Opinion by Chief Judge KOZINSKI; Dissent by Judge REINHARDT.
OPINION
KOZINSKI, Chief Judge:With fifteen strokes of his knife, Richard Leavitt slashed and stabbed Danette Elg to death in her bedroom. Then, as Ms. Elg lay dying on top of her punctured waterbed, Leavitt hacked out her womanhood — -just as his ex-wife had seen him do to “play[ ] with the female sexual organs of a deer.” State v. Leavitt (Leavitt I), 116 Idaho 285, 775 P.2d 599, 602 (1989). We decide whether Leavitt’s lawyer rendered ineffective assistance of counsel while trying to have him acquitted of the death penalty.
Facts
Our first opinion in this case recounts the facts of Leavitt’s crime and trial. See Leavitt v. Arave (Leavitt III), 383 F.3d 809 (9th Cir.2004). We repeat only those relevant to this appeal. Jay Kohler and Ron Hart represented Leavitt at trial and sentencing. After the jury convicted Leavitt of murder, Kohler and Hart moved for appointment of a mental health expert to evaluate Leavitt for sentencing purposes. The trial court granted the motion and appointed Dr. David Groberg, a forensic psychologist, to perform the evaluation.
Dr. Groberg diagnosed Leavitt with antisocial personality disorder and intermittent explosive disorder. He reported that Leavitt was otherwise “of average intelligence with no serious deficits in his cognitive abilities.” Although he opined that these disorders rarely have a physiological cause, Dr. Groberg recommended that Leavitt receive neurological testing to be sure. Kohler and Hart moved for such an examination, which the trial judge granted.
Dr. Jaynes’s neurological examination of Leavitt revealed “no evidence of higher cerebral dysfunction” nor any “objective neurological deficit.” Nevertheless, Dr. Jaynes believed that Leavitt’s CT scan showed a “very slight cortical cerebral atrophy.... [that] may or may not have an effect on his cognative [sic] function.” Based on this finding, Dr. Jaynes suggested further testing. The trial judge denied the motion for an MRI, stating that additional mental health evidence would not be a significant factor in sentencing. At the conclusion of the hearing, the trial judge found that the aggravating factors outweighed the mitigating evidence and sentenced Leavitt to death.
David Parmenter then replaced Kohler and Hart as Leavitt’s counsel. Parmenter represented Leavitt in his appeal to the Idaho Supreme Court and succeeded in having the death sentence vacated. On remand, at the second sentencing hearing, Parmenter made a strategic decision to focus on convincing the judge that Leavitt was a “good guy” rather than pursue the mental health angle that had proven unsuccessful at the first sentencing. Despite this change in strategy, the trial court again sentenced Leavitt to death, and this time the state supreme court affirmed. State v. Leavitt (Leavitt II), 121 Idaho 4, 822 P.2d 523 (1991).
After exhausting his direct appeals and state collateral review, Leavitt petitioned for a writ of habeas corpus claiming Parmenter was ineffective for failing to investigate his mental health. The district court granted Leavitt’s request for the *608MRI that the state court had denied, and the experts found that it looked normal.
The district court dismissed Leavitt’s claims as procedurally defaulted, but we reversed and remanded for the district court to consider Leavitt’s ineffectiveness claims. Leavitt III, 383 F.3d at 814. On remand, the state moved for an additional MRI because the defense never disclosed the results of its earlier test. The new MRI showed white matter hyperintensities (WMHs) in Leavitt’s brain, which could indicate an organic cause of his personality disorders. Based on this evidence, the district court concluded that Parmenter had been ineffective in failing to investigate Leavitt’s mental health before the second sentencing hearing, specifically in failing to renew the request that the court obtain an MRI. The district court granted a conditional writ of habeas and the state appeals.
Analysis
We review de novo the district court’s grant of his petition for writ of habeas corpus. See Martinez-Villareal v. Lewis, 80 F.3d 1301, 1305 (9th Cir.1996). “To the extent it is necessary to review findings of fact made in the district court, the clearly erroneous standard applies.” Silva v. Woodford, 279 F.3d 825, 835 (9th Cir.2002). Clear error review is “significantly deferential” and requires us to accept the district court’s findings absent a “definite and firm conviction that a mistake has been committed.” Rhoades v. Henry, 596 F.3d 1170, 1177 (9th Cir.2010) (quoting Silva, 279 F.3d at 835) (internal quotation marks omitted). Because Leavitt filed his original habeas petition in the district court before the effective date of AEDPA, its provisions do not apply. Alcala v. Woodford, 334 F.3d 862, 868 (9th Cir.2003). To establish ineffective assistance of counsel, Leavitt must show both that his counsel’s performance was objectively deficient and that it prejudiced his sentencing. See Edwards v. Lamarque, 475 F.3d 1121, 1126 (9th Cir.2007) (en banc) (citing Strickland v. Washington, 466 U.S. 668, 687, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984)). We review the district court’s determination as to both of these issues de novo. United States v. Birtle, 792 F.2d 846, 847 (9th Cir.1986).
1. Deficient Perfomance
Judicial scrutiny of counsel’s performance is highly deferential. Strickland v. Washington, 466 U.S. 668, 690, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) (“[Counsel is strongly presumed to have rendered adequate assistance and made all significant decisions in the exercise of reasonable professional judgment.”). A defense lawyer must make reasonable investigations that, at a minimum, permit informed decisions about how best to represent his client. Sanders v. Ratelle, 21 F.3d 1446, 1456 (9th Cir.1994). But “strategic 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.
Here, Parmenter made a thorough investigation in preparation for the sentencing hearing. He met with Leavitt just a few days after he was appointed. He discussed the case with prior counsel and reviewed all transcripts and records from the prior proceedings. He spoke with Leavitt’s mother and father “many more than 25” times. He interviewed Leavitt’s brothers and sister, and “had many conversations” with Leavitt’s ex-wife. Parmenter also interviewed several prison guards to gather information about Leavitt’s behavior while incarcerated. This case thus does not present the typical capital case ineffectiveness situation where counsel scrambled to prepare just before *609the penalty phase, or failed to investigate entire areas of mitigation. See, e.g., Hamilton v. Ayers, 583 F.3d 1100 (9th Cir.2009).
Petitioner nonetheless argues Parmenter was ineffective for failing to gather additional mental health evidence for the second sentencing hearing. We reject this argument because the decision to forego further investigation into Leavitt’s mental health condition was reasonable in light of counsel’s knowledge of what had transpired at, and in preparation for, the initial sentencing hearing.
To begin with, Parmenter knew that Kohler and Hart had already spent significant time investigating mental health evidence during the course of their representation. All of Leavitt’s prior mental health records, two from licensed psychologists and one from a psychiatrist, diagnosed him with a personality disorder, without suggesting any possibility of an organic brain injury. Dr. Groberg, a psychologist retained specifically to gather mitigation evidence, also diagnosed Leavitt with a personality disorder. Original counsel requested a second expert, Dr. Jaynes, to conduct more testing. Leavitt’s EEG came back completely normal, but the CT scan revealed a “very slight” cortical atrophy that suggested a “possibility” of disease. But even Dr. Jaynes, who identified this abnormality, concluded that there was no “objective neurological deficit on examination” nor any “evidence of higher cerebral dysfunction.”
“While a lawyer is under a duty to make reasonable investigations, a lawyer may make a reasonable determination that particular investigations are unnecessary.” Babbitt v. Calderon, 151 F.3d 1170, 1173 (9th Cir.1998); see also Cullen v. Pinholster, — U.S. —, 131 S.Ct. 1388, 1406-07, 179 L.Ed.2d 557 (2011). Parmenter’s review of prior counsel’s efforts provided an informed basis for him to decide not to investigate further. The dissent dismisses this conclusion as “appellate factfinding.” Dissent at 619. But our job is not to divine Parmenter’s actual reasons for failing to seek additional testing, but rather to “affirmatively entertain the range of possible ‘reasons[Leavitt’s] counsel may have had for proceeding as [he] did.’ ” Pinholster, 131 S.Ct. at 1407 (emphasis added).
Leavitt argues that Parmenter could not have made a reasonable strategic decision to forgo further investigation because Dr. Jaynes’s findings constituted a promising lead that required follow up. Our recent decision in West v. Ryan, 608 F.3d 477 (9th Cir.2010), suggests otherwise. In West, we held that counsel was not deficient for failing to unearth additional mental health evidence where the doctor’s report contained purported “red flags.” Id. at 488-89. These flags included slow movement of his right hand, which could reflect neurological deficits. Id. at 489. Based on this finding, the doctor commented that he could not “rule[] out” a cognitive impairment absent further testing. Id. We held that this did not create a duty to investigate further because earlier tests revealed that the defendant’s “intellectual memory, language and perceptual functioning” were normal. Id. at 488. The doctor ultimately concluded that “the results ... [of the evaluation were] more consistent with an individual of low educational status ... than with any cognitive impairment.” Id. (alteration in original) (internal quotation marks omitted).
Similarly, Leavitt’s doctor may have found some evidence suggesting a cognitive impairment, but he ultimately concluded that the results — including his performance on personality, psychological and cognitive tests — were more consistent with a diagnosis of a personality disorder. As in West, we conclude that “such an equivo*610cal finding ... is not the kind of ‘powerful mitigating evidence’ sufficient to overcome Strickland’s presumption that counsel acted reasonably in declining to investigate further the possibility that [the defendant] might suffer from a cognitive impairment.” Id. (citing Bobby v. Van Hook, — U.S. —, 130 S.Ct. 13, 19, 175 L.Ed.2d 255 (2009)).
Second, even had Parmenter wanted to investigate further, he had good reason to believe a motion for another court-appointed doctor would be denied. Leavitt was not constitutionally entitled to a third court-appointed psychiatric expert under Ake v. Oklahoma, 470 U.S. 68, 105 S.Ct. 1087, 84 L.Ed.2d 53 (1985). Contra dissent at 621-22. Leavitt himself has never tried to argue otherwise, and with good reason: By its own terms, Ake “limit[ed] the right [it] recognize[d]” to “provision of one competent psychiatrist.” Ake, 470 U.S. at 79, 105 S.Ct. 1087 (emphasis added). Given this unambiguous language, we’ve held that the defendant “lacks the right to appointment of a second psychiatrist,” Pawlyk v. Wood, 248 F.3d 815, 824 (9th Cir.2001), even where the first psychiatrist is alleged to be incompetent or reaches a diagnosis unfavorable to the defense, see Harris v. Vasquez, 949 F.2d 1497, 1516-17 (9th Cir.1990). We’ve recognized that Ake’s “limitation to a single, independent psychiatrist is critical given that ‘[p]syehiatry is not ... an exact science, and psychiatrists disagree widely and frequently ... on the appropriate diagnosis.’ ” Pawlyk, 248 F.3d at 823 (quoting Ake, 470 U.S. at 80, 105 S.Ct. 1087) (alteration and first omission in original). Accordingly, neither we, nor the Supreme Court, has ever held that a trial court violated Ake by refusing to appoint a second, let alone third, mental health expert. E.g., Harris, 949 F.2d at 1516 (Ake did not require appointment of a third psychiatrist); see also, e.g., Granviel v. Lynaugh, 881 F.2d 185, 191 (5th Cir.1989) (Ake did not require appointment of an additional psychiatrist); Martin v. Wainwright, 770 F.2d 918, 934 (11th Cir.1985) (Ake did not require appointment of a second neurologist). Leavitt had not one, but two court-appointed experts, and so was not entitled to an additional evaluation.
Sensing the walls of precedent closing in on its conclusion, the dissent resorts to arguing that Leavitt was denied his right to “a competent psychiatrist who will conduct an appropriate examination.” Dissent at 621. First, it’s far from clear that such a right exists, see Vickers v. Stewart, 144 F.3d 613, 615 (9th Cir.1998), or that, if it does, we’d be able to review it on habeas, see Wilson v. Greene, 155 F.3d 396, 400-01 (4th Cir.1998) (refusing to recognize an Ake claim “solely based on whether [the first mental health expert] conducted an ‘appropriate’ examination”); cf. Harris, 949 F.2d at 1516-17 (refusing to recognize an Ake claim based on the argument that the first expert was incompetent). Second, there’s no indication that the examinations in this case were in any way inappropriate. The doctors reviewed Leavitt’s files, conducted a battery of psychological tests and administered both an EEG and a CT scan to detect neurological abnormalities. Due process does not require a state to fund every technologically conceivable test to rule out the possibility of an organic mental disorder. Third, even putting aside the issue of adequacy, Leavitt wasn’t entitled to additional testing because he couldn’t have “made a preliminary showing that his [mental health was] likely to be a significant factor” in his sentencing in light of the judge’s express indication to the contrary. Ake, 470 U.S. at 74, 105 S.Ct. 1087; see also Williams v. Stewart, 441 F.3d 1030, 1048 (9th Cir.2006) (upholding trial court’s denial of an expert under Ake because the defendant “failed to *611establish that his sanity was likely to be a significant factor in his defense”). Ake did not give Leavitt the right to another expert; had the state trial court granted him one, it would have been a matter of judicial grace, not constitutional right.
Given that Leavitt was not entitled to a third expert, the judge’s previous hostility towards appointing another doctor became all the more relevant in Parmenter’s development of a mitigation strategy. Parmenter knew about prior counsel’s failed attempt to obtain an MRI. Not only did the judge deny Kohler and Hart’s request, he had an order ready and waiting before they had even presented it. The order stated that “any further evidence of the mental condition of the defendant ... will not be a significant factor in the sentencing ... [and so the MRI] shall not be ordered.” Given the judge’s emphatic statement that mental health evidence would not be significant, it was perfectly reasonable for Parmenter to believe that he “might pretty routinely deny” a second request for an MRI. Counsel need not file motions that are likely to lose, because doing so may cost the defendant “some of his lawyer’s credibility with the judge.” Lowry v. Lewis, 21 F.3d 344, 346 (9th Cir.1994).
Preserving credibility was particularly important because the judge was not just presiding over the hearing, but deciding the ultimate issue of whether to impose the death penalty. Parmenter recognized that, in general, filing frivolous motions “may have some effect on [the judge’s] fact finding.” In this situation, it would have been reasonable for him to fear that renewing the motion would irritate the judge and hurt his client’s case. We review counsel’s decision solely to determine whether it fell within the “wide range of reasonable professional assistance.” Strickland, 466 U.S. at 689, 104 S.Ct. 2052. A decision to avoid annoying the judge — at least when it comes at the small price of forgoing the filing of a motion that was once denied and will likely be denied again — falls well within Strickland’s range of reasonableness.
Leavitt’s direct request to the sentencing judge for more psychological testing did not obligate Parmenter to renew the motion. “The decisions on ... what trial motions should be made, and all other strategic and tactical decisions[, is within] the exclusive province of the lawyer....” ABA Standards for Criminal Justice 4-5.2 (2d ed.1980) (emphasis added). In fact, because the judge indicated that he “was going to consider” Leavitt’s request, Parmenter could reasonably have believed it was unnecessary to submit a motion. Even with the benefit of hindsight, Parmenter testified that he wasn’t sure what new information he could have added to the original motion to convince the trial judge to change his mind. Parmenter reviewed all of Leavitt’s medical records and the judge’s denial of the original motion, and exercised his independent judgment that “it wouldn’t do much good to take another run at Judge George [for] additional testing.” Leavitt’s disagreement with counsel’s decision did not render it unreasonable.
Third, Parmenter’s decision to steer clear of the mental health issue was reasonable because the judge had already decided the mental health evidence was an aggravating factor. The judge stated that the personality disorder diagnosis was “not a mitigating factor, but rather a condemning factor [because i]t is the catalyst to provoke another possible homicide or serious physical injury.” Parmenter “didn’t want Judge George to have additional ammunition” and acknowledged that this concern “may have been part of the reason that ... Mr. Leavitt and I decided not to *612pursue that angle.” His concern was reasonable because “in some cases, presenting evidence of ... mental disorders to create empathy ... might actually cause ... worry and concern that the defendant is an ‘irreparable monster.’ ” Edwards v. Ayers, 542 F.3d 759, 776 (9th Cir.2008) (quoting John M. Fabian, Death Penalty Mitigation and the Role of the Forensic Psychologist, 27 Law & Psychol. Rev. 73, 90 (2003)). The trial judge had considered Leavitt’s mental health aggravating at the first sentencing, and it was thus reasonable for Parmenter to fear that he would treat it as aggravating again.
Instead of continuing with a mitigation strategy Parmenter knew had been rejected by the trial judge, he reasonably decided to switch gears. Parmenter’s goal was to humanize Leavitt by portraying him as something other than the monster the prosecution made him out to be. Parmenter wanted to develop a theme “that Rick Leavitt is a pretty good guy and not the kind of guy that should be put to death.” Evidence of mental health may have detracted from, or even conflicted with, this strategy. See Cox v. Ayers, 613 F.3d 883, 897 (9th Cir.2010). Parmenter acknowledged that raising a mental condition at mitigation “might [have] be[en] somewhat inconsistent with Mr. Leavitt’s defense at trial.” In Cox, we held that “counsel reasonably decided not to present, and not to look further for, evidence concerning Petitioner’s character and emotional state [where t]hat decision reflected counsel’s strategic choice to emphasize their primary argument at the penalty phase.” Id. Parmenter likewise made a conscious and informed decision to focus the mitigation case on portraying the defendant as a “good guy,” rather than try to excuse the crime by presenting evidence that he might have had mental health problems.
Counsel is not required to undertake all possible investigations. See Strickland, 466 U.S. at 690, 104 S.Ct. 2052. There will always be more documents that could be reviewed, more family members that could be interviewed and more psychiatric examinations that could be performed. But, as the Supreme Court recently reminded us, “[t]here comes a point where a defense attorney will reasonably decide that another strategy is in order, thus ‘makfing] particular investigations unnecessary.’ ” Pinholster, 131 S.Ct. at 1407 (quoting Strickland, 466 U.S. at 691, 104 S.Ct. 2052) (alteration in original); see also Bobby v. Van Hook, — U.S. —, 130 S.Ct. 13, 19, 175 L.Ed.2d 255 (2009) (per curiam) (holding that at some point, additional evidence would be only cumulative “and the search for it distractive from more important duties”). Parmenter had developed a knowledgeable foundation— from reading expert trial testimony, reviewing and marking up medical reports and meeting with Leavitt — on which to base his decision to take the mitigation case in a different direction.
While Parmenter’s chosen strategy failed, we must avoid the temptation to evaluate his decision through the “fabled twenty-twenty vision of hindsight.” Brown v. Uttecht, 530 F.3d 1031, 1035 (9th Cir.2008) (internal quotation marks omitted). We must evaluate his performance only based on whether he made reasonable, informed decisions based on what he knew at the time.
So what did Parmenter know once he took over the second sentencing? He knew that Leavitt had been examined by at least five mental health professionals, all of whom diagnosed him with personality disorders. He knew that the trial judge had summarily denied a request for more testing. And he knew that the judge who would ultimately decide whether to impose *613the death penalty didn’t consider such evidence to be mitigating. Having seen former counsel try and fail with a mental impairment mitigation strategy, it was not unreasonable for counsel to try a different tack. Such strategic decisions are precisely of the kind we vest in the discretion of informed, experienced trial counsel.
2. Prejudice
Even if we assume that Parmenter’s performance was deficient, Leavitt suffered no prejudice. Petitioner has the burden of showing a reasonable possibility that, but for counsel’s deficient performance, the death sentence would not have been imposed. Wong v. Belmontes, — U.S. —, 130 S.Ct. 383, 386, 175 L.Ed.2d 328 (2009). Where the defendant claims ineffective assistance for failure to file a particular motion, he must “not only demonstrate a likelihood of prevailing on the motion, but also a reasonable probability that the granting of the motion would have resulted in a more favorable outcome.” Styers v. Schriro, 547 F.3d 1026, 1030 n. 5 (9th Cir.2008). Because it is unlikely that Judge George would have granted the motion for additional testing, or that the results of any such testing would have changed the outcome of the sentencing, we cannot say the alleged deficiency was prejudicial.
Leavitt has not established a reasonable probability that the motion would have been granted had Parmenter presented it. Although Parmenter believed that Judge George might be “a little more wishywashy than other judges,” he also testified that the judge had “already made himself clear in the first sentencing hearing” and was therefore likely to deny the request. Parmenter might have supported a renewed motion with evidence of Leavitt’s behavior while imprisoned over the four years, but as the prison officers testified, Leavitt was a model prisoner who had no incidents of explosive behavior. This undermined the theory that he had an organic brain disorder that rendered him uncontrollably violent from time to time.
It would certainly not have been an abuse of discretion for the judge to deny the motion. See United States v. George, 85 F.3d 1433, 1437-38 (9th Cir.1996). As discussed above, the defendant had no right to an additional expert under Ake. See pp. 609-11 supra. Moreover, although the Idaho Supreme Court vacated the original sentence, it did not do so on the ground that there should have been more mental health testing. To the contrary, the court “[did] not disagree with the findings of the trial court that the defendant herein is possessed of an ‘intermittent explosive disorder’ ” or that the mental health evidence should be considered aggravating. Leavitt I, 775 P.2d at 608. It disagreed only with the “trial court’s misperception of the alternatives available to him” and authorized him, “in [his] discretion, ... [to] obtain additional information and/or testimony.” Id. (emphasis added). Having already appointed two experts, Judge George’s exercise of that discretion to deny the motion would have been appeal-proof.
Most damaging to Leavitt’s prejudice claim, however, is the fact that Judge George actually considered his request for additional testing but did not grant it. At the beginning of the second sentencing hearing, Leavitt told the judge that he “would like to have ... a[nother] psychological evaluation done on [him].” Judge George could have denied the request as untimely or required Parmenter to submit a motion as Leavitt’s counsel, but instead he said he would “consider what might need to be done.” At the close of the hearing, Judge George again indicated his intent to consider Leavitt’s request: “Mr. *614Leavitt in the beginning indicated that the court might consider ... some further psychological report. I’m not passing judgment on that right now. I’m certainly going to consider that.” A little over a month later, the court sentenced Leavitt to death. There is no reason to believe the judge didn’t take Leavitt’s request into consideration, as he promised he would. If the judge considered it, but did not grant it, Parmenter’s failure to file a formal motion to the same effect cannot have been prejudicial.
Even if the motion had been granted, Leavitt must show that technology was sufficiently advanced in 1989 to enable a doctor to detect the abnormalities. The district court acknowledged that the abnormalities in the 1996 MRI were overlooked by Leavitt’s doctor, and were not discovered until 2006, but ultimately credited his neurological expert who testified that a reasonable medical examiner would have been able to spot the abnormalities on an MRI in 1989. Because the district court’s resolution of the factual dispute was not clearly erroneous, we proceed on the assumption that the abnormalities would have been detected at that time had an MRI been performed. See Bonin v. Calderon, 59 F.3d 815, 823 (9th Cir.1995).
Starting with that assumption, “we [must] reweigh the evidence in aggravation against the totality of available mitigating evidence.” Wiggins v. Smith, 539 U.S. 510, 534, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003). Thus, we must examine what these abnormalities were and how they would have impacted the mitigation case. White Matter Hyperintensities, like those on Leavitt’s brain scan, are “simply bright spots that appear on an MRI.” But the WMHs, previously called UBOs, or “unidentified bred [sic] objects,” just don’t tell us very much. In trying to explain what the bright spots might mean, Dr. Bigler testified that WMHs represent a disruption in the normal flow of electrical impulses, which “can” affect behavioral regulation and “may” slow down brain responses. Dr. Beaver testified that he would “hypothesize” that the WMHs were an underlying cause of Leavitt’s personality disorders. Such opinions, which couch results in tentative language, are simply not enough to show prejudice. Rhoades, 596 F.3d at 1193 (finding no prejudice where expert reports “talk in terms of conditions that [the defendant] ‘likely’ has or ‘may’ have”); cf. Sears v. Upton, — U.S. —, 130 S.Ct. 3259, 3263, 177 L.Ed.2d 1025 (2010) (finding prejudice where “the expert’s opinion [of brain pathology] was unequivocal”).
The experts’ testimony was not only tentative, but also highly speculative. Dr. Bigler opined that the location of the WMHs “seems to have a bearing” on the type of problem manifested. Dr. Beaver then testified that because the bright spots are located in an area associated with emotion and behavior, they might have caused Leavitt’s violent outbursts. But Dr. Mar-tell testified, and the district court accepted, that brain injury anywhere in the brain could lead to violent behavior. And, although Dr. Bigler testified that WMHs are more likely to be present in people with psychological disorders, he conceded that they are also seen in the MRIs of some perfectly healthy people. This kind of speculative mitigation evidence is not entitled to significant weight. See Bible v. Ryan, 571 F.3d 860, 871 (9th Cir.2009) (finding no prejudice because argument that “brain dysfunction ... can be an explanation for violent behavior” was speculative).
Moreover, there’s a significant possibility that, had this highly speculative evidence been given any weight at all, it would have been treated as aggravating *615rather than mitigating. See Atkins v. Virginia, 536 U.S. 304, 321, 122 S.Ct. 2242, 153 L.Ed.2d 335 (2002). In Cullen v. Pin-holster, — U.S. —, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011), for example, the Supreme Court found that new evidence of an “organic personality syndrome” and “brain damage” was “by no means clearly mitigating, as the jury might have concluded that [the defendant] was simply beyond rehabilitation.” Id. at 1396-97, 1410. The Court concluded that, in light of the ambiguous nature of the brain injury, “[t]here [wa]s no reasonable probability that the additional evidence ... would have changed the jury’s verdict.” Id. at 1409. Here, too, there’s no way of knowing which way evidence of a biological mental impairment would have cut; it very well may have counted against Leavitt.
Even assuming the evidence of WMHs would have been treated as mitigating, we must discount its weight because any evidence of a brain dysfunction causing uncontrolled, sudden, violent impulses would not explain the most disturbing aspect of the murder here — the surgical mutilation of the victim’s body. Cf. Belmontes, 130 S.Ct. at 389. We agree with the district court that “Leavitt’s removal of Ms. Elg’s sexual organs is more strongly associated with simple depravity than a continuation of rage or anger.” Even had the trial judge been convinced by Leavitt’s experts that the repeated, violent stabbing of the victim could be explained by some abnormalities in his brain, the sexual mutilation, for which no expert testimony was available, pointed to a different cause altogether. See Mickey v. Ayers, 606 F.3d 1223, 1248 (9th Cir.2010) (finding no prejudice in counsel’s failure to adequately prepare mental health expert whose testimony “suffered from [the] fundamental weakness ... that a jury was unlikely to believe that a defendant suffering as [the expert] diagnosed could act as the facts of the crime showed [the defendant] did”); Pizzuto v. Arave, 280 F.3d 949, 962 (9th Cir.2002) (finding no prejudice in counsel’s failure to request neurological testing because “a[n organic brain] disorder could not account for, or have any bearing upon, the Herndon murders which the evidence demonstrates were premeditated, planned out, and part of a consecutive series of complex acts”).
The evidence is less weighty still because it merely adds to what had already been presented. See Bible, 571 F.3d at 871-72. The trial judge knew that Leavitt had a slight atrophy in his cerebral cortex. This was more than just a “lead,” it was itself evidence of some physiological problem. True, it was not clear exactly what the brain atrophy meant — Dr. Jaynes had testified that it “may or may not” lead to cognitive impairment — but neither is it clear what the WMHs mean. The WMHs are thus additional, cumulative evidence of the brain disorder the sentencing judge already knew Leavitt had. For the prejudice analysis, cumulative evidence is given less weight because it is not as likely to have affected the outcome of the sentencing. Babbitt, 151 F.3d at 1175.
We must look at the effect of the omitted evidence in light of all the mitigation evidence presented. Wiggins, 539 U.S. at 534, 123 S.Ct. 2527. The sentencing judge was aware of all the evidence, including mental health evidence, that was given at the first trial. The court already treated evidence of Leavitt’s intermittent explosive disorder as mitigating. To this, Parmenter added the testimony of several witnesses from Leavitt’s past, including family members who spoke very highly of him. Parmenter also presented the testimony of several prison guards, who testified that Leavitt had been a model prisoner in the intervening years, that he thrived in the structured prison environment and that he *616did not pose a threat of harm to other prisoners. Parmenter additionally introduced evidence that Leavitt was an accomplished artist and poet and had won awards for his talent. The sentencing judge also considered mitigating evidence that Leavitt was married, had reestablished contact with his son, had been steadily employed and had no prior felony convictions. The addition of an ambiguous result from Leavitt’s third mental health evaluation is not enough to raise a reasonable possibility that the outcome would have been different.
Given the exceptional depravity of this murder, it is unlikely that additional evidence of a brain abnormality would have made a difference. See, e.g., Woodford v. Visdotti, 537 U.S. 19, 25-26, 123 S.Ct. 357, 154 L.Ed.2d 279 (2002) (per curiam); Campbell v. Kincheloe, 829 F.2d 1453, 1464 (9th Cir.1987) (“[G]iven the overwhelming aggravating] factors and the heinous nature of the crime there is no reasonable likelihood that the jury’s verdict would have been different had mitigating evidence been introduced.”). The Supreme Court’s recent decision in Wong v. Belmontes, — U.S. —, 130 S.Ct. 383, 175 L.Ed.2d 328 (2009), is instructive. Belmontes was sentenced to death after bludgeoning his victim to death with a steel dumbbell during the course of a robbery. Id. at 384. The Supreme Court held that excluded evidence of “impairment of the neurophysiological mechanisms for planning and reasoning” was not prejudicial because it was “hard to imagine expert testimony ... outweighing the [gruesome] facts of [the] murder.” Id. at 389, 391. Belmontes’s crime was gruesome, to be sure, but it pales in comparison to Leavitt’s murder of Danette Elg:
Leavitt’s repeated and pitiless stabbing and cutting of his victim in all parts of her body, including even a thrust through her eye and into her brain, was vicious and remarkable enough for the most jaded reviewer of this genre of crimes. The added organ-removing mutilation of the victim “as part of the death dealing attack or as a grisly aftermath” is yet another marker of the unnecessary tortuousness of this crime.
Leavitt III, 383 F.3d at 837 (quoting Leavitt II, 822 P.2d at 526). The details of this crime are simply too atrocious for the exclusion of tentative, cumulative evidence to undermine confidence in the sentence.
The dissent would have us believe that an MRI would have revealed critical new mitigating evidence sufficient to shake our confidence in the judge’s imposition of a capital sentence. Dissent at 623-26. But this is simply not so. The state trial judge^ — who made the ultimate life or death decision — described the mitigating evidence as “feathers on the scale” when weighed against the heinousness and brutality of Leavitt’s crime. Reweighing all the evidence presented at the second sentencing hearing, the additional feather provided by the MRI evidence would not have been nearly enough to tip the scale in Leavitt’s favor, so there was no prejudicial error.
One definition of insanity is repeating the same course of action twice and expecting a different result. Parmenter’s decision to cease further investigation into Leavitt’s already heavily analyzed mental health was entirely rational. Leavitt has not made out his claim that Parmenter’s assistance was constitutionally deficient. Even if he had, the gruesome nature of the crime, coupled with the relatively weak additional evidence that might have been revealed had an MRI been granted, leads us to conclude that any ineffectiveness was not prejudicial.
REVERSED.