Russell Coleman is a California prisoner under sentence of death. He filed a petition for a writ of habeas corpus pursuant to 28 U.S.C. § 2254. The district court denied relief from the conviction but granted relief from the sentence, because the state court jury instructions violated Hamilton v. Vasquez, 17 F.3d 1149 (9th Cir.1994). We have jurisdiction under 28 U.S.C. § 2253. We agree with the district court and we affirm.

On September 5, 1979, between 5 p.m. and 9 p.m., Shirley Hill was raped, sodomized and strangled to death inside a small bungalow next to a football field within the grounds of a high school. In 1981, Coleman was tried and convicted of Hill's murder.

Three parts of the evidence against Coleman strongly support his conviction. These consist of fingerprints, serology tests, and a lie.

The police matched Coleman's palm print to a print left on a windowsill in the bungalow. They also matched his thumbprint to a print left on a chair near Hill's body. The thumbprint tested presumptively positive for blood. Hill's fingerprints were also found on the windowsill on either side of Coleman's palm print. It was not possible to determine whether Hill's fingerprints had been laid down at the same time as Coleman's palm print.

The serology evidence was presented through a forensic serologist, Edward Blake. Blake tested Hill's blood and the semen found in her vagina. Based on these tests, Blake placed Coleman in eight percent of the population that could have deposited the semen in Hill.

The lie occurred when Coleman was questioned by police. He denied ever having been in the bungalow. At trial, however, he changed his story and testified he had been in the bungalow a few days before the murder. This accounted for his palm and thumbprints at the scene of the crime.

The jury found Coleman guilty of first degree murder, with special circumstances of rape and sodomy. In the penalty phase of the trial, the jury returned a verdict of death. The trial court upheld that verdict.

The California Supreme Court rejected Coleman's direct appeal, with two judges dissenting. People v. Coleman, 46 Cal. 3d 749, 759 P.2d 1260, 251 Cal. Rptr. 83 (1988). Coleman then filed a petition for a writ of habeas corpus in the district court. Over the next several years, he exhausted in state court unexhausted claims included in his initial and various amended petitions which he filed in the district court. When he completed exhausting his claims in state court and amending his petition in federal court, the district court held an evidentiary hearing. In 1997, that court denied Coleman's petition as to his conviction, but granted the petition as to Coleman's death sentence. This appeal followed.

We turn first to the three parts of the evidence that became the centerpiece of Coleman's conviction: the fingerprints, the serology tests, and the lie.

Police Inspector Ihle canvassed the crime scene for fingerprints. He found Coleman's thumbprint on the back side of a chair, along with four unidentifiable smudged fingerprints on the front side of the chair. Because all of these prints appeared slightly discolored, Ihle performed a Hemastix test on them. A Hemastix test is used as a presumptive test to detect blood in urine. The test showed that the discolored material on the chair back was presumptively blood. After the test was completed, there was not enough discolored material left on the chair back for further testing.

The discovery of Coleman's thumbprint, presumptively in blood on the chair back, triggered other police procedures. Standard police practice required a notation on both the latent fingerprint card and on the object where the fingerprint was found. In this case, Ihle wrote the word "bloody" on the chair back to indicate a bloody print had been found, but failed to note "blood" on the latent print card. Ihle, however, told two police inspectors and the original prosecutor about the test results prior to Coleman's preliminary hearing. The police reports failed to mention either the thumbprint on the chair back or the results of the Hemastix test.

Charles Gush, Coleman's initial counsel, was shown the chair back. At the preliminary hearing, Gush asked Ihle if there were any bloody fingerprints. Ihle responded, "No, there were not." Gush left it at that and did not pursue the matter further.

Gush then resigned and the State appointed Public Defender Peter Keane to represent Coleman. Gush did not discuss discovery matters with Keane and did not tell Keane about the bloody print. After the preliminary hearing, the state court ordered the prosecution to produce "all physical evidence" and "all laboratory, technician's and other reports concerning the testing and examination of [the] physical evidence." Notwithstanding this court order, the prosecution did not produce the chair back or the results of the Hemastix test. Keane did not see the chair back before trial, and he knew nothing of the bloody print or the Hemastix test.

Keane first learned about the bloody print on the first day of trial, when the coroner said the prosecutor had told him about it. Keane first learned of the Hemastix test when Ihle began to testify. Keane moved to strike Ihle's Hemastix testimony. Keane argued there was no proper foundation for the test and that the prosecution had violated the court's discovery order by not disclosing the bloody print evidence or the Hemastix test results. The prosecution responded by arguing that Keane had sufficient notice of the bloody print because that fact was mentioned in a transcribed police interview with Coleman and because the notation "bloody" was on the chair back.

Although Keane had not seen the chair back, he had seen the transcribed police interview. He did not regard the mention of bloody prints in that interview as significant, however, because he believed the police had used the word "bloody" as hyperbole when questioning Coleman to get him to admit that he murdered Hill.

The trial judge allowed Keane to question Ihle on voir dire in the judge's chambers outside the presence of the jury. During that questioning, Ihle explained that he did not know the chemical processes that caused the Hemastix strip to change color. He said he had tried the test on a few other substances (semen, urine and catsup) and had found that only blood caused the strip to change color. Ihle explained that the test would not differentiate between human blood and other types of blood. At Keane's request, during the voir dire examination, Ihle conducted Hemastix tests on two discolored spots on the front of the chair back; he obtained a negative reaction for blood from both tests.

The trial court determined that Keane had received adequate notice based upon the notation "bloody prints" on the chair back and a report from the Department of Corrections (which made no mention of bloody prints). This was incorrect. Keane had no such notice. The trial court also found the Hemastix test reliable based upon Ihle's three years of experience with it.

Ihle then testified before the jury. He stated that the Hemastix test is a "presumptive test for blood," which does not distinguish human blood from animal blood. He did not, however, explain what it meant to label a test "presumptive." He admitted that he had no knowledge of the chemistry behind the test. (Later that day, forensic serologist Blake described the chemical reaction upon which the test is based.) Ihle also explained that not enough blood was left to conduct further tests.

On cross-examination, Ihle admitted that he had tested only a few items with the Hemastix test: blood, urine, catsup, tomato sauce and coffee. He had not tested sweat, whiskey or coke. Ihle had performed the original test in the presence of a police officer, but no chemist checked the results. He also testified that when he tested the front of the chair back in the judge's chambers, the Hemastix test did not show the presence of blood.

Ihle further testified that he did not know if fingerprint powder would cause the Hemastix strip to react. On redirect, however, he explained that the Hemastix test he had just performed in chambers showed that the Hemastix strip would not react to fingerprint powder.

The subject of the Hemastix test and the bloody print came up again during argument to the jury. In the prosecutor's opening argument, he explained that Ihle did not need to be a chemist to effectively use the Hemastix test. He characterized the Hemastix test as "selective for blood." He explained in detail how Coleman would have gotten blood on his hand if he had put his hand over Hill's mouth to silence her while he was strangling her. He ended his depiction of the crime by stating: "[Coleman] left his calling card in blood."

In Keane's jury argument, he emphasized that Ihle was not a chemist and that the blood might not have been human blood. Then, in the prosecutor's closing argument, the prosecutor returned to the bloody print. He told the jury: "Russell Coleman, Russell Coleman, whose fingerprints we have in blood; Russell Coleman, whose fingerprint in blood is on this chair."

The prosecution's expert forensic serologist, Dr. Edward Blake, analyzed Hill's blood, the vaginal wash taken from her body, and Coleman's blood and saliva. He used three different classifications: (1) the ABO blood typing system; (2) secretor status and (3) the phosophoglucomutase ("PGM") enzyme type.

Blake found that Hill was a non-secretor, with type A blood and PGM type 1/1. He also found that the semen donor must have been a non-secretor with PGM type 2/1 or 2/2. Based upon his analysis of Coleman's blood and saliva, Blake concluded that Coleman was a non-secretor with PGM type 2/1 and type O blood. This meant Coleman could have been the source of the semen.

Blake prepared two reports which summarized these conclusions. His reports were furnished to the defense. Blake's laboratory notes, however, were not part of the prosecution's trial file and were not turned over to the defense. Keane gave a copy of Blake's reports to his investigator, Russell Stetler, who interviewed Blake and prepared for Keane a summary of the tests, the results of the tests, and his conversation with Blake.

On October 30, 1980, four days before trial,1 Stetler delivered copies of Blake's reports to Brian Wraxall, a defense serological consultant. Wraxall did not ask for or receive Blake's notes. Keane consulted with Wraxall to prepare to cross-examine Blake. In total, Wraxall estimates that he spent one hour with the defense team discussing Blake's reports. Although Wraxall was not available to testify, Keane did not consult with any other expert about the serological evidence.

At trial, Blake testified that twenty percent of the population are non-secretors. Forty percent of the population are PGM types 2/1 and 2/2. Only eight percent of the population would have both characteristics (twenty percent times forty percent). Thus, according to Blake, Coleman was included in only eight percent of the male population that could have deposited the semen found in Hill's vagina.

On cross-examination, Keane established that there were other tests Blake could have performed, which might have eliminated Coleman completely. Keane questioned Blake about the peptidase A test, which is useful for typing the semen of African-Americans. Blake had not performed the peptidase A test and said that he should have. Blake explained that he did not perform additional tests, because the number of tests he was able to perform was limited by the small amount of vaginal wash. He performed the tests he believed were most important. Keane also established on cross-examination that Blake had used up all of the vaginal wash, so no more tests could be performed. In addition, Keane got Blake to admit that additional tests might have eliminated Coleman as the semen donor, or substantially decreased the likelihood that he could have been the donor.

When Coleman was arrested, he lied and told the police he had never been on the football field or in the nearby bungalow where Hill was murdered. At trial, Coleman changed his story. He testified that the day before the murder, he had jogged on the football field, as was his habit, and then stopped in the bungalow to smoke a joint of marijuana. Coleman explained that he lied to the police, because he knew it was a murder investigation and he felt "scared, paranoid, and nervous."

Coleman supported the story he told at trial with an alibi. He testified that on the day of the murder, he was in college classes from 7:30 a.m. until approximately 3:30 p.m. (Hill was murdered between 5:00 p.m. and 9:00 p.m.) He left school with his friend, Carlton McAllister, and went to his ex-wife Clara Roper's home. McAllister stayed a few hours. Coleman and Roper stayed there until the next morning.

Coleman presented McAllister and Roper as witnesses to corroborate his alibi. McAllister testified that he met Coleman in Cloud Hall at the college at 2:30 or 3:30 p.m. The two of them took the bus to Roper's home, where McAllister stayed with Coleman for two hours. This testimony, however, was undercut somewhat when McAllister stated on cross-examination that he thought he met Coleman on the day of the murder, because that was the first day of classes, but it could have been any day that week.

Roper testified that on the day of the murder (which occurred on the first day of classes) Coleman came to her home with McAllister between 3:30 and 4:00 p.m. Coleman did not leave from the time he arrived until the next morning. On cross-examination, however, Roper testified that, although she had spoken with the police and the district attorney, she had never told them about Coleman being at her home on the day of the murder, because, as Coleman's ex-wife, she did not "want to contribute anything to the case."

Louis Lindsey, a teacher at the college, testified that he taught a class which met from 1:10 to 2:00 p.m. on the day of the murder. Lindsey's attendance records for that day showed that Coleman did not attend the class, but McAllister did. Lindsey marked Coleman's absence "excused," which meant that Coleman came to Lindsey sometime after class (that day or a later day) and requested to be excused.

Coleman retook the stand and explained that he had missed class because of a class conflict, although he could not remember which class caused the conflict.

We review de novo the district court's decision to grant or deny a petition for a writ of habeas corpus. Dyer v. Calderon, 122 F.3d 720, 726 (9th Cir.), reh'g granted (Oct. 9, 1997). We review for clear error the district court's findings of fact underlying its decision to grant or deny relief. Id. Because Coleman filed his petition before enactment of the Antiterrorism and Effective Death Penalty Act of 1996, 28 U.S.C.A. § 2261 (Supp.1998), the Act does not apply to this case. Jeffries v. Wood, 114 F.3d 1484, 1494 (9th Cir.), cert. denied, --- U.S. ----, 118 S. Ct. 586, 139 L. Ed. 2d 423 (1997).

Coleman attacks the constitutionality of his conviction by challenging the prosecution's use of the Hemastix test and bloody print evidence. He also argues his counsel was ineffective in responding to this evidence and to the serology evidence, and in not using to his advantage other evidence of a hair fiber found on the victim's hand. Additionally, he contends the police withheld information that they had other suspects. Finally, he contends he was denied due process because the foreman of his jury was a convicted felon and statutorily ineligible under California law to serve on the jury. We consider these challenges in order.

Coleman contends the prosecution's failure to disclose the Hemastix test results and bloody print evidence in advance of trial violated the Due Process and Confrontation Clauses of the Constitution.

Criminal defendants are constitutionally assured "a meaningful opportunity to present a complete defense." California v. Trombetta, 467 U.S. 479, 485, 104 S. Ct. 2528, 81 L. Ed. 2d 413 (1984). Although due process requires that prosecutors disclose exculpatory evidence, Brady v. Maryland, 373 U.S. 83, 83 S. Ct. 1194, 10 L. Ed. 2d 215 (1963), there is no constitutional requirement that defendants in state court proceedings be given notice of inculpatory evidence the prosecution plans to introduce at trial. Gray v. Netherland, 518 U.S. 152, 168-170, 116 S. Ct. 2074, 135 L. Ed. 2d 457 (1996). In Gray, the Due Process Clause was not violated even though the prosecution had misled the defense by stating that it would not introduce the evidence it then used at trial. Id. at 156-57, 116 S. Ct. 2074; see also Weatherford v. Bursey, 429 U.S. 545, 559-61, 97 S. Ct. 837, 51 L. Ed. 2d 30 (1977) (surprise testimony of undercover agent accomplice did not violate due process even though agent had told defendant and his counsel that he would not testify).

In the present case, Keane was misled because the transcript of the preliminary hearing reflected that Ihle had testified there were no bloody prints found at the crime scene, and because the latent print card did not note that the thumbprint on the back of the chair was bloody. The effect of this, coupled with the prosecution's violation of the discovery order, was that Keane did not learn of the bloody thumbprint or the Hemastix evidence until the time of trial.

The district court assumed without deciding that the prosecution did indeed violate the state court discovery order. We make the same assumption. This violation of state law, however, does not alter the holding of Gray. The evidence withheld from the defense was inculpatory evidence, and there is no constitutional requirement that a state prosecutor provide the defense with inculpatory evidence. Gray, 518 U.S. at 168, 116 S. Ct. 2074. And, absent some constitutional violation, a violation of state law does not provide a basis for habeas relief. Estelle v. McGuire, 502 U.S. 62, 67-68, 112 S. Ct. 475, 116 L. Ed. 2d 385 (1991).

Coleman argues that the prosecutor's failure to disclose the results of the Hemastix test, in violation of the state court discovery order, prevented him from having an adequate opportunity to cross-examine the state's witness, Ihle, and this violated the Confrontation Clause of the Sixth Amendment.

The Confrontation Clause prevents restrictions on cross-examination that "effectively ... emasculate the right of cross-examination itself." Delaware v. Fensterer, 474 U.S. 15, 19, 106 S. Ct. 292, 88 L. Ed. 2d 15 (1985) (quoting Smith v. Illinois, 390 U.S. 129, 131, 88 S. Ct. 748, 19 L. Ed. 2d 956 (1968)). The Confrontation Clause does not require that the prosecution disclose evidence that would help the defense effectively cross-examine a prosecution witness; the Confrontation Clause is not "a constitutionally compelled rule of pretrial discovery." Pennsylvania v. Ritchie, 480 U.S. 39, 52, 107 S. Ct. 989, 94 L. Ed. 2d 40 (1987) (plurality opinion) (holding that State did not have to turn over its confidential investigative files). "In short, the Confrontation Clause only guarantees 'an opportunity for effective cross-examination, not cross-examination that is effective in whatever way, and to whatever extent, the defense might wish.' " Id. at 53, 107 S. Ct. 989 (quoting Fensterer, 474 U.S. at 20, 106 S. Ct. 292 (emphasis in original)).

In this case, Coleman's right to confront witnesses was not violated. He exercised his right to confront and cross-examine Ihle at trial. There was no Confrontation Clause violation.

To prevail on a claim of ineffective assistance of counsel, Coleman must first show that "counsel's representation fell below an objective standard of reasonableness." Strickland v. Washington, 466 U.S. 668, 688, 104 S. Ct. 2052, 80 L. Ed. 2d 674 (1984). We "must indulge a strong presumption that counsel's conduct falls within the wide range of reasonable professional assistance." Id. at 689, 104 S. Ct. 2052. We conduct an "extremely limited" review of counsel's performance:

The test has nothing to do with what the best lawyers would have done. Nor is the test even what most good lawyers would have done. We ask only whether some reasonable lawyer at the trial could have acted, in the circumstances, as defense counsel acted at trial.

Dyer, 122 F.3d at 732 (quoting White v. Singletary, 972 F.2d 1218, 1220 (11th Cir.1992)).

These standards apply to the duty to investigate. Strickland, 466 U.S. at 690-91, 104 S. Ct. 2052. "[A] particular decision not to investigate must be directly assessed for reasonableness in all the circumstances, applying a heavy measure of deference to counsel's judgements." Id. at 691, 104 S. Ct. 2052.

Coleman must also show that any deficient performance prejudiced the defense. For errors that were committed during the guilt phase, "the question is whether there is a reasonable probability that, absent the errors, the factfinder would have had a reasonable doubt respecting guilt." Id. at 695, 104 S. Ct. 2052.

Coleman contends that he was denied effective assistance of counsel because his trial counsel, Keane, failed to properly investigate the physical evidence, failed to request a continuance to deal with the Hemastix test results, failed to investigate and deal with the serology evidence, and failed to take advantage of evidence favorable to the defense contained in a laboratory report concerning a hair fragment found on the victim's hand.

"The proper measure of attorney performance remains simply reasonableness under prevailing professional norms." Strickland, 466 U.S. at 688, 104 S. Ct. 2052. These norms are reflected in American Bar Association ("ABA") standards. Id. The ABA standards state that a defense attorney's "investigation should always include efforts to secure information in the possession of the prosecution and law enforcement authorities." ABA Standards for Criminal Justice, Standard 4-4.1 (approved 1979); see also United States v. DeCoster, 487 F.2d 1197, 1204 (D.C.Cir.1973). In any given case, then, one would expect an attorney to examine the physical evidence, especially in a prosecution involving the death penalty. Sims v. Livesay, 970 F.2d 1575, 1580-81 (6th Cir.1992) (counsel ineffective for failing to investigate exculpatory physical evidence for no apparent reason); see also Sanders v. Ratelle, 21 F.3d 1446, 1460 (9th Cir.1994) (counsel ineffective, in part, for not familiarizing himself with the physical evidence).

In the present case, however, Keane's performance was not constitutionally deficient. He did not examine the chair back, but he hired an expert who examined the latent print card. There was no notation on the card that the thumbprint on the chair back tested positive for blood, even though standard police practice called for such a notation. Keane also read the transcript of the preliminary hearing. That transcript revealed that at the preliminary hearing Ihle testified there were no bloody prints. Finally, Keane believed that the police officer's mention of bloody prints during Coleman's questioning was mere hyperbole, because the police were trying to get Coleman to confess.

Perhaps, notwithstanding the foregoing, the best lawyers, or most good lawyers, would have examined the chair back. In the circumstances of this case, however, we are satisfied a reasonable lawyer could have acted as Keane did. His performance was not constitutionally deficient. See Dyer, 122 F.3d at 732.

Coleman also argues that Keane's performance was deficient because, when the trial court allowed the introduction of the Hemastix test results, Keane, who was surprised by this evidence, did not request a continuance to consult an expert to help him deal with it.

Had Keane obtained a continuance and consulted with an expert, he could have more effectively rebutted the test results. The test, as Ihle performed it, was unreliable because Ihle failed to include positive, negative and substrate controls. More importantly, the Hemastix test reacts positively to such common substances as cleaning agents, bacteria and plant material. Cf. Clark v. Moran, 942 F.2d 24, 31 (1st Cir.1991) (effective cross examination of test showed that confirmatory test should have been performed and that the test reacted positively to a detergent that was used in the case). An expert would have helped Keane reinforce these points.

On the other hand, Ihle had testified that the Hemastix test was only a presumptive test. Although he did not explain what it meant for a test to be labelled "presumptive," he stated that the test did not distinguish between human blood and animal blood. He also admitted he was not a chemist, he did not know the chemical reaction which was responsible for the Hemastix test's results, and he had tested only a few substances for false positives.

Keane assessed the impact of this testimony at the time of trial. The record of the evidentiary hearing in the district court indicates Keane believed that his cross-examination had substantially weakened Ihle's testimony. Keane testified that after he had cross-examined Ihle, he thought Ihle's testimony was "absurd."

Our reading of the record persuades us that Keane's decision not to request a continuance does not indicate ineffective assistance. Instead, it indicates a trial counsel who met the Hemastix evidence as best he could and presented a cross-examination which cast doubt on the test's results. Seeking a continuance to confer with a defense expert, or to put a defense expert on the stand, might have reinforced the evidence. Moreover, the trial court had just ruled that the defense had sufficient notice of the test results, negating Keane's claim of surprise. In the face of this ruling, it is unlikely the trial court would have granted a motion for a continuance even if the motion had been made. See United States v. Fish, 34 F.3d 488, 494-95 (7th Cir.1994) (counsel not deficient for failing to make a futile motion for continuance). We conclude Keane did not render ineffective assistance by failing to move for a continuance.

Coleman contends that Keane was ineffective for failing to adequately investigate and deal with the serology evidence. Specifically, he argues Keane should have obtained the forensic serologist Blake's laboratory notes, which explained which tests Blake did and how those tests were conducted; and he should have presented his own serology expert as a witness at trial.

As we stated earlier, Keane's investigator gave Brian Wraxall, the defense expert, copies of Blake's reports. Neither Keane nor his investigator obtained Blake's notes. Wraxall testified at the evidentiary hearing in the district court that without those notes he could not give an opinion as to the correctness of the conclusions set forth in Blake's reports. On the other hand, Wraxall believed that he was hired to explain the meaning of the reports, rather than to give an opinion as to their validity. This may explain why Wraxall did not ask Keane or his investigator for Blake's notes.

In any event, in the absence of a specific request, counsel does not have a duty to gather background information which an expert needs. Hendricks v. Calderon, 70 F.3d 1032, 1038 (9th Cir.1995). But see Clabourne v. Lewis, 64 F.3d 1373, 1384-85 (9th Cir.1995) (deficient performance where counsel consulted expert a few days before expert testified and failed to provide both expert and the state's experts with defendant's mental health records). Part of the skill of an expert is to recognize what information he needs in order to render an opinion. Hendricks, 70 F.3d at 1038. "To require an attorney, without interdisciplinary guidance, to provide a psychiatric expert with all information necessary to reach a mental health diagnosis demands that an attorney already be possessed of the skill and knowledge of the expert." Id. at 1039. Keane was not required to know what information his serology expert needed.

Additionally, Keane was not required to request an expert opinion from his serology expert. Counsel may consult with an expert to uncover the weaknesses in the state's case; counsel need not hire an expert to give an opinion. Battle v. Delo, 64 F.3d 347, 353-54 (8th Cir.1995) (discussing a motion's court ruling concerning serological evidence).

Keane consulted with his expert, Wraxall, to prepare his cross-examination of Blake. That consultation was productive. It enabled Keane to establish that Blake failed to perform certain common tests which might have eliminated Coleman as a suspect. Because the tests Blake performed used up the entire sample, other exculpatory tests could not be performed. Keane was also able to bring out that, based upon Blake's tests, Coleman was one of five million men who could have deposited the semen. Keane's handling of the serology evidence was not constitutionally deficient. Compare Driscoll v. Delo, 71 F.3d 701, 708-09 (8th Cir.1995) (deficient performance where counsel merely read a three-page laboratory report, but failed to understand the serological tests performed and to rebut the prosecution's theory that conflicted with the report), cert. denied sub nom., Bowersox v. Driscoll, --- U.S. ----, 117 S. Ct. 273, 136 L. Ed. 2d 196 (1996).

Coleman contends that Keane was ineffective because he failed to investigate or present to the jury evidence that a one-half inch strand of hair found on the victim Hill's hand was "inconsistent" with Coleman's hair. We agree that this aspect of Keane's performance was deficient, but hold that Coleman suffered no prejudice as a result.

The State's criminologist, Robert Ogle, conducted two separate analyses of hair found at the crime scene. Ogle's first analysis was included in the report of the prosecution's forensic serologist, Blake. Ogle's analysis showed that hair found in Hill's pubic area could not be distinguished from Hill's own hair. Ogle's second analysis was set forth in a supplemental report. This supplemental report involved a hair fragment found on Hill's hand. That hair fragment, which lacked a root, was presented in a slide which included three cat hairs and some wool fibers. Ogle compared the hair fragment with numerous samples of Coleman's hair. He concluded that the hair fragment was "inconsistent" with Coleman's hair, although he could not "definitely eliminate[ ] [Coleman] as a source of the questioned hair."

The prosecution delivered a copy of Ogle's supplemental report to an investigator working in the Public Defender's office that represented Coleman. Neither Keane nor his co-counsel nor his investigator saw a copy of the report. During Keane's cross-examination of Blake, however, Blake referred to Ogle's supplemental report. Keane asked to see the supplemental report, looked at a copy and said he had never seen it before. The supplemental report was never admitted into evidence.

At the evidentiary hearing in the district court, Keane testified that when he was shown the supplemental report at trial, he thought it was another copy of Blake's earlier report, which discussed the analysis of the pubic hairs. This testimony is at odds with Keane's statement in the trial court that he had never before seen the supplemental report. In any event, Keane never read the supplemental report. If he had read it, he would have learned it contained an analysis of the fragment of hair found on Hill's hand, and that this hair fragment was inconsistent with Coleman's hair.

Given these facts, Keane's performance was constitutionally deficient. He clearly breached his duty to investigate. As the district court stated: "It is unreasonable for counsel in a capital case to ignore a two-page report that was handed to him, at his request, by a prosecution witness he was examining, when he insisted that he had not previously seen the report." See Baylor v. Estelle, 94 F.3d 1321, 1323-24 (9th Cir.1996) (ineffective assistance where counsel failed to follow up on a laboratory report that "tend[ed] to eliminate" the defendant as the source of the semen).

To determine if this deficient performance prejudiced Coleman, we must decide whether "there is a reasonable probability that, absent the error[ ], the factfinder would have had a reasonable doubt respecting guilt." Strickland, 466 U.S. at 695, 104 S. Ct. 2052. Had Keane introduced Ogle's report, the jury would have learned that the hair fragment found on Hill's hand was "inconsistent" with it being Coleman's hair. Had Keane consulted an expert, an analysis could have been developed that would have definitely excluded Coleman as a source of the hair fragment.

Coleman argues Keane's deficient performance prejudiced him because the hair fragment, which did not come from Coleman, most likely came from Hill's attacker. We disagree. The floor of the bungalow where Hill was murdered was covered with "debris and dust and dirt." While it is true the hair fragment did not come from Coleman, neither did the cat hairs or wool fibers also found on Hill's hand. The most logical source for all of this debris was the floor.

Moreover, the inconclusive nature of the hair fragment evidence stands in contrast to the probative force of the evidence the prosecution relied upon to establish Coleman's guilt. The jury heard the uncontroverted evidence that Coleman's palm print was found on a windowsill, bracketed by Hill's fingerprints, and that Coleman's thumbprint was found on a chair next to Hill's body. The serology evidence showed that Coleman was included in eight percent of the population that could have deposited the semen found in Hill's body. Finally, Coleman initially lied to the police when he said he had never been in the bungalow in which Hill was murdered, and he did not change that story until he testified at trial after the prosecution had presented its case.

In sum, there is no reasonable probability that had the supplemental report pertaining to the hair fragment been introduced into evidence, or used by Coleman to present the testimony of an expert witness on behalf of the defense, the jury would have had a reasonable doubt as to Coleman's guilt. Coleman suffered no prejudice within the meaning of Strickland.

Coleman contends the State's failure to disclose information about other suspects violated his due process rights under Brady v. Maryland, 373 U.S. 83, 83 S. Ct. 1194, 10 L. Ed. 2d 215 (1963). Under Brady, due process is violated when the prosecution suppresses evidence that is favorable to the defense and material to the issues of guilt or punishment. Moore v. Illinois, 408 U.S. 786, 794-95, 92 S. Ct. 2562, 33 L. Ed. 2d 706 (1972).

The police investigated four other suspects: Hill's current boyfriend; a man who had been bothering Hill and had served time for rape; a man who had committed a rape and kidnapping near the school where Hill was killed; and a former boyfriend who had been following Hill around, and who frightened Hill because he was so violent. The police eliminated these suspects because their fingerprints did not match the fingerprints found at the crime scene. The police did not check the suspects' blood types, nor did they check the suspects' hair against the hair fragment found on Hill's hand.

The district court found, and the State does not dispute, that the prosecution was aware of these other suspects, but did not disclose this information to Coleman. Nor does the State dispute that information about these other suspects was favorable to Coleman. The success of Coleman's Brady claim depends, therefore, on whether the information about the other suspects was material.

To be material, evidence must be admissible or must lead to admissible evidence. United States v. Kennedy, 890 F.2d 1056, 1059 (9th Cir.1989). In order for evidence of another suspect to be admissible, "there must be direct or circumstantial evidence linking the third person to the actual perpetration of the crime." People v. Hall, 41 Cal. 3d 826, 833, 718 P.2d 99, 104, 226 Cal. Rptr. 112, 116 (1986). Motive or opportunity is not enough. Id.

In this case, the information about the other suspects consisted of (1) police files on the convicted kidnapper/rapist and (2) police notes of two telephone calls reporting that Hill was bothered by both a violent ex-boyfriend and a convicted rapist. Coleman argues that had he been given this information he could have conducted a timely investigation which might have uncovered witnesses who saw these other suspects bothering Hill or being violent towards her.

What is missing in Coleman's argument is any evidence linking any of these other suspects to the crime, or any showing that disclosure of their existence would have led to the discovery of admissible evidence. This missing link compels the conclusion that the information about the other suspects was not material. There was no Brady violation.

Coleman contends that his right to due process was violated because the foreman of his jury was a convicted felon. In fact, the foreman had been convicted of at least three felonies.2

The Sixth Amendment does not bar ex-felons from jury service. United States v. Boney, 977 F.2d 624, 633 (D.C.Cir.1992); see also Coughlin v. Tailhook Ass'n, 112 F.3d 1052, 1058-59 (9th Cir.1997) (approving Boney ). Under California law, however, felons may not serve on juries. Cal. Const. Art. VII § 8(b) (West 1996); Cal.Civ.Proc.Code § 203(a)(5) (West Supp.1998) (formerly codified at Cal.Civ.Proc.Code § 199(b)). Coleman contends California law has created a federal due process liberty interest in a jury composed of non-felons, and the state violated this liberty interest. We disagree.

State law may create a liberty interest protected by the Due Process Clause of the Fourteenth Amendment. Sandin v. Conner, 515 U.S. 472, 483-84, 115 S. Ct. 2293, 132 L. Ed. 2d 418 (1995). In order for state law to create a constitutionally protected liberty interest, however, that interest must be of "real substance." Id. at 478, 115 S. Ct. 2293 (quoting Wolff v. McDonnell, 418 U.S. 539, 557, 94 S. Ct. 2963, 41 L. Ed. 2d 935 (1974)). A juror disqualification "defect is not fundamental as affecting the substantial rights of the accused, and the verdict is not void for want of power to render it." Kohl v. Lehlback, 160 U.S. 293, 302, 16 S. Ct. 304, 40 L. Ed. 432 (1895) (noncitizen juror in habeas case); see also Hogue v. Scott, 874 F. Supp. 1486, 1529 (N.D.Tex.1994) (Texas law that disqualifies felons from jury service did not create a constitutionally protected interest), aff'd sub nom., Hogue v. Johnson, 131 F.3d 466 (5th Cir.1997), cert. denied, --- U.S. ----, 118 S. Ct. 1297, 140 L. Ed. 2d 334 (1998).

Although California law precludes felons from serving on juries, this does not create a liberty interest protected by the Constitution. The right to a jury of non-felons is not so fundamental that it affects "the substantial rights of the accused...." Kohl, 160 U.S. at 302, 16 S. Ct. 304. Accordingly, Coleman was not denied due process because a felon served on his jury, or because that juror was the foreman.

Primary among Coleman's challenges to his death sentence is his contention that the state trial judge committed constitutional error in instructing the jury on the power of the governor of California to commute Coleman's sentence to life imprisonment with the possibility of parole. The state trial court instructed the jury:

You are instructed that under the State Constitution, a governor is empowered to grant a reprieve, pardon or commutation of a sentence following conviction of the crime. Under this power, a governor may in the future commute or modify a sentence of life imprisonment without the possibility of parole to a lesser sentence that would include the possibility of parole. So that you will have no misunderstandings relating to a sentence of life without possibility of parole, you have been informed generally as to the Governor's commutation power.

You are now instructed, however, that the matter of a Governor's commutation power is not to be considered by you in determining the punishment for this defendant. You may not speculate as to, if, or when a governor would commute the sentence to a lesser one which includes the possibility of parole.

The district court correctly held that this instruction as applied to Coleman gave the jury inaccurate information about the governor's commutation power and potentially diverted the jury's attention from the mitigation evidence presented. For that reason, we affirm the district court's grant of the writ as to Coleman's sentence.

A commutation instruction is unconstitutional when it is inaccurate. Hamilton, 17 F.3d at 1160 (modified Briggs instruction);3 see also McLain v. Calderon, 134 F.3d 1383 (9th Cir.1998) (modified Briggs instruction) (petition for cert. filed June 1, 1998); Gallego v. McDaniel, 124 F.3d 1065 (9th Cir.1997), cert. denied, --- U.S. ----, 118 S. Ct. 2311, 141 L. Ed. 2d 169, and cert. denied, --- U.S. ----, 118 S. Ct. 2299, 141 L. Ed. 2d 159, 1998 WL 99750. This rule relies upon the Supreme Court's emphasis on the importance of accuracy in the description of sentencing alternatives in a death penalty proceeding. California v. Ramos, 463 U.S. 992, 1009, 103 S. Ct. 3446, 77 L. Ed. 2d 1171 (1983).

The instruction in this case, just like the instruction held unconstitutional in McLain,4 was inaccurate. Because Coleman had numerous prior felony convictions, the governor alone did not have the power to commute a sentence of life imprisonment without the possibility of parole to a sentence of life imprisonment with the possibility of parole. The governor had to consult with the Board of Prison Terms (although he need not abide by their recommendation) and obtain the approval of four justices of the California Supreme Court. See Cal. Const. Art. 5, § 8 (West 1996); Cal.Penal Code §§ 4802, 4813, 4850-52, 4852.16 (West 1982).

The value of an accurate commutation instruction is its ability to clarify for the jury the meaning of terms such as "life without the possibility of parole." See Ramos, 463 U.S. at 1009, 103 S. Ct. 3446. An accurate description of sentencing alternatives helps the jury focus on its task of evaluating whether the circumstances of the case and the characteristics of the defendant warrant the imposition of a sentence of death. Where, as here, the court misinforms the jurors about the nature of one sentencing option, the jurors no longer understand the choice they are asked to make. Here, the court inaccurately informed the jury that, unless they sentenced Coleman to death, he could be released into the community on the authority of one person, the Governor. This assertion dramatically overstates the possibility of commuting the life sentence of a person such as Coleman, who had numerous prior felony convictions. The jury was "invited to speculate" that Coleman could be effectively isolated from the community only through a sentence of death. Hamilton, 17 F.3d at 1162. Such an instruction inappropriately minimizes the viability of all sentencing options other than death and prevents the jury from deliberating with the guided discretion required by the Eighth Amendment. See Gregg v. Georgia, 428 U.S. 153, 189, 96 S. Ct. 2909, 49 L. Ed. 2d 859 (1976).

The State argues that the commutation instruction given to Coleman's jury is not inaccurate; it is merely incomplete. Because a state decides "how much (if at all) juries [should] be informed about the postsentencing legal regime," O'Dell v. Netherland, 521 U.S. 151, ----, 117 S. Ct. 1969, 1977, 138 L. Ed. 2d 351 (1997), a state may choose to provide the jury with incomplete information.

The information in the commutation instruction given to Coleman's jury, however, is not simply incomplete; it is inaccurate as applied to Coleman. We held in McLain that a commutation instruction is prejudicially inaccurate when it creates the false impression that the Governor, acting alone, can commute a life sentence. See McLain, 134 F.3d at 1386; cf. Gallego, 124 F.3d at 1074-76 (Nevada commutation instruction inaccurate as applied to defendant because defendant may not be paroled if he must answer for a crime in another jurisdiction).

The State also argues that imposing a constitutional requirement to provide more complete commutation information is impermissibly retroactive. Cf. O'Dell v. Netherland, 521 U.S. 151, 117 S. Ct. 1969, 138 L. Ed. 2d 351 (1997) (holding the requirement to inform juries that a defendant was not eligible for parole if the prosecution argues future dangerousness was a new rule under Teague v. Lane, 489 U.S. 288, 109 S. Ct. 1060, 103 L. Ed. 2d 334 (1989)). We disagree. Hamilton imposes no such requirement. Hamilton does not require states to give any particular information about commutation. It simply requires that any information that a state chooses to give be accurate. The instruction in this case was not accurate.

Finally, the State argues that even if the instruction was inaccurate, this error did not have a "substantial and injurious effect or influence" on the jury's sentence of death. Brecht v. Abrahamson, 507 U.S. 619, 637, 113 S. Ct. 1710, 123 L. Ed. 2d 353 (1993) (quoting Kotteakos v. United States, 328 U.S. 750, 776, 66 S. Ct. 1239, 90 L. Ed. 1557 (1946)).

To decide this question, we look to Boyde v. California, 494 U.S. 370, 110 S. Ct. 1190, 108 L. Ed. 2d 316 (1990). McDowell v. Calderon, 130 F.3d 833, 838 (9th Cir.1997) (en banc), cert. denied, --- U.S. ----, 118 S. Ct. 1575, 140 L. Ed. 2d 807 (1998). When the inaccuracy undermines the jury's understanding of sentencing options, "there is a reasonable likelihood that the jury has applied the challenged instruction in a way that prevents the consideration of constitutionally relevant evidence." Boyde, 494 U.S. at 380, 110 S. Ct. 1190.

We conclude the district court did not err in holding that Coleman was denied due process by the state trial court's inaccurate commutation instruction. Because we affirm the district court's issuance of the writ of habeas corpus as to Coleman's death sentence on the ground that the state trial court's commutation instruction was constitutionally deficient, we do not reach Coleman's other challenges to the penalty phase of his trial.

AFFIRMED.