Alcala v. Woodford

334 F.3d 862, 2003 WL 21479370

• Court: Court of Appeals for the Ninth Circuit
• Decided: June 27, 2003
• Docket: Nos. 01-99005, 01-99006
• Status: Published

Opinion

OPINION

D.W. NELSON, Senior Circuit Judge.

Jeanne Woodford, Warden of California’s San Quentin State Prison (“California”), appeals the district court’s conditional grant of habeas relief to petitioner Rodney J. Alcala. Alcala was sentenced to death following his conviction for first-degree murder. He is currently in prison.

California argues that the district court (1) incorrectly found that Alcala’s trial counsel had been constitutionally ineffective in presenting Alcala’s alibi, (2) improperly found that the state trial court committed constitutional error in excluding the testimony of defense witness Dr. Ray London, (3) erred in concluding that the state trial court’s denial of Alcala’s request for an independent medical examination of prosecution witness Dana Crappa violated the Sixth Amendment, and (4) erroneously aggregated non-constitutional errors in its cumulative error analysis.

Alcala cross-appeals, challenging the district court’s conclusions that (1) Alcala’s constitutional rights were not violated when the state trial court admitted Crap-pa’s prior testimony; (2) the exclusion of defense witnesses Tim Fallen, Gerald Crawford, and Raul Vasquez did not deny Alcala a fair trial; (3) the admission of the two sets of knives seized from Alcala’s home did not deny him a fair trial; (4) trial counsel did not render ineffective assistance in failing to investigate and rebut crime scene evidence, failing to investigate and present evidence of the value of a pair of earrings found in Alcala’s possession, and calling David Vogel as a witness without preparation; and (5) these failures to investigate were not constitutional deficiencies that could be included in the cumulative error analysis.

We conclude that Alcala’s trial suffered from multiple constitutional errors that had a substantial and injurious effect on the jury’s determination of guilt. Accordingly, we affirm the district court’s conditional grant of Alcala’s habeas petition.

Factual and Procedural Background

This case concerns the 1979 death of twelve-year-old Robin Samsoe after her sudden disappearance in the area of Huntington Beach, California. Samsoe left the Huntington Beach apartment of her friend, Bridget Wilvert, just after 3:00 p.m. on June 20, 1979, to attend a ballet lesson. She never arrived at her dance class and none of her family or friends saw her alive again. Police discovered Samsoe’s partially decomposed body in a remote mountain ravine about fifty miles away from her home almost two weeks after she disappeared. The state of her remains prevented the coroner from determining the cause of death or whether Samsoe suffered sexual molestation.

Police also found Samsoe’s beach towel within a mile of where authorities recovered her remains. A criminalist testified that blood stains on the towel indicated “wipe marks,” suggesting that someone had used the towel to wipe clean a bloody instrument such as a straight-edged weapon. Detectives also uncovered a knife caked with mud and covered in debris in the same general location as Samsoe’s body; the criminalist found a very small spot of human blood on the knife. The test for human blood consumed the entire sample of blood, precluding more specific blood typing.

Various pieces of circumstantial evidence prompted police to arrest Alcala on July 24, 1979, a little more than one month after Samsoe’s disappearance. Alcala was convicted of first degree murder and sentenced to death. The California Supreme Court reversed this conviction based on the erroneous admission of Alcala’s prior offenses and granted Alcala a new trial. People v. Alcala (“Alcala I”), 36 Cal.3d 604, 205 Cal.Rptr. 775, 685 P.2d 1126 (1984).

In 1986, nearly seven years after Sam-soe disappeared, California retried Alcala before a different judge. It is this trial that is at issue before us. Again a jury convicted Alcala of first degree murder; he again was sentenced to death. The Supreme Court of California affirmed his conviction. People v. Alcala (“Alcala II”), 4 Cal.4th 742, 15 Cal.Rptr.2d 432, 842 P.2d 1192 (1992), cert. denied, 510 U.S. 877, 114 S.Ct. 215, 126 L.Ed.2d 171 (1993).

The prosecutor relied on various forms of circumstantial evidence in securing both of Alcala’s convictions; no physical evidence directly connected him to Samsoe’s death. This circumstantial evidence that Alcala murdered Samsoe included various eyewitness identifications. Two young women, Lorraine Werts and Patty Elmen-dorf, testified that on the afternoon of June 20, 1979, a man approached them at Sunset Beach, a few miles north of Huntington Beach, and asked if he could photograph them for a class contest. Werts consented. Police later discovered a slide photo of Werts in a Seattle storage locker that Alcala rented a few weeks after Sam-soe disappeared. At trial, Elmendorf identified Alcala as the Sunset Beach photographer.

Samsoe and Wilvert also spent June 20, 1979, at the beach. They were at Huntington Beach at approximately 2:00 or 3:00 that afternoon when a man asked if he could take their pictures for a school contest. They agreed, and he took one photo each of Samsoe and Wilvert and one of the two of them together. As the man photographed them, an adult neighbor, Jackelyn Young, mistook Samsoe for her niece and approached the group. The man hurried away as Young got close. Wilvert and Young helped police prepare a composite sketch, which, according to the district court, bore a “moderate resemblance to Alcala.” Wilvert never identified Alcala as the man at Huntington Beach. Although Young could not identify Alcala in a photographic lineup just one week after Sam-soe’s disappearance, she unhesitatingly identified him as the Huntington Beach photographer at trial seven years later. She testified that he was wearing a striped, collarless shirt, and at the first trial she also had stated, in addition to this description, that it was a long-sleeved shirt.

In addition, Richard Sillett, a city surveyor, contacted police after Samsoe’s disappearance. He informed them that he, too, had been at Huntington Beach on June 20, 1979. After Alcala was arrested, Sillett identified him as the man he saw taking photographs there that day. Before this identification, Sillett had seen the composite sketch created with the help of Wilvert and Young, as well as pictures of Alcala in the local media and in a police interview. He testified that he was certain that the man had been wearing a blue Hawaiian shirt and had the impression that the man had on cut-off shorts and sandals.

Two other young women, Joanne Murchland and Toni Esparza, testified at trial that they were at Huntington Beach the day before Samsoe’s disappearance, when a man sought their permission to take photographs of them for an alleged bikini-of-the-month contest. The man left when the young women declined to give him their phone numbers. Both Murch-land and Esparza told police that the composite drawing of the suspect in Samsoe’s disappearance depicted the man who took their pictures. Only Murchland selected Alcala’s photograph out of a photo lineup. At trial, however, both women positively identified Alcala as the man from the beach. Neither woman could remember what Alcala’s car looked like when they testified at the second trial. A police officer who interviewed the women during the investigation testified that Murchland described Alcala’s car as “an older red car” and Esparza said it was “an older bigger car,” unlike Alcala’s one-year-old blue Dat-sun F-10.

The prosecution also introduced evidence that Alcala straightened his hair three days after Samsoe’s disappearance, cut his hair a few days after that, and planned to move away from the Southern California area. In defense, Alcala presented evidence that his girlfriend had been pressuring him to change his hairstyle consistently for about a month before he straightened his hair. He also offered testimony that he had purchased the necessary products for straightening his hair before Samsoe disappeared.

In the Seattle storage locker that Alcala rented after Samsoe disappeared, police found, in addition to the slide photo of Werts, a pair of gold-ball earrings that Samsoe’s mother testified belonged to her. The defense rebutted with evidence that Alcala usually wore one earring that a coworker identified as “exactly like” the ones police found in the locker.

Also at trial, jailhouse informant Freddie Williams testified that Alcala claimed to have kidnapped and killed Samsoe. The defense attempted to rebut this contention with the testimony of David Vogel, who also had been in jail with Williams. Vogel testified that Williams was desperate to testify against someone — anyone—in order to secure a deal with the prosecutor. Vo-gel significantly undermined his own credibility, however, when he admitted that he, too, once had told police that Alcala confessed to kidnapping and murdering Sam-soe.

The testimony of prosecution witness Dana Crappa proved most damaging to Alcala’s case. Crappa, a twenty-year-old forest service worker, met with the authorities twelve times before testifying at Alca-la’s first trial. During this time period, her knowledge about the crime evolved from volunteering nothing at all about the murder to placing Alcala at the crime scene with Samsoe, visiting the decomposing body twice at night, and “interacting” with the corpse twice before the police discovered the remains.

At Alcala’s second trial, Crappa testified that she did not recall the kidnapping, her visits to the crime scene and Samsoe’s body, or even testifying against Alcala at his first trial. The trial court denied Alca-la’s motion for a court-appointed, independent psychiatric evaluation of Crappa, found Crappa unavailable as a witness, and allowed the prosecutor to read Crappa’s prior testimony into the record. The trial court then refused to allow Alcala to put on Dr. Ray London, who would have testified that Crappa’s knowledge of the murder may have been the product of suggestive interview techniques.

After his second conviction and death sentence, Alcala unsuccessfully pursued direct appeals and state post-conviction remedies. In 1994, Alcala sought federal habeas corpus relief. The district court conducted an evidentiary hearing on his claims, and in 2001 conditionally granted his petition, issuing a writ ordering California to release him or grant him a new trial.

Standard of Review

Because Alcala filed his federal ha-beas petition in 1994, the Anti-Terrorism and Effective Death Penalty Act of 1996 (AEDPA) does not apply to his petition. See Lindh v. Murphy, 521 U.S. 320, 327, 336, 117 S.Ct. 2059, 138 L.Ed.2d 481 (1997) (holding that in general AEDPA applies only to habeas petitions filed after the statute’s effective date of April 24, 1996, and noting that AEDPA’s special procedures for 28 U.S.C. § 2254 petitions in capital cases apply also to petitions pending on April 24, 1996). We review de novo the district court’s decision to grant Alca-la’s 28 U.S.C. § 2254 habeas petition. Benn v. Lambert, 283 F.3d 1040, 1051 (9th Cir.2002).

We stated the standard of review for ineffective assistance of counsel and for district and state court findings of fact in the pre-AEDPA case Silva v. Woodford, 279 F.3d 825 (9th Cir.2002):

[C]laims alleging ineffective assistance of counsel are mixed questions of law and fact and are reviewed de novo. To the extent it is necessary to review findings of fact made in the district court, the clearly erroneous standard applies. Our review for clear error is significantly deferential, in that we must accept the district court’s factual findings absent a definite and firm conviction that a mistake has been committed.

Although less deference to state court factual findings is required under the pre-AEDPA law which governs this case, such factual findings are nonetheless entitled to a presumption of correctness unless they are not fairly supported by the record.

Id. at 835 (citations and internal quotation marks omitted). We may affirm the decision to grant a petition “on any ground supported by the record, even if it differs from the rationale of the district court.” Paradis v. Arave, 240 F.3d 1169, 1175-76 (9th Cir.2001).

Discussion

I. Alcala’s Trial Counsel’s Presentation of the Knott’s Berry Farm Alibi

The district court found merit in Alcala’s claim that his counsel provided ineffective assistance in failing to present an alibi defense adequately, determining that this error both prejudiced Alcala and should be included in the cumulative error analysis. To show ineffective assistance, Alcala first “must show that counsel’s performance was deficient.... Second, [he] must show that the deficient performance prejudiced the defense.” Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Alcala must prove all facts underlying his claims of ineffective assistance by a preponderance of the evidence. See, e.g., McKenzie v. McCormick, 27 F.3d 1415, 1418-19 (9th Cir.1994). We agree that Alcala has met his burden and that deficiency and prejudice are both present here.

A Facts

Alcala’s trial counsel attempted to show that, on the afternoon of June 20, 1979, Alcala was seeking freelance photography work at Knott’s Berry Farm, a theme park in Buena Park, California, and therefore could not have been in Huntington Beach at that time. Trial counsel presented the testimony of four witnesses, all employees of Knott’s Berry Farm, who established only that Alcala had visited their office sometime in the afternoon of a day in late June, around the middle of the week. One of these witnesses, Carolyn Carey, testified that she would have seen anyone who entered the office, and that she did not see Alcala, but that on a day that she “assume[d]” was June 20 she and other managers left Knott’s Berry Farm between 2:30 and 3:30 p.m. for a tour of other local theme parks.

In closing argument, the prosecutor made every effort to highlight the alibi witnesses’ failure to establish the time or date of the Knott’s Berry Farm visit:

[Defense counsel] told you that people put him, the defendant, at Knott’s Berry Farm on June 20th of 1979. And it absolutely is not true....

I kept waiting. We had four people.... [T]hree live people, and one stipulated witness where we agreed what she would say about Knott’s Berry Farm....

But recall what the testimony was? Four people came in here and talked about Knott’s Berry Farm. People who have worked there. And counsel wrote it up there on his chart as a proven fact.

The defendant was at Knott’s Berry Farm, I think, at 3:15 to 3:45, on the afternoon of June 20th, 1979. He wrote that up as if: Hey, that’s proven, four people. Terry McDowell was here in person, Robin Humphrey was a stipulation, Joanne Sutch was here in person, Carolyn Carey was here in person.

Carolyn Carey and Joanne Sutch add zero to the alibi....

They never saw Rodney Alcala in their life at anytime before they came to court....

I am left scratching my head: What the heck are they calling them as defense witnesses for? They don’t know anything about this case. They never saw this man before.

Robin Humphrey, who wasn’t here, but again we stipulated to what she would say — -she worked at Knott’s Berry Farm.

Sometime during the week of June 20th, which is the 18th, 19th, 20th, 21st, and 22nd, sometime during that week, she saw somebody who looked something like the defendant. No time of day, no date....

The last one, Terry McDowell. And I think she’s the one counsel said puts him there.

She said he was there — she’s pretty sure this was the guy, and he was there sometime during the week.

Now, what kind of alibi is that? That he was at Knott’s Berry Farm? They are assuming that that’s been proven....

* * *

There’s nothing. There’s not a doggone thing in terms of alibi at Knott’s Berry Farm....

And those are the only four people who came in here and told you anything about Knott’s Berry Farm. And they don’t help him a bit.

There’s no evidence that he was at Knott’s Berry Farm on June 20th, 1979. There is zero evidence of that.

The California Supreme Court also noted the lack of a specific date and time for the alibi:

[Alcala] presented an alibi defense, attempting to establish that he was at Knott’s Berry Farm in Buena Park during the early to midafternoon of Robin’s disappearance, seeking employment as a photographer .... [S]everal employees of Knott’s Berry Farm testified that they remembered seeing [Alcala] at the park near the date of Robin’s disappearance, although none could testify specifically to having seen him there on June 20.

Alcala II, 15 Cal.Rptr.2d 432, 842 P.2d at 1201 (emphasis added). We doubt that the alibi helped Alcala’s case.

Alcala’s counsel did in fact have access to evidence of the date and time, which would have placed Alcala at Knott’s Berry Farm on the afternoon of June 20, 1979. At the evidentiary hearing before the district court, Alcala introduced the statements of Tina Dodwell, another Knott’s employee, and various business records in Carey’s possession. Dodwell told police that Alcala arrived at her office in Knott’s Berry Farm on the day of the managers’ tour around 2:30 or 3:00 p.m. and that the tour departed around 3:00 or 3:30 p.m. She gave the same information to a defense investigator, who tape-recorded the interview with her consent. After the interview, she called the investigator and stated that Alcala might have been there at 1:30 p.m. Carey’s records established that the managers’ tour occurred on June 20, sometime after 2:10 p.m. The district court found that the failure to introduce this evidence constituted ineffective assistance of counsel.

B. Deficient Performance

Alcala’s trial counsel’s presentation of the alibi was plainly deficient; Alcala has “show[n] that counsel’s representation fell below an objective standard of reasonableness.” Strickland, 466 U.S. at 688, 104 S.Ct. 2052. Even when we “indulge a strong presumption that counsel’s conduct. falls within the wide range of reasonable professional assistance,” we conclude that Alcala has “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)).

Trial counsel made a sound strategic choice to present an alibi defense, but nonetheless failed in his duty to present that defense reasonably and competently. The district court found that Dodwell’s testimony and Carey’s records would have been far more helpful than the testimony of the alibi witnesses who did testify.¹ This finding is not clearly erroneous, and it compels the conclusion that a competent attorney would have presented this evidence unless the attorney was unaware of its existence or had a reasonable strategic reason for not doing so.

We recognize that “[f]ew decisions a lawyer makes draw so heavily on professional judgment as whether or not to proffer a witness at trial,” Lord v. Wood, 184 F.3d 1083, 1095 (9th Cir.1999), but trial counsel here offered no strategic reason for failing to call Dodwell or to present Carey’s records. The record shows that trial counsel identified Dodwell as a trial witness and intended to call her. At the evidentiary hearing, he reaffirmed that “we fully intended to call her,” but could not recall why she was not called. The record also discloses that Carey told a defense investigator that her personal calendar, one of the documents submitted at the evidentiary hearing, might be of use in establishing the date of the managers’ tour at Knott’s Berry Farm. Although trial counsel’s lack of recollection as to why he did not present this evidence does not, in and of itself, rebut the presumption that counsel acted reasonably, see Harris v. Pulley, 885 F.2d 1354, 1368 (9th Cir.1988), neither does it compel us to conclude that his actions were reasonable where all of the other record evidence suggests otherwise. See Ainsworth v. Calderon, 138 F.3d 787, 791 (9th Cir.1998).

California suggests that Dodwell’s equivocation about Alcala’s arrival time at Knott’s motivated trial counsel’s decision not to call her, and that we must therefore defer to the decision as a strategic choice. Not only is this argument contrary to our caselaw, because it would have us find a strategic basis for trial counsel’s actions in the absence of any evidence, it is inconsistent with the evidence in the record. Dod-well’s alleged “recantation” occurred long before Alcala’s second trial and long before trial counsel told the trial court that he intended to call Dodwell. Further, even if she had testified that Alcala was in the office around 1:30 p.m., her prior identification of the time as 3:00 p.m. would have been admissible under California law as a prior inconsistent statement, see Cal. Evid. Code § 1235, and her testimony still would have been far more useful than that of the witnesses who did testify. Finally, this argument does not address the failure to introduce Carey’s business records. We will not assume facts not in the record in order to manufacture a reasonable strategic decision for Alcala’s trial counsel.

Even if Alcala’s trial counsel did offer a basis for his decision not to present alibi evidence, that basis would be unreasonable if it were unsupported by objective evidence because Dodwell’s testimony and Carey’s records were consistent with the alibi defense that counsel chose. See Lord, 184 F.3d at 1095 (holding that counsel’s decision not to call witnesses was unreasonable because counsel’s stated reasons for disputing the witnesses’ credibility were not supported by objective evidence). Absent an objectively reasonable basis to undermine the credibility or utility of Dod-well’s testimony and Carey’s records, “a competent attorney would not have failed to put” on this evidence. Id.

When defense counsel undertakes to establish an alibi, but does not present available evidence of the time or even the date of the alibi, or offer a strategic reason for failing to do so, his actions are unreasonable. Alcala has overcome the presumption that his trial counsel’s actions were reasonable strategic decisions. Trial counsel’s failure to call Dodwell or to present information regarding the date and time of the managers’ tour was deficient.

C. Prejudice

We agree with the district court that Alcala was prejudiced by trial counsel’s deficient presentation of his alibi because the alibi would have challenged the eyewitness identification placing Alcala with Samsoe. Following Strickland, we find prejudice because “there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different”; the deficient presentation of the alibi “undermine[s our] confidence in the outcome.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. Considering “the totality of the evidence” before the jury, id. at 695, 104 S.Ct. 2052, we conclude that the case against Alcala was “only weakly supported by the record” and therefore “more likely to have been affected by errors than one with overwhelming record support.” Id. at 696, 104 S.Ct. 2052.

We agree with the district court that the prosecution’s case was far from compelling. The evidence that Alcala murdered Samsoe was entirely circumstantial. In Rios v. Rocha, 299 F.3d 796 (9th Cir.2002), we found that “the State’s case ... was at best a close one” where there was no physical evidence linking the defendant to the crime: “There was no weapon found, no fingerprints, no gunpowder residue, no DNA evidence.” Id. at 810. Similarly, here, there was no evidence linking Alcala to a murder weapon,² no matching fingerprints or hair, no DNA. Apart from the dubious testimony of Dana Crappa, discussed at length below, the only eyewitness to testify that she saw Alcala with Samsoe was Jackelyn Young, who said that she saw them at Huntington Beach around 3:00 p.m. on June 20, 1979.

Alcala has met his burden of proving that the absence of the alibi evidence prejudiced his case. The fact that Dodwell was under subpoena to testify is sufficient to establish, by a preponderance of the evidence, that she could have been called to testify, and the interviews submitted at the evidentiary hearing were sufficient to establish what her testimony would have been.³ If this testimony had been presented along with Carey’s records, the alibi defense would have accounted for Alcala’s whereabouts during a critical period of time. The travel time between Huntington Beach and Buena Park would have required at least a half-hour each way, in addition to any time actually spent at the Knott’s Berry Farm office. Dodwell’s statements and Carey’s records suggest that Alcala was present at Knott’s Berry Farm around 2:30 or 3:00 p.m., and thus would not have been able to return to Huntington Beach until after 3:00 p.m. at the earliest.⁴ In contrast, Young stated that she saw Alcala on the beach with Samsoe around 3:00 p.m., and Bridget Wil-vert stated that Samsoe left her house around 3:10 p.m.

The alibi evidence would have given the jury a choice between believing the testimony of apparently disinterested employees of Knott’s Berry Farm or that of Young. Although Young’s identification was confident at trial, it was not unimpeachable. Prior to the first trial, Young had been unable to identify Alcala from a photo lineup, and Young’s description of Alcala’s clothing was inconsistent with other witnesses’ testimony. Furthermore, Wilvert, who was present with Samsoe when Young allegedly saw the girls with Alcala, never identified him.

If trial counsel had presented the evidence establishing that Alcala was at Knott’s Berry Farm, in Buena Park, on June 20, 1979, around 3:00 p.m., there is a reasonable likelihood that the jury would have discounted Young’s testimony and concluded that Alcala could not have encountered Samsoe as she left Wilvert’s house at 3:10 p.m. Such conclusions would have significantly weakened, if not wholly undermined, the prosecution’s case.

Not only was the deficient presentation of the alibi far less helpful than a competent presentation would have been, it was probably actually harmful to Alcala’s case. Trial counsel told the jurors that he would prove that Alcala was at Knott’s Berry Farm on the afternoon of June 20, 1979, and utterly failed to do so, harming the credibility of Alcala’s entire defense. The prosecutor’s rebuttal highlighted the weakness of the alibi evidence; indeed, the prosecutor devoted more of his closing argument to the alibi than Alcala’s trial counsel did.

The district court did not err in finding prejudice from the deficient presentation of the alibi and in granting the writ on this basis.

II. The Exclusion of Dr. Ray London’s Testimony

The trial court also found constitutional error in the exclusion of defense expert Dr. Ray London, a psychologist who would have testified that Crappa had been hypnotically influenced in various interviews with police investigators. We agree that the exclusion of Dr. London’s testimony violated Alcala’s due process right to a fundamentally fair trial and to present crucial witnesses in his defense.

A. Facts

Dana Crappa was the prosecution’s key witness. In 1979, she was a twenty-year-old firefighter with the United States Forest Service. Her knowledge and memory of the murder continuously evolved. Crappa’s Forest Service crew discovered Samsoe’s body on July 2, 1979, near Mile Marker 11 on Santa Anita Canyon Road. Crappa volunteered nothing about the crime or the corpse at that time. One month later on August 2, 1979, after being shown photographs of Alcala, Samsoe, and Alcala’s Datsun F-10, Crappa told the police that she did not recognize either Alca-la or Samsoe. She claimed that she had nearly collided with the Datsun while driving near Mile Marker 11 between 9:30 and 10:00 p.m. on the night that another firefighter had prepared a pizza dinner, either June 7 or June 14. Five days later, she revised her story, asserting that she had seen the vehicle on the evening of June 21. At the preliminary hearing in September 1979, Crappa revised her story a third time, testifying that she saw the Datsun parked on the side of the road around 10:00 or 10:30 p.m. Crappa testified that she did not see anyone in or near the car.

Five months later on February 7, 1980, Crappa was introduced to Art Droz, who unbeknownst to her was a police detective trained in hypnosis. Droz claimed that he could help Crappa deal with the incredible stress she was experiencing if she told him her dreams and anxieties. Crappa related a similar story about seeing the Datsun on June 21, 1979, adding that there was a full moon that night and that her dreams were like movies that she saw a little more of each time she awakened. Crappa told Droz that in one of her dreams she saw a man, who may have been wearing Levis and a white shirt, sitting on a wall near the Datsun F-10. She emphasized, however, that she did not “know if I really saw it or if I just think I saw it.” Crappa also told Droz that on the night of June 29, 1979, she had seen the decomposed body of a child near Mile Marker 11, with clothes strewn about the area, a “crusty” knife in a hole, and six .22 caliber bullet casings on the ground that she picked up and threw away. During this interview, Crappa was under the misimpression that the police had established the cause of death.

Four days later, on February 11, 1980, Crappa again met with Droz, who was accompanied this time by psychologist Larry Blum. Both men encouraged Crappa to discuss her “feelings,” “impressions,” and “dreams.” In doing so, she could not recall seeing a child near the Datsun F-10 or having seen Samsoe and remained unsure about whether she had seen a man next to the car. Crappa confirmed, however, that she saw six .22 shells that were still “pretty” and not rusted next to the body and a knife “in the hole” near the body. When Crappa insisted that she could not recall anything further, Blum warned Crappa “when you start talking and then say I don’t know, I know that’s B.S. You understand that, I know that’s B.S.” Crappa also explained that one of the investigators even assured her of Alcala’s guilt, saying that he was “a hundred percent sure this guy is guilty, a hundred percent without any doubt.”

On February 15, 1980, Crappa met with prosecutor Richard Farnell and police officer Craig Robison, both of whom were trained in hypnosis. Although Crappa had told the police that she had not seen the Datsun F-10 prior to June 21, 1979, she revised her story a fifth time. In the recorded portion of the interview,⁵ Crappa claimed that on June 20, the day before she almost collided with the Datsun, she saw the same car parked on the side of the road and a man nearby “pushing” or “steering” a young blond-haired girl into the ravine. She stated that the man was wearing a white t-shirt and Levis and that she thought he was “the same guy that’s ... suspected of killing the little kid.” Crappa claimed that she never told anyone about this incident because she “just felt guilty like [she’d] done something wrong.” Crappa also reiterated that she had gone to the murder scene on the night of June 29, 1979,' and had seen a knife and children’s clothes near the corpse.

Eleven days later on February 26, 1980, Crappa met with Robison again, revising her story a sixth time. Crappa had told investigators that she never saw Samsoe’s corpse before June 29, 1979, but Robison told her that he found this contention implausible. Because Crappa could not account for her activities on the evening of June 25, 1979, Robison suggested that she visited the scene that night; he proceeded to paint a hypothetical picture of what the scene would have looked like at that time, suggesting that the body would have “smelled foul” and been easy to find. Crappa denied having visited the scene on June 25 the first few times Robison asked her about it, but eventually stated that, “Well, it’s a real possibility” that she visited the corpse prior to June 29 as well.

Crappa continued to talk to investigators before trial. On March 19, 1980, at Alcala’s first trial, Crappa testified that she saw a man “forcefully steering” a girl with long blond hair towards the ravine. He was wearing a white t-shirt and Levis and was near a vehicle that resembled Alcala’s car. Crappa further testified that she saw that same vehicle parked nearby at the side of the road on June 21. This time she claimed that she saw the Datsun between 8:00 and 8:30 p.m. and that her earlier estimate of 10:00 or 10:30 p.m. had been mistaken. According to Crappa, there was a man standing near the vehicle, again wearing Levis and a white t-shirt that “appeared to be sort of dirty or have a stain.”

Crappa also repeated her story that she visited Samsoe’s corpse on the night of June 29, 1979. Unlike in her previous versions of this event, Crappa denied seeing a knife near the corpse, although she repeated her earlier statement that she picked up six .22 shells and discarded them. She said that this visit took place around 7:00 p.m. Crappa also testified for the first time at trial that she had made another nocturnal visit to the murder scene on June 25, 1979, that it “smelled pretty foul,” and that she saw a child’s tennis shoe and some clothing near the body. She also testified that she saw Sam-soe’s corpse and that it was “cut up pretty bad.”

On April 30, 1986, the prosecution called Crappa as a witness in Alcala’s second trial. Four days earlier, however, Crappa had told the prosecutor that she could no longer remember any of the facts or circumstances relating to Samsoe’s murder. She had told them she could not even recall testifying at the first trial. The prosecutor advised the trial court:

Miss Crappa, who would be the People’s next witness, is present outside in the hallway ready to testify under subpoena. However, she has essentially informed me that she ... is not going to testify because she doesn’t have any recollection about the events in this case, essentially.

She has further told me, in response to my direct question essentially, I have asked her, is it a situation where you can’t remember, or you don’t want to remember. And essentially it seems to be a situation where she just doesn’t want to remember so she is going to say she doesn’t remember. But it’s not a situation where she can’t remember.

The prosecutor argued that Crappa was thus unavailable as a witness, and requested admission of Crappa’s testimony from Alcala’s first trial. Crappa then testified, first to the court outside the presence of the jury and then before the jury, that she could not remember testifying at Alcala’s first trial or any of the events relating to the case against him.

Between Alcala’s two trials, the California Supreme Court found hypnotically-induced testimony inherently unreliable, and thus, per se inadmissible. People v. Shirley, 31 Cal.3d 18, 181 Cal.Rptr. 243, 723 P.2d 1354, 1355-56 (1982). Accordingly, Alcala wanted to offer Dr. London at his second trial to prove that investigators had hypnotized Crappa. In determining the admissibility of Crappa’s testimony pursuant to Shirley, the trial court made a preliminary finding of fact that Crappa had not been hypnotized, which is a question for the court, not the jury, under California law. Cal. Evid.Code § 310(a). Alcala did not offer Dr. London solely for the purpose of proving that Crappa was hypnotized as that question relates to admissibility under Shirley, however, but also to impeach Crappa’s testimony as tainted and unreliable. Dr. London’s testimony was offered to demonstrate the influence that hypnotic and suggestive techniques had on Crappa’s memory as well as Crappa’s adoption of investigators’ suggestions, her increasingly certain memory over time, and her purported amnesia at Alcala’s second trial.⁶

Dr. London had reviewed all of the transcripts of police interviews with Crappa preceding Alcala’s first trial and listened to the available tape recordings. He concluded that Crappa was hypnotized at the interviews on February 7 and 11, 1980, and that she may have been hypnotized on February 15. Dr. London found that Detective Droz and' others used specific techniques and suggestions to put Crappa into a hypnotic state and to aid Crappa in “remembering” what she saw in the mountains. •

He opined that Crappa adopted the investigators’ suggestions. For example, Crappa originally told police that she never saw Alcala and Samsoe together. During the course of the investigation, however, detectives encouraged Crappa to piece different clues together, such as the man she saw on the mountain road and the girl who was kidnapped. Crappa ultimately adopted this suggestion and, at the first trial, testified that she saw Alcala “forcefully steering” Samsoe into the ravine on June 20. Crappa also adopted the investigator’s suggestions that she saw Samsoe’s body prior to June 29, 1979.

Dr. London testified that people who undergo hypnosis commonly experience progressively increased certainty in their recollections, much like Crappa’s evolving confidence in her memory. He added that Crappa’s behavior at Alcala’s first trial, including substantial pauses in her testimony and rocking her body back and forth while speaking, suggested a hypnotic trance. Other behavior indicated “dissociation,” or separation from reality, and “vi-vification,” or making an image or dream so real that it seems to take place presently — both of which indicate an altered state of consciousness. Additionally, Dr. London took note of negative post-hypnotic suggestions, which communicated to Crap-pa that she should not remember the content of or techniques used in the interviews. He also pointed out that Crappa’s certainty in her testimony waned between the February 15 interview and a March 12 conversation with an investigator, explaining that hypnotically-induced suggestions must be reinforced with some frequency to last. This phenomenon explained Crap-pa’s amnesia at Alcala’s second trial.

B. Error

“The right of an accused in a criminal trial to due process is, in essence, the right to a fair opportunity to defend against the State’s accusations. The right[ ] ... to call witnesses in one’s own behalf ha[s] long been recognized as essential to due process.” Chambers v. Mississippi, 410 U.S. 284, 294, 93 S.Ct. 1038, 35 L.Ed.2d 297 (1973); see also Crane v. Kentucky, 476 U.S. 683, 690, 106 S.Ct. 2142, 90 L.Ed.2d 636 (1986) (holding that “an essential component of procedural fairness is an opportunity to be heard”); Washington v. Texas, 388 U.S. 14, 19, 87 S.Ct. 1920, 18 L.Ed.2d 1019 (1967) (“The right to offer the testimony of witnesses ... is in plain terms the right to present a defense, the right to present the defendant’s version of the facts.... [The accused] has the right to present his own witnesses to establish a defense. This right is a fundamental element of due process of law.”).

That the Constitution affords Alcala the right to present witnesses in his defense does not mean that this right is absolute. “Even relevant and reliable evidence can be excluded when the state interest is strong.” Perry v. Rushen, 713 F.2d 1447, 1450 (9th Cir.1983). While “[t]he right to present a defense is fundamental,” id., “the state’s legitimate interest in reliable and efficient trials is also compelling.” Id. at 1451. Where evidence has been excluded pursuant to a state eviden-tiary law, we use a balancing test:

In weighing the importance of evidence offered by a defendant against the state’s interest in exclusion, the court should consider the probative value of the evidence on the central issue; its reliability; whether it is capable of evaluation by the trier of fact; whether it is the sole evidence on the issue or merely cumulative; and whether it constitutes a major part of the attempted defense. A court must also consider the purpose of the [evidentiary] rule; its importance; how well the rule implements its purpose; and how well the purpose applies to the case at hand. The court must give due weight to the substantial state interest in preserving orderly trials, in judicial efficiency, and in excluding unreliable or prejudicial evidence.

Miller v. Stagner, 757 F.2d 988, 994-95 (9th Cir.1985) (citation omitted).

The weight of the Miller factors compels us to conclude that the trial court’s exclusion of Dr. London violated Alcala’s due process rights. First, Dr. London’s testimony was highly probative because of its remarkable impeachment value. It would have emphasized the suggestive nature of police interview tactics, the evolution of Crappa’s testimony and its vulnerability to and incorporation of investigators’ suggestions, and Crappa’s increased certitude. This impeachment testimony would have explained Crappa’s bizarre and disturbing demeanor at Alcala’s first trial, her subsequent amnesia, and the parallel relationship between her testimony and both the prosecution’s theory and the physical evidence.

Second, California did not object to Dr. London’s expert qualifications or the bases for his conclusions, giving us no reason to question its reliability. That prosecution witnesses disagreed with Dr. London’s conclusion that Crappa was hypnotized does not make his expert opinion unreliable; rather, it provides the very basis for admitting and relying on Dr. London’s testimony — the presentation of the defense theory of the case. Third, California does not assert that the jury could not evaluate Dr. London’s testimony and consider it with the totality of the evidence. Fourth, Dr. London provided the sole evidence for impeaching Crappa on the basis of hypnosis, suggestion, brainwashing, cajoling, or improper influence; his testimony was not cumulative but was critical to Alcala’s case.

Finally, the testimony of Dr. London proved integral — vital even — to Alcala’s case. His testimony would have provided a formidable defense tool because his expert opinion would cast serious doubt on the most damning portions of Crappa’s testimony and on its overall believability. As the star prosecution witness, Crappa was the only person to place Alcala at the murder scene with the murder victim close in time to the theorized murder date. The district court’s compelling analysis demonstrates the import of Dr. London’s testimony to Alcala’s case:

[W]ithout Dr. London, the defense was prevented from rebutting the prosecutor’s claim that Crappa’s inconsistencies were attributable to the traumatic nature of the events that she was describing (and had allegedly witnessed). Without Dr. London, the defense was precluded from presenting evidence that the substance and evolution of Crappa’s testimony at the first trial, and subsequent claim of amnesia, were indicative of hypnosis. Without Dr. London, the defense was precluded from proving that hypnosis could instill a greater degree of certainty and that the hypnotic subject was incapable of distinguishing implanted memories from actual recollection. Without Dr. London, the defense could not dispel the aura of credibility that Crappa’s testimony received by virtue of its consistency with the remaining evidence. Through Dr. London, however, the defense proposed to offer an explanation for this “fit” — because the memories had been constructed through hypnosis, Crappa’s testimony had been manufactured to fit.

In short, Dr. London provided the only available means for Alcala to impeach Crappa’s testimony with a coherent theory. This theory would not only have undermined Crappa’s accounts of what she saw, but also would have diminished the overall weight, if any, the jury afforded her testimony.

California’s interest in excluding Dr. London does not outweigh Alcala’s strong interest in the admission of his testimony. The trial court excluded Dr. London’s testimony under California Evidence Code § 352, finding that presentation of this evidence, though probative, would confuse the issues and waste an undue amount of time. While the policy underlying this rule is to allow the exclusion of otherwise probative evidence if an undue consumption of time or confusion of the issues would substantially outweigh its relevance, that is not the case here. We cannot conclude on the record before us that Dr. London’s testimony would consume even a considerable amount of time, let alone an undue amount of time, or that it would confuse the jury.

The balance of these factors supports a conclusion that the trial court unconstitutionally excluded Dr. London. Therefore, We must consider whether the exclusion was prejudicial. Calderon v. Coleman, 525 U.S. 141, 145, 119 S.Ct. 500, 142 L.Ed.2d 521 (1998). We agree with the district court that it was.

Crappa’s testimony provided the bedrock of the prosecution’s case by placing Alcala and his distinctive car at the murder scene with the victim just hours after she disappeared. Dr. London’s testimony, if accepted by the jury, would have damaged Crappa’s credibility severely. It also would have helped place the cold transcript of Crappa’s testimony in context by showing the dramatic evolution of her purported recollections. Thus, the exclusion of Dr. London likely had a substantial and injurious effect or influence on the jury’s verdict.

III. The Trial Court’s Denial of Alcala’s Motion for an Independent Medical Examination of Dana Crappa

The district court held that the trial court’s denial of Alcala’s motion for an independent medical examination of Crap-pa constitutes a violation of the Sixth Amendment’s Compulsory Process Clause. We disagree. While the Compulsory Process Clause guarantees a criminal defendant the right to present relevant and material witnesses in his defense, see Washington, 388 U.S. at 17-19, 87 S.Ct. 1920, the trial court never barred Alcala from exercising this right.

A. Crappa’s Alleged Unavailability

The prosecution asserted that Crappa was “unavailable” because of her purported amnesia. After hearing Crappa’s testimony on April 30, 1986, the trial court observed that Crappa appeared to be suffering from an “existing ... mental illness or infirmity” and, therefore, “probably qualifie[d] under unavailability, in view of [her] present mental status.”

When Alcala requested additional time to consider Crappa’s purported claim of amnesia, the trial court set a hearing for May 5, and informed Alcala,

But I really think you gentlemen can do all the research in the world, and everybody knows I am never hampered by the law anyway.

Basically, what we are going to do is put her on, let her have her say-so, give it a shot, and get into the reading....

And there won’t be anything further until tomorrow. And then you have the rest of the night to show me I am wrong, and everybody knows I never am.

On May 5, Alcala filed a motion asking the trial court “to appoint, for the information of the court, a mental health professional to conduct a clinical interview of [Crappa].” Alcala argued that due process required the trial court to perform an independent medical examination of Crap-pa’s current mental condition before ruling on Crappa’s unavailability.

Two witnesses testified at the hearing: Dr. Anthony Staiti, a psychiatrist called by the prosecution, and Superior Court Judge Phillip Schwab, who had presided over Al-cala’s first trial. Dr. Staiti had met with Crappa three times for a total of two to three hours to assess whether “she was capable of returning to her position as a police dispatcher.” Dr. Staiti explained that his “working diagnosis” of Crappa was that she was suffering from “post traumatic stress disorder chronic delayed.”

The defense called Judge Schwab. He commented that Crappa’s demeanor during Alcala’s first trial was “unusual,” and often punctuated with long delays of one minute or longer between questions and answers. Judge Schwab also testified that Crappa’s behavior was peculiar and that he even conferred with the parties about the possibility of terminating her testimony.

At the close of the hearing, the trial court denied Alcala’s motion for an independent medical examination and ruled that Crappa was “unavailable” because of a “pre-existing mental infirmity.” The court based its decision on, inter alia, “the history of the case” as well as the “obvious frailties of the witness.”

Accordingly, Crappa was excused from testifying in person and the prosecution was permitted to read a transcript of Crappa’s previous testimony to the jury.

B. Alcala’s Right to Compulsory Process

The right of an accused to have compulsory process for obtaining and calling witnesses in his favor is guaranteed under the Sixth and Fourteenth Amendments. Washington, 388 U.S. at 17-19, 87 S.Ct. 1920. “Just as an accused has the right to confront the prosecution’s witnesses for the purpose of challenging their testimony, he has the right to present his own witnesses to establish a defense. This right is a fundamental element of due process of law.” Id. at 19, 87 S.Ct. 1920. Alcala’s right to compulsory process was violated if he was barred by the trial court from presenting “testimony [that] would have been relevant and material, and ... vital to [his] defense.” Selam v. Warm Springs Tribal Corr. Facility, 134 F.3d 948, 952 (9th Cir.1998) (quoting Washington, 388 U.S. at 16, 87 S.Ct. 1920).

At the hearing on Crappa’s purported unavailability, Alcala was never barred from presenting any witnesses or evidence, nor did Alcala ever request an opportunity to examine Crappa with his own medical experts. Alcala points out that if the prosecution was permitted to present a medical expert, due process requires that the defense be allowed to do the same. But, Alcala never requested such an opportunity; he only asked that the trial court appoint a mental health professional to conduct an independent medical examination of Crappa for the court’s benefit in assessing her unavailability.

If the trial court had barred Alcala from presenting defense witnesses or from conducting a medical examination of Crappa, Alcala would be correct in arguing that his constitutional rights were violated. Washington, 388 U.S. at 17-19, 87 S.Ct. 1920. However, that did not happen here. Rather, the trial court merely refused to exercise its discretion to appoint an expert to conduct an independent investigation of a disputed matter.

While Alcala’s suggestion was probably the more prudent path for the trial court to adopt, the court’s decision not to perform an independent investigation of the facts cannot be characterized as a violation of the Sixth Amendment. California Evidence Code § 240(a)(3) states in pertinent part that a witness is “unavailable” if she is “[d]ead or unable to attend or to testify at the hearing because of then existing physical or mental illness or infirmity.” California law defines a “mental infirmity” as “a defect of personality or weakness of the will.” People v. Rojas, 15 Cal.3d 540, 125 Cal.Rptr. 357, 542 P.2d 229, 235 (1975). In order to establish the existence of a mental infirmity, expert medical testimony, while potentially relevant, is not essential. Alcala II, 15 Cal.Rptr.2d 432, 842 P.2d at 1213. Furthermore, California Evidence Code § 730 provides that a trial court’s decision to appoint such an expert is a purely discretionary one. Torres v. Mun. Ct. of Los Angeles Jud. Dist., 50 Cal.App.3d 778, 123 Cal.Rptr. 553, 556 (1975) (“Although Evidence Code section 730 provides for court appointed experts, the decision to comply with the defendant’s request remains within the sound discretion of the trial court.”); see also In re Jennifer J., 8 Cal.App.4th 1080, 10 Cal.Rptr.2d 813, 816 (1992) (“Although the court in a proceeding ... is assuredly empowered to appoint one or more factfinding expert witnesses ... such action is a matter of discretion.”).

At any point between April 30, when Crappa informed the court of her amnesia, and May 5, when the court found her unavailable, Alcala could have requested that he be allowed to examine Crappa with a defense expert or, if additional time was needed, moved for a continuance. Alcala offers no explanation as to why he never made such requests.

Accordingly, we hold the district court erred in granting Alcala’s claim that his Sixth Amendment rights were violated by the trial court’s denial of his motion for an independent medical examination.

IV. The Admission of Crappa’s Previous Trial Testimony

The district court rejected Alcala’s claim that the admission of Crappa’s previous testimony violated his rights under the Confrontation Clause of the Sixth Amendment. In his cross-appeal, Alcala argues that Crappa’s testimony was not sufficiently reliable to be deemed admissible as evidence.

The Sixth Amendment’s Confrontation Clause, made applicable to the states through the Fourteenth Amendment, provides that, “In all criminal prosecutions, the accused shall enjoy the right ... to be confronted with the witnesses against him.” U.S. Const. amend. VI. In Mattox v. United States, 156 U.S. 237, 15 S.Ct. 337, 39 L.Ed. 409 (1895), the Supreme Court explained:

The primary object of the constitutional provision in question [is] to prevent depositions or ex parte affidavits ... being used against the prisoner in lieu of a personal examination and cross-examination of the witness, in which the accused has an opportunity, not only of testing the recollection and sifting the conscience of the witness, but of compelling [her] to stand face to face with the jury in order that they may look at [her], and judge by [her] demeanor upon the stand and the manner in which [she] gives [her] testimony whether [she] is worthy of belief.

Id. at 242-43, 15 S.Ct. 337. The admission of hearsay testimony, however, does not violate the Confrontation' Clause, so long as the witness is unavailable and the testimony bears adequate “indicia of reliability.” Ohio v. Roberts, 448 U.S. 56, 65, 100 S.Ct. 2531, 65 L.Ed.2d 597 (1980). Such “indicia” can be inferred if “the evidence falls within a firmly rooted hearsay exception.” Id. at 66, 100 S.Ct. 2531.

Crappa’s demeanor during Alcala’s first trial was odd, if not bizarre. Her conduct was so peculiar that it is next to impossible not to question the trustworthiness of her testimony. The record is replete with examples of Crappa behaving in a manner that calls into question her credibility, mental stability, psychiatric health, and veracity as a witness.

• Crappa waited up to one minute or more before answering certain questions and, at times, did not answer questions unless they were repeated multiple times. Judge Schwab testified that, “There were, at least in portions of her testimony, substantial delays between the question and an answer, sometimes running perhaps close to a minute, perhaps even longer.”

• At one point during her testimony, Crappa was “rocking back and forth, her eyes closed.... ” for several minutes. According to Alcala’s trial counsel, Crappa “seem[ed] to be undergoing some kind of psychiatric or psychotic break.”

• For one period lasting fifteen minutes, Crappa’s only responsive utterance to the prosecutor’s questions was to mutter repeatedly, “It was a.”

• The prosecution asked Crappa to describe what she saw on the evening of June 25, 1979, multiple times. Crap-pa, however, either remained silent, muttered unresponsively to the questions, or would continue to rock her body back and forth without speaking. It was only after the trial court called for a recess and Crappa spoke with Robison, a police officer and trained hypnotist, during the break, that she was able to return to the witness stand and testify in a somewhat coherent fashion about having seen Samsoe’s corpse on the evening of June 25.

• Judge Schwab, a veteran California state trial judge with more than twenty years of experience on the bench, agreed that Crappa’s behavior was “unusual.” In particular, the length of time it took for her to begin responding to a question after it was asked was not normal. Judge Schwab even called counsel into chambers to discuss whether it was appropriate for Crappa to continue testifying. Although he said this was not the only time he had considered terminating a witness’s testimony, he could not recall any other specific cases where he had done so.

• Judge Schwab advised Crappa of her constitutional right to assert her privilege against self-incrimination and consult an attorney, after it was suggested to him that Crappa was committing perjury.

Crappa’s behavior as a witness was so extraordinarily odd that it strikes at the very core of her reliability as a witness. Moreover, reading the cold transcript of her testimony may have had the effect of transforming an incredibly bizarre performance into a credible presentation.

We are asked to determine whether, under these unusual circumstances, Crap-pa’s past trial testimony nonetheless bears sufficient indicia of reliability to have been admissible. California argues that the reliability of prior trial testimony given under oath and subjected to cross-examination should be inferred without further inquiry because such testimony is recognized as a firmly rooted exception to the hearsay rule. We do not opine on whether prior trial testimony is a firmly rooted hearsay exception, although we acknowledge some indications in the case law that it is. See Roberts, 448 U.S. at 66 n. 8, 100 S.Ct. 2531; Mancusi v. Stubbs, 408 U.S. 204, 216, 92 S.Ct. 2308, 33 L.Ed.2d 293 (1972); Barber v. Page, 390 U.S. 719, 722, 725-26, 88 S.Ct. 1318, 20 L.Ed.2d 255 (1968). In any event, the Supreme Court has noted that in “extraordinary cases,” further inquiry into the reliability of prior trial testimony may be required. Roberts, 448 U.S. at 73 n. 12, 100 S.Ct. 2531.

Although this may be such a case where further inquiry would be appropriate, we ultimately need not decide whether Crap-pa’s testimony bears adequate indicia of reliability. The trial court’s other constitutional errors, combined with the ineffectiveness of Alcala’s trial counsel, are more than sufficient to warrant the denial of California’s appeal and the granting of Al-cala’s petition. Accordingly, we decline to rule on this issue as it is unnecessary in our ultimate assessment of the merits of Alcala’s petition.

V. Cumulative Error

The district court granted Alcala’s habe-as petition in part due to cumulative error. We hold that the district court did not err in finding that the combined prejudice of the multiple errors committed in this case deprived Alcala of a fundamentally fair trial and constitutes a separate and independent basis for granting his petition.

California contends on appeal that some of the purported errors considered by the district court in its cumulative error analysis were not constitutional errors and that others were not prejudicial.⁷ Alcala argues that the district court erred in failing to find error in the exclusion of certain defense witnesses, holding that his trial counsel’s failure to investigate the crime scene was not deficient, and in ruling that some adverse evidentiary rulings were not sufficiently prejudicial to grant the petition on those grounds alone.

In cases where “there are a number of errors at trial, ‘a balkanized, issue-by-issue harmless error review’ is far less effective than analyzing the overall effect of all the errors in the context of the evidence introduced at trial against the defendant.” United States v. Frederick, 78 F.3d 1370, 1381 (9th Cir.1996) (quoting United States v. Wallace, 848 F.2d 1464, 1476 (9th Cir.1988)). In other words, “ ‘[e]rrors that might not be so prejudicial as to amount to a deprivation of due process when considered alone, may cumulatively produce a trial setting that is fundamentally unfair.’ ” Thomas v. Hubbard, 273 F.3d 1164, 1180 (9th Cir.2001) (quoting Matlock v. Rose, 731 F.2d 1236, 1244 (6th Cir.1984)), overruled on other grounds by Payton v. Woodford, 299 F.3d 815, 829 n. 11 (9th Cir.2002) (en banc). “In those cases where the government’s case is weak, a defendant is more likely to be prejudiced by the effect of cumulative errors.” Frederick, 78 F.3d at 1381.

A. The Exclusion of Certain Defense Witnesses

The prosecutor theorized that Alcala kidnapped Samsoe on the afternoon of June 20, 1979, some time after she left Wilvert’s home and before she should have arrived at her ballet class. Crappa’s testimony placed Alcala in the mountains with Samsoe later that very day, suggesting that Alcala killed Samsoe on the same day that he kidnapped her, June 20, 1979.

The testimony of Tim Fallen would poke at least one hole in this theory. Fallen testified at Alcala’s first trial that he saw Samsoe in Huntington Beach on June 21, 1979 — the day after she disappeared. A police detective showed Fallen a picture of Samsoe on June 21 and asked if he had seen her. Fallen told the detective that he had seen Samsoe just minutes before the detective’s arrival. He offered, without prompting, that she was riding a yellow, ten-speed bike. When Samsoe left Wil-vert’s home the day before, she had borrowed her friend’s yellow ten-speed.

Late the next night, on June 22, 1979, police officer Gerald Crawford saw a car parked at a turnout just north of Mile Marker 11 on the Santa Anita Canyon Road, about 100 feet from where authorities ultimately found Samsoe’s remains. Crawford then saw Raul Vasquez walking toward the car. Crawford questioned Vasquez about his purpose for being in the area. He responded that he had relieved himself in the woods while waiting for his girlfriend. When Crawford asked Vasquez his girlfriend’s name, he either did not know her name or did not respond.

Crawford patted Vasquez down and found a pair of heavy pliers in his back pocket. Crawford questioned Vasquez about the tool, and he responded, “you never know what can happen up here in these mountains.” Crawford also searched Vasquez’s car, where he found a towel or blanket in the back seat and a six pack of beer on the floorboard of the front seat. Crawford noted that a passenger window was shattered, and broken glass lay strewn across the back seat. Crawford testified that Vasquez seemed upset, extremely nervous, and very shaky; he acted as if he were hiding something.

Alcala wanted to offer Crawford and Vasquez to advance his theory that someone other than Alcala kidnapped Samsoe on June 21 — after Fallen saw her — and that Vasquez murdered Samsoe on June 22. Crawford would have testified, as he did at Alcala’s first trial, to his encounter with Vasquez, while Vasquez would have testified that he was on parole from a murder conviction.

At Alcala’s first trial, Fallen had identified a photograph of a different young blond girl as the one he saw; the prosecutor argued at the second trial that Fallen’s testimony was neither probative nor reliable. Despite Alcala’s arguments to the contrary, and considering only Alcala’s offer of proof and the prosecutor’s objection, the trial court excluded Fallen’s testimony as confusing and irrelevant.⁸

The trial court also excluded the testimony of Crawford and Vasquez. After stating on the record that “[criminals] usually stay away from the scene of the crime,” which would be inconsistent with Vasquez “go[ing] back on the 22nd to see if he did a good job,”⁹ the trial judge found that “the probative value of this type of evidence is zero” and that presenting it would be “a waste of time.”

Even if the trial court erroneously excluded Fallen, Crawford, and Vasquez as a matter of state law, we cannot afford him habeas relief unless the exclusion violated his due process right to a fair trial. See Estelle v. McGuire, 502 U.S. 62, 67-68 & n. 2, 112 S.Ct. 475, 116 L.Ed.2d 385 (1991). We employ a balancing test for determining whether the exclusion of testimony violates due process. Courts should weigh the probative value of the evidence, its reliability, whether the trier of fact can evaluate the evidence, whether the evidence is cumulative, and whether the evidence proves integral to the defense theory in evaluating whether admissible evidence was unconstitutionally excluded. Miller, 757 F.2d at 994. In addition, we must consider California’s interest in excluding the evidence. Id. at 994-95.

i. Exclusion of Fallen’s Testimony

While California concedes that the trial court erred in excluding Fallen’s testimony, some analysis of whether this exclusion violated Alcala’s due process rights is helpful. First, Fallen’s sighting of the girl he believed to be Samsoe is probative to a central issue in this case: whether Alcala kidnapped and killed Samsoe on June 20. Fallen’s testimony could create reasonable doubt that Samsoe was kidnapped and killed on June 20, the day on which eyewitnesses claimed to have seen Alcala at Huntington Beach, rather than June 21, the day Fallen claimed to have seen Samsoe. Fallen could also rebut Crappa’s testimony placing Alcala with the victim close to the murder scene.

Fallen’s testimony also bears indicia of reliability. He identified the girl he saw within five or ten minutes of viewing her; his memory was recent. He also volunteered that the girl was riding a yellow ten-speed bicycle, before the officer could indicate that Samsoe was riding such a bike when she disappeared. Moreover, that the prosecutor impeached Fallen at the first trial did not make his testimony unreliable but instead raised questions about his credibility and the weight his testimony should be accorded. These are issues to be weighed by the jury, not the judge. See, e.g., United States v. Scheffer, 523 U.S. 303, 312-13, 118 S.Ct. 1261, 140 L.Ed.2d 413 (1998).

No question exists that the jury could evaluate Fallen’s testimony. His insistence that he saw Samsoe the day after the prosecutor argued that she was kidnapped introduces an alternate exculpatory theory. The evidence is inconsistent with the prosecutor’s case, but that it conflicts is the very point of presenting Fallen’s testimony. If courts prohibit the introduction of any evidence that conflicts with the prosecution’s case because it might “confuse” the jury, the right of the accused to present a defense would exist only in form. In addition, Fallen’s testimony was not cumulative; it provided the only evidence that Samsoe was seen alive after June 20.

Lastly, Alcala focused on a misidentification theory via his efforts to rebut the various eyewitness identifications, to undermine Crappa’s credibility, and to put on an alibi defense. Fallen’s testimony would have facilitated this theory by allowing Alcala to undermine the prosecution’s assertion that Samsoe was kidnapped and killed on June 20. In other words, if Samsoe were kidnapped on or after June 21, and the prosecution could not link Alca-la to the area on June 21, Alcala could further his misidentification defense.

The trial court excluded Fallen because his testimony was both confusing and irrelevant. These concerns do not outweigh Alcala’s interest in putting on this testimony. That Fallen’s testimony weakened the prosecution’s case made it probative, not confusing, and that it contradicted California’s theory demonstrates its great relevance.

Because the factors weigh in Alcala’s favor, the trial court committed constitutional error in excluding Fallen’s testimony. This error likely affected the jury’s verdict. Fallen’s testimony would have given the defense an eerie coincidence for the jury to weigh against the many bizarre coincidences that the prosecution presented. It also would contradict the prosecution’s theory that Alcala kidnapped and murdered Samsoe on June 20. Regardless of whether this error was sufficiently prejudicial in itself to grant Alcala’s petition, the district court correctly included this error in its cumulative error analysis.

ii Exclusion of Crawford’s and Vasquez’s Testimony

The exclusion of Crawford’s and Vasquez’s testimony is subject to the same balancing test. See Miller, 757 F.2d at 994. Their combined testimony about the events of June 22, 1979, is arguably more probative than Fallen’s testimony to the central issue: Alcala’s guilt. Vasquez was found in a remote mountain area, after he literally emerged from the bushes 100 feet from where authorities ultimately discovered Samsoe’s body. Crawford testified that Vasquez was , upset, nervous, and shaky and acted like he was hiding something. Vasquez provided Crawford with a flimsy excuse for his presence in the area. He possessed a heavy tool, which he suggested he would use as a weapon if necessary.

This testimony also is reliable. Crawford was a police officer with no motive to lie. Indeed, he was called by the prosecution to testify about finding the body. As for Vasquez, his testimony only concerned his prior homicide conviction. Even if he did testify to the events of the night of June 22, 1979, California has suggested no reason that Vasquez would lie about that night in order to incriminate himself.

Just as Fallen’s testimony was capable of evaluation by the trier of fact, so too was the testimony of Crawford and Vasquez. Again, their testimony presented an alternate explanation for Samsoe’s murder that the jury could accept or reject. Again, their testimony would not have been cumulative; it provided the sole evidence that Vasquez may have murdered Samsoe.

Finally, the exclusion of testimony from Crawford and Vasquez precluded Alcala from presenting a third-party culpability defense; he instead relied on a misidentifi-cation theory. Crawford and Vasquez would have helped Alcala pursue a third-party culpability defense, a theory not inconsistent with the defense presented and which — had the evidence been admitted— would have proven a strong defense theory.

The trial court excluded Crawford’s and Vasquez’s testimony as irrelevant and a “waste of time,” and suggested Alcala’s third-party culpability theory was untenable. Even still, the above factors clearly weigh in favor of Alcala’s interest in having the testimony admitted. The testimony was relevant, and no record evidence indicates that it would have consumed an undue amount of time.

This evidence helps demonstrate reasonable doubt as to Alcala’s guilt by suggesting that Vasquez may have murdered Samsoe. The exclusion of this evidence prejudiced Alcala and belongs in the cumulative error analysis.

B. The Admission of the Kane Kutlery Knives

As mentioned above, police discovered a knife obscured by some vegetation and covered with debris and caked mud in the same general area as Samsoe’s remains. The prosecutor theorized that Alcala used the knife as the murder weapon because of blood found on the knife, wipe marks on Samsoe’s towel, and Crappa’s testimony that the body was “pretty cut up.”

To support this hypothesis, the prosecutor admitted into evidence two complete, unused sets of kitchen knives that police seized from Alcala’s home where he lived with his mother and stepfather. Kane Kutlery manufactured both these knives and the carving knife found in the ravine near Samsoe’s remains. Alcala’s mother testified that the knife sets were gifts from her husband’s former employer. She also asserted that her husband’s employer never gave her a separate knife like the carving knife found near Samsoe’s body, that she was not missing any knives from her kitchen pantry, and that the alleged murder weapon found in the ravine differed from the knife sets she owned.

The trial court judge remarked on the many differences between the knife sets and the purported murder weapon. He noted the difference in the studs on the handles, the type of wood used for the handles, and the shape of the knives. He stated that “[t]here’s no way that this particular weapon could be tied in [ ] the remotest to any of these particular sets [seized from Alcala’s home] — I mean, the whole knife set is not even close to the knife [found in the ravine].”

To buttress its murder weapon theory, the prosecutor called Clella Schneider, a Kane Kutlery representative, as a witness. Schneider testified that although the two knife sets seized from Alcala’s home and the alleged murder weapon both were distributed through Kane Kutlery to the same six western states and marketed around the same time, the knives differed in design. Schneider also testified that during a period of five years, Kane Kut-lery sold about 15,000 of each of the two knife sets in six western states and about 4,000-5,000 of the individual carving knives in the same general geographic area. She further testified that the carving knife found in the ravine was sold separately from the seized knife sets and that all of the knives in question were sold at major supermarket chains and drug stores.

The trial court admitted the knives and Schneider’s testimony over Alcala’s objection that they were irrelevant and highly prejudicial.

A conclusion that the admission of the Kane Kutlery knife sets violated Alcala’s right to a fundamentally fair trial requires that the knife sets and Schneider’s testimony were irrelevant to the prosecution’s case and that the “erroneously admitted evidence was of such quality as necessarily prevents a fair trial.” McKinney v. Rees, 993 F.2d 1378, 1384 (9th Cir.1993) (internal quotation marks omitted). “Evidence is considered irrelevant if it fails to make any fact of consequence more or less probable.” Id. at 1380. Moreover, “[o]nly if there are no permissible inferences the jury may draw from the evidence can its admission violate due process.” Jammal v. Van de Kamp, 926 F.2d 918, 920 (9th Cir.1991).

The knife sets, and the accompanying testimony of Schneider, are irrelevant. The evidence showed that two unused sets of Kane Kutlery knives were found in Al-cala’s home. These knives belonged to Alcala’s mother and stepfather and had been given to them by a former employer; they were not found in Alcala’s direct possession, nor did he purchase them. Furthermore, even as the trial court admitted, the knife sets differed in many material respects from the knife found at the murder scene.

The jury could draw no permissible inference from this evidence. To infer that Alcala used the knife in the ravine to murder Samsoe, the jury would have had to speculate — and could have done no more than speculate — that because the purported murder weapon and the unused knife sets shared the same brand name, Alcala was connected to the murder weapon. That the same company manufactured both the purported murder weapon and the knife sets fails to make any fact pf consequence to the prosecution’s case more or less probable.

California argues that the knife sets were relevant to show that Alcala had access to or familiarity with Kane Kutlery. The fact that Alcala’s home contained substantially different types of knives of the same brand as the purported murder weapon, however, does nothing to advance the argument that Alcala had special or increased access to Kane Kutlery knives. As Schneider testified, both the knife sets and the carving knife were readily available by the thousands in major supermarkets and drug stores in six western states. This evidence fails to show that Alcala had any more access to Kane Kutlery knives than any other person in the general public with access to stores that sold this brand of knife. Admittedly, the fact that Alcala’s mother and stepfather had Kane Kutlery knives in their kitchen may show that Alcala was familiar with this particular brand. This purported familiarity, however, does not tend to make any fact of consequence to California’s case more or less probable. Alcala’s general access to or familiarity with Kane Kutlery has no relevance to connecting Alcala to the alleged murder weapon.

The admission of the knife sets amounts to constitutional error, and the prejudicial effect of this evidence likely influenced the jury. California’s entire case rested on “strange coincidences.” That the alleged murder weapon and the knives seized from Alcala’s home were both of the Kane Kut-lery brand fits neatly into that “strange coincidences” theme, and the prosecutor spent a good deal of his closing argument — more than three pages of the court reporter’s transcripts — framing the issue that way.¹⁰ Thus, there is a reasonable likelihood that some jurors linked Alcala to the knife in the mountains based on brand commonality with the seized knives. Because this evidence caused some prejudice to Alcala, the district court properly included it in its cumulative error analysis.

C. Ineffective Assistance of Counsel

As noted above, one of the grounds on which the district court granted Alcala’s petition was that his trial counsel had been ineffective in presenting his alibi. The district court also included this deficiency, along with the deficient preparation of defense witness David Vogel, in its cumulative error analysis. Alcala suggests that, in addition to these deficiencies, his counsel was deficient in failing to investigate the crime scene and in failing to investigate the value of the gold-ball earrings found in Alcala’s storage locker. California argues that none of these acts and omissions amounts to deficient performance.

We agree with the district court’s inclusion of the deficient presentation of the alibi in the cumulative error analysis. We further agree that counsel was deficient in failing to prepare Vogel and that the prejudice resulting from this deficiency was properly included in the cumulative error analysis. In addition, the district court should have included in the cumulative error analysis trial counsel’s deficiency in failing to investigate the crime scene. Finally, the district court correctly concluded that there was no deficiency in failing to investigate the value of the earrings.

i. Preparation and Presentation of David Vogel

Vogel’s testimony, which was intended to discredit jailhouse informant Freddie Williams’s testimony, was completely undermined by the prosecutor on cross-examination. Vogel, who had himself acted as a jailhouse informant in several previous cases, denied ever giving police or prosecutors false information about another inmate. He then admitted talking to the police about Alcala, with whom he had also been in jail, but claimed that he did not remember what he told the police. The prosecutor then confronted him with an interview he had given to the police in 1979, in which Vogel stated that Alcala had confessed to him that he murdered Sam-soe. Vogel ultimately testified that he had lied to the police about Alcala’s confession to him, but that he was not lying in his testimony under oath, in part because he had converted to Christianity in the interim.

In addition to this evidence, the district court made several factual findings to support its conclusion that Alcala’s trial counsel’s performance was constitutionally deficient; none of these findings is clearly erroneous. First, trial counsel knew about Vogel’s 1979 police interview prior to putting Vogel on the stand. Vogel was a former client of Alcala’s trial counsel, and when the prosecutor sought to play a tape of the interview at trial, trial counsel volunteered, “Your honor, I have a transcription of the tape.... If the court wishes we can Xerox this. It’s my understanding this was prepared by ... the district attorney’s office, back in 1979.”

The district court also found that “prior to testifying at trial, Vogel had given little thought to the fact that he had previously reported that Alcala had made incriminating statements.” Vogel’s own testimony, in which he stated that he could not remember telling police that Alcala had confessed to him, supports this finding.

Finally, the district court found that Al-cala’s trial counsel failed “to forewarn Vo-gel that this topic would likely be covered during cross-examination” and failed “to ascertain for himself how Vogel would likely respond.” Again, Vogel’s own testimony is evidence of a lack of preparation. Furthermore, at the evidentiary hearing, trial counsel confirmed that his general practice was not to interview witnesses himself — let alone prepare them for specific topics of cross-examination — but merely to say a few words to them in the hallway immediately prior to calling them.

Given these facts, we must determine whether Alcala’s trial counsel’s conduct “falls within the wide range of reasonable professional assistance.” Strickland, 466 U.S. at 689, 104 S.Ct. 2052. Although an attorney’s performance is generally entitled to a strong presumption of competence, see id., deference to the attorney’s strategic decisions is diminished where the attorney has not done the preparation necessary to make informed decisions; in particular, the decision not to call a witness is entitled to less deference if the attorney has not interviewed the witness. See Lord, 184 F.3d at 1095 & n. 8. The same holds true for an attorney’s decision to call a witness whom he has not interviewed or otherwise prepared.

Here, we fail to see, and California fails to explain, how the decision to call Vogel without preparing him for cross-examination could possibly “ ‘be considered sound trial strategy.’ ” Strickland, 466 U.S. at 689, 104 S.Ct. 2052 (quoting Michel, 350 U.S. at 101, 76 S.Ct. 158). Any competent attorney would have made an effort to find out what Vogel would say when asked about his statements to the police incriminating Alcala. There is no suggestion that trial counsel had limited access to Vogel before the trial; indeed, Vogel was his former client. We need not determine whether, after proper preparation, the decision to call Vogel might have been a reasonable exercise of professional judgment. In the absence of such preparation, Alcala’s trial counsel’s performance was clearly deficient.

We agree that this deficiency prejudiced Alcala and should be included in the cumulative error analysis. If Alcala’s trial counsel had adequately prepared Vogel to testify, two outcomes are likely: either Vogel would have been forthright in his testimony about his 1979 interview with the police, or trial counsel would have realized that the 1979 interview was a serious liability and elected not to call Vogel at all. Either way, we can assume that Vogel’s initial evasive answers, which greatly damaged his credibility, were likely the result of defense counsel’s incompetence. See, e.g., Johnson v. Baldwin, 114 F.3d 835, 840 (9th Cir.1997) (noting that if attorney had confronted witness with evidence that his testimony was false, the witness “probably would have elected not to lie to the jury”).

We have no trouble concluding that Al-cala would have been better off if Vogel had either not testified at all or testified credibly. If Vogel had not testified, his statement to the police that Alcala had confessed to murdering Samsoe would not have been admitted. If he had testified candidly, he would have been more useful as an impeachment witness against Williams, lessening the impact of Williams’s testimony that Alcala had confessed to him. Because a “defendant’s own confession is probably the most probative and damaging evidence that can be admitted against him,” Arizona v. Fulminante, 499 U.S. 279, 296, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991) (internal quotation marks omitted), either of these outcomes would have been more favorable than the damaging testimony that Vogel actually gave. The district court properly included the prejudice from this deficiency in the cumulative error analysis.

ii Failure to Investigate the Crime Scene

At trial, two expert witnesses testified for the prosecution about the crime scene and Samsoe’s remains: a criminalist, Margaret Kuo, and a pathologist, Dr. Sharon Schnittker. Kuo testified that she found a single, tiny drop of human blood on the blade of the knife recovered from the scene. As noted above, the prosecutor used this testimony, along with Dana Crappa’s testimony that Samsoe’s body was “pretty cut up,” to argue that the knife was the murder weapon. Schnittker testified that there was no evidence of knife wounds on the skeletal remains but stated that it was possible for fatal stab wounds to leave no mark on a skeleton.

Alcala claims that his trial counsel was ineffective for failing to investigate the crime scene and to introduce competing expert testimony to show that Crappa’s observations were factually impossible. He suggests that a full investigation by a forensic pathologist and a criminalist would have revealed additional details to impeach Crappa and discredit the prosecution’s theory. The district court found that Alcala’s trial counsel was not deficient in failing to investigate the crime scene. We disagree and, contrary to the district court, include the prejudice resulting from this deficiency in our cumulative error analysis.

Alcala has met his burden of demonstrating that an adequate investigation could have resulted in additional evidence favorable to the defense. At the evidentia-ry hearing, Alcala presented the testimony of a forensic pathologist that, due to the condition of Samsoe’s skeleton, “it is highly unlikely that Robin Samsoe died as a result of multiple stab wounds,” as well as a criminalist, who testified that the condition of the knife was inconsistent with its purported use as a murder weapon. The criminalist noted that, even after disassem-bly, there was no blood in any crack or crevice of the knife as would typically be found if it had been used as a murder weapon. Moreover, the spot of blood on the knife was consistent with a blood splatter rather than a wipe or smear mark.

“[D]efense counsel must, ‘at a minimum, conduct a reasonable investigation enabling him to make informed decisions about how best to represent his client.’ ” Rios, 299 F.3d at 805 (quoting Sanders v. Ratelle, 21 F.3d 1446, 1456 (9th Cir.1994)). “While a lawyer is under a duty to make reasonable investigations, a lawyer may make a reasonable decision that particular investigations are unnecessary.” Babbitt v. Calderon, 151 F.3d 1170, 1173 (9th Cir.1998). The district court found trial counsel’s performance adequate primarily because he managed to elicit enough information from the prosecution’s own experts to rebut the prosecution’s case.

Although it may have been reasonable for trial counsel not to retain specific experts, such as a pathologist or a criminalist, it cannot have been reasonable for him not to investigate the crime scene at all.¹¹ Strickland v. Washington itself was in part a duty-to-investigate case and sheds some light on the applicable analysis. Washington, a death row inmate, asserted that his lawyer was ineffective for failing to investigate psychological and character evidence for his sentencing trial. 466 U.S. at 675-77, 104 S.Ct. 2052. The Supreme Court found that the decision not to investigate was a strategically reasonable one because the attorney “could reasonably surmise from his conversations with [Washington] that character and psychological evidence would be of little help.” Id. at 699, 104 S.Ct. 2052.

Here, by contrast, trial counsel apparently had decided that evidence to impeach Dana Crappa would be of help; he went to the trouble of calling an astrophysicist to refute Crappa’s testimony as to the phase of the moon on one of the nights in question. Impeaching Crappa was central to Alcala’s defense, and without an investigation, trial counsel could not reasonably have known whether the crime scene evidence was consistent with her testimony.

As to the argument that trial counsel reasonably relied on the prosecutor’s own witnesses, we find that such reliance, if any, was unreasonable. In Rios, we considered the government’s argument that defense counsel’s failure to investigate was not deficient because he reasonably relied upon the investigation conducted for a co-defendant. 299 F.3d at 807-08. In rejecting this argument, we held that “it would have been unreasonable for him to rely solely on the investigation performed for a co-defendant, because the co-defendant’s interests in the case might well conflict with [the defendant’s].” Id. at 808. There is no question but that the prosecution’s interests in this case conflicted with Alca-la’s, and so trial counsel’s reliance on the prosecution’s investigation is no more reasonable than the reliance on a co-defendant’s investigation in Rios.12

In Holsomback v. White, 183 F.3d 1382 (11th Cir.1998), the Eleventh Circuit considered an analogous failure to investigate in the context of sexual abuse. The prosecution’s case rested entirely on the testimony of one witness, the purported abuse victim. Id. at 1384. The prosecution’s own investigation revealed that there was no medical evidence of sexual abuse. Id. The court determined that it was error for defense counsel “simply to rely on the prosecutor’s references to the lack of physical evidence as the sole source of information on the subject,” id. at 1387-88, and that defense counsel should have conducted his own investigation into the medical evidence or lack thereof. Id. at 1388. Having made no investigation, defense counsel “could not have made an informed tactical decision” not to call medical expert witnesses. Id. Similarly, the prosecution’s case against Alcala rested largely on the credibility of Dana Crappa. The prosecution’s own witnesses established that there was no physical evidence to establish the cause of death as testified to by Crappa, and Alcala’s trial counsel elected “simply to rely on” the prosecution’s investigation as “the sole source of information on the subject.” Id. at 1387-88. He could not have made an informed decision about whether the inability to establish a cause of death and other objective crime scene evidence should be used to impeach Crap-pa’s testimony.

Following Rios and consistent with Holsomback, we hold that the failure to investigate the crime scene was deficient. Alca-la has shown some prejudice resulting from this deficiency, suggesting that if his trial counsel had investigated the crime scene, he would have retained the services of a criminalist and a forensic pathologist. The resulting evidence would have helped to discredit Crappa’s testimony and the prosecution’s inferences drawn therefrom. The district court should have included the prejudice flowing from this deficiency in the cumulative error analysis, and we will do so here.

iii. Failure to Investigate the Value of the Earrings

At trial, the prosecution presented evidence that a pair of gold-ball earrings had been seized from a storage locker rented by Alcala. Marianne Frazier, Samsoe’s mother, testified that the earrings were similar to a pair of cheap $3 earrings that she owned and that Samsoe sometimes wore, earrings that she had not seen since Samsoe’s disappearance.

Alcalá urges that his trial counsel was constitutionally ineffective in failing to call an expert gemologist to demonstrate that the earrings found in the storage locker were not cheap $3 earrings, such as those that Samsoe’s mother testified that she had owned, but instead were mid-priced custom jewelry with a higher gold content. Alcala presented the testimony of a gemologist at the evidentiary hearing to establish the gold content and value of the earrings. This testimony was inconsistent with Frazier’s testimony, substantially undercutting the assertion that the earrings found in the storage locker were taken from Samsoe. Although the district court recognized the potential impact of an expert gemologist’s testimony in “neutral-iz[ing] the damaging inferences the jury might otherwise draw from the similarity between the earrings in Alcala’s storage locker and Ms. Frazier’s lost earrings,” the district court found no deficiency because trial counsel had no way of knowing that the earrings were inconsistent with Frazier’s description.

We agree with the district court’s conclusion that this alleged deficiency was properly excluded from the cumulative error analysis. Alcala’s trial counsel reasonably could have expected that if the earrings were more valuable than Frazier suggested, Alcala, as the owner of the earrings, would have informed him of this fact. See Strickland, 466 U.S. at 691, 104 S.Ct. 2052 (“The reasonableness of counsel’s actions may be determined ... by the defendant’s own statements or actions. Counsel’s actions are usually based, quite properly, ... on information supplied by the defendant.”). There is no indication in the record that Alcala did so, or that competent counsel would have realized that the earrings were more valuable than the earrings described by Frazier.¹³ Trial counsel’s performance was not deficient, and the prejudice from the failure to investigate the value of the earrings cannot be included in our cumulative error analysis.

D. Cumulative Prejudice

We now turn to the issue of whether the cumulative effect of these errors had “a substantial and injurious effect” on the jury’s verdict. Coleman, 525 U.S. at 145, 119 S.Ct. 500 (internal quotation marks omitted). We agree with the district court that the cumulative effect of these errors “operated to deprive Alcala of a fundamentally fair trial.” The district court correctly weighed all of the errors together in order to assess their cumulative impact on Alcala’s constitutional rights. “[E]ven if no single error were [sufficiently] prejudicial, where there are several substantial errors, ‘their cumulative effect may nevertheless be so prejudicial as to require reversal.’ ” Killian v. Poole, 282 F.3d 1204, 1211 (9th Cir.2002) (quoting United States v. de Cruz, 82 F.3d 856, 868 (9th Cir.1996)).

Here, the cumulative impact of these errors goes to the heart of the prosecution’s theory of the case and undermines every important element of proof offered by the prosecution against Alcala. Indeed, after reviewing the errors in this case, we are left with the unambiguous conviction that the verdict in this case was not the result of a fair trial.

The trial court precluded Alcala from effectively challenging or excluding the testimony of the prosecution’s key witness — Dana Crappa. The trial court’s exclusion of Dr. London’s expert testimony deprived Alcala of an important opportunity to discredit the only eyewitness who allegedly could place Alcala with Samsoe at the scene of the crime on the evening of June 20. Had Dr. London been permitted to testify, the jury may have discounted Crappa’s testimony as not credible in light of the fact that it was obtained through improper and dubious means, as well as the obvious and apparent instability of Crappa’s mental condition. Dr. London’s expert testimony, combined with Crappa’s bizarre demeanor, would have seriously called into question her reliability as a witness. Had the credibility of the prosecution’s star witness been effectively challenged, the case against Alcala would have been undoubtedly weaker.

Second, aside from Crappa, the only witness who could place Alcala with Samsoe was Jackelyn Young, who claimed to have seen Alcala with Samsoe at Huntington Beach on the afternoon of June 20 at 3:00 p.m. Had Alcala’s counsel adequately presented Alcala’s Knott’s Berry Farm alibi, Alcala could have directly challenged Young’s testimony. Alcala’s alibi also would have challenged the theory that Al-cala abducted Samsoe after she left Wil-vert’s house at 3:10 p.m.

Third, Fallen’s, Crawford’s, and Vasquez’s testimony would have further weakened the prosecution’s theory of the case. The combined testimony of these witnesses would have challenged Crappa’s version of the events and presented a colorable third-party culpability theory for the jury to assess.

Fourth, the erroneous admission of the Kane Kutlery knives seized from Alcala’s home permitted the jury to draw an impermissible connection between Alcala and the purported murder weapon — the key and only piece of physical evidence discovered at the scene of the crime that could be linked to Alcala. The prosecution’s placement of undue emphasis on this link compels us to conclude that the jury’s verdict was tainted by the admission of this evidence.

Fifth, Alcala’s trial counsel presented a fatally impeachable witness, Vogel, without first assessing how Vogel would respond on cross-examination to evidence that he told police that Alcala confessed to him. As a result, Vogel actually served to reinforce the informant’s tale that Alcala confessed to murdering Samsoe. Trial counsel’s performance not only deprived Alcala of a potentially meritorious defense, but his actions and omissions also substantially impaired Alcala’s efforts to demonstrate reasonable doubt.

Lastly, Alcala’s trial counsel completely failed to conduct any investigation of the crime scene. This failure likely resulted in the loss of an opportunity to challenge Crappa’s purported observations as inconsistent with the objective evidence.

Therefore, we conclude that the cumulative impact of these errors is more than sufficient to demonstrate prejudice. The cumulative weight of the above errors deprived Alcala of a fundamentally fair trial.

Conclusion

We conclude, as did the district court, that the deficient presentation of Alcala’s alibi and the exclusion of Dr. London’s testimony was each on its own an error sufficiently prejudicial to grant Alcala’s petition. When combined with the erroneous exclusion of Fallen, Crawford, and Vasquez, the erroneous admission of the Kane Kutlery knives, the deficient failure to prepare Vogel to testify, and the deficient failure to investigate the crime scene, the cumulative impact of these errors severely undermines our confidence in the jury’s verdict. We affirm the conditional grant of Alcala’s petition.

AFFIRMED.

1 . Dodwell's statements, although perhaps hearsay, nonetheless constitute probative evidence because California failed to object to their admission at the evidentiary hearing before the district court. See United States v. Armstrong, 48 F.3d 1508, 1518 n. 8 (9th Cir.1995), rev'd on other grounds, 517 U.S. 456, 116 S.Ct. 1480 (1996).

2 . As discussed below, the prosecutor did attempt to connect Alcala to the knife found near Samsoe’s body, but the evidence relied on by the prosecutor was immaterial to such a connection.

3 . California cites cases denying ineffective assistance claims where the defendant failed to identify what excluded witnesses would have said, United States v. Berry, 814 F.2d 1406, 1409 (9th Cir.1987), or to identify any witnesses that should have been called for the defense, United States v. Murray, 751 F.2d 1528, 1535 (9th Cir.1985); where the witness had a Fifth Amendment right not to testify and had credibility problems because he was a convicted felon, United States v. Harden, 846 F.2d 1229, 1231-32 (9th Cir.1988); where the witness did not testify because counsel could not locate him, Tinsley v. Borg, 895 F.2d 520, 532 (9th Cir.1990); and where the only evidence that the witness existed was the petitioner's own affidavit, Dows v. Wood, 211 F.3d 480, 486 (9th Cir.2000). None of these situations is presented here.

4 . Our determination of prejudice is not diminished by the fact that Dodwell indicated to a defense investigator that Alcala might have been at Knott's Berry Farm around 1:30 p.m. instead of 3:00 p.m. "We have previously found prejudice when counsel failed to ... present the testimony of alibi witnesses, even though their testimony was 'vague with regard to time.’ " Luna v. Cambra, 306 F.3d 954, 961 (9th Cir.2002) (quoting Brown v. Myers, 137 F.3d 1154, 1157 (9th Cir.1998)).

5 . Robison and Farnell did not record the entire interview, only the final twenty-four minutes.

6 . California presents a strong argument that Alcala never offered Dr. London for impeachment purposes, but we are unconvinced. Al-cala offered Dr. London's testimony after Detective Droz testified that he never hypnotized Crappa, putting her hypnosis at issue and opening the door for impeachment. Moreover, the record supports that Alcala meant to use Dr. London to undermine Crappa's testimony. Finally, the trial court imposed a blanket prohibition on Dr. London's testimony pursuant to California Evidence Code § 352 when it made its preliminary finding of fact that Crappa had not been hypnotized. By excluding Dr. London pursuant to § 352, the trial court understood the probative value of his testimony, independent of whether Crappa's testimony was admissible, but excluded it anyway given other concerns. This ruling indicated that Dr. London’s testimony would be excluded no matter how Alcala intended to use his testimony.

7 . California originally argued that the district court erred in cumulating the prejudice from both trial court errors and deficiencies of trial counsel. At oral argument, California conceded that our case law allows such cumulation. See Mak v. Blodgett, 970 F.2d 614, 622 (9th Cir.1992).

8 . The trial court also stated that Fallen "looked like a drifter,” but went on to say that Fallen’s appearance "doesn’t influence me.... So I don’t want to hear that come back and haunt me from that super court that is up there with God alone.”

9 . This theory also would be inconsistent with Alcala returning to Mile Marker 11 on June 21 after taking Samsoe there and killing her on June 20, as the prosecution argued.

10 . The prosecutor made the following statements about the Kane Kutlery knives in his closing argument:

The Kane Kut knives. Again, interesting coincidences. Interesting coincidences. There's a Kane Kut knife found in the mountains. The defense, as I mentioned, says that’s not even the murder weapon in this particular case. It’s just a coincidence.

But lying there in the brush right next to remnants of Robin Samsoe’s blond hair is a butcher knife. A Kane Kut butcher knife with human blood on it. That’s a coincidence, folks.

The knife was dirty or had some mud on it or something. That couldn’t be the knife. Is that a coincidence? Is that one of the most extraordinary coincidences you ever heard in your life that lying at the scene of the murder of a young child is a knife with human blood on it and they want you to believe that’s a coincidence.

In the defendant’s home, there are knives. What kind of knives are they? Kane Kut knives. Again, the defendant says that's a coincidence. It’s just coincidental that the very brand that is found in the mountains is also in the defendant’s home. And that’s certainly not enough to convict him, obviously. If that was all the evidence we had ... we wouldn’t be here. But it’s a very interesting coincidence, especially when you hear from Mrs. Schneider, from Kane Kut, about the number of knives we’re talking about.

And again, my memory for numbers isn’t that great, but I am sure you all wrote it down. The knife found in the mountains, my recollection is she said something like 3,000 of those knives had been sold in the six western states over a period of something like five years. And they are sold in supermarkets. They are cheap knives.

It is not as if the market was saturated with Kane Kut knives. Three thousand sold in six states during five years.

We don't want to argue statistics, and I am not going to, but it makes me wonder. That's an interesting coincidence.

With respect to the other knife sets that were Kane Kut knives, certainly this is not a leading knife brand. Probably none of us ever heard about it before. It was something like 15,000 of each one of those sets had been sold, again over a five-year period in the six western states, so in terms of their availability that’s an interesting coincidence. It’s not exactly like if we walked into all of our homes and looked through the drawers, we'd come across a bunch of Kane Kut knives.

11 . California cites an Eighth Circuit case, Battle v. Delo, 19 F.3d 1547 (8th Cir.1994), which held that "trial counsel is not required to employ the services of experts, provided that counsel prepares an adequate defense for a client through careful investigation of facts surrounding the case.” Id. at 1557. Battle is inapplicable here for the simple reason that Alcala’s trial counsel did not conduct an investigation of the crime scene.

12 . The district court relied upon Hall v. Sumner, 682 F.2d 786 (9th Cir.1982), in which we held that counsel was not ineffective in declining to present a diminished capacity defense where he had elicited enough information from the prosecutor’s expert witness to argue diminished capacity. Id. at 789. In Hall, however, we found that defense counsel had made a “reasonable tactical decision[ ]" not to present a diminished capacity defense and that he had in fact argued such a defense, as an alternative to his factual innocence claim, based on evidence elicited from prosecution witnesses. Id. Because Alcala’s counsel did not make a tactical decision not to impeach Crappa, but instead made every effort to impeach Crappa, Hall is inapposite.

13 . Alcala argues that trial counsel should have known that the earrings were valuable because one of the friction post nuts on the earrings was stamped "14k," indicating 14 karat gold. Although the expert gemologist who testified at the evidentiary hearing stated that this stamp "was clearly visible with a magnifying glass,” there is no evidence that it would have been obvious upon an ordinary examination of the earrings.