Aaron REYNOSO, Petitioner-Appellee, v. George J. GIURBINO, Warden, Respondent-Appellant, 462 F.3d 1099

REINHARDT, Circuit Judge: The State, through Warden Giurbino, appeals the district court’s decision to grant Aaron Reynoso’s petition for a writ of habeas corpus. It asserts that the claim on which relief was granted was unex-hausted and that the state court’s rejection of the claim was reasonable. We conclude that Reynoso’s claim was properly exhausted and that, on the merits, he has demonstrated ineffective assistance of counsel under Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984).

We agree with the district court that the state court’s decision to the contrary constituted an unreasonable application of clearly established Supreme Court law. Accordingly, we affirm its grant of the writ of habeas corpus. I. On July 11, 1995, Jyotsna Prajapati was shot once in the head and died from her wounds shortly afterwards. At the time of the shooting, Prajapati was working alone behind the counter of the Top Produce Market, a convenience store that she and her husband owned.

After investigating Prajapati’s murder for a week, the Los Angeles Police Department asked the Los Angeles City Council to approve a $25,000 reward “for information leading to the arrest and conviction of the person or persons responsible for the murder of Jyotsna Prajapati.” On July 18, 1995, the City Council approved the reward for sixty days, during which time no witnesses responded. Two years later, on April 30, 1997, at the investigating officer’s request, the City Council renewed the reward for another sixty days.

In July of 1997, after seeing advertisements about the reward on television, Luis Alberto Lopez contacted police and told them that he had heard Reynoso confess to involvement in the murder. In September of 1997, police questioned Luis Hinojo-sa, a fellow gang member of Reynoso’s, about the murder. Hinojosa was questioned at Centinela State Prison where he was serving a sentence for burglary, possession for sale, and a probation violation.

After first unsuccessfully attempting to implicate his cousins, Hinojosa implicated Reynoso, saying he had heard him admit his participation in the murder at a party two years earlier. The detective leading the Prajapati murder investigation, David Escoto, believed that the case was “close,” but that the evidence was not sufficient to support a successful charge against Reynoso. He believed that he needed “more information” and “witnesses to come forward” to make the case. Escoto asked the City Council to renew the reward once again, which it did on October 21, 1998.

The reward was publicized both on television and in newspapers. In November of 1998, Detective Escoto tracked down Robert Mendoza, a witness with whom police had spoken on the night of Prajapati’s murder. In December of 1998, Javier Terrones, another witness, contacted the police. Each witness identified Reynoso as one of the two men that he claimed to have seen inside the convenience store moments before and moments after Prajapati’s murder.

Although the investigating officers were unable to find any physical evidence tying Reynoso to the murder, he was arrested and indicted almost three years after the shooting had occurred. A. The Trial At trial, in January 2000, only four witnesses were called to testify on the State’s behalf: two were the eyewitnesses identified above, one of whom died before the evidentiary hearing in district court, and the other two were the witnesses who claimed to have heard admissions made by Reynoso, one of whom recanted his statement at trial. i.

Luis Alberto Lopez Luis Alberto Lopez was the first of the four witnesses to contact police with incriminating evidence against Reynoso. In 1995, he had been incarcerated in juvenile hall with Reynoso, and he claimed to have heard Reynoso describe the Prajapati murder at that time. According to Lopez, Reynoso said that he and a few friends had gone to the Top Produce Market to do a “beer run” and, during the run, a woman was shot. Even though Reynoso never explicitly confessed to shooting Prajapati, Lopez testified, he physically acted out the shooting as if he were the shooter.

Even though Lopez purportedly heard Reynoso’s account of the shooting in 1995, he testified at trial that he did not report this information to police until 1997 because it was not in his interest to do so before then. In June of 1997, he said, he experienced a religious conversion and decided to contact police because he believed it “would be the right thing to do.” Also around that time, Lopez said, he saw a television broadcast that discussed the unsolved shooting of Prajapati, showed a composite of the robber, and included a description of the reward.

Lopez said that he also saw an advertisement describing the reward and stating that police had not yet found the suspect. Lopez testified at trial that the broadcast had refreshed his memory and had motivated him to come forward. He also conceded that, in addition to his new-found religious convictions, the offer of a reward motivated him to report Reynoso’s incriminating statements. Lopez could not remember whether he initially asked about the reward when he contacted the police, but his testimony suggested he had inquired about it before testifying.

At trial, Reynoso’s defense counsel attempted to undermine Lopez’s credibility by pointing out that he remembered few details about Reynoso or Reynoso’s account of the shooting. Lopez acknowledged that he had seen television reports about the murder, and defense counsel suggested that Lopez had learned of the facts to which he testified by watching those reports. Defense counsel also questioned Lopez about the $25,000 reward; he responded that the reward had played a role in his decision to contact police, but that the money had not been his main concern.

At the evidentiary hearing held in federal court in 2003, Lopez explained that when he contacted the police, he wanted only to leave an anonymous tip, but that the officers forced him to provide answers to their questions: “[T]hese people pressured me into saying all this.” He testified that the detectives had told him what he had to say regarding what Reynoso had said to him and that they had “pretty much walked [him] through the process of what to say and how to say it.” Even so, he maintained that he had testified truthfully at Reynoso’s trial.

At the evidentiary hearing, Lopez stated that he did not remember whether he was told that he was guaranteed reward money if he testified, but he did vaguely remember being told that the reward would be distributed “after everything was over.” Lopez repeatedly denied any interest in the reward, although he did admit that seeing the reward on television prompted him to come forward as a witness. He also testified at the evidentiary hearing that he was ultimately paid $10,000 for providing the information that helped convict Reynoso. ii.

Luis Hinojosa Hinojosa spoke with the police in September of 1997, after the reward had been renewed and while he was in prison. Having been told that he was a suspect in the murder and fearing another conviction, he first tried, unsuccessfully, to incriminate his cousins, stating that they looked like him. Subsequently, he told police that he had been at a party approximately one week after the incident at which Reynoso was present. According to Hinojosa, Rey-noso told him that he and another gang member went on a beer run and that when the victim tried to stop him, the gun slipped and shot her in the head.

Specifically, Hinojosa told police, Reynoso was pistol-whipping Prajapati when the gun went off. At trial, Hinojosa recanted his prior statements, testified that he did not know Reynoso, and that he had learned about the facts surrounding the murder through news reports on television. He also recanted his statement that he and his cousins looked alike, explaining that he feared being convicted for the murder himself. During direct examination, Hi-nojosa denied any knowledge about the reward being offered; defense counsel did not question him about the reward on cross-examination.

Despite the recantation, the prosecutor urged the jury to credit Hinojosa’s report of Reynoso’s purported admissions. iii. Javier Terrones Javier Terrones was one of the two witnesses who placed Reynoso at the Top Produce Market moments before Prajapa-ti’s murder. Although aware of the shooting, Terrones did not contact police at the time and spent the next couple of years in Denver, Colorado.

Upon his return to California in 1998, Prajapati’s husband— who had known Terrones in the past, when he had frequented their store — showed him several articles regarding the apprehension of a suspect in his wife’s murder, one of which had a picture identifying Rey-noso as a suspect. Terrones testified at trial that he recognized the picture of Rey-noso as the individual he saw at the store on the night of the murder. Instead of immediately contacting the police, however, Terrones left his contact information with Prajapati’s husband to give to them.

On December 11, 1998, Detective Escoto interviewed Terrones, who subsequently identified Reynoso in both a photographic and a live lineup. At trial, Terrones testified that he left the store before the shooting occurred and was only able to describe vaguely the person he saw the night of the murder. He testified that on the night in question, he saw a man walk past him and stand three or four feet away from him. Aside from remembering that the person had a dirty face and wore “eholo”-type clothes, he could not remember any other details about his appearance.

Terrones testified that he saw the man for only a “matter of seconds,” from only a “sideway glance” from approximately twenty feet away. Nevertheless, he identified the man as Reynoso. On cross-examination, defense counsel challenged Terrones’s identification. Through her questioning, she suggested that Terrones’s exposure to the newspaper photograph had tainted his subsequent identification.

Further, on cross-examination, Terrones admitted that he “didn’t have much time to examine [the man at the store].” Defense counsel did not question Terrones as to whether he knew of the reward and whether he expected to receive a portion of it in exchange for his testimony. In fact, as the magistrate judge noted in her Report and Recommendation, the reward was not discussed at all during the entire time that Terrones testified. At the evidentiary hearing in district court, Terrones explained that he learned of the reward through Prajapati’s husband upon his return to California.

Terrones testified that he initially contacted police because he was angry and upset about the murder and that he did not want a share of the reward. When he did speak with the police, however, Terrones asked about the reward and told officers, “if you are going to give me some compensation or something, I will accept it,” with the belief that he would get a share “when everything was done.” The Deputy District Attorney testified at the evidentiary hearing that she had discussed the reward with him before trial, and Terrones testified that he had also discussed the reward with the police after the trial ended.

Terrones ultimately received $7,500 for his testimony. He testified at the evidentiary hearing that at the time of trial, he was making $8.00 an hour as a painter, work was slow, and he needed the money. iv. Robert Mendoza Robert Mendoza was the other witness to place Reynoso at the scene of the murder. Like Terrones, he did not see the shooting, but testified that he saw Reynoso in the store before it occurred and saw him running out of the store afterwards. Mendoza also testified at trial that he had been smoking crack cocaine that morning and had been drinking alcohol all day.

According to Mendoza, as he was walking out of the store, he saw Reynoso and another man walk in and, shortly thereafter, he saw the two leave the store quickly. From outside the store, he testified, he observed Reynoso leave with a pack of beer in his hand and his companion holding what appeared to be a gun. Mendoza returned to the store and found Prajapati lying on the floor; however, instead of calling the police, he said, he left. He testified that he tried to call the police but the number had been busy.

He did not speak to the police until later that evening, at which point he told them that the taller of the two men had tattooed lettering across his back. On July 25, 1995, about two weeks after the murder, Mendoza singled out two individuals other than Reynoso from a photographic lineup and said that they looked like the men whom he had seen in the store. Three years later, in November of 1998, Detective Escoto tracked down Mendoza and asked him to look at another photographic lineup. This time, Mendoza identified Reynoso as the man who he had seen with beer in his hand as the two men ran out of the store.

In a subsequent live lineup, he identified someone other than Reynoso as the man who had beer in his hand before changing his mind and again identifying Reynoso. At the time of Rey-noso’s trial, immediately before testifying, Mendoza was accidentally placed in the same cell with Reynoso and did not recognize him. Mendoza testified that he did not recognize Reynoso for approximately 30 minutes, even though he saw his face, until he saw his distinctive ears.

Later in his testimony, however, Mendoza admitted that it was only after the prosecutor asked him whether he recognized anyone in the cell that he learned that he had been in the same cell as Reynoso. On cross-examination, defense counsel challenged Mendoza’s credibility on multiple grounds. First, she questioned Mendoza regarding his use of crack cocaine and alcohol.

Mendoza admitted that he had smoked “a lot” of crack on the morning of the shooting and had been drinking continuously from the morning until after Prajapati’s murder; he further testified that on the day before the shooting, he drank two to three six-packs of beer and shared approximately $120 worth of crack cocaine with two other people. Additionally, he conceded that he was addicted to cocaine and was under the influence of alcohol when he saw Reynoso and his companion enter the store.

Second, defense counsel focused attention on Mendoza’s repeated misidentifications of Reynoso and his other inconsistent statements. Defense counsel questioned Mendoza about his inconsistent identifications of lettering tattooed across Reynoso’s back, statements he made to police regarding whether he was “messy” or “cleaned up” on the morning of the shooting, and whether he had previously known the person with whom he was drinking on the day of Praja-pati’s murder. Even though defense counsel challenged Mendoza’s credibility on various grounds, she did not question him about the reward.

Although it is not clear exactly when Mendoza learned about the financial offer, he was aware of it when he was contacted three years after the murder by Detective Escoto, after it had been renewed by the City Council in October 1998, and before he testified at Reynoso’s trial. Detective Escoto testified at the federal evidentiary hearing that when he contacted Mendoza in 1998, Mendoza was aware that a reward had been offered and “wasn’t interested in any of the reward.” He further testified that he did not tell Mendoza that his eligibility for the reward would be contingent upon Reynoso’s conviction.

In contrast, the prosecutor recalled speaking to Mendoza about the reward and remembered that he had said that he wanted his share to go to his sister because he was in INS custody. Detective Escoto testified at the evidentiary hearing that, ultimately, Mendoza was given $7,500 for his testimony. On January 21, 2000, the jury convicted Reynoso of one count of first degree murder and one count of second degree robbery. Additionally, the jury found that he committed the murder during the commission of a robbery or burglary and that a principal was armed with a firearm during the commission of the offense.

It did not find that Reynoso personally used a firearm. The trial court sentenced him to twenty-five years to life. B. Reynoso’s Trial Counsel Rose Regios, a Deputy Public Defender, represented Reynoso at his trial for the murder of Prajapati and knew before trial that a reward had been offered. The prosecutor provided discovery documents to Regios, which contained a copy of the original reward and a newspaper

article discussing the renewal of the reward. Regios acknowledged that the prosecutor had disclosed the existence of the reward before trial. There is disagreement between the State and Regios as to whether the prosecutor told her that each of the witnesses knew about the reward.

The prosecutor testified at the evidentiary hearing that she and Regios “talked about the fact that the witnesses knew about the rewards,” and that the reward was “such common knowledge” that “it wasn’t an issue.” In contrast, Regios testified that although the prosecutor had disclosed that Lopez and Hinojosa both knew of the reward, she was never informed that either Terrones or Mendoza knew about it, and she could not remember whether either she or her investigator had interviewed Terrones or Mendoza before trial.

She stated that as a result she was unaware before and during trial that Terrones and Mendoza knew of or had applied for the reward. Even though Regios did know that a reward had been offered, at trial she questioned only Lopez about it. Reglos’s Declaration of March 11, 2003 asserted that even if she had been informed that Terrones and Mendoza knew about the reward, she would not have questioned them about it.

She explained that she had planned to attack their credibility through various factors — specifically, the fact that Terrones had seen a photograph of Reynoso before identifying him and Mendoza’s drug use at the time of the murder — and declared that “challenging [Terrones and Mendoza] based on their knowledge of the reward money would have diluted what [she] fe[lt] were strong issues for the defense.” At the evidentiary hearing in July of 2003, however, Regios changed her position.

The magistrate found that “[although she initially maintained she would not have questioned either Terrones or Mendoza about the reward, she later admitted that questioning them about the reward would have been consistent with her other cross-examination. In fact, she conceded that cross-examining Mendoza about the reward would have further undermined his credibility.” Regios also admitted that she could not have made a strategic decision about cross-examining Terrones and Mendoza about the reward without first knowing that they knew of it.

Specifically, she stated that knowledge of the reward would have put a “different spin on things in [her] investigation, in [her] preparation, and just strategically on what [she] would have been doing.” C. Post-Conviction Proceedings Reynoso appealed his conviction to the California Court of Appeal on February 15, 2000. In his appeal, Reynoso focused on two instructional errors in his direct appeal, neither of which was raised in the district court. While the appeal was pending, Reynoso filed a petition for writ of habeas corpus in the California Court of Appeal.

In that petition, he argued that the prosecution had violated Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), by failing to disclose that two witnesses were aware of the reward, that he received ineffective assistance of counsel, and that California’s reward system violated his due process rights. The California Court of Appeal affirmed Reynoso’s conviction and denied his habeas petition without providing any reasoning for its denial. Reynoso filed a petition for review in the California Supreme Court, which denied the petition without comment or citation to authority.

Reynoso next filed a petition for a writ of habeas corpus in the United States District Court for the Central District of California, arguing the same three issues raised in his state habeas petition. The State filed a return and Reynoso subsequently filed a traverse. The magistrate judge held an evidentiary hearing on Rey-noso’s Brady and ineffective assistance of counsel claims. The magistrate judge issued a Report and Recommendation, recommending that the district court deny the Brady and due process claims but grant relief on the ineffective assistance of counsel claim.

The district judge adopted the magistrate judge’s Report and Recommendation and entered judgment granting the writ of ha-beas corpus. An amended judgment was entered upon the State’s request. The State filed a timely notice of appeal. II. The district court’s grant or denial of a habeas petition is reviewed de novo. Leavitt v. Arave, 383 F.3d 809, 815 (9th Cir.2004) (per curiam) (as amended). Factual findings made by the district court are reviewed for clear error. Lambert v. Blodgett, 393 F.3d 943, 964 (9th Cir.2004).

Claims of ineffective assistance of counsel are mixed questions of law and fact, and are therefore reviewed de novo. Beardslee v. Woodford, 358 F.3d 560, 569 (9th Cir.2004) (as amended). Reynoso’s federal habeas petition was filed after April 24, 1996, and is therefore governed by the Antiterrorism and Effective Death Penalty Act (“AEDPA”). Woodford v. Garceau, 538 U.S. 202, 204, 210, 123 S.Ct. 1398, 155 L.Ed.2d 363 (2003).

Under AEDPA, we may grant habeas relief only when the state court’s decision was “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States” or “was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” 28 U.S.C. § 2254(d); Wiggins v. Smith, 539 U.S. 510, 520, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003). In Williams v.

Taylor, 529 U.S. 362, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000), the Supreme Court held that a state court’s application of federal law is unreasonable only if it is “objectively unreasonable.” Id. at 409, 120 S.Ct. 1495. When, as in the instant case, “no reasoned state court decision denying a habeas petition exists,” this court must assume that the state court has decided all the issues and “perform an ‘independent review of the record’ to ascertain whether the state court decision was objectively unreasonable.” Pham v. Terhune, 400 F.3d 740, 742 (9th Cir.2005) (per curiam) (quoting Himes v.

Thompson, 336 F.3d 848, 853 (9th Cir.2003) (quoting Delgado v. Lewis, 223 F.3d 976, 982 (9th Cir.2000))) (internal quotation marks omitted). When it is clear, however, that the state court has not decided an issue, we review that question de novo. Rompilla v. Beard, 545 U.S. 374, 377, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005). III. AEDPA’s exhaustion requirement, encompassed by 28 U.S.C. § 2254(b), “ensures that the state courts have the opportunity fully to consider federal-law challenges to a state custodial judgment before the lower federal courts may entertain a collateral attack upon that judgment.” Duncan v.

Walker, 533 U.S. 167, 178-79, 121 S.Ct. 2120, 150 L.Ed.2d 251 (2001); see also Duncan v. Henry, 513 U.S. 364, 365, 115 S.Ct. 887, 130 L.Ed.2d 865 (1995) (holding that claims may be reviewed by a federal habeas court only if they have been “fairly present[ed]” to the state courts so as to give the state the “opportunity to pass upon and correct alleged violations of its prisoners’ federal rights” (citing Picard v. Connor, 404 U.S. 270, 275, 92 S.Ct. 509, 30 L.Ed.2d 438 (1971)) (internal quotation marks omitted)). The § 2254(

b) exhaustion requirement means that petitioners must present to the state court both the facts necessary to state a claim for relief and the federal legal theory on which that claim is based. Gray v. Netherland, 518 U.S. 152, 162-63, 116 S.Ct. 2074, 135 L.Ed.2d 457 (1996).

The State characterizes the ineffective assistance claim on which the district court ultimately granted relief as a failure “to either interview or cross-examine either Terrones or Mendoza about the reward” and the claim that Reynoso presented in state court as a failure “to investigate the reward file and the timing of the identifications vis-a-vis the renewal of the reward.” Given these varying characterizations, the State argues that Reynoso has failed to exhaust the claim on which the district court granted his petition.

We see no merit either to the dichotomy the State attempts to create or to the State’s ultimate position on exhaustion. Because Rey-noso’s state habeas petition and the subsequent filings by both Reynoso and the State demonstrate that his claim of ineffective assistance of counsel, as addressed by the district court, was in fact “fairly presented” to the state court, we conclude that he has exhausted his ineffective assistance of counsel claim.

Contrary to its current position, the State conceded in the district court that Reynoso’s ineffective assistance claim—as it was presented to the state court—included the issue of counsel’s failure to elicit through cross-examination evidence from Terrones and Mendoza that they knew about the reward and expected to receive reward money in exchange for their testimony.

In its reply to one of petitioner’s filings in the district court, the State acknowledged: “In the portion of the Memorandum of Points and Authorities alleging ineffective assistance of counsel in the state habeas petition, Petitioner did request an evidentiary hearing on trial counsel’s tactical reasons for not cross-examining witnesses Mendoza and Ter-rones about the possibility of a reward.” As further confirmation, in its order denying one of the State’s motions, the district court reported that the State “belatedly concede[d] that Petitioner did request an evidentiary hearing in [his] state habeas petition on trial counsel’s tactical reasons for not cross-examining witnesses Mendoza and Terrones about the possibility of a reward.” The State is bound by its concessions.

See Russell v. Rolfs, 893 F.2d 1033, 1038-39 (9th Cir.1990). Aside from the State’s concessions, it is clear that the failures of counsel the district court addressed were fairly presented to the state court.

In the state habeas petition he filed in the California Court of Appeal, Reynoso asserted that “[i]f the jury had known the other two prosecution witnesses were eligible to receive the award, this would have cast the prosecution’s case in an entirely different light” and that such information “would have cast considerable doubt on the credibility of the two identification witnesses by suggesting they were promised a reward for identifying petitioner.” In that petition, Reynoso requested, in the alternative, that the state court hold “an evidentiary hearing to determine whether counsel had a tactical reason for failing to elicit this evidence or investigate the reward file.” In sum, a review of the state court proceedings demonstrates that the issue the district court decided was fairly presented to the state court. rv.

A. Deficient Conduct Reynoso claims that his trial counsel’s failure to investigate the issue of the reward—including the failure to interview the two eyewitnesses, Terrones and Mendoza, about their knowledge of the offer and their expectation of receiving a substantial payment if their testimony helped convict him—constituted ineffective assistance of counsel. Similarly, he contends that he was denied effective representation when his counsel failed to cross-examine the two witnesses about the reward, and their expectations of receiving it, for the purposes of impeachment and demonstrating bias.

It is unclear from the record whether Reynoso’s trial counsel, Rose Regios, knew at the time of trial that Mendoza and Terrones were aware of and had inquired about the reward. At the evidentiary hearing, the prosecutor insisted that the reward was “common knowledge” and that she and Regios “discussed the fact that all three witnesses knew about the reward.” “[Regios] knew the witnesses knew,” the prosecutor asserted — “I knew the witnesses knew; the witnesses knew.

It wasn’t any kind of issue at the time.” At the same time, the prosecutor could not recall whether she had told Regios specifically about Mendoza and Terrones’s knowledge of the reward, let alone their interest in it. She testified that she had “no memory” of telling Regios that Ter-rones had discussed the reward with a detective before trial and explained that maybe she “just forgot to tell [Regios] or figured that it was in the report” that she had turned over to the defense.

Similarly, she had no “recollection, one way or another,” whether she had told Regios that she had personally discussed the reward with Mendoza. At no point did she contend that she had informed Regios that either of the two eyewitnesses had expressed an interest in receiving the proceeds of the reward. Defense counsel testified that before and during trial she was unaware that all of the State’s witnesses knew about the reward.

Specifically, she insisted that she was never informed that Mendoza and Terrones knew about the reward and was ignorant of the fact that they had both inquired about their eligibility for it.

She testified that knowing that Mendoza and Terrones had knowledge of the reward— information she could have obtained by interviewing them prior to trial — would have put “a whole different spin on things in [her] investigation, in preparation, and just strategically what [she] would have been doing.” She further testified that she could not have made a tactical decision regarding whether to cross-examine Mendoza and Terrones about the reward because she did not know that they were aware of it.

In view of the testimony at the eviden-tiary hearing, the most likely explanation of what transpired is that the prosecutor and defense counsel had some discussion about the reward, but for whatever reason — perhaps because the conversation lacked sufficient specificity — Regios either failed to receive or failed to understand the full message; in short, whether the prosecutor or defense counsel was at fault, and it may have been both, Regios did not comprehend that Mendoza and Terrones had acknowledged their awareness of the reward, and she had no knowledge that they had expressed an interest in receiving it.

This explanation also appears to be most consistent with the magistrate judge’s findings. Magistrate Judge Lum found the prosecutor to be a “particularly credible witness.” At the same time, throughout her analysis of the issue, the magistrate judge accepted as true defense counsel’s testimony that she was unaware of Mendoza and Terrones’s knowledge of and interest in the reward.

Ultimately, the factual misunderstanding or disagreement makes no difference here, and we need not resolve it — or, as the dissent suggests, remand to the district court to resolve it — “as a predicate to our legal analysis,” dis. op. at 1121. While some of the facts may not be entirely clear, the law is. As we explain, counsel’s performance was constitutionally ineffective regardless of the extent of her knowledge of Mendoza and Terrones’s awareness of the reward, and of their financial motivations, as a result of her conversation with the prosecutor.

Whether she had direct or specific knowledge of their awareness of the reward, or whether she knew only in the most general sense of such a possibility, her failure to investigate the matter more fully, given the information she possessed, rendered her performance deficient. In addition, her failure at trial to question Mendoza and Terrones about the reward cannot under any theory be deemed a “sound trial strategy.” Strickland, 466 U.S. at 689, 104 S.Ct. 2052 (quoting Michel v. Louisiana, 350 U.S. 91, 101, 76 S.Ct. 158, 100 L.Ed. 83 (1955)) (internal quotation mark omitted).

Defense counsel “has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.” Id. at 691, 104 S.Ct. 2052. “A lawyer who fails adequately to investigate, and to introduce into evidence, [information] that demonstrates his client’s factual innocence, or that raises sufficient doubts as to that question to undermine confidence in the verdict, renders deficient performance.” Lord v. Wood, 184 F.3d 1083, 1093 (9th Cir.1999) (quoting Hart v. Gomez, 174 F.3d 1067, 1070 (9th Cir.1999)) (internal quotation marks omitted and second alteration in original).

In particular, if counsel’s failure to investigate possible methods of impeachment is part of the explanation for counsel’s impeachment strategy (or a lack thereof), the failure to investigate may in itself constitute ineffective assistance of counsel. See Tucker v. Ozmint, 350 F.3d 433, 444 (4th Cir.2003) (“Trial counsel have an obligation to investigate possible methods for impeaching a prosecution witness, and failure to do so may constitute ineffective assistance of counsel.”).

Although trial counsel is typically afforded leeway in making tactical decisions regarding trial strategy, counsel cannot be said to have made a tactical decision without first procuring the information necessary to make such a decision. See Riley v. Payne, 352 F.3d 1313, 1324 (9th Cir.2003) (holding that, under clearly established Supreme Court law, when defense counsel failed to contact a potential witness, counsel could not “be presumed to have made an informed tactical decision” not to call that person as a witness); see also Williams v.

Washington, 59 F.3d 673, 681 (7th Cir.1995) (“Because investigation [of the witnesses] might have revealed evidence bearing upon credibility (which counsel believed was the sole issue in the case), the failure to investigate was not objectively reasonable.”); United States v. Tucker, 716 F.2d 576, 583 (9th Cir.1983) (holding that the failure to interview or to attempt to interview key prosecution witnesses constitutes deficient performance); Baumann v.

United States, 692 F.2d 565, 580 (9th Cir.1982) (“We have clearly held that defense counsel’s failure to interview witnesses that the prosecution intends to call during trial may constitute ineffective assistance of counsel.”); cf. Sanders v. Ratelle, 21 F.3d 1446, 1457 (9th Cir.1994) (“Ineffectiveness is generally clear in the context of complete failure to investigate because counsel can hardly be said to have made a strategic choice when s/he [sic] has not yet obtained the facts on which such a decision could be made.” (citations, emphasis, and internal quotation marks omitted)).

The duty to investigate is especially pressing where, as here, the witnesses and their credibility are crucial to the State’s case. See Huffington v. Nuth, 140 F.3d 572, 580 (4th Cir.1998) (collecting cases).

Moreover, although matters such as counsel’s approach to impeachment are often viewed as tactical decisions, and such decisions do not constitute deficient conduct simply because there are better options, a poor tactical decision may constitute deficient conduct if “the defendant [can] overcome the presumption that, under the circumstances, the challenged action [or lack of action] ‘might 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).

If, as she testified, Reynoso’s trial counsel was ignorant of the fact that Mendoza and Terrones knew and had inquired about obtaining the reward, counsel’s failure to investigate the issue was not objectively reasonable. It is undisputed that counsel knew prior to trial that a reward had been offered, and that she possessed police interview transcripts in which the reward was discussed. Counsel also knew that the State intended to call both Mendoza and Terrones as witnesses. Further, she testified that the prosecutor had told her that both Lopez and Hinojosa sought rewards.

As the magistrate judge concluded, “[a]rmed with knowledge of the reward and the fact that two other witnesses knew of the reward, defense counsel should have, at a minimum, determined if the only two eyewitnesses to the robbery also knew about the reward.” Unlike other impeachment evidence presented at trial — for example, evidence attacking the witnesses’ general credibility or demonstrating the inconsistency in their statements — such information would have provided the jury with a reason ivhy the witnesses may have had a motive to lie, especially as they had inquired as to their ability to collect the reward.

See Stephens v. Hall, 294 F.3d 210, 224 (1st Cir.2002) (“A colorable showing of bias can be important because, unlike evidence of prior inconsistent statements — which might indicate that the witness is lying — - evidence of bias suggests why the witness might be lying.”). Such cross-examination, as Regios herself ultimately conceded, at least with respect to Mendoza, would not have been inconsistent with her defense strategy and would have exposed a strong motive for witness bias on the part of the State’s only two eyewitnesses.

For these reasons, if counsel had no more than general information about the existence of the reward, her failure to investigate the issue with respect to Mendoza and Terrones rendered her performance deficient under Strickland. The same is true even if Regios did have some direct knowledge that Mendoza and Terrones knew about the reward. Such a limited understanding would not have relieved Regios of her duty to investigate; it would have heightened that duty.

Just as, according to the magistrate judge, counsel’s knowledge of the existence of the reward and Lopez and Hinojosa’s interest in it made it more unreasonable for her to fail to determine whether Mendoza and Terrones had a similar financial interest, knowledge that Mendoza and Terrones were aware of the reward would have made it all the more important for Regios to determine whether they had actively sought it and whether they believed that their ability to obtain the financial compensation depended upon their testimony inculpating or convicting Reynoso.

Given the inadequacy of Reglos’s investigation into Mendoza and Terrones’s motives for testifying against Reynoso, her failure at tidal to cross-examine Mendoza and Terrones about the reward further rendered her performance deficient. The State argues that counsel’s decision to pursue a theory of mistaken identification, and not one of deliberately false identification, was a strategic decision that cannot constitute deficient conduct.

The dissent echoes this argument, concluding that “it is manifestly and demonstrably probable that defense counsel made a reasonable strategic choice to focus on fertile ground, and not dilute her attack by opening up another front.” Dis. op. at 10723. We reject this argument for two reasons. First, the record directly refutes the contention that Regios made a strategic decision. Indeed, Regios testified to the contrary at the evidentiary hearing, stating quite clearly that she did not make a strategic decision not to question Mendoza and Terrones about the reward.

Second, any such decision would in any event have been completely unsupportable and, under the circumstances, could not have been considered a “sound trial strategy.” Strickland, 466 U.S. at 689, 104 S.Ct. 2052 (quoting Michel, 350 U.S. at 101, 76 S.Ct. 158) (internal quotation mark omitted). “[Strategic choices made after less than complete investigation are reasonable precisely to the extent that reasonable professional judgments support the limitations on investigation.” Id. at 690-91, 104 S.Ct. 2052.

As we explained, Reglos’s inadequate investigation, as a result of which she failed to uncover the financial motives of the State’s only two purported eyewitnesses, fell outside “the wide range of reasonable professional assistance.” Id. at 689, 104 S.Ct. 2052. Any decision not to obtain this information at trial through cross-examination following so deficient an inquiry was, as a matter of law, equally unreasonable. Moreover, during the evidentiary hearing, trial counsel was unable to suggest a credible reason for failing to question Mendoza and Terrones about the reward.

She admitted that cross-examination about the reward would have been entirely consistent with her decision to cross-examine another similarly situated witness, Lopez, about the subject and would not have been “mutually exclusive” with her strategy of attacking Mendoza and Terrones’s credibility. The magistrate judge agreed with that assessment, finding that “further questioning [of] Terrones and Mendoza would not have diluted or negatively impacted her attempt to discredit them.” Regios also conceded that such questioning likely would have weakened Mendoza’s credibility substantially.

The inconsistent statements in counsel’s earlier declaration that cross-examining Terrones and Mendoza about the award would have “diluted” her other, stronger impeachment evidence, are plainly without merit. Not only did counsel acknowledge that fact at the hearing, at least with respect to Mendoza, but any claim of inconsistency is clearly undermined by the fact that, at trial, counsel cross-examined the two witnesses on relatively minor points, such as their appearance on the day of the shooting.

That is, the “fertile ground” for cross-examination to which the dissent refers, dis. op. at 10723, was actually quite barren. In addition, counsel managed to cross-examine Lopez about the reward even while attacking his credibility on other grounds. Most important, the questioning in which counsel engaged and the questioning in which she failed to engage are entirely compatible, and the latter could only have helped explain, to her client’s benefit, the reasons for the weakness in the witnesses’ answers to the former.

We reject as well the State’s argument that eliciting information about the reward would have served no purpose because some of Mendoza and Terrones’s statements that were not helpful to the prosecution tended to support their credibility— for example, Mendoza’s testimony about his use of drugs and alcohol on the day of the alleged identification and Terrones’s admissions of his own uncertainty about the identification.

The State suggests that Mendoza and Terrones’s candor about these potentially damaging facts was logically inconsistent with any theory of deliberate misidentification; that is, if the two witnesses were identifying Reynoso only in order to obtain the reward, they would have had no reason to sound anything less than sure of themselves on the stand because they were making it all up anyway. Contrary to the State’s argument, Mendoza and Terrones’s financial motivations actually would have helped to explain why they would testify against Reynoso despite the shaky foundations for their testimony.

It twists logic to conclude that the (quite possibly truthful) weaknesses in their testimony precluded the jury from finding that their identifications of Reynoso were incredible in part because the testimony was financially motivated. Finally, the State argues that it “would be a phenomenal coincidence for four independent witnesses to falsely identify the same person from photographic lineups,” and thus, the State suggests, it was entirely reasonable for Reynoso’s counsel not to question the veracity of Mendoza and Ter-rones’s identifications. The States overlooks two important points.

First, the witnesses were, at various times, unable to identify Reynoso or identified individuals other than Reynoso as the perpetrator. Second, it would be much more of a “coincidence” for all four to have eventually identified Reynoso as the perpetrator had there not been general knowledge of the reward and, at least in some instances, public identification of Reynoso as the primary suspect, including in a press photograph. Also, either one or two of the witnesses had known Reynoso previously.

In short, counsel’s investigation and preparation for trial were objectively unreasonable in light of her knowledge about the reward and the absence of any explanation for her failure to interview the two crucial eyewitnesses on that subject. Counsel’s ensuing failure to cross-examine the witnesses about their motivation for testifying as they did was, accordingly, equally unreasonable and cannot 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) (internal quotation mark omitted).

Moreover, such cross-examination would not have been inconsistent with the defense strategy counsel followed and would have provided a strong motive for witness bias on the part of the only two eyewitnesses. For these reasons, counsel’s conduct was deficient under Strickland. B.

Prejudice Having examined the deficiencies in trial counsel’s conduct, we now consider whether, but for such deficiencies, “there is a reasonable probability that ... the result of the proceeding would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. “A reasonable probability is a probability sufficient to undermine confidence in the outcome.” Id. In making this determination, the court “must consider the totality of the evidence before the judge or jury.” Id. at 695, 104 S.Ct. 2052.

A defendant need not show “that counsel’s deficient conduct more likely than not altered the outcome in the case.” Id. at 693-94, 104 S.Ct. 2052. We start from the premise that the case for the prosecution was extremely weak. The State presented no physical evidence tying Reynoso to the Prajapati murder. Its case consisted solely of four witnesses, only two of whom were purportedly eyewitnesses to the events surrounding Prajapati’s murder. The other two witnesses reported that they had heard or overheard confessions that Reynoso had allegedly made.

None of the four suggested that Reynoso was involved in the crime until two or three years after the shooting. One of the witnesses, Lopez, a jailhouse informant, contacted the police to report a confession allegedly made two years earlier, but only after he had seen a television broadcast and a composite of the suspect; further, he admitted that the availability of a reward had motivated him to get in touch with the police. One of the other witnesses was Hinojosa, who himself was initially a suspect in the murder.

After failing to implicate his cousins, Hinojo-sa told police that he had overheard Rey-noso discussing the events of the Prajapati murder at a party two years earlier. At trial, however, he recanted his statements, admitted that he had learned the facts of the murder through news reports on television, and denied knowing Reynoso. ' The two eyewitnesses the State presented were Javier Terrones and Robert Mendoza. There were inconsistencies between their stories — for example, their different descriptions of the make and condition of the getaway car, whether it had been left running, and where it was parked.

Ter-rones and Mendoza also testified that they never saw each other in the market that has been described as a small grocery or liquor store, even though they both testified that they were in the store when Reynoso walked in and that they left before the shooting occurred. Mendoza admitted that he had engaged in heavy alcohol and drug use both before and during the day of the murder. Just two weeks after the shooting, Mendoza singled out from a photographic lineup two individuals other than Reynoso as the perpetrators.

Although he eventually selected Reynoso from a photographic lineup in 1998, three years after the murder, he subsequently identified someone other than Reynoso from an in-person lineup, before changing his mind again and identifying Reynoso. Further, at the time of Reynoso’s trial, Mendoza and Reynoso were placed in the same cell, and even though they were sitting just five or six feet apart, Mendoza did not recognize Rey-noso. He later admitted that it was only after speaking with the prosecutor that he discovered that he had been in the same cell as the person he had identified as the murderer.

Although Mendoza did not testify at the evidentiary hearing because he had died, that hearing revealed that Mendoza had known about the reward when he testified at Reynoso’s trial and that he ultimately received $7,500 for his testimony. Terrones, too, did not contact police until three years after the shooting, and after he also had obtained knowledge of the reward. He initially testified at the evi-dentiary hearing that he did not want any part of the reward money, but later admitted that he had immediately asked about the reward when he contacted the police.

At trial, Terrones was able to describe Reynoso only vaguely and said that he had seen him only briefly. Before identifying Reynoso from both a photographic and live lineup, Terrones had seen an

article with his picture that identified him as a leading suspect in the Prajapati murder. Ultimately, Terrones received $7,500 for his testimony. At trial, defense counsel primarily stressed the witnesses’ inconsistent statements, sought to discredit Terrones’s identification of Reynoso, and emphasized Mendoza’s general lack of credibility.

Although defense counsel questioned Lopez about the reward — in addition to attempting to impeach his credibility more generally — she did not question either of the two purported eyewitnesses, Terrones or Mendoza, about the funds or their knowledge that a substantial amount of money was available for those whose testimony helped secure a conviction of Reynoso. In closing at trial, the prosecutor emphasized defense counsel’s failure to impeach Terrones’s credibility, stressed that neither Terrones nor Mendoza demonstrated any bias, and argued forcefully that neither had any reason or motive to lie.

She went so far as to state: “They don’t have a bias in this case. There’s no reason for them to come forward and say it’s the defendant if it’s not the defendant” (emphasis added). This last statement demonstrates the substantial impact that establishing the financial motives of the two purported eyewitnesses would have had. See Horton v. Mayle, 408 F.3d 570, 580 (9th Cir.2005) (holding that when the prosecution emphasizes a witness’s testimony, impeachment of that witness may significantly damage the prosecution’s case).

Unlike the other evidence used to impeach the eyewitnesses — the two who claimed to have seen Reynoso at the scene of the murder — such as inconsistent statements and general attacks on their credibility, evidence of their financial motives would have established a real incentive to lie, explaining why their testimony may have been fabricated.

The failure to adduce such evidence, or even to question the witnesses regarding their financial interests in their testimony, undermines our confidence in the jury’s verdict and establishes a reasonable probability that, but for counsel’s failure to elicit the reason for the witnesses to fabricate evidence, the result would have been different. See Silva v.

Brown, 416 F.3d 980, 987 (9th Cir.2005) (“Impeachment evidence is especially likely to be material when it impugns the testimony of a witness who is critical to the prosecution’s case.”); Horton, 408 F.3d at 580-81 (holding that “where a witness is central' to the prosecution’s case, the defendant’s conviction demonstrates that the impeachment evidence presented at trial likely did not suffice to convince the jury that the witness lacked credibility” and that, therefore, any impeachment evidence not introduced at trial takes on greater significance).

Had defense counsel investigated and questioned Terrones and Mendoza about their expectation of reward money in return for their testimony inculpating Rey-noso, she would have been able to provide the jury an explanation of the eyewitnesses’ incentive to identify him, regardless of their lack of knowledge, and would have effectively demonstrated witness bias. She would have answered directly the open question that the prosecution’s closing argument posed for the jury — what was the witnesses’ motive to lie?

In the absence of the missing cross-examination, the defendant was unable to provide an answer to this critical question. Given those facts and circumstances, we conclude that, but for counsel’s deficient performance, there is a reasonable probability that the outcome of Reynoso’s trial would have been different. Therefore, we conclude that Reynoso has demonstrated prejudice under Strickland. C. Objectively Unreasonable Under AEDPA The State mistakenly asserts that the district court applied a “clear error” standard of review in deciding to grant Reyno-so habeas relief.

In her Report and Recommendation, the magistrate judge stated: The Supreme Court has admonished courts against equating the term “unreasonable application” with “clear error”: “These two standards ... are not the same. The gloss of clear error fails to give proper deference to state courts by conflating error (even clear error) with unreasonableness.” Lockyer v. Andrade, 538 U.S. 63, 75, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003). Instead, in this context, habeas relief may issue only if the state court unreasonably applied firmly established federal law.

Later in her Report and Recommendation, the magistrate judge concluded: [T]he proposed impeachment evidence held great value. Unlike the other evidence that the prosecutor used to impeach Terrones or Mendoza, it would have explained why the key prosecution witnesses may have fabricated their testimony. As the prosecutor’s closing argument demonstrates, this evidence would have had a substantial impact on the jury’s decision to credit or discredit Terrones’s and Mendoza’s credibility.

Moreover, given the weaknesses of the prosecution’s case and the importance of the witnesses’ testimony, defense counsel’s failure to impeach the witnesses with evidence of bias ‘undermine[s] confidence’ in the jury’s verdict. Strickland, 466 U.S. at 694, 104 S.Ct. 2052. Thus, Petitioner has satisfied both prongs of the Strickland two-part test to show ineffectiveness of counsel. Accordingly, the state court’s conclusion to the contrary constituted an unreasonable application of clearly established federal law. 28 U.S.C. § 2254(d)(1); Williams v.

Taylor, 529 U.S. 362, 405-406, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). The State offers two arguments in support of its position. First, it seizes on the single word “[ajccordingly” and asserts that its use demonstrates that the district court failed to apply the standard that it announced it was required to apply. The State is wrong. It is important to note that this appeal is not subject to typical AEDPA analysis because here there is no reasoned state court decision to assess. See Pham, 400 F.3d at 742.

In this situation, the district court must conduct an independent review of the record; if after such review, it concludes that controlling Supreme Court law, unless applied in an unreasonable manner, would preclude the result reached by the state courts, it must grant relief to the petitioner. In this case, the magistrate judge and district court properly concluded that Reynoso had established that only an unreasonable application of Strickland would allow them to hold that petitioner had not been deprived of effective assistance of counsel.

The magistrate judge’s use of the term “[a]c-cordingly” does not indicate a belief that clear error is enough to constitute an unreasonable application, but rather that in light of her independent review of the law and the facts, she was compelled to conclude that the state court’s determination was, in actuality, an unreasonable application of clearly established Supreme Court law. In its second argument, the State contends that by citing Delgado v. Lewis, 223 F.3d 976, 982 (9th Cir.2000), the magistrate judge showed that she was relying on the inapplicable “clear error” standard.

At one point in her discussion, the magistrate judge included a quote from Delgado referring to the standard in question— “whether the state court clearly erred”— but that quote was relied on only to support thé unquestioned proposition that when the state court has issued a silent denial, courts must perform an independent review of the record.

At the outset of her discussion of Reynoso’s claims, the magistrate judge set forth the correct AEDPA standard of review, specifically distinguishing between “unreasonable application” and “clear error.” Viewed in the context of the entire Report and Recommendation, it is evident that the magistrate judge applied the correct standard and that her inclusion of the “clearly erred” language in a parenthetical quotation from a case that was properly cited for a different proposition was inadvertent. Such a happenstance does not constitute a basis for concluding that the court has applied the wrong standard.

See Wood ford v. Visciotti, 537 U.S. 19, 23-24, 123 S.Ct. 357, 154 L.Ed.2d 279 (2002); see also White v. Roper, 416 F.3d 728, 732-33 (8th Cir.2005) (holding that even though the district court erred in using “reasonable possibility” language, thereby misstating the Strickland “reasonable probability” standard, because it had earlier stated the correct standard three times and quoted the Supreme Court’s

summary of that standard, it had correctly applied the Strickland prejudice standard). Because the magistrate judge, and later the district judge, applied the correct legal standard, the district court did not commit the procedural error the State asserts. Nor did it err with respect to its substantive ruling that the state court’s decision that Reyno-so received effective assistance of counsel was objectively unreasonable under AED-PA.

We review the district court’s substantive decision de novo, and based on our own independent review of the record, we reach the same conclusion as the magistrate judge and the district judge. We affirm the district court’s decision that Reynoso was denied effective assistance of counsel because, as we explain below, the record clearly demonstrates that a contrary decision would constitute an unreasonable application of clearly established Supreme Court law. V. Reynoso’s trial counsel failed to interview the only two witnesses who placed Reynoso at the scene of the murder.

She also failed to cross-examine these critical witnesses effectively, neglecting to examine the relationship of the reward to their pivotal testimony in this case. The consequence was that the defense did not provide the jury with the motive to lie that would have explained why the State’s eyewitnesses’ identifications of the defendant were not worthy of credence.

Given that the testimony of the two witnesses was central to the prosecution’s case and that the case against Reynoso was otherwise exceedingly weak, such a performance unquestionably fell outside “the wide range of reasonable professional assistance” contemplated by Strickland, 466 U.S. at 689, 104 S.Ct. 2052. Moreover, defense counsel’s deficient performance was extremely prejudicial to petitioner. Counsel’s failure to elicit essential impeachment evidence at trial through cross-examination was critical to the outcome. The credibility of the eyewitnesses was determinative.

Had the jurors believed that the two eyewitnesses were motivated to identify Reynoso by their desire for money rather than a willingness to aid the truth-seeking process, it is unlikely that any reasonable juror would have voted to convict. We do not find the question before us to be a close one. Upon an independent review of the record, we conclude that given so ineffective a performance with so adverse a consequence, it would constitute an unreasonable application of Strickland to hold that Reynoso received effective assistance of counsel.

Thus we conclude that the state court determination that Reynoso was afforded his Sixth Amendment rights was objectively unreasonable. Accordingly, we affirm the district court’s grant of habeas relief. AFFIRMED. . The State argues that because Lopez's version of Reynoso’s confession involved Reyno-so "pointing a finger like a gun at the center of his forehead, the precise location of Mrs. Prajapati's gunshot wound,” and because the exact location of the gunshot wound had not been revealed in news reports, he must have been told about the shooting by Reynoso.

It was public information, however, that Praja-pati had been shot in the head, and published reward notices specifically stated that the victim was "shot in the head.” Given that Praja-pati was running the store on the day of the murder, it is not unlikely that she would have been standing behind the counter, facing forward, making her forehead a likely location for the shooting. . Mendoza was deceased at the time of the evidentiary hearing in district court and therefore did not testify at that hearing. .

In November of 1998, Mendoza identified Reynoso as the shorter of the two men and said that it was he who had lettering tattooed across his back. At trial, he changed his story again and said that he did not remember Reynoso having any lettering on his back. . Mendoza was in the custody of the INS, awaiting deportation, and had been brought to the county jail on an unrelated charge of felony cocaine possession. .

Although the State argues that Terrones and Mendoza identified.the same car because they both described a black vehicle with a rear spoiler and lights extending to the back, the magistrate judge noted that their testimony conflicted in several ways. Terrones described the robbers’ car as old and noisy while Mendoza described it as fairly new and quiet. While Terrones said that the robbers left the car running, Mendoza said the engine was turned off. Terrones said that the car was parked in a lot in front of the store; Mendoza said it was parked in an alley behind the store.

Terrones said the car was a Camaro, while Mendoza said it could not have been a Camaro, but was a Berretta. Additionally, the magistrate judge noted that although Terrones and Mendoza testified that they were both in the store when Reynoso walked in, and that they both left the store before the shooting, they testified that they never saw one another. . Regios testified that it was her practice to have her investigator interview witnesses pri- or to trial — particularly when they were identification witnesses.

Although she acknowledged that the identity of the perpetrator was a primary issue in Reynoso's case, she could not recall whether she asked her investigator to interview Terrones or Mendoza or indeed whether they had ever been interviewed.

Reglos testified that because she did not know about Mendoza or Terrones’s knowledge of the reward, she "couldn’t have made a strategic decision not to question [them] about that." Based on all of the testimony given at the evidentiary hearing, the magistrate judge concluded that “[n] either defense counsel nor her investigator interviewed either Terrones or Mendoza before trial.” .

The amended judgment did not materially alter the original judgment; it added the words “becomes final” to the original judgment: “Respondent is ORDERED to release Petitioner and discharge him from all other adverse consequences in Los Angeles County Superior Court Case, No. LA032732, unless charges are refiled against him in the Superi- or Court and counsel is appointed within ninety (90) days of the date judgment herein becomes final " (emphasis added). .

In view of this history, it is remarkable that the dissent criticizes the federal courts, and in particular the magistrate judge, for making inadequate findings and failing to do her job properly. The state courts failed not only to make any findings but also to give any reason for denying a petition that raised serious questions as to the actual innocence or guilt of the defendant. The magistrate judge, in contrast, wrote a 59-page report carefully examining each of Reynoso's claims and finding merit in a claim the state courts had brushed aside.

Had the state courts done a proper job, a federal petition might have been totally unnecessary. . It is unclear how the State can now represent that "Petitioner never alleged in state court or his federal Petition that trial counsel should have cross-examined either Mendoza or Terrones about their knowledge of the reward,” when it conceded in the district court that he did. Perhaps the Attorney General will wish to review this question further. .

Counsel also failed to cross-examine the third of the four witnesses, Hinojosa, about the reward and its possible effect on his ultimately retracted testimony inculpating Rey-noso. Counsel testified at the evidentiary hearing that the prosecutor had told her that Hinojosa knew about the reward. Nevertheless, at trial, the prosecutor elicited testimony from Hinojosa denying knowledge of the reward and Reynoso’s counsel failed to challenge the denial on cross-examination.

This omission was not insignificant; despite Hino-josa's disavowal of his accusations, the State continued to rely on them and urged the jury to disregard the disavowal. . In Tucker, this court found that defense counsel's "ability to cross-examine the government's witnesses effectively was seriously compromised by his failure to interview them, since he would have little idea as to the specific areas of testimony which could be challenged." Id. at 583. . Reglos's testimony on this point was unequivocal: Q: [Y]ou didn't know about [Mendoza’s] knowledge of the reward— A: Right.

Q: — so you couldn't have made a strategic decision at the time not to question him about that; is that right? A: That’s correct. Q: And that’s the same for Mr. Terrones, as well? A: Right. . The State argues that the district court "essentially disregarded” the confessions Rey-noso allegedly made to Lopez and Hinojosa. The magistrate judge's Report and Recommendation did not ignore the purported confessions, however; it discussed them in detail.

In doing so, it considered the substantial credibility problems of the witnesses claiming to have heard the confessions and examined the context in which the confessions were allegedly made. Finally, it considered the fact that one of the two witnesses later recanted his testimony regarding hearing the purported confession. . We have found prejudice even when the prosecution's case has been stronger than in the instant case. See Rios v.

Rocha, 299 F.3d 796, 810-13 (9th Cir.2002) (holding that counsel’s failure to investigate and present witnesses was prejudicial when no physical evidence tied the defendant to the shooting and the state’s case rested on the testimony of three eyewitnesses); Lord, 184 F.3d at 1094-96 (holding that counsel’s failure to interview witnesses who had claimed to see the victim alive after the murder was prejudicial even though physical evidence tied the defendant to the murder and two inmates testified that the defendant had confessed to them); Brown v.

Myers, 137 F.3d 1154, 1155-57 (9th Cir.1998) (holding that counsel's failure to investigate and to locate and produce witnesses was prejudicial when three witnesses identified the defendant as the shooter). . A state court’s decision is an "unreasonable application” of Supreme Court law if "the state court correctly identifies the governing legal principle ... but unreasonably applies it to the facts of the particular case.” Bell v. Cone, 535 U.S. 685, 694, 122 S.Ct. 1843, 152 L.Ed.2d 914 (2002). .

The magistrate judge wrote: "In this case, both the California Court of Appeal and the California Supreme Court issued silent denials of Petitioner's current claims for relief. The Court, therefore, gives less deference to those courts’ decisions. Delgado, 223 F.3d at 982 ('Thus, we accordingly concluded that, in such circumstances, the state court decisions do "not warrant the deference we might usually apply” ....’) (citations omitted).

Instead, Hue Court must independently review the record to determine whether 'the state court clearly erred in its application of controlling federal authority.’ Id." .Moreover, the magistrate judge’s earlier citation to Lockyer v. Andrade, in which the Supreme Court overruled in

part Van Tran v. Lindsey, 212 F.3d 1143, 1152-54 (9th Cir.2000) (relied upon in Delgado, 223 F.3d at 981, for the proposition that in determining whether a state court's decision was objectively unreasonable, we must determine "whether the state court clearly erred”), makes it clear that she was aware of, and applied, AEDPA’s "objectively unreasonable” standard of review.