OPINION

GRABER, Circuit Judge:

Petitioner George H. Wharton appeals the district court's denial of habeas relief in this capital case. Police officers arrested Petitioner after finding the body of his live-in girlfriend stuffed in a barrel in their kitchen. Petitioner admitted killing her but claimed at trial, in California state court, that he had been provoked into the killing and, therefore, was guilty only of second-degree murder. The jury disagreed and convicted him of first-degree murder. In this habeas proceeding, Petitioner asserts that his due process rights were violated when jurors saw him shackled and that his trial lawyer provided ineffective assistance. We affirm the district court's denial of those claims. The district court correctly held that, although some jurors occasionally saw Petitioner in shackles while being transported through the halls of the courthouse, those sporadic sightings outside the courtroom did not rise to the level of a constitutional violation. The district court also correctly held that Petitioner's trial lawyer chose a constitutionally permissible guilt-phase strategy of forgoing certain defenses for fear of opening the door to the jury's learning about Petitioner's significant criminal history, which included a prior murder and rape.

At the penalty phase, the prosecutor introduced evidence of Petitioner's earlier convictions for both murder and rape, and Petitioner introduced evidence of mental illness and of his physically abusive and deprived childhood. The same jury deliberated over the course of three days but ultimately returned a verdict of death, which the trial judge imposed. Petitioner now claims that his lawyer provided ineffective assistance in investigating and presenting the mitigation evidence. We affirm in part and vacate in part the district court's denial of Petitioner's ineffective assistance of counsel claim at the penalty phase. The district court held that no prejudice resulted from the failure of Petitioner's trial lawyer to call Petitioner's half-brother, Gerald Crawford, as a witness. Accordingly, the court declined to decide — or make the necessary factual findings related to — Petitioner's claim that his trial lawyer was ineffective in investigating what Crawford knew. We hold that, if Crawford was available to testify or otherwise provide evidence, and trial counsel was ineffective in his investigation, then Petitioner has demonstrated prejudice. We vacate that portion of the district court's judgment and remand for further proceedings, including factual findings related to the investigation of Crawford.

FACTUAL AND PROCEDURAL HISTORY

The California Supreme Court, whose factual findings are "entitled to a presumption of correctness," Rhoades v. Henry, 598 F.3d 495, 500 (9th Cir.2010) (internal quotation marks omitted), described the evidence at the guilt phase as follows:

People v. Wharton, 53 Cal.3d 522, 280 Cal.Rptr. 631, 809 P.2d 290, 299-301 (1991) (citations omitted).

In light of the overwhelming evidence that Petitioner killed Smith, there was little hope of an acquittal on all charges. Petitioner's trial lawyer, William Duval, sought to convince the jury that Petitioner was guilty of only second-degree murder or manslaughter. Duval argued that Petitioner lacked the malice required for a first-degree murder conviction because Petitioner's actions were the result of provocation. See People v. Williams, 71 Cal.2d 614, 79 Cal.Rptr. 65, 456 P.2d 633, 638 (1969) ("Evidence of adequate provocation overcomes the presumption of malice."). The jury was unpersuaded and returned a verdict of guilt on the first-degree murder charge after deliberating for little more than a day.

The same jury then heard evidence in a separate penalty phase. The jury had not learned during the guilt phase about Petitioner's 1975 crimes of murder and forcible rape. During the penalty phase, those crimes were the focus of the prosecutor's case in aggravation. The California Supreme Court described the penalty-phase evidence:

Wharton, 280 Cal.Rptr. 631, 809 P.2d at 301-02.

On the third day of deliberations, the jury returned its verdict of death. The California Supreme Court affirmed the conviction and sentence. Id. 280 Cal.Rptr. 631, 809 P.2d at 299. The United States Supreme Court denied certiorari. Wharton v. California, 502 U.S. 1038, 112 S.Ct. 887, 116 L.Ed.2d 790 (1992).

Petitioner then filed this habeas action. The district court stayed the case pending exhaustion of state remedies. The California Supreme Court summarily denied habeas relief. The district court then granted an evidentiary hearing on, among other things, the claims now on appeal: Petitioner's shackling claim and his claims of ineffective assistance of counsel at the guilt and penalty phases. After a lengthy evidentiary hearing in 2006, the court issued an order in 2009 denying the claims addressed by the evidentiary hearing. The court later denied all remaining claims in a separate order.

Petitioner's shackling claim, labeled claim 18, arises from the fact that Petitioner was tried in the historic courthouse in Santa Barbara, California. At the time of Petitioner's trial, the building's design required less than optimal arrangements for the transportation of prisoners. Petitioner arrived each morning in a prison bus and was led, in what witnesses described as a "chain gang," to a holding facility in the courthouse. Like the other prisoners brought to the courthouse on the bus, Petitioner was shackled both independently and to other prisoners while in transit. To reach the holding facility, the chain gang walked through the courthouse's public hallways — within sight of the public, including jurors who happened to arrive early for trial.

The district court found that, although some jurors occasionally saw Petitioner being transported in the chain gang, Petitioner was never shackled in the courtroom. The court held that the jurors' occasional sightings of Petitioner in shackles, outside the courtroom and while being transported with other prisoners, did not rise to the level of a due process violation.

Petitioner's claim of ineffective assistance of counsel at the guilt phase, claim 41 subclaim 4,1 concerns Duval's trial strategy of arguing the defense of provocation only and not also pursuing the available defenses of intoxication and mental health. Evidence of intoxication and mental disease is admissible to demonstrate that a defendant did not form the specific intent required for a first-degree murder conviction. See Cal.Penal Code § 29.4 (intoxication); id. § 28 (mental health). The district court held that Duval's trial strategy was constitutionally adequate largely because Duval reasonably feared that introducing evidence of intoxication and mental illness would have opened the door to the jury's learning about Petitioner's 1975 crimes of murder and rape.

Finally, Petitioner's claim of ineffective assistance of counsel at the penalty phase, encompassing claim 37 and claim 41, subclaims 16, 17, 19, 20, and 22, challenges the adequacy of Duval's investigation and presentation of Petitioner's case in mitigation. With three exceptions, the district court rejected Petitioner's theories on those claims because Duval's performance was constitutionally adequate. On two of Petitioner's subclaims — cultural mitigation and Petitioner's positive adjustment to prison — the court held that Duval provided ineffective assistance but that the resulting prejudice was very small and did not warrant relief. Finally, the district court declined to decide whether Duval performed deficiently in investigating the potential testimony of Petitioner's half-brother, Gerald Crawford. Instead, the court held that, even if Crawford had been available to testify, no prejudice resulted from the fact that he did not testify.

Petitioner timely appeals. The district court granted a certificate of appealability on the shackling claim. We ordered supplemental briefing on the claims of ineffective assistance of counsel mentioned above. Because the standard in 28 U.S.C. § 2253(c) is met with respect to those claims, we now grant a certificate of appealability on claim 37 and claim 41, subclaims 4, 16, 17, 19, 20, and 22.

STANDARDS OF REVIEW

"Because [Petitioner's] first federal habeas petition was filed before the effective date of the Antiterrorism and Effective Death Penalty Act of 1996 (`AEDPA'), pre-AEDPA standards apply to his claims." Hamilton v. Ayers, 583 F.3d 1100, 1105 (9th Cir.2009). We review de novo the district court's denial of habeas relief. Arnold v. Runnels, 421 F.3d 859, 862 (9th Cir.2005). We review for clear error the district court's factual findings. Buckley v. Terhune, 441 F.3d 688, 694 (9th Cir.2006) (en banc).

DISCUSSION

A. Shackling While Being Transported

Petitioner argues that jurors' viewing of him in shackles while being transported deprived him of a fair trial under Deck v. Missouri, 544 U.S. 622, 125 S.Ct. 2007, 161 L.Ed.2d 953 (2005). "[T]he Fifth and Fourteenth Amendments prohibit the use of physical restraints visible to the jury absent a trial court determination, in the exercise of its discretion, that they are justified by a state interest specific to a particular trial." Id. at 629, 125 S.Ct. 2007. Three reasons support that rule: the need for a defendant to assist counsel, "[t]he courtroom's formal dignity," and the presumption of innocence. Id. at 630-31, 125 S.Ct. 2007. In the absence of a particularized determination that shackling is justified, visible shackling in the courtroom is "`inherently prejudicial.'" Id. at 635, 125 S.Ct. 2007 (quoting Holbrook v. Flynn, 475 U.S. 560, 568, 106 S.Ct. 1340, 89 L.Ed.2d 525 (1986)). That is, "where a court, without adequate justification, orders the defendant to wear shackles that will be seen by the jury, the defendant need not demonstrate actual prejudice to make out a due process violation. The State must prove `beyond a reasonable doubt that the shackling error ... did not contribute to the verdict obtained.'" Id. at 635, 125 S.Ct. 2007 (brackets omitted) (quoting Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967)).

Deck concerned visible shackling in the courtroom. We have held that visible shackling outside the courtroom — at least when the viewing is brief and accidental — is not inherently prejudicial; instead, a due process violation occurs only if the criminal defendant demonstrates actual prejudice. See Williams v. Woodford, 384 F.3d 567, 593 (9th Cir.2004) (holding that one "juror's viewing of Williams in handcuffs with a coat draped over his handcuffed hands as he went to or from the courtroom was not inherently or presumptively prejudicial"); Ghent v. Woodford, 279 F.3d 1121, 1133 (9th Cir.2002) (holding that there was no inherent prejudice where "a few jurors at most glimpsed Ghent in shackles in the hallway and as he was entering the courtroom"); United States v. Olano, 62 F.3d 1180, 1190 (9th Cir.1995) (holding that "a jury's brief or inadvertent glimpse of a defendant in physical restraints is not inherently or presumptively prejudicial" where, "on the sixth day of trial, the jury briefly witnessed [the defendant] in handcuffs as he entered the courtroom"); Castillo v. Stainer, 983 F.2d 145, 148 (9th Cir.1992) (holding that, concerning "a brief and accidental viewing of the defendant in a corridor, chained [at the waist]," "[n]o harm that rises to a constitutional level is done by such an unintended, out-of-court occurrence"); United States v. Halliburton, 870 F.2d 557, 559-61 (9th Cir.1989) (holding that a "brief and inadvertent display of Halliburton in handcuffs" when "he was observed handcuffed to a codefendant by at least two jurors as the elevator doors opened" was not inherently prejudicial).

We explained long ago the reasons for the distinction between shackling in open court and shackling during transportation. "[E]ven the `most unsophisticated juror' knows that defendants may have to post bail and that some lack the resources to do this." Halliburton, 870 F.2d at 561 (quoting Dupont v. Hall, 555 F.2d 15, 17 (1st Cir.1977)). "`Under these circumstances we cannot think that the emotional impact of seeing the defendant in custody is necessarily hostile-it may be quite the reverse.'" Id. (alteration omitted) (quoting Dupont, 555 F.2d at 17). "`It is a normal and regular as well as a highly desirable and necessary practice to handcuff prisoners when they are being taken from one place to another, and the jury is aware of this.'" Id. (alteration omitted) (quoting United States v. Leach, 429 F.2d 956, 962 (8th Cir.1970)).

The distinction is also consistent with the Supreme Court's more recent reasoning in Deck, 544 U.S. at 630-31, 125 S.Ct. 2007. Unlike shackling in the courtroom, shackling during transport does not affect the defendant's ability to assist counsel during trial. Id. at 631, 125 S.Ct. 2007. Nor does it have any effect on the dignity of the courtroom; indeed, it could be perceived as increasing the dignity of the courtroom because a prisoner's shackles are removed for open-court proceedings. Id. Admittedly, visible shackling during transportation might affect the jury's perception of the presumption of innocence, id. at 630, 125 S.Ct. 2007, but that concern is mitigated greatly by the reasons discussed above — jurors know that, as a matter of routine, some defendants are in custody during trial and that security needs during transport demand restraints. In this case, too, Petitioner was part of a group of prisoners being moved through the courthouse; he was not singled out.

Here, after conducting a lengthy evidentiary hearing, the district court found that "Petitioner was not visibly restrained while in the courtroom" but that "at least some of the jurors observed Petitioner being transported through the courthouse's public areas while in visible restraints." At the evidentiary hearing, almost everyone testified that Petitioner did not appear in shackles in the courtroom in front of the jury: Petitioner's lawyer, William Duval; Petitioner's investigator, Craig Stewart; the bailiff; the state-court trial judge; the prosecutor; and Juror Reginald C.

Only two jurors contradicted that observation, both with ambivalent testimony. Juror George B testified that he saw Petitioner shackled in some form or another in the courtroom — suggesting at times that Petitioner's hands were chained to a table or to his waist, or that his legs were shackled. At other times, however, George B contradicted that testimony or hedged: "[I]t may be that I saw the actual chains in the hall and not in the courtroom and just assumed. I can't guarantee you I saw the chains in the courtroom." George B confirmed at the evidentiary hearing that his memory was "a little hazy."

Juror Shelley T testified that she saw Petitioner once in handcuffs. She generally testified that this occurred in the courtroom, as Petitioner was led either into or out of the courtroom. When pressed, however, she also hedged: "I could be wrong." Her memory of the trial was "pretty bad."

In light of the overwhelming testimony in support of the district court's factual finding and only weak testimony to the contrary, we conclude that the district court did not clearly err in finding that Petitioner was not shackled in the courtroom. See, e.g., Rodriguez v. Holder, 683 F.3d 1164, 1176 (9th Cir.2012) ("Although an appellate court or other reviewing body may find clear error in a fact-finder's credibility determination if a witness's story is contradicted by the evidence or is internally inconsistent or implausible, a factfinder may nevertheless credit one witness's testimony over another's if both have related coherent and facially plausible stories that are not contradicted by extrinsic evidence.").

The district court likewise did not clearly err in finding that "at least some of the jurors observed Petitioner being transported through the courthouse's public areas while in visible restraints." The witnesses testified consistently that prisoners arrived each morning in a bus and were led — shackled — through the courthouse's public hallways in a "noisy" chain gang. Because the courthouse is open and lacks separate hallways for transporting prisoners, jurors arriving early for trial easily could see a defendant in shackles. The presiding judge testified that the lack of separate hallways was an "ongoing problem." According to the prosecutor, the problem was "endemic" at the time. Additionally, two of the three testifying jurors stated that they saw Petitioner in shackles in the hallway. Juror Reginald C saw Petitioner in leg shackles "maybe once or twice" in the hallway in the mornings before trial started for the day, and Juror George B testified with confidence that he saw Petitioner shackled in the hallway.

The legal question that we confront, then, is whether some jurors' viewing of Petitioner being transported in shackles through the courthouse's public areas deprived him of a fair trial in violation of his due process rights.2 We agree with the district court that Petitioner's due process rights were not violated.

Because jurors saw Petitioner shackled only occasionally and only while being transported, Petitioner must demonstrate actual prejudice. See, e.g., Ghent, 279 F.3d at 1132-33 (requiring that the petitioner demonstrate prejudice where the petitioner "was transported to and from the courtroom in shackles and ... on some of these occasions jurors observed him under restraint"). Only three out of the twelve jurors testified. Shelley T testified that she saw Petitioner in shackles once. Reginald C saw Petitioner in leg shackles "[m]aybe once or twice." (Emphasis added.) George B testified that he saw Petitioner in shackles an unspecified number of times: "Between one and 20. I have no recollection." Those jurors saw Petitioner in shackles only while being transported in a group, outside the courtroom.3 The jurors' sightings of Petitioner were not so pervasive or harmful that we must presume inherent prejudice.

Nor has Petitioner demonstrated actual prejudice. As the district court found, "[t]here is no testimony indicating prejudice." Petitioner clearly was implicated in the death of Linda Smith; jurors likely understood that the transportation shackling was a regular part of his custody — just as it was for all the other prisoners being transported. Moreover, the fact that Petitioner was not shackled in the courtroom, even though he was shackled entering and exiting the courthouse, suggested that Petitioner was not a dangerous person. We agree with the district court that "Petitioner was not singled out for special treatment, hence he suffered no particular prejudice as a result of his treatment."

Petitioner makes much of a statement by one juror, while explaining his inability to remember precisely where he saw Petitioner shackled, that "[s]ince you've seen it [Petitioner in shackles] in the hallways, it's not going to make a major impression when you see the same thing in the courtroom."4 Petitioner reads this statement as somehow proving that the out-of-courtroom viewing was as prejudicial as an incourtroom viewing would have been. Read most naturally, however, the statement amounts to the reverse: the juror speculated that an in-courtroom viewing would have had only the same minimal effect as the out-of-courtroom viewing. In any event, the statement is pure speculation. As the district court found, jurors actually saw Petitioner shackled only outside the courtroom. We decline to upend established caselaw that recognizes legally significant differences in the effect on jurors of in-courtroom shackling versus out-of-courtroom shackling simply because of one juror's personal, hypothetical speculation. That juror's speculation falls well short of demonstrating actual prejudice.

We thus agree with the district court that Petitioner's shackling claim fails.

B. Ineffective Assistance of Counsel at the Guilt Phase

Petitioner next argues that his trial lawyer, Duval, was ineffective under Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), by choosing to present the provocation defense only and choosing not to present evidence of intoxication and mental illness. To prevail on an ineffective assistance of counsel claim, Petitioner must establish both deficient performance and prejudice. Id. at 687, 104 S.Ct. 2052. We undertake a "highly deferential" assessment of Duval's performance. Id. at 689, 104 S.Ct. 2052. Duval "is strongly presumed to have rendered adequate assistance." Id. at 690, 104 S.Ct. 2052. "[S]trategic choices made after thorough investigation of law and facts relevant to plausible options are virtually unchallengeable; and strategic choices made after less than complete investigation are reasonable precisely to the extent that reasonable professional judgments support the limitations on investigation." Id. at 690-91, 104 S.Ct. 2052.

Duval investigated the intoxication and mental-health defenses by consulting with many experts. At Duval's request, Drs. Donald Patterson, Michael Stulberg, Richard Steinberg, Robert Sbordone, and William Rack evaluated Petitioner. Duval also consulted with Dr. Ronald Siegel, who had met with Petitioner and gathered hair samples at the request of Duval's predecessor. He reviewed a copy of a report by Dr. James Wells, who had examined Petitioner at the request of the prosecution. Because Petitioner's memory was unclear concerning the murder, he arranged for the administration of sodium amytal, or "truth serum."

The investigation into the cocaine intoxication defense yielded little support for that theory. The district court found "a substantial hole" in the intoxication defense because "the exact date of the murder cannot be fixed" and a "taped jailhouse interview with Petitioner provides no support for it." See also Mayfield v. Woodford, 270 F.3d 915, 931 n. 17 (9th Cir.2001) (en banc) ("[J]uries are unlikely to favor defenses based on abuse of dangerous drugs in evaluating a defendant's culpability for violent behavior."). The district court concluded that there was "no value in pursuing an intoxication defense." We agree. Duval's decision not to present the intoxication defense passes scrutiny under Strickland, particularly in light of the fact, discussed in detail below, that pursuing the defense likely would have opened the door to evidence of Petitioner's past crimes.

By contrast to the intoxication defense, the investigation did yield some support for the mental-health defense, although the results were far from conclusive. Duval had available to him several medical experts who could have testified about Petitioner's mental health at the time of the murder. Petitioner had a long history of mental illness, and the psychiatrists concluded that Petitioner suffered from a variety of mental conditions, including schizophrenia, atypical impulse disorder, hallucinations, delusions, and paranoia.

Some of the mental-health evidence was in tension with the defense of provocation. For example, evidence that Petitioner had atypical impulse control tended to undermine Duval's strategy of showing that Petitioner's response to provocation was objectively reasonable. But we need not dwell on the merits of presenting a mental-health defense versus the strength of presenting the provocation defense alone. As the district court found, Duval had an independent and very strong reason not to introduce testimony by the mental-health experts: Duval was concerned that introducing their testimony would lead to the jury's learning about Petitioner's prior convictions for rape and murder.

Duval's concern was justified. Under well-established California law, a mental-health expert may be cross-examined about a patient's criminal history in order to impeach the expert. People v. Doolin, 45 Cal.4th 390, 87 Cal.Rptr.3d 209, 198 P.3d 11, 45 (2009); People v. Panah, 35 Cal.4th 395, 25 Cal.Rptr.3d 672, 107 P.3d 790, 853-54 (2005); People v. Osband, 13 Cal.4th 622, 55 Cal.Rptr.2d 26, 919 P.2d 640, 698 (1996); People v. Hendricks, 44 Cal.3d 635, 244 Cal.Rptr. 181, 749 P.2d 836, 838-39 (1988); People v. Nye, 71 Cal.2d 356, 78 Cal.Rptr. 467, 455 P.2d 395, 405-08 (1969); see, e.g., Hendricks, 244 Cal.Rptr. 181, 749 P.2d at 838 ("Other-crimes evidence may be used to impeach the testimony of an expert witness."); Nye, 78 Cal.Rptr. 467, 455 P.2d at 406 (holding that cross-examination of an expert witness, including concerning the defendant's prior crimes, is permissible on "the extent of his knowledge, the reasons for his opinion including facts and other matters upon which it is based, and which he took into consideration" (citation and emphasis omitted)). Given that Petitioner's prior crimes involved heinous acts, it was eminently reasonable for Duval to choose a trial strategy that avoided the serious risk that the jury would learn about those past crimes. See Mickey v. Ayers, 606 F.3d 1223, 1238-39 (9th Cir. 2010) (holding that the decision by the petitioner's lawyer not to introduce evidence of a mental-health defense was adequate in part because it "was likely to open the door to evidence of Mickey's deviant sexual behaviors"); Hendricks v. Calderon, 70 F.3d 1032, 1037 (9th Cir.1995) (holding the same because "the jury would have learned, during [the experts'] cross-examination, of Hendricks' other murders for which he was tried separately").

We reject Petitioner's arguments to the contrary. At the evidentiary hearing, Duval testified repeatedly — both in general and with respect to specific doctors — that he decided to forego a mental-health defense out of concern for exposing the jury to Petitioner's past crimes. He testified in general that "I wanted to defend [the case] in a manner that kept that [prior] murder and the rape, by the way, from the jury." Petitioner's prior murder conviction "posed some very serious problems in defending him," as Duval described:

The rape also troubled Duval: "[I]t was a brutal crime involving a ... 63-year old [sic: 61-year-old] woman. And to me it was almost as bad as the jury learning about the prior murder." Duval had recently defended a client at whose trial the jury learned that the defendant had committed a prior murder, so Duval "was real leery about that as a possibility." He believed that there was a "decent chance" of a second-degree murder verdict if he could "keep[] the jury from learning about the prior convictions."

Duval also testified repeatedly that, because of Petitioner's prior convictions, Duval feared calling as witnesses many of the specific doctors — Drs. Stulberg, Steinberg, Sbordone, and Patterson — who could testify about Petitioner's mental health: "Q. Why? A. Because that would be a way to get what I was most fearful of in front of the jury, or a possible way to get it there[, his prior convictions]." Indeed, Duval interrupted the line of questioning about Dr. Sbordone to clarify that the prior murder and rape were an overarching concern throughout the trial: "Look, the 1975 priors, the prior murder and the prior rape, were always on the table as far as I was concerned in defending this case. It made it a difficult case to defend because of those priors...." The district court was well within its parameters as a fact-finder to credit Duval's testimony, and Petitioner is plainly mistaken in asserting that Duval's strategic decision had nothing to do with the 1975 crimes.

Petitioner next argues that Duval conducted an inadequate investigation into mental-health issues and that, accordingly, counsel's decision to forego a mental-health defense is not entitled to deference. We disagree. As noted above, Duval hired many experts to assess Petitioner's mental health, and he reviewed additional reports and medical records. With the benefit of hindsight and the absence of the time pressure of preparing for trial, Petitioner points to many ways in which Duval could have followed up on alleged leads, provided more records to certain experts, or asked experts to conduct additional testing or analysis.

But "the Sixth Amendment does not guarantee the right to perfect counsel." Burt v. Titlow, ___ U.S. ___, 134 S.Ct. 10, 18, 187 L.Ed.2d 348 (2013). Duval's trial preparation was limited by time and resources. Given that a mental-health defense carried with it the considerable detriment of opening the door to Petitioner's prior convictions, Duval's already extensive investigation of that defense was more than adequate. Even if we considered his investigation "less than complete," we would conclude that "reasonable professional judgments" supported his decision not to investigate further. Strickland, 466 U.S. at 690-91, 104 S.Ct. 2052.

Finally, Petitioner argues that Duval could have filed an "in limine motion," before trial, to seek to exclude Petitioner's past crimes from evidence. But Petitioner has provided no authority at all suggesting that the trial court actually would have granted such a motion. Indeed, all the cases that we have found, cited above, point in the opposite direction: the trial court likely would have denied the motion to the extent that Duval sought to introduce testimony by mental-health experts. Similarly, Petitioner has not pointed to any evidence in the record suggesting that the trial judge here would have granted the motion.5 Given the exceedingly low likelihood of success, it was a reasonable decision for Duval to focus his limited time and resources on strategies with a greater likelihood of success than to file a motion that almost certainly would have been denied.

In sum, the district court correctly denied relief on Petitioner's claim of ineffective assistance of counsel at the guilt phase.

C. Ineffective Assistance of Counsel at the Penalty Phase

Petitioner argues that Duval did not provide constitutionally adequate assistance at the penalty phase. "To perform effectively in the penalty phase of a capital case, counsel must conduct sufficient investigation and engage in sufficient preparation to be able to present and explain the significance of all the available mitigating evidence." Correll v. Ryan, 539 F.3d 938, 942 (9th Cir.2008) (internal quotation marks and alterations omitted). Duval had "an obligation to present and explain to the jury all available mitigating evidence." Hamilton, 583 F.3d at 1113. He also "had a duty to conduct `a thorough investigation of the defendant's background.'" Id. (quoting Williams v. Taylor, 529 U.S. 362, 396, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000)). "It is imperative that all relevant mitigating information be unearthed for consideration at the capital sentencing phase." Caro v. Calderon ("Caro I"), 165 F.3d 1223, 1227 (9th Cir. 1999). "To that end, trial counsel must inquire into a defendant's social background, family abuse, mental impairment, physical health history, and substance abuse history; obtain and examine mental and physical health records, school records, and criminal records; consult with appropriate medical experts; and pursue relevant leads." Hamilton, 583 F.3d at 1113 (citations, internal quotation marks, and alteration omitted). "[S]trategic choices made after thorough investigation of law and facts relevant to plausible options are virtually unchallengeable; and strategic choices made after less than complete investigation are reasonable precisely to the extent that reasonable professional judgments support the limitations on investigation." Strickland, 466 U.S. at 690-91, 104 S.Ct. 2052. "In any ineffectiveness case, a particular decision not to investigate must be directly assessed for reasonableness in all the circumstances, applying a heavy measure of deference to counsel's judgments." Id. at 691, 104 S.Ct. 2052.

In addition to showing unprofessional judgment, Petitioner also must establish prejudice, by demonstrating "that there is a reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different. A reasonable probability is a probability sufficient to undermine confidence in the outcome." Id. at 694, 104 S.Ct. 2052.

Hamilton, 583 F.3d at 1131 (citations, internal quotation marks, and brackets omitted).

"A convicted defendant making a claim of ineffective assistance must identify the acts or omissions of counsel that are alleged not to have been the result of reasonable professional judgment." Strickland, 466 U.S. at 690, 104 S.Ct. 2052. In the certified claims on appeal, Petitioner asserts that Duval: (1) failed to investigate and present evidence of Petitioner's mental illness; (2) failed to present evidence of Petitioner's positive adjustment to prison; (3) failed to present testimony by potential mitigation witness Robert Short; and (4) failed to investigate and present testimony by potential mitigation witness Gerald Crawford.6

1. Failure to Investigate and Present Evidence of Petitioner's Mental Health

As discussed above, Duval made a strategic decision not to introduce evidence of Petitioner's mental health at the guilt phase in order to shield the jury from learning about Petitioner's prior convictions. But that consideration became moot at the penalty phase, because the prosecutor could — and did — introduce evidence of Petitioner's prior convictions as aggravating circumstances supporting the death penalty. As part of Petitioner's mitigation case, Duval introduced evidence of Petitioner's mental health through expert testimony by Dr. Patterson and Dr. Hamilton.7

Dr. Patterson saw Petitioner several times before the penalty phase and reviewed many reports by other treating and examining doctors over the span of a decade. He testified that Petitioner presented a "very complex" psychiatric case. Dr. Patterson found support for "a number of diagnostic possibilities," including "mental disorder," "personality disorder ... with a marked potential for aggressive behavior," "thought disorder or a more serious mental illness in the form of a schizophrenia," "schizophrenia" as evidenced by "auditory hallucinations," and "atypical impulse disorder."

Petitioner argues that Duval provided inadequate assistance at the penalty phase because Duval should have presented better evidence of Petitioner's mental illness. At the evidentiary hearing in district court, Petitioner presented testimony by Dr. Patterson, via a 2002 video deposition, and testimony by Dr. Richard Dudley, Jr., an expert hired by Petitioner in 1993. Dr. Patterson testified that, even though he requested all background information related to Petitioner, Duval did not provide him with several important reports and pieces of information, including doctors' reports, jail records, and background information about Petitioner. Had Dr. Patterson seen that information before his penalty-phase testimony, he would have testified that Petitioner had even more severe mental-health issues.

Dr. Dudley is a professor at New York University School of Law and also a practicing physician whose specialty is psychiatry. He examined a wide range of reports and information in preparation for the evidentiary hearing. He concluded that, at the time of the 1986 murder, Petitioner was "suffering from a major psychiatric disorder." The district court found, and the parties do not challenge, that, "[i]n Dr. Dudley's view, Petitioner suffered from `a borderline personality disorder' with cognitive defects." "As he passed through adolescence, he acquired a `schizoaffective disorder' with which he attempted to cope through substance abuse."

In assessing Petitioner's ineffective assistance of counsel claim, the district court carefully compared Dr. Patterson's testimony at the penalty phase with the habeas testimony by Drs. Patterson and Dudley. The court held:

We have conducted our own review, and we agree with the district court's analysis. All of the doctors who examined or treated Petitioner concluded that he suffered from severe mental illness; they just disagreed on the precise diagnosis. Petitioner has not explained how the specific diagnosis by Dr. Dudley in 2006 (or by Dr. Patterson in 2002) differed in any material way from Dr. Patterson's original diagnosis in 1987. Cf. Miles v. Ryan, 713 F.3d 477, 493-94 (9th Cir.) ("The newly uncovered portion of Petitioner's social history simply does not have significant mitigating value in view of what was already available to the sentencing judge."), cert. denied, ___ U.S. ___, 134 S.Ct. 519, 187 L.Ed.2d 374 (2013). We agree with the district court that "Petitioner's proposed expert testimony is only better than that which was actually presented and not significantly different in kind."

Petitioner's finer-grained arguments fail for the same reason. For example, Dr. Patterson may not have been the wisest choice for an expert witness; Duval likely could have provided Dr. Patterson with better supporting documentation; and Duval possibly could have done a better job of eliciting testimony from Dr. Patterson tying Petitioner's mental health more directly to the 1975 murder and rape. But Petitioner has not shown how a more diligent lawyer would have produced materially different expert testimony. Accordingly, Petitioner has not overcome the strong presumption that Duval provided constitutionally adequate assistance in presenting evidence of Petitioner's mental health.

2. Positive Adjustment to Prison

Petitioner next argues that Duval was ineffective by failing to introduce evidence of Petitioner's allegedly positive adjustment to prison, such as his history of only minor, nonviolent disciplinary violations while in prison. Petitioner's argument fails for the simple reason that Duval did introduce evidence of his positive adjustment to prison.

A custodian of prison records, Lucy Bross, testified that Petitioner was housed at Vacaville, a psychiatric treatment facility, from 1976 to 1980. On cross-examination, she testified that Petitioner had eight disciplinary violations while he was in prison: one for possession of marijuana and seven for failing to appear for work assignments and disobeying orders. On redirect, in response to questioning by Duval, she explained that the violations were minor and non-violent. From this evidence of only occasional, minor, and nonviolent offenses while in prison, the jury could infer that Petitioner would pose little risk of future dangerousness in prison.

On appeal, Petitioner does not argue that there were additional details that should have been brought to light or additional records that were not mentioned. Nor does he argue that introduction of the records themselves — as opposed to the custodian's description of them — would have aided his case. In any event, even if Duval could have introduced slightly better evidence, such as a mental-health expert's testimony of Petitioner's allegedly positive adjustment to prison, we fail to see how Duval was constitutionally ineffective for not dwelling on this aspect of the case: we agree with the district court that the evidence was "at best weakly positive."

3. Failure to Present Testimony by Robert Short

Petitioner argues that Duval provided ineffective assistance of counsel by failing to present testimony by Robert Short. Short was a childhood friend and neighbor of Petitioner's and, for a time, was married to Petitioner's sister Claudia. Petitioner told Investigator Craig Stewart, whom Duval had hired to investigate Petitioner's background, that Short might have helpful information. Stewart successfully contacted Short, interviewed him, and prepared a nine-page report for Duval.

The report contained information that was helpful to Petitioner's mitigation case. For example, the district court found that the report "supported the other witnesses' testimony about the chaos and abuse present in Petitioner's childhood home and added new details about Petitioner's mother." Those new details included that Petitioner's mother had a boyfriend approximately the same age as Petitioner, which Petitioner resented, and that she had a sexual relationship with a woman, which upset Petitioner. But the report also contained information that was harmful to Petitioner's mitigation case. The district court found that the report "added uncomfortable details about Petitioner's thefts from Mr. Short's clients, how he would frequent gay bars to prostitute himself, and about how, prior to his arrest, Petitioner told him that Linda Smith had died suddenly due to a concussion or a brain tumor." Additionally, "Short would have contradicted witnesses who stated that Petitioner drank and used drugs as a youngster."

After receiving the report from Stewart, Duval wrote on the top: "Great reading, but a disaster." At the evidentiary hearing, Duval explained — and the district court credited — that he made a strategic decision not to call Short because of the potential for testimony harmful to Petitioner's case. We conclude that Duval's strategic decision did not run afoul of Strickland. "[S]trategic choices made after thorough investigation of law and facts relevant to plausible options are virtually unchallengeable...." Strickland, 466 U.S. at 690, 104 S.Ct. 2052; see, e.g., Toliver v. McCaughtry, 539 F.3d 766, 775 (7th Cir. 2008) ("After conducting an investigation (or making a reasonable decision that investigation is unnecessary), counsel may make a legitimate strategic decision not to call a witness if he makes a determination that the testimony the witness would give might on balance harm rather than help the defendant." (internal quotation marks and brackets omitted)).

On appeal, Petitioner argues that the harmful testimony may not have come out during cross-examination and that the helpful testimony was very helpful. As an initial matter, even if we agreed with Petitioner's assessment that perhaps Short may have done more good than harm, Strickland demands more: Duval's strategic choice is "virtually unchallengeable." 466 U.S. at 690, 104 S.Ct. 2052. In any event, Petitioner's arguments fail on their own terms.

Petitioner makes much of the fact that Duval had no duty to give the report to the prosecution and therefore had no reason to think that the harmful testimony would come out. But Duval reasonably feared that the harmful testimony would come out through ordinary cross-examination. This is not a situation in which the harmful testimony concerned some subject far afield from the witness' testimony or the crime itself, such that it was unreasonable for Duval to seek to avoid the harmful testimony. The harmful subject areas related directly to either the basis of Short's testimony — his childhood and adult friendship with Petitioner — or the crime itself — the murder of Linda Smith. Cross-examination easily could have uncovered the harmful topics.

As for the helpful testimony, Short would have provided testimony, for example, about the sexual activity of Petitioner's mother and its effect on Petitioner. But Duval already introduced extensive testimony about Petitioner's family background and history. The helpful testimony would not have framed a new perspective on a critical issue in the case. Cf., e.g., Hamilton, 583 F.3d at 1135-36 ("It is difficult to imagine a more significant discrepancy than that between the portrait painted at the penalty phase of a man whose childhood was `unfortunate' but largely unmarred, and that of a child who was raised in the presence of incest, rape, and violence, suffered from mental illness, and was shuffled from home to home."); Boyde v. Brown, 404 F.3d 1159, 1177-78 (9th Cir.2005) (finding ineffective assistance at the penalty phase where the evidence suggested that the petitioner "had a normal, non-violent childhood" when, in fact, he had been "violently abused" and exposed to "physical and sexual abuse by [his] mother and stepfather").

Accordingly, Duval's decision not to call Short as a witness was "a judgment call within the range of competent counsel." Jackson v. Calderon, 211 F.3d 1148, 1157 (9th Cir.2000).

4. Failure to Investigate and Present Testimony by Gerald Crawford

Petitioner argues that Duval provided ineffective assistance of counsel by failing to investigate and present testimony by Petitioner's half-brother, Gerald Crawford.8 Petitioner argues that Crawford would have testified about facts not otherwise presented, most notably, sexual abuse by Petitioner's father and step-grandfather of many victims, including Petitioner himself, his grandmother, his sisters, and the family's pet dog.

The district court declined to decide whether Duval performed deficiently9 because, in the court's view, Petitioner had not established prejudice. If we agree with the district court's assessment of prejudice then we, too, need not reach the performance prong. For that reason, we begin by analyzing prejudice. To do so, we must weigh the evidence actually presented along with the evidence that Petitioner asserts a competent lawyer would have presented. Hamilton, 583 F.3d at 1131. "Prejudice is established if there is a reasonable probability that at least one juror would have struck a different balance between life and death." Id. (internal quotation marks omitted).

As detailed by the California Supreme Court, Wharton, 280 Cal.Rptr. 631, 809 P.2d at 301-02, and quoted in the factual background section above, both parties submitted extensive evidence at the penalty phase. The prosecution's focus was evidence of the 1975 murder and the separate 1975 forcible rape. Duval's primary mitigation evidence concerned the abysmal childhood that Petitioner had suffered, which was caused by, among other things, extreme poverty and repeated physical abuse by his step-grandfather, Big Daddy.

At the evidentiary hearing, the district court admitted Crawford's testimony from his 2002 video deposition. Some of Crawford's testimony merely repeated or added a few inconsequential details to the penalty-phase testimony already introduced by Duval. For example, Crawford could have verified, and possibly added a few details concerning, the poverty and physical abuse that Petitioner endured. Because those topics were adequately covered by the penalty-phase witnesses, we agree with the district court that Crawford's testimony on those subjects was "cumulative" and that its exclusion did not prejudice Petitioner's mitigation case. See, e.g., Pizzuto v. Arave, 280 F.3d 949, 956 (9th Cir.2002) ("Every weakness or discrepancy that Pizzuto now says should have been cited and argued at sentencing was already before the court."), amended, 385 F.3d 1247 (9th Cir.2004).

But some of Crawford's testimony was not cumulative. Crawford testified about three subjects that differed materially from the penalty-phase testimony: (1) Petitioner's cultural background;10 (2) his odd personal behavior and medical symptoms as a child and, most importantly; (3) extensive sexual abuse by Petitioner's father and step-grandfather.

Crawford testified at his 2002 video deposition that the Ku Klux Klan regularly set up burning crosses in Hammond, where he and Petitioner grew up. He testified that an African-American "wasn't nothing during that time." The police would not respond to calls by an African-American; reporting crimes to the police was ineffective, even in the case of a murder: "You know, if a black person got killed, he's just killed." No penalty-phase witness testified on the general subject of cultural oppression that African-Americans experienced in Louisiana at the time of Petitioner's childhood, so Crawford's testimony on this subject was not cumulative.11 We nevertheless agree with the district court's assessment that the mitigating value of this testimony is "slight."

Crawford also testified that Petitioner occasionally had trance-like spells, would talk to himself in what sounded like a foreign language, had bad headaches and dizzy spells, and sometimes cut himself on purpose. No penalty-phase witness testified on these topics, so the testimony would not have been cumulative. But we find the mitigating value of this testimony, standing alone, to be small. There was ample evidence of Petitioner's long-standing mental illness. Crawford's testimony does suggest that the mental illness arose early in Petitioner's life rather than, for example, solely from Petitioner's later abuse of drugs. So it is possible that the jury would have viewed Petitioner in a slightly more sympathetic light. But we are unpersuaded that Petitioner's odd childhood behavior, on its own, would have played more than a small role in the jury's deliberations.

The prejudicial effect of the two topics just discussed — cultural mitigation and Petitioner's childhood medical problems — was modest. But we reach a very different conclusion with respect to Crawford's testimony of sexual abuse. Crawford testified about sexual abuse ubiquitous in Petitioner's family. He testified that both Petitioner's father and step-grandfather raped Petitioner at a very young age.12 He also testified that Big Daddy molested two of Petitioner's sisters; beat and raped his wife when she was dying of cancer; and had sex with the pet dog, Lassie.

The jury did not hear any evidence of that extensive sexual abuse. Indeed, the only evidence of sexual abuse admitted during the penalty phase was the attempted molestation of Petitioner's mother by Big Daddy when she was a child — well before Petitioner was born — and an isolated instance of sexual abuse of Petitioner by a non-family member when he was 11 years old. The jury knew that Petitioner grew up in a poor family and had been physically abused. But they did not know that Petitioner had been raped — first by his father, then by his step-grandfather — and that his step-grandfather had raped Petitioner's sisters, grandmother, and pet dog.

Childhood sexual abuse can be powerful evidence in mitigation, particularly when it is not an isolated event. See Boyde, 404 F.3d at 1176 (holding that "the family history of sexual abuse [the petitioner] had known about growing up[] is the sort of evidence that could persuade a jury to be lenient"); cf. Wiggins v. Smith, 539 U.S. 510, 534-35, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003) (describing as "powerful" the mitigating evidence that counsel failed to find, which included "repeated rape during his ... years in foster care"). But see Schurz v. Ryan, 730 F.3d 812, 815 (9th Cir.2013) (rejecting the mitigating value of a bare allegation that the petitioner was "likely sexually abused by a priest"), petition for cert. filed, ___ U.S.L.W. ___ (U.S. June 6, 2014) (No. 13-10596); Samayoa v. Ayers, 649 F.3d 919, 929 (9th Cir.2011) (holding that failure to present evidence of a one-time possible childhood sexual abuse by an uncle did not prejudice the defendant, in light of the more powerful mitigating evidence actually presented to the jury). Evidence that Petitioner had been raped by both his father and his step-grandfather — the two father figures during his childhood — "could have engendered sympathy" from the jury. Boyde, 404 F.3d at 1180. Although the jury heard evidence that Petitioner grew up in a poor family and suffered physical abuse, the jury would have had no reason to think that intra-family sexual abuse had occurred — in stark contrast to Crawford's testimony about extensive sexual abuse not only of Petitioner himself but also of other family members. See id. at 1178 ("But the evidence [counsel] elicited from the parents suggested — in stark contrast to what [competent counsel could have presented] — that Boyde had a normal, non-violent childhood.").

Nor is sexual abuse unrelated to the crimes before the jury. Both of Petitioner's 1975 crimes — the aggravating factors presented by the prosecution — concerned sex. The 1975 forcible rape of Petitioner's 61-year-old neighbor clearly concerns sexual abuse, and the 1975 murder of Pierce occurred after Pierce propositioned Petitioner for homosexual sex, enraging Petitioner. See Wharton, 280 Cal.Rptr. 631, 809 P.2d at 301 (noting that Petitioner "admitted killing Robert Pierce after the latter solicited a homosexual act from him"). Obviously, the childhood sexual abuse that Petitioner suffered in no way excuses those crimes. But the evidence of childhood sexual abuse may have cast those past crimes in a different light. A juror may have seen the 1975 murder as stemming from repressed anger about the homosexual rapes of Petitioner by his father and step-grandfather. A juror also may have seen the 1975 rape — of a 61-year-old woman by a much younger Petitioner — as stemming from a childhood filled with rape and sexual abuse across generations, committed with impunity. Yet from the evidence actually presented at the evidentiary hearing, the jury had no information from which to consider those possibilities.

Even without the evidence of sexual abuse, and with the knowledge of Petitioner's terrible past acts, the jury struggled to reach a unanimous verdict. On the first day of deliberations, the jury sent the following note to the judge: "If the jury does not reach unanimous concurrence on any of the three verdicts — what will be the verdict?" The court responded that "there will be no verdict." At the end of the second day of deliberations, the jury sent the following note to the judge: "Judge Dodds: It appears we are at an impass[e] on reaching a unaminous [sic] decision as to the penalty in this case. What are your directions to us at this time?" The judge declined to respond to the note, after being informed by the bailiff that the jury wanted to continue deliberating. The jury returned its verdict shortly before 11:00 a.m. on the third day of deliberations. The jury's notes and the fact that it deliberated over the course of three days suggest that the verdict was not an easy one to reach. See Thomas v. Chappell, 678 F.3d 1086, 1103 (9th Cir.2012) (holding that deliberations lasting almost five days, in combination with the jury's requests for readbacks of testimony, "strongly suggest that the case was close").

In the final analysis, we conclude that, had Crawford testified, "`there is a reasonable probability that at least one juror would have struck a different balance' between life and death." Hamilton, 583 F.3d at 1135 (quoting Wiggins, 539 U.S. at 537, 123 S.Ct. 2527). Although evidence of sexual abuse is sometimes not enough to tip the scales, e.g., Schurz, 730 F.3d at 815; Samayoa, 649 F.3d at 929, the sexual abuse at issue here was personal to Petitioner, from more than one source (his father and step-grandfather), and extensive throughout Petitioner's family (rapes of Petitioner, his sisters, grandmother, and pet dog). Moreover, the jury did not reach its verdict easily, possibly out of recognition that, although his crimes were heinous, Petitioner himself clearly suffers from serious mental illness and came from a disadvantaged and abusive home. The totality of the evidence discussed above — cultural mitigation, childhood behavior, and sexual abuse — gives rise to a reasonable probability that the jury may not have rendered a verdict of death.

It is true that the aggravating factors against Petitioner were strong. But the missing sexual-abuse testimony could have mitigated those factors. Moreover, we have emphasized that relief must be granted "`even in the face of ... strong aggravating evidence' ... `if we cannot conclude with confidence that the jury would unanimously have'" reached the same decision, had it heard the evidence that competent counsel would have presented. Caro v. Woodford, 280 F.3d 1247, 1257 (9th Cir. 2002) (brackets omitted) (quoting Mayfield, 270 F.3d at 929). Accordingly, we conclude that, if Duval performed deficiently by failing to investigate and present testimony by Crawford, then Petitioner has established prejudice. We turn, then, to the question of performance.

Duval "had a duty to conduct `a thorough investigation of [Petitioner's] background.'" Hamilton, 583 F.3d at 1113 (quoting Williams, 529 U.S. at 396, 120 S.Ct. 1495). "It is imperative that all relevant mitigating information be unearthed for consideration at the capital sentencing phase." Caro I, 165 F.3d at 1227. The sexual abuse described by Crawford is plainly relevant information that competent counsel would want to discover during a background investigation. See Hamilton, 583 F.3d at 1113 (listing general categories of inquiry for a penalty-phase investigation, including "family abuse," required by Strickland).

As noted above, Duval had hired Investigator Craig Stewart to investigate Petitioner's background. Stewart visited Louisiana in order to interview Petitioner's family members and acquaintances. Consistent with Petitioner's right to a thorough investigation, Stewart wrote at the beginning of his notes a list of topics to ask Petitioner's family members, and the list included an entry for inquiring about sexual, as well as physical, abuse of Petitioner. At the evidentiary hearing, Stewart confirmed that sexual abuse, if any, was one of the topics he sought to learn about while in Louisiana.

On his trip, Stewart successfully interviewed many people. He often tape-recorded the interviews and, afterwards, either wrote a written report summarizing the interview or had the interview transcribed. For example, as noted above, Stewart wrote a nine-page, single-spaced report summarizing his interview with Robert Short, which he provided to Duval.

The record does not provide a clear answer to why Stewart's otherwise thorough investigation failed to uncover Crawford's highly relevant accounts of sexual abuse. Crawford was one of the family members whom Stewart sought to interview, in part because Petitioner had asked specifically that Stewart contact Crawford. Indeed, Stewart testified at the evidentiary hearing that "Gerald [Crawford] is very important. It was his brother. George asked me over and over, `Did you get in touch with Gerald?'"

Yet Stewart did not interview Crawford. The record contains neither a written report concerning Crawford nor a transcript from an interview with Crawford. And Stewart confirmed at the evidentiary hearing that he did not, in fact, interview Crawford.

The reasons why Stewart did not interview Crawford and the extent of Stewart's contacts with Crawford are the subject of a clear factual dispute. In the 2002 video deposition of Crawford, admitted at the evidentiary hearing, Crawford stated definitively that Stewart had never contacted him and that he would remember if he had. He further stated that, had someone contacted him about Petitioner's childhood, he would have provided the same information that he provided in the deposition. By contrast, Stewart's notes from the Louisiana trip contain a few sparse entries suggesting that Stewart did contact Crawford and that Crawford told Stewart that prison officials had used Petitioner as a "guinea pig" in drug tests while he was incarcerated. Stewart's notes are otherwise silent about Crawford. The notes do not reveal whether Stewart asked Crawford about topics such as sexual abuse, and they do not explain why Stewart failed to interview Crawford.

At the evidentiary hearing, Stewart testified that the gap in his investigative records concerned him: "It bothered me, both the state public defender and the state attorney general said it looks like you create memos and reports, and how come nothing was written on this. That bothered me." A few months before the evidentiary hearing in 2006, Stewart awoke one morning with a vague memory of having phoned Crawford. According to Stewart's early morning revelation, Stewart now remembered contacting Crawford: Crawford told him that he moved away from the family home at an early age, is older than Petitioner, had nothing to share about Petitioner's childhood, did not have any present-day contact with Petitioner, and knew only about Petitioner's having been used as a "guinea pig." Stewart "didn't know if [his recollection] was a dream or if it was whatever." So he called one of the lawyers on the case and told him that "I have some information that I woke up with. I know it sounds crazy." The lawyer confirmed that Crawford was indeed older than Petitioner and that Crawford had moved out of the house. Stewart testified that the lawyer's confirmation "prove[d] in my mind" that what he "seemed to recall" was not "fantasy or dream." The record also contains evidence that Crawford had reasons to dismiss Stewart's inquiries: Stewart recalls that there were outstanding warrants for Crawford's arrest in California, where Petitioner's trial took place, and Crawford had health problems that could have made it difficult for him to fly to California to testify.

Without factual findings on the extent of Stewart's efforts to contact Crawford and Crawford's availability to testify or otherwise supply evidence or leads to evidence, we cannot determine whether Petitioner has established deficient performance. For example, if Stewart never contacted Crawford at all, as Crawford testified, then Petitioner almost certainly has established deficient performance. The Warden has not offered any reason — strategic or otherwise — that would justify a decision simply not to contact Petitioner's half-brother. See, e.g., Hamilton, 583 F.3d at 1123 ("Counsel also acted deficiently in not contacting Hamilton's other sister, Carolyn, who could have provided the most poignant and revealing mitigating evidence, as her declaration demonstrates."); Correll, 539 F.3d at 946 (holding that defense counsel's investigation was deficient where he "was aware that a chaplain ..., Reverend Curry, might have been willing to testify on Correll's behalf, but the attorney never even attempted to contact Reverend Curry").

Even if Stewart did contact Crawford, as Stewart's notes suggest, the inquiry becomes whether there was a constitutionally sufficient effort to learn relevant information about Petitioner's upbringing. In this regard, it is important that Crawford is not a passing acquaintance — Crawford is Petitioner's half-brother and lived in the same house as Petitioner for many years when they were young. Moreover, Petitioner emphasized the importance of Crawford's testimony by repeatedly questioning Stewart as to whether he had interviewed Crawford yet. Stewart himself testified that Crawford was a "very important" person to the background investigation.

At this procedural stage, we cannot engage in depth with Petitioner's claim. The district court declined to make the preliminary factual findings necessary to evaluate this issue. Accordingly, we vacate the district court's decision on this claim and remand for further factual development and for the court's assessment, in the first instance, of whether Petitioner has established deficient performance. See, e.g., Reyes v. Brown, 399 F.3d 964, 965 (9th Cir.2005) (remanding for further fact-finding by the district court); see also United States v. Prieto-Villa, 910 F.2d 601, 602 (9th Cir.1990) ("Since the findings which would permit review of this determination are absent here, we remand for factual findings by the district court."). In particular, the district court should determine on remand whether Stewart contacted Crawford; and, if so, whether Stewart made sufficient efforts to find out what Crawford could say about Petitioner's childhood, whether Crawford denied having useful information, and whether Crawford would have made himself available as a witness or otherwise available to provide evidence or leads to evidence. In making those determinations, the court may take additional evidence at its discretion. If the court rules that Petitioner has established deficient performance and finds that Crawford would have made himself available to testify, the court should grant the writ with respect to the sentence.13 Otherwise, the court should deny relief.

CONCLUSION

We affirm the district court's denial of relief on Petitioner's shackling claim (claim 18), his claim of ineffective assistance of counsel at the guilt phase (claim 41, subclaim 4), and most subclaims of his claim of ineffective assistance of counsel at the penalty phase (claim 41, subclaims 16, 17, 20, and the Robert Short portion of subclaim 19). We vacate the denial of relief on Petitioner's claim of ineffective assistance of counsel with respect to the alleged failure to investigate and present testimony by Gerald Crawford (claim 37 and claim 41, subclaim 22 and the Crawford portion of subclaim 19). We remand to the district court for further proceedings consistent with this opinion.

AFFIRMED in part, VACATED in part, and REMANDED. The parties shall bear their own costs on appeal.