unpredictable behavior began after [a] tragic [motorcycle] accident”). Thus, al*1172though counsel’s deficiency would be prejudicial in any case, it was particularly so here, where Mann’s capital sentence was automatically reviewed by the Arizona Supreme Court.
Not only did evidence relating to the 1985 accident potentially reduce Mann’s moral culpability, but that evidence .also undermined factual and legal findings made by the state trial court. The record developed in the post-conviction proceeding eradicated the two factual pillars upon which the sentencing court based its legal conclusion that Mann did not establish the non-statutory mitigating factor of remorse, thus per se altering the balance of aggravating and mitigating factors established at Mann’s trial.
At sentencing, the trial court concluded that Mann failed to establish the non-statutory mitigating factor of remorse. The court specifically noted Mann’s description of the 1985 accident contained in his biographical letter, and stated that in the letter Mann “indicate[d] no remorse” for the death of his two passengers. The court then concluded that Mann was “incapable of remorse.” That conclusion collapses in light of the post-conviction record.
The post-conviction record demonstrates that Mann was capable of remorse and, indeed, expressed significant remorse following the 1985 accident. The medical records following the accident describe Mann as experiencing a “grief reaction.” Mann described himself as “devastated” by the death of his passengers, especially the younger girl, whom he considered “like a sister.” Miller said that Mann entered a depression after the accident, and that the death of his passengers “destroyed him.” This evidence directly contradicts the sentencing court’s factual finding that Mann did not experience remorse for the death of his passengers following the 1985 accident.
The court’s reliance on psychological testimony was similarly flawed because the psychologist was not provided evidence indicating that Mann may have received a head injury and therefore lacked “the information necessary to make an accurate evaluation” of Mann’s neuropsychological health. Caro I, 165 F.3d at 1226-27. The psychologist’s diagnosis of anti-social personality disorder required ruling out organic brain damage, which he was unable to do because he lacked the critical information. The psychologist’s opinion that Mann was an anti-social “psychopath” with “no conscience” was founded on a mistaken assumption of Mann’s neuropsychological health. Thus, the post-conviction record entirely undermines the basis of the sentencing court’s conclusion that Mann was “incapable of remorse,” and its rejection of remorse as a non-statutory mitigating factor. Therefore, there is no question that Mann was directly prejudiced by counsel’s failure to investigate and present evidence of organic brain injury and Mann’s actual remorse about the accident.
In sum, the post-conviction evidence materially altered the sentencing profile that was presented to the sentencing judge and to the Arizona Supreme Court on direct review. The mitigating evidence introduced in the post-conviction proceeding was not “largely duplicat[ive]” of the mitigating evidence introduced at trial, or otherwise of “questionable mitigating value.” Pinholster, 563 U.S. at 200-01, 131 S.Ct. 1388. Rather, it was pure, undiluted mitigation, different in kind from the “lay background and character evidence” introduced at trial. Caro II, 280 F.3d at 1257. Thus, particularly in light of Arizona law at the time of Mann’s conviction, there is a reasonable probability that evidence relating to the 1985 accident would have altered the balance of aggravating and mitigating factors and resulted in Mann receiving a sentence *1173other than death. Counsel’s deficiency prejudiced Mann.
IV
For these reasons, I respectfully disagree that Mann is not entitled to relief on his claim of ineffective assistance of counsel at sentencing. Accordingly, I concur in part and dissent in part.
. Mann exhausted his Strickland claim by presenting it to the state post-conviction court and Arizona Supreme Court. The Arizona Supreme Court effectively adopted the post-conviction court’s flawed Strickland analysis by summarily denying Mann’s petition for review. Ylst v. Nunnemaker, 501 U.S. 797, 803, 111 S.Ct. 2590, 115 L.Ed.2d 706 (1991) ("Where there has been one reasoned state judgment rejecting a federal claim, later unexplained orders upholding that judgment or rejecting the same claim rest upon the same ground.”). Mann was not required to exhaust his arguments regarding the standard of review this Court should apply to that *1163Strickland claim. Indeed, Mann lacked "the right under the law of [Arizona] to raise” the question of the appropriate federal standard of review under AEDPA, as such an inquiry is only relevant in a federal habeas proceeding. See 28 U.S.C. § 2254(c) (requiring exhaustion of claims that can be adjudicated in the state court). Thus, because Mann properly exhausted his Strickland claim, we must assess whether the state court’s Strickland analysis was contrary to federal law and thereby determine whether to apply AEDPA deference to our review of Mann’s Strickland claim. See Williams v. Taylor, 529 U.S. 362, 397, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (concluding that state court analysis was "unreasonable application of” Strickland based on embedded legal error that was neither raised nor briefed in the state court).
. The majority suggests that the post-conviction judge, who also presided over Mann’s trial, did not apply the causal nexus test during Mann’s trial and thus likely did not do so during Mann's post-conviction proceeding. However, the appropriate standard of review under AEDPA turns on the legal analysis of the court that adjudicated the claim at issue, not analyses in prior and distinct proceedings. Moreover, during the sentencing phase of Mann’s trial, the prosecution repeatedly invoked the causal nexus test, arguing that unless a mitigating factor "ha[d] a direct causal tie” to the crime, the court did not "have to consider it.” Although brief, the trial court's sentencing memorandum suggests that it agreed with the prosecution and, in accord with binding Arizona precedent, applied the causal nexus test to Mann’s non-statutory mitigating evidence. Thus, consideration of the trial court record suggests that the post-conviction judge was aware of, and had previously applied in Mann’s case, Arizona's unconstitutional causal nexus test.
. The Supreme Court's recent statement that "§ 2254(d) applies when a 'claim,' not a component of one, has been adjudicated” does not require a different analysis. Harrington v. Richter, 562 U.S. 86, 98, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011). Here, unlike the summary denial in Richter, the court knows which component of Mann’s ineffective assistance of counsel claim — namely, prejudice— was addressed by the state court. This distinction between AEDPA review of summary de*1168nials and partial adjudications is apparent in post -Richter Supreme Court caselaw, which applies de novo review to unanalyzed portions of multi-prong tests. See, e.g., Brumfield v. Cain,-U.S.-, 135 S.Ct. 2269, 2282-83, 192 L.Ed.2d 356 (2015) (applying de novo review to unanalyzed portion oí Atkins claim). Indeed, in Brumfield, the Court analogized the partial adjudication at issue to Wiggins v. Smith, 539 U.S. 510, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003), where the state court’s reasoned decision was "premised solely” on one of Strickland’s prongs, and distinguished the case from Richter, where there was no "opinion explaining the reason relief [was] denied.” 135 S.Ct. at 2282-83.
. Counsel requested a continuance because it was not until Mann was convicted that counsel "kind of realized ... that it normally takes quite a few months to prepare a case for a sentencing.” This concession provides further evidence of counsel’s deficiency. "The failure to start the mitigation investigation until after the guilt phase flies directly in the face of the 1989 Guidelines, which directs an attorney to start preparing for sentencing as soon as he starts working on the case.” Jones v. Ryan, 583 F.3d 626, 644 (9th Cir. 2009), vacated for reconsideration in light of Pinholster, 563 U.S. 170, 131 S.Ct. 1388, 179 L.Ed.2d 557, by 563 U.S. 932, 131 S.Ct. 2091, 179 L.Ed.2d 886 (2011); see ABA Guidelines for the Appointment and Performance of Counsel in Death Penalty Cases 11.4.1(A) (1989) (stating that the penalty phase investigation should "begin immediately upon counsel's entry into the case”).