dissenting:
I concur in Part A of the majority opinion. I conclude, however, that Brodit’s counsel was ineffective under the standard established in Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), and that habeas relief is available for that constitutional injury. I *943therefore respectfully dissent from Part B of the majority opinion.
BACKGROUND
The majority opinion provides no factual background against which to evaluate the ineffective assistance claim. As I do not believe the contention can be adjudged outside of its factual context, I provide that context here.
Charging documents accused Brodit of a violation of California Penal Code section 288.5, alleging that he committed at least three lewd and lascivious acts with his stepniece Jane Doe.1 These incidents purportedly occurred while Brodit was staying in the home of Jane’s mother, Arcel, and stepfather, John Brodit, his brother, between June 12, 1992 and December 31, 1994. The course of the ensuing trial was considerably influenced by the set of special rules that California has evolved for use in child abuse prosecutions. The existence of these rules, which I agree are valid under our precedents and in light of our deferential role under 28 U.S.C. § 2254(d), makes defending such cases exceedingly difficult by greatly limiting the options available to defense counsel. For that reason, failing completely to investigate a critical line of defense available in child abuse cases constitutes, in my view, ineffective assistance of counsel.
A. The Trial
1. The Prosecution Case
At Brodit’s trial in April 1996, ten-year-old Jane testified that he had repeatedly engaged in anal and vaginal intercourse with her and had also initiated other sexual touching while living in her family’s home. Jane asserted that the abuse had almost always taken place at night in Bro-dit’s bed or in the bed she shared with her stepsister, while her stepsister was asleep.
Jane testified that she did not tell anyone about the abuse until January 1995, when she stayed home sick one day from school under the supervision of her maternal grandmother, Estrelita. According to Jane, she was watching television with Es-trelita when a public service announcement about “good touching” and “bad touching” came on, prompting Estrelita to ask Jane if anyone had touched her in a bad way. Jane then told Estrelita that Brodit had sexually molested her.
On cross-examination, Jane alleged that John Brodit, her stepfather, had physically abused her, and added that she did not like living with John. She also stated that her mother was angry at John. The cross-examination brought out inconsistencies between details of the abuse that Jane had recounted to police and her testimony before the court. Jane also admitted on direct and cross-examination that she had viewed parts of sexuahy-explicit movies with some of her young relatives.
Although there were no eyewitnesses to the abuse, the prosecution supported Jane’s account by presenting hearsay testimony from several witnesses pursuant to California Evidence Code sections 1253 and 1360, as discussed in the majority opinion. A Child Protective Services investigator, a police detective, and Jane’s grandmother, mother, and aunt, each testified at trial about what Jane had told them about the alleged abuse. Susan Kuhn, Jane’s therapist, testified as to what Jane said about the alleged abuse during their therapy sessions. Nancy Jo Elliff, a nurse-practitioner who had examined Jane for physical evidence of abuse, also re*944counted what Jane told her about the alleged abuse.
This hearsay testimony largely reiterated and supported the details of Jane’s testimony, but both direct and cross-examination revealed inconsistencies among the various accounts of abuse Jane had given and shed light on her possible motives to fabricate allegations of abuse.
In particular, therapist Kuhn testified that Jane disliked and feared her stepfather John. Jane had told Kuhn that John physically abused her on several occasions. Jane’s grandmother, Estrelita, and her mother, Arcel, testified that there was considerable marital discord between Arcel and John during the period in which the abuse allegedly occurred. The discord was so severe that Arcel and her children moved in and out of the family home several times. Estrelita’s testimony, as well as Arcel’s, also revealed that during the weekend immediately prior to Jane’s first airing of the abuse allegation, Arcel and John took a trip together to attempt to resolve their marital problems. Although Estrelita denied that she was perturbed by John and Arcel’s possible reconciliation, she admitted that she was concerned that the couple’s frequent fights made the children unhappy.
Estrelita testified that Jane told her about the abuse in the context of a discussion about Jane’s recurrent genital pain. When Estrelita noticed that Jane appeared upset during the discussion, Estrelita asked if anyone had hurt her. Jane then told Estrelita that Brodit had abused her.
To bolster Jane’s credibility in light of some inconsistencies between the hearsay accounts of her previous statements and her trial testimony, the prosecution presented the testimony of Dr. Theresa Schu-man, an expert on Child Sexual Abuse Accommodation Syndrome (CSAAS). Schuman stated that CSAAS describes various emotional stages experienced by sexually abused children that may explain an abused child’s sometimes piecemeal and contradictory manner of disclosing abuse.
Schuman testified that, in the context of CSAAS, inconsistencies in a child’s accounts of abuse do not necessarily mean that a child is lying. Such inconsistencies, Schuman asserted, could simply mean that a child is telling different parts of what happened to different adults, based on the child’s comfort level with each adult or the developmental immaturity of the child’s memory.
The prosecution presented minimal physical evidence of abuse. According to Elliff, the nurse-practitioner who examined Jane, Jane’s hymen was still intact and the size of her vaginal opening was within the normal range of measurements for a girl her age. Dr. James Carpenter, the physician who consulted with Elliff on Jane’s examination, testified that although he had initially thought that small bumps found in Jane’s vaginal area could be genital warts, he now believed that they could also be normal lymphoid follicles not indicative of sexual abuse.
Both Elliff and Carpenter testified that it was not unusual for sexual abuse victims to have normal genital examinations. On cross-examination, however, Carpenter admitted that “full penetration through a seven- to ten-year-old hymen usually does leave findings” of hymenal disruption, although the findings could be minor in nature. Elliff also conceded on cross-examination that while the examination findings were not inconsistent with abuse, they were also consistent with the absence of abuse.
2. The Defense Case
Brodit testified in his own defense, denying that he had ever abused Jane.
*945According to Brodit, he began living in John and Arcel’s home in June 1993, moved out in November 1993, and returned in August 1994. He offered a partial alibi by testifying that during late 1994 he worked a graveyard shift security job that took him away from John and Arcel’s home most nights, the time during which the abuse allegedly occurred.
Brodit further testified that in November 1993 he and Arcel fought over his view that her parenting skills were sub-par, and Arcel demanded that he move out. Brodit subsequently lived in the homes of two different Mends until August 1994, when he and Arcel resolved their differences and Arcel permitted him to move back into the family’s home. Toward the end of 1994, Brodit asserted, Arcel and John “constantly fought in front of their kids.” Brodit also stated that in early 1995, before Jane made her allegations, he spoke to Estrelita about his concern that Arcel’s children were adversely affected by their frequent moves in and out of the family home.
In response to questions about his sexual history, Brodit testified that he had sexual relationships with adult girlMends in 1992, 1993, and 1994. He also stated that he was involved with Arcel’s sister for a few months in 1991.
John Brodit testified on his brother’s behalf. According to John, Brodit lived continuously with the family from August or September of 1994 until he moved out in January 1995, soon after Jane aired her allegations. Brodit worked the graveyard shift as a security officer during most of those months and so was not around the house after 10 at night. John also testified that he kept X-rated videos in the family home and subscribed to cable channels that carried sexually explicit programming.
John depicted his marriage to Arcel as tumultuous. He recounted that, when the two were making plans to go away together for the reconciliation weekend in January, Arcel did not want the plans discussed in front of the children because she feared they would tell Estrelita. Despite this, John said, he and Arcel did discuss their plans in front of his stepdaughter Jane.
Mica Brodit, John’s daughter and Jane’s stepsister, testified that she and Jane routinely slept together in the bottom bunk bed in the children’s room at the family home, the same bed in which Brodit allegedly engaged in intercourse with Jane on multiple occasions. Mica stated that she woke up easily at night but never saw Brodit in the bunk bed. Nor did she ever see Jane sleeping in Brodit’s bed. In addition, Mica testified that she, Jane, and other young relatives had watched parts of sexually explicit movies together. Mica’s account was supported in this regard by the testimony of Brodit’s 12 year old nephew (Jane’s stepcousin), who recounted that he, Jane, and some other cousins watched a pornographic movie together on at least one occasion during the summer of 1994.
Defense counsel also presented character witnesses, including friends and relatives of Brodit who testified about Brodit’s relationships with adult women and his general trustworthiness and helpfulness. Two of Brodit’s friends testified that he lived with them and their young children in close quarters for several months in 1993 and 1994 without incident.
To attack the CSAAS testimony and possible physical evidence of abuse presented by the prosecution, the defense presented its own child sex abuse expert, Dr. Stewart Coleman. Coleman discussed various “red flags” that cast doubt on the reliability of a child’s account of sexual abuse, including whether the child first mentioned abuse in the context of an adult asking her if she had been abused, and whether the adult who solicited the child’s *946account of abuse “may have an agenda, may have an ax to grind, may have some feelings about certain people in the family.” Coleman testified that these “red flags” existed with respect to the accounts of abuse Jane gave various adults, noting in particular that there was family conflict and that Estrelita broached the topic of abuse with Jane. Coleman also addressed the physical evidence, asserting that the small size of Jane’s vaginal opening was inconsistent with repeated penetration.
During deliberations, the jury sent the court a note asking for clarification of El-liff s testimony regarding possible physical evidence of abuse. The record does not reflect whether the court responded. The jury deliberated from approximately the morning of April 25th to about 3 p.m. on April 26th, when it delivered its verdict of guilty. Brodit was sentenced to six years in prison and ordered to pay restitution, undergo HIV testing, and register as a sex offender for the rest of his life pursuant to California Penal Code section 290.
B. Prior Proceedings
1. State Proceedings
Brodit argued in the state courts that his trial counsel was constitutionally ineffective because he failed to present the psychological evidence authorized in People v. Stoll, 49 Cal.3d 1136, 265 Cal.Rptr. 111, 783 P.2d 698 (1989). See People v. Brodit, 61 Cal.App.4th 1312, 72 Cal. Rptr.2d 154 (1998) (published portion of California Court of Appeal decision).
Stoll held that a judge cannot exclude an expert opinion, offered by a psychologist who has examined and tested a defendant charged with committing lewd and lascivious acts upon a child, that shows the defendant’s personality does not fit the profile of a child molester. 783 P.2d at 707-14. The Stoll court reasoned that the testimony is admissible as character evidence tending to show that the defendant did not commit the crime of child molestation. Id. at 707-10.
Without holding an evidentiary hearing on Brodit’s ineffective assistance of counsel claim, the California Court of Appeal applied Strickland to hold that Brodit’s counsel made a reasonable strategic choice not to introduce any expert opinion that Brodit lacked the personality characteristics associated with child molesters. The court noted that defense counsel may have feared that proffering Stoll evidence would open the door to the admission of damaging character evidence in rebuttal. Even if the Stoll evidence had been introduced, the court concluded, there was no reasonable probability that the outcome of the trial would have differed. The court agreed with the state’s argument that it was “highly unlikely that the subjective opinion of one defense-retained psychologist ... that [Brodit] did not fit the profile of a ‘typical’ residential child molester would have changed the result in this case.” Brodit, 72 Cal.Rptr.2d at 167.
2. District Court Proceedings
On habeas review, the district court did hold an evidentiary hearing on the Stoll ineffective assistance claim. At the hearing, Brodit’s trial counsel, Robert Beles, testified as a witness for the state. Beles asserted that his trial strategy was to present Brodit as “a normal, regular guy,” and to show that Jane, “caught in the middle of a break up between John and Arcel,” had reason to fabricate a story of sexual abuse in order to “attempt[ ] to draw attention to herself in perhaps getting her parents back together or possibly strike back at her mother.”
Although Beles stated that he had presented witnesses who attested to Brodit’s good character and trustworthiness around *947young children, he added that he “shied away from the pure character question in general,” because he did not want the prosecution to introduce evidence concerning an affair Brodit had with Arcel while he was living at the family home, including evidence that Jane once saw the two having sex in a closet. Beles testified that such evidence might suggest to the jury that “[Brodit] is not a good person, and he might do some other bad things.” Beles tried to think of ways that evidence of the affair could be used to Brodit’s advantage — for example, to show that his sexual interest focused on adult women, not children — but concluded that the bad character shown by the affair “couldn’t be overcome by any technical points that we might try and make ... the whole case relied on Goodwin making a nice impression with the jury.”
Beles indicated, through his testimony and by alluding to the existence of some continuing education materials discussing Stoll evidence that he placed in Brodit’s case file before the trial, that he knew of the possibility of pursuing Stoll evidence. Fear of opening the door to evidence of the affair, Beles maintained, was the.primary motivating factor in his decision not to have a psychologist conduct an evaluation of Brodit to generate potential Stoll evidence for the trial. In Beles’s view, had he proceeded that way, evidence of the affair would have been admissible on the theory that Jane looked much like her mother. A jury could reasonably conclude, Beles feared, that when Arcel broke off her relationship with Brodit he moved on to her daughter.
Beles mentioned that he was also concerned about other unhelpful information that might have become relevant if a Stoll expert had testified, namely, Brodit’s membership in FERET, an organization devoted to finding missing children, which “could be misinterpreted.... We didn’t want anybody to think that Goodwin had an inordinate interest in small children.” Beles recognized, however, that evidence of Brodit’s FERET membership could, conversely, have been helpful.
Beles asserted that, in his experienced opinion, the judge would definitely have permitted presentation of the FERET and affair evidence. On cross-examination, however, Beles admitted that he had never filed any in limine motions or sought any ruling to suppress evidence of the affair. He vaguely recalled that he may have agreed with the prosecutor in a pre-trial conference that information about the affair would not be introduced unless the door was opened to this type of rebuttal evidence, but the conference was not on the record and Beles did not remember clearly whether or not there was an agreement.
Experienced defense attorney Kathleen Coyne testified on Brodit’s behalf as a Strickland, and Stoll expert. Coyne stated that Stoll evaluations are routinely employed by defense practitioners handling child sex abuse cases. She stressed that the prosecution’s evidence against Brodit was thin. Stoll evidence would therefore have been “extraordinarily potent,” because testimony that “[Brodit] does not possess the character traits of deviants [sic] necessary to commit the offense is something that has an enormous comfort factor for a jury, and is some evidence which should always be considered in any case where the client’s denying the guilt.”
Coyne further opined that Beles had an incorrect understanding of the law regarding the possible adverse effects of introducing Stoll evidence. She noted, first, that Beles’s concern that admitting Stoll evidence would lead to the admission of general character evidence was misplaced because Beles had already opened this *948door by introducing character witnesses; and, second, that Stoll evidence concerns a particular character trait, proclivity toward sexual activity with children, and thus only permits rebuttal evidence regarding that specific trait, not general character evidence.
Brodit also called psychologist Dr. David Stein as a witness. Stein testified that he was retained to evaluate Brodit at the time of Brodit’s state appeal and habeas petition. For that purpose, Stein conducted a four-hour clinical interview with Brodit in 1997, during which he asked questions regarding Brodit’s background, childhood, and sexual history. The sexual history part of the interview revealed “nothing characteristic of somebody who would show some intense interest in ... child molestation, his history seemed entirely consistent with somebody who would not be interested in that.” Stein also recounted that he gave Brodit four different personality tests. The results of the tests, Stein asserted, “do not show anything that would suggest ... anything about sexual deviancy;” the personality characteristics revealed by the tests “are not typical of child molesters.”
The state cross-examined Stein about Brodit’s failure to disclose during the sexual history portion of the interview that he had an affair with Arcel in the period when the child abuse was allegedly occurring. Although Stein admitted that Brodit’s withholding of this information “raises a flag,” on redirect he stated that a person’s non-disclosure of an affair with an adult does not indicate sexual deviance. The affair itself indicates moral lapses, Stein said, but not a proclivity to engage in deviant sexual acts.
To counter Stein’s testimony, the state called psychology professor Dr. Judith Becker, an expert in the psychology of child molesters. Becker contended that child molesters are a heterogeneous group and rejected the idea of a fixed personality profile associated with a broad category of child molesters. She also asserted that the personality tests given to Brodit can be “faked” or may generate non-deviant results when taken by some child molesters, particularly incest offenders.
The district court found that Brodit had not shown that Beles’s performance was deficient under Strickland, because introduction of Stoll evidence would necessarily have permitted damaging counter-evidence concerning the affair and the fact that Jane looked like Arcel. The district court also observed that, given the debate between Drs. Stein and Becker at the eviden-tiary hearing, Brodit could have lost a battle of Stoll experts. Accordingly, the district court concluded that Beles made a reasonable tactical decision not to investigate the presentation of Stoll evidence, because counsel believed that the costs of presenting such evidence out-weighed the benefits.
The district court found, in addition, that even if Beles’s representation was deficient, Brodit had not shown the requisite prejudice. Because any Stoll evidence, in the court’s view, would have been accompanied by evidence of the affair and Bro-dit’s FERET membership — and would have involved simultaneously trying to present expert psychological testimony as legitimate (the Stoll evidence), while attacking other expert psychological testimony (the CSAAS evidence introduced by the prosecution) — there was no reasonable probability that the outcome of the proceedings would have been altered by the inclusion of Stoll evidence.
C. Standard of Review
The majority maintains that full deference under the Antiterrorism and Effective Death Penalty Act (“AEDPA”), Pub.L. *949No. 104-132, 110 Stat. 1214 (1996), applies to the ineffective assistance of counsel claim despite the failure of the California Court of Appeal to hold an evidentiary hearing. I do not agree, but, as will appear, even if the majority were correct in this regard, I would find the AEDPA standard met.
This court has held that AEDPA’s requirement of deference to the determinations of state courts does not apply to claims for which the state court did not hold an evidentiary hearing and therefore could not properly decide the merits. See Killian v. Poole, 282 F.3d 1204, 1208 (9th Cir.2002), cert. denied, 537 U.S. 1179, 123 S.Ct. 992, 154 L.Ed.2d 927 (2003). These claims are reviewed de novo. Id.2
Although the state Court of Appeal ruled on Brodit’s ineffective assistance of counsel claim, it did not hold an evidentia-ry hearing on the issue. The Court of Appeal’s discussion of the ineffective assistance claim reflects that it saw Dr. Stein’s declaration stating that Brodit did not have the psychological profile of a child molester. Although the Court of Appeal speculated that Beles’s failure to present Stoll evidence was a tactical choice because he “could have feared that opening this door on appellant’s personality would have permitted the prosecution to introduce damaging rebuttal character evidence,” Brodit, 72 Cal.Rptr.2d at 166, there is no indication that the court was informed about the actual rebuttal evidence Beles sought to keep out: the affair and FERET membership.
Nor did the Court of Appeal have before it all the facts pertinent to determining whether Brodit was prejudiced by the alleged ineffective assistance. Details of the uninvestigated Stoll evidence that could have been adduced in Brodit’s favor and the precise nature of the adverse evidence Beles feared would come in were lacking. The Court of Appeal was therefore in no position to determine whether adverse evidence could have been averted and whether, on balance, there is a reasonable probability that if the Stoll evidence had been admitted, a juror would have had reasonable doubt concerning Brodit’s guilt. See Rios v. Rocha, 299 F.3d 796, 813 (9th Cir.2002); Lord v. Wood, 184 F.3d 1083, 1095 (9th Cir.1999).
The majority opinion maintains that we should defer to the Court of Appeal’s “factual” conclusions even though that court was speculating about possible tactical reasons, not finding facts. I disagree. While the facts developed by the district court may have been the same “kind of reasons” about which the Court of Appeal speculated, all detail was lacking in the state court proceedings. It was therefore impossible for the state court to evaluate the propriety of the reasons given with respect to either representation or, especially, prejudice. As will appear, evaluating each Strickland prong requires some means of determining whether Beles’s guess that admission of adverse evidence could be the result of Stoll testimony was a blind stab or had a basis in the factual context considered as a whole, as well as in California law. Without full knowledge of *950the evidence at issue, the Court of Appeal had no way to evaluate that question and did not do so.
My difference from the majority on this AEDPA standard of review point is not, however, determinative. Even on the majority opinion’s assumption that AEDPA deference applies, its conclusion on ineffective assistance of counsel is wrong, as I develop below.
D. The Merits of the Ineffective Assistance of Counsel Claim
Under the standard announced by the Supreme Court in Strickland, a defendant’s claim that counsel was constitutionally ineffective cannot succeed unless he makes two showings: that counsel’s representation was deficient; and that there is “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” 466 U.S. at 687, 694, 104 S.Ct. 2052.
1. Deficient Representation3
Brodit contends that as a practical matter the California procedural rules applicable to residential child abuse cases require the defense to present some effective evidence beyond the defendant’s denial of abuse. Stoll evaluations were therefore routinely obtained by defense practitioners at the time of Brodit’s trial. According to Brodit, Beles’s failure both to investigate Stoll evidence and to consider possible means of avoiding introduction of adverse character evidence constituted deficient representation.
Counsel’s representation need only be reasonably effective to comport with a defendant’s constitutional right to effective counsel. See Strickland, 466 U.S. at 687, 104 S.Ct. 2052. Judges must review counsel’s performance deferentially, “indulging] a strong presumption that counsel’s conduct falls within the wide range of reasonable professional assistance; that is, the defendant must overcome the presumption that, under the circumstances, the challenged action ‘might be considered sound trial strategy.’ ” Id. at 689, 104 S.Ct. 2052 (quoting Michel v. Louisiana, 350 U.S. 91, 101, 76 S.Ct. 158, 100 L.Ed. 83 (1955)).
Where, as here, the ineffectiveness claim turns on a failure to investigate certain lines of defense at all, evaluation of the reasonableness of counsel’s decision must “focus on whether the investigation supporting counsel’s decision ... was itself reasonable.” Wiggins v. Smith, — U.S. -, 123 S.Ct. 2527, 2536, 156 L.Ed.2d 471 (2003). Strickland put forth the following guidelines for determining whether counsel’s strategic choices in failing to investigate might constitute ineffective assistance:
[Strategic 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. In other words, counsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary. In any ineffectiveness case, a particular decision not to investigate must be directly assessed for reasonableness in all the circumstances, applying a heavy mea*951sure of deference to counsel’s judgments.
466 U.S. at 690-91, 104 S.Ct. 2052.
Moreover, “[a]s we noted in United States v. Span, 75 F.3d 1383, 1389 (9th Cir.1996), an attorney’s performance is not immunized from Sixth Amendment challenges simply by attaching to it the label of ‘trial strategy.’ Rather, ‘certain defense strategies may be so ill-chosen that they may render counsel’s overall representation constitutionally defective.’ United States v. Tucker, 716 F.2d 576, 586 (9th Cir.1983).” Silvd v. Woodford, 279 F.3d 825, 846 (9th Cir.2002).
My evaluation of Beles’s representation necessarily takes place against the backdrop of the generic nature of the offense with which Brodit was charged and the California evidentiary rules specific to child sexual abuse trials. By allowing convictions based on generalized assertions of three or more instances of sexual abuse without identification of the dates and specifics of each alleged incident of abuse, California Penal Code section 288.5 represents a policy decision to aid the state’s ability to prosecute residential child molesters in light of the difficulties posed by a young victim’s inability to recall details regarding the time, place, and circumstances of repeated assaults. See People v. Jones, 51 Cal.3d 34, 270 Cal. Rptr. 611, 792 P.2d 643 (1990). California Evidence Code sections 1253 and 1360 create hearsay exceptions that allow the state to bolster a firsthand victim account by permitting adults to repeat (and potentially refine) the child’s allegations of abuse, even when the child testifies herself. Because these provisions enable. allegations to be repeated multiple times through authoritative adult voices, a child’s account of abuse may be significantly amplified.
The prosecution is further aided by the use of CSAAS testimony, which is offered to explain inconsistencies and inaccuracies in a child’s narrative of abuse. CSAAS testimony may effectively neutralize the crucial defense strategy of highlighting testimonial lapses in order to attack the accuser’s credibility. See United States v. Bighead, 128 F.3d 1329, 1331 (9th Cir. 1997) (CSAAS testimony functioned to “rehabilitate[ ]” a victim’s credibility after she was cross-examined about inconsistencies in her testimony and delays in reporting abuse).
In all these ways, California’s system for defining and trying charges of child sexual abuse facilitates the state’s ability to prove its case. In a prosecution where physical evidence of abuse is scant and eyewitnesses other than the victim are lacking— not an unusual circumstance — the trial boils down to a contest between the victim’s account, as reinforced by prosecution-assisting aspects of California law, and the denial of the accused. Cf. id. at 1334 (Noonan, J., dissenting) (describing how relaxed prosecutorial requirements in sexual abuse cases leave a defendant with “only two defenses: to deny that the crime happened and to cast doubt on the credibility of his accuser”).
In the context of these constraints, which have individual validity but can cumulatively suffocate a defense, California courts have recognized the import and utility of Stoll evidence to a defendant. See Jones, 270 Cal.Rptr. 611, 792 P.2d 643 (relying .in part on the availability of Stoll evidence to uphold generic charges of child molestation against a due process challenge to California Penal Code section 288.5); Stoll, 265 Cal.Rptr. 111, 783 P.2d at 715 (reversing child molestation convictions where judge failed to admit Stoll evidence). We are thus presented with a rare case in which' adequate investigation and presentation of character evidence was vital to a successful defense, contrary to *952Beles’s uninformed understanding. This is a situation, in other words, in which counsel’s options were necessarily limited. Closing off one of the few available options raises a strong likelihood of a professional error of constitutional magnitude. Compare Yarborough v. Gentry, — U.S.-, 124 S.Ct. 1, 6, 157 L.Ed.2d 1 (2003) (per curiam) (“The issues counsel omitted were not so clearly more persuasive than those he discussed that their omission can only be attributed to a professional error of constitutional magnitude.”).
More specifically: Given inconclusive physical evidence of abuse, Mica Brodit’s testimony that she never saw Brodit in the bed she shared with Jane, and the absence of eyewitnesses, the prosecution’s case rested entirely on Jane’s testimony and the hearsay repetition of her allegations by several witnesses. Expert testimony that Brodit’s psychological profile did not demonstrate a capacity to abuse children sexually could have lent critical credibility to Brodit’s denials of abuse. See Brown v. Myers, 137 F.3d 1154, 1157 (9th Cir.1998) (omitted evidence “would have altered significantly the evidentiary posture of the case”).
We accord less deference to counsel’s strategic choices not to present potentially exculpatory testimony where counsel has not fully investigated whether a witness might be helpful to the defendant’s case. See Lord, 184 F.3d at 1095; Sanders v. Ratelle, 21 F.3d 1446, 1456-60 (9th Cir. 1994). Beles did not even take the threshold exploratory step of engaging a Stoll expert to interview Brodit. He was, as a result, unable to discover what an evaluation might reveal and whether evidence gleaned from a Stoll investigation might be helpful to his client’s case. Without undertaking this first level of investigation, Beles could not make an informed decision about whether to present Stoll evidence at Brodit’s trial. See Wiggins, 123 S.Ct. at 2538 (“counsel chose to abandon their investigation at an unreasonable juncture, making a fully informed decision with respect to ... strategy impossible”); Sanders, 21 F.3d at 1457 (‘Whatever decision[counsel] might have made ... was not an informed one and thus could not be deemed ‘strategic.’ ”); id. (noting the Third Circuit’s statement in United States v. Gray, 878 F.2d 702, 711 (3d Cir.1989), that “counsel can hardly be said to have made a strategic choice against pursuing a certain line of investigation when s/he has not yet obtained the facts on which such a decision could be made”).
In concluding otherwise, the majority relies exclusively on Beles’s asserted reason for not pursuing a Stoll evaluation, namely, that he feared the admission of potentially damaging evidence of the affair and Brodit’s membership in FERET. Although Beles may have legitimately feared this potentially adverse evidence, he admitted that he did not consider or attempt to make a pre-trial motion in limine to prevent its introduction. Instead, he incorrectly assumed that he would first have to introduce Stoll evidence at trial without knowing whether the judge would agree to keep out the other, potentially harmful evidence.
The majority ignores counsel’s failure to file a motion in limine. Yet, “[a] decision not to investigate must be directly assessed for reasonableness in all the circumstances.” Wiggins, 123 S.Ct. at 2541 (quotation marks and citation omitted) (emphasis added). Those circumstances must include the possibility of avoiding feared negative consequences of otherwise helpful evidence.
California state courts recognize pretrial motions in limine as useful tools precisely because such motions allow parties to resolve evidentiary disputes ahead of trial, *953without first having to present potentially prejudicial evidence in front of a jury. See, e.g., Edwards v. Centex Real Estate Corp., 53 Cal.App.4th 15, 61 Cal.Rptr.2d 518, 524 (1997) (“A motion in limine is made to exclude evidence before the evidence is offered at trial, on grounds that would be sufficient to object to or move to strike the evidence.”); Kelly v. New W. Fed. Savs., 49 Cal.App.4th 659, 56 Cal. Rptr.2d 803, 808 (1996) (noting that pretrial motions in limine to preclude the introduction of prejudicial evidence “avoid the obviously futile attempt to ‘unring the bell’ ” once the evidence is aired before the jury). Contrary to Beles’s testimony at the evidentiary hearing, motions in limine are apparently widely used by defense practitioners in child sexual abuse cases. See, e.g., People v. Otto, 26 Cal.4th 200, 109 Cal.Rptr.2d 327, 26 P.3d 1061, 1063 (2001) (motion in limine to exclude prior child sexual abuse convictions from proceedings to determine whether defendant should be adjudicated a Sexually Violent Predator); People v. Smith, 98 Cal.App.4th 1182, 120 Cal.Rptr.2d 185, 194 (2002) (motion in li-mine arguing that certain sexual abuse charges fell outside applicable statute of limitations); People v. McFarland, 78 Cal. App.4th 489, 92 Cal.Rptr.2d 884, 887-88 (2000) (motion in limine to restrict improper expert witness testimony regarding defendant’s sexual proclivities); People v. Callahan, 74 Cal.App.4th 356, 87 Cal. Rptr.2d 838, 843-44 (1999) (motion in li-mine to exclude evidence of misdemeanor conviction for unlawful sex with a minor); People v. Harlan, 222 Cal.App.3d 439, 271 Cal.Rptr. 653, 655-56 (1990) (motion in limine to exclude evidence that male defendant wore women’s underwear); People v. Barney, 143 Cal.App.3d 490, 192 Cal.Rptr. 172, 175 (1983) (motion in limine to exclude evidence of uncharged incest incidents). Beles therefore had a well-established route for addressing his evidentiary fears before deciding whether to present Stoll evidence: He could have gathered Stoll evidence and argued, in a motion in limine, why potentially damaging character evidence should not be admitted if the proposed Stoll evidence were presented.
Defense counsel’s mistakes of law do not qualify as reasonable strategic choices under Strickland. See Williams v. Taylor, 529 U.S. 362, 395-96, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (finding deficient performance where defense counsel mistakenly believed that state law barred access to capital defendant’s juvenile records, which contained useful mitigating evidence). It was therefore not a reasonable strategic decision for Beles completely to forego an investigation of Stoll evidence under the mistaken belief that there was no way to determine prior to trial whether the results of a Stoll evaluation could be admitted and potentially damaging character evidence kept out.
As in Williams, counsel
failed to conduct an investigation that would have uncovered extensive records ..., not because of any strategic calculation but because [he] incorrectly [understood] state law.... [N]ot all of the additional evidence was favorable to [the defendant].... But ... the failure to introduce the comparatively voluminous amount of evidence that did speak in [his] favor was not justified by a tactical decision.... Whether or not those omissions were sufficiently prejudicial ... they clearly demonstrate that trial counsel did not fulfill [his] obligation to conduct a thorough investigation of the defendant’s background.
529 U.S. at 395-96, 120 S.Ct. 1495.
In sum, Beles simply did not fulfill his duty to investigate either the facts or the law under the accepted professional standard. See Wiggins, 123 S.Ct. at 2536; *954ABA Standards for Criminal Justice 4-4.1(a) (3d ed. 1993) (“Defense counsel should conduct a prompt investigation of the circumstances of the case and explore all avenues leading to facts relevant' to the merits of the case and the penalty in the event of conviction.”). Where, as here, character evidence was a crucial component of a reasonably effective defense and there was a low-risk way to litigate in advance whether potentially damaging evidence could be kept from the jury after a proposed Stoll presentation, Beles’s failure to investigate minimally whether useful Stoll evidence would be available and to consider competently whether a motion in limine would preclude his concerns reflected unreasonable professional judgment. The potential Stoll evidence was simply too important, given the constraints under which a child molestation defense must operate, to be cast aside without full factual and legal investigation of both its merits and its hazards.
2. Prejudice
To show prejudice, a defendant must demonstrate “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different,” Strickland, 466 U.S. at 694, 104 S.Ct. 2052, but “need not show that counsel’s deficient conduct more likely than not altered the outcome in the case.” Id. at 693, 104 S.Ct. 2052. The Supreme Court has defined a reasonable probability as “a probability sufficient to undermine confidence in the outcome.” Id. at 694, 104 S.Ct. 2052.
This court has emphasized that “a probability sufficient to undermine confidence in the outcome” is a fairly low threshold. See Brown, 137 F.3d at 1157; Sanders, 21 F.3d at 1461. Where the deficient representation alleged is counsel’s failure to introduce evidence potentially helpful to the defense, we ask whether the omitted evidence might have created reasonable doubt in the mind of a reasonable juror. See, e.g., Rios, 299 F.3d at 813; Lord, 184 F.3d at 1095.
Prejudice must be considered in light of the strength of the prosecution’s case. Johnson v. Baldwin, 114 F.3d 835, 838 (9th Cir.1997). See also Strickland, 466 U.S. at 696, 104 S.Ct. 2052 (“[A] verdict or conclusion only weakly supported by the record is more likely to have been affected by errors than one with overwhelming record support.”); Alcala v. Woodford, 334 F.3d 862, 872 (9th Cir.2003) (where case was only “weakly supported by the record” it was more likely to have been affected by counsel’s failure to introduce alibi evidence).
The state’s case against Brodit was quite vulnerable. Jane’s physical condition was consistent with lack of abuse and, the medical testimony suggested, hard to square with the details of the abuse to which she testified. Aspects of Jane’s testimony at trial were at times implausible and inconsistent with previous statements she made. Although Jane’s accounts of the sexual abuse contained some level of precision, several witnesses established that she could have gained her precocious knowledge of sexual activities from watching sexually explicit movies. Jane’s dislike of John Brodit and her possible desire for John and her mother Arcel to separate provided a compelling motive for concocting allegations of abuse against John’s brother, Brodit, who lived in John’s home. Moreover, Estrelita, the first adult to whom Jane reportedly recounted the abuse, may have had her own motives to break up John and Arcel. In both Jane’s and Estrelita’s accounts, which otherwise differed substantially, the abuse was disclosed after leading questions by Estrelita. Finally, Mica Brodit, Jane’s stepsister and *955bedmate, testified that, despite Jane’s claims of repeated abuse by Brodit in her shared bed when Mica was sleeping, she had never seen or been awakened by Bro-dit in the children’s bedroom.
The jury did not arrive at its verdict quickly, taking a day and a half to deliberate. Cf Jennings v. Woodford, 290 F.3d 1006, 1019 (9th Cir.2002) (noting that long jury deliberations — lasting two full days— confirmed that a different outcome was reasonably probable). The note written from the jury to the court during deliberations indicates juror concerns about the sufficiency of the physical evidence.
At trial, it was difficult for Brodit to mount a complete alibi defense, as the abuse charged in the indictment encompassed two and a half years and Jane’s testimony was not specific as to the dates of the abuse. Accordingly, Brodit’s defense relied heavily on the lack of physical evidence and a depiction of himself as a person unlikely to abuse a child sexually. Brodit’s character presentation was therefore a crucial component of his defense. Augmentation of this defense could well have made a difference, as emphasized at the evidentiary hearing by attorney Kathleen Coyne. Cf Bloom v. Calderon, 132 F.3d 1267, 1278 (9th Cir.1997) (where the defense case rested in part on a psychiatric defense, failure of counsel properly to prepare psychiatric expert witness prejudiced defendant).
In evaluating prejudice, however, we must consider the totality of the evidence before the jury, Strickland, 466 U.S. at 695, 104 S.Ct. 2052, including any rebuttal evidence that could have been admitted had Beles introduced the Stoll evidence. See Strickland, 466 U.S. at 699-700, 104 S.Ct. 2052 (noting with respect to prejudice that admission of potentially mitigating evidence that defense counsel failed to offer might have resulted in admission of harmful rebuttal evidence). Even if Beles had attempted to keep out potentially damaging character evidence through a motion in limine, the question remains whether such a motion would have succeeded.
Under California Evidence Code section 721(a), an expert witness may be fully cross-examined as to “the matter upon which his or her opinion is based and the reasons for his or her opinion.” The state asserts that the prosecution would have been able to question a Stoll expert (just as the state questioned Dr. Stein in the evidentiary hearing) about whether Brodit told him about the affair and FERET membership, and how that information — or the non-disclosure of such information— might affect the expert’s opinion. In addition, because the Stoll expert testimony could possibly be considered broadly as evidence of a character trait of disinclination to engage in sexual misconduct, it is conceivable that evidence of the affair would have been allowed in as rebuttal evidence. See Cal. Evid.Code § 1102(b) (prosecution may rebut character evidence presented by defendant).
Stein, however, adamantly resisted broad characterization of his Stoll testimony, maintaining that the psychological evidence he presented bears only on Brodit’s propensity or lack of propensity to engage in deviant sexual activity, or, even more narrowly, deviant sexual activity with children. An adult’s affair, Stein testified, may be immoral, but it is not deviant and would not be inconsistent with the normal profile Brodit presented. It is therefore unlikely that evidence of the affair was relevant to rebut Stein’s Stoll testimony.
Even if negative character evidence would have been admissible, the state’s assertion that it would have introduced such evidence is highly speculative at best. After all, Beles did put on character wit*956nesses for Brodit, but the state declined to attack these witnesses’ testimony through use of the affair and FERET membership evidence.
The state now proffers the theory that Jane and Arcel looked alike, so Brodit may have abused Jane as revenge for Arcel breaking off the affair. But this theory presumably could have been propounded, yet was not, as rebuttal to the non -Stoll good character evidence that was in fact presented at trial.
During his evidentiary hearing testimony, Beles vaguely recalled that the state’s restraint may have been due to a gentlemen’s agreement between the parties that such evidence would not come in unless the door was opened to it. Beles admitted his memory on this point was foggy, however. And the trial transcript indicates that the door to such evidence was opened: A friend of Brodit’s specifically testified at trial regarding Brodit’s dating relationships with adult women.
Finally, even if the Stoll evidence had been introduced at trial and the potentially damaging evidence admitted in rebuttal, it is reasonably probable that for at least one juror any negative character impressions generated by evidence of the affair and Brodit’s FERET membership would have been out-weighed by the positive effects of a Stoll presentation. Stein, as noted, testified that a consensual sexual relationship with an adult — even when that relationship runs counter to dominant social mores— has no bearing on a person’s propensity to engage in sexual misconduct with children. That Jane apparently saw her mother and Brodit having sex could provide a reason for her to focus on Brodit rather than his brother, John, in inventing an abuse story. This line of argument might have supported rather than detracted from the defense’s theory of the case. In addition, evidence concerning Brodit’s FERET membership might well have supported Beles’s attempted depiction of Brodit as a caring individual who would not abuse a child rather than marking him as a pedophile.
In a case like this one, where the victim’s allegations were unsupported by physical evidence of abuse and where alibis were difficult to construct due to the generic nature of the charges, it is reasonably probable that Stoll testimony could have provided an outcome-determinative edge by bolstering the defense’s portrait of Brodit as a “normal, average guy” who did not have sexual interest in children. Even if the strength of the Stoll evidence were somewhat sapped by concomitant revelations of the affair and Brodit’s FERET membership, as well as by attack from a prosecution expert, “ ‘weak’ prejudice is prejudice nonetheless.” Caro v. Calderon, 165 F.3d 1223, 1228 (9th Cir.1999).
At a minimum, the Stoll presentation “would have given the jury a choice between believing” the defense’s psychological profile of Brodit as a man unlikely to abuse children and the activities alleged by the prosecution. Alcala, 334 F.3d at 873. Given this choice, at least one juror could well have harbored reasonable doubt that Brodit abused Jane.
In sum, I conclude that if Beles had not erred by failing to investigate Stoll evidence and litigate in limine the admissibility of potentially damaging character evidence, it is reasonably probable that the outcome of Brodit’s trial would have been different. Were I to analyze the issue under AEDPA standards, as does the majority opinion, I would reach the same result.
The California Court of Appeal did not reasonably analyze the facts before it under Strickland and its ruling was objectively unreasonable under 28 U.S.C. § 2254. In Stoll, the California Supreme *957Court found that exclusion of what is now called Stoll evidence prejudiced the defendant. 783 P.2d at 714. The Court of Appeal attempted to distinguish Stoll from Brodit’s case as follows: “Here, there was no question about the identity of the molester, there was no apparent reason to invent the accusations, and there were [sic] no conflicting eyewitness testimony by other participants.” Brodit, 72 Cal. Rptr.2d at 167 n. 14. Two of the three reasons the Court of Appeal gave for distinguishing this case from Stoll are inaccurate. There was, in fact, “apparent reason to invent the accusations,” as Brodit' stressed. Additionally, Mica, who claimed that she would have woken up in the bed she shared with Jane and witnessed any abuse, testified that she never saw Brodit there, despite Jane’s allegation that abuse had taken place on several occasions while Mica was present.
The California Court of Appeal also relied on language from Stoll indicating that jurors would not regard Stoll evidence as independently conclusive of whether a defendant was a child molester. See 72 Cal. Rptr.2d at 167. Stoll, however, mentioned that consideration in the course of explaining why such evidence is admissible, not as a basis for holding that its exclusion cannot be prejudicial. Indeed, as noted, the exclusion was deemed prejudicial in Stoll. See Stoll, 265 Cal.Rptr. 111, 783 P.2d at 713-15. In Stoll, as here, there were other reasons for the jury to doubt the victim’s account, and the Stoll evidence could have convinced one or more jurors to give voice to that doubt in reaching a verdict. See id. at 715. The Court of Appeal’s analysis in this case is tantamount to a conclusion that exclusion of Stoll evidence can never be prejudicial under Strickland because it is not independently determinative. It therefore misapplies a binding state law precedent and is an objectively unreasonable application of Strickland. Accordingly, I would reverse the district court’s denial of habeas relief on Brodit’s ineffective assistance of counsel claim.
CONCLUSION .
California Penal Code section 288.5 allows conviction for sexual abuse of a child following generic charges. Other provisions of state law permit the amplification of victim allegations through hearsay and expert testimony. California thereby assists prosecutors in overcoming the formidable problems of proof they often face in child sexual abuse cases. Although the Court of Appeal did not act contrary to or unreasonably apply clearly established Supreme Court law in finding that California Penal Code section 288.5, California Jury Instruction 2.20.1, California Evidence Code sections 1253 and 1360, and the use of CSAAS testimony are consistent with due process, the prosecution’s use of these tools in Brodit’s trial has framed my analysis of his ineffective assistance of counsel claim.
The prosecution’s deployment of this full arsenal on behalf of the victim’s credibility, coupled with the generic nature of charges under section 288.5, heightened the importance of a vigorous character defense in Brodit’s case. Defense counsel was constitutionally ineffective when he completely failed to investigate whether potentially exculpatory Stoll evidence existed and did not bring a motion in limine to determine whether Stoll evidence could be presented without opening the door to unfavorable character evidence. Despite the prosecution’s systemic advantages, the weak evidence of Brodit’s guilt compels the conclusion that it is reasonably probable that he would not have been convicted had Stoll evidence been investigated and introduced.
I therefore respectfully dissent.
. The charging documents and much of the trial transcript refer to the child in this case by the pseudonym “Jane Doe.” Accordingly, I will also use this pseudonym.
. The majority’s citation of United States v. Hanoum, 33 F.3d 1128 (9th Cir. 1994), does not support its contrary conclusion. As we stated in Hanoum, "[t]his court usually declines to reach ineffectiveness challenges on direct appeal, because the claim cannot be advanced without development of facts outside the record.” Id. at 1131 (emphasis added). Here, it was the evidentiary hearing held by the district court on habeas that produced the facts at the core of Brodit’s ineffective assistance claim. The California Court of Appeal simply did not have before it a "sufficiently complete [record] to allow [it] to decide the issue.” Id.
. See Franklin v. Johnson, 290 F.3d 1223, 1233 n. 5 (9th Cir.2002) ("Cases applying the Strickland standard tend to refer to counsel's deficient 'performance.' We prefer to discuss the quality of representation,' to assure that we focus on the attorney's true obligation. That obligation, of course, is to forward one’s client's interest as legally appropriate.'').