Dennis v. Secretary, Pennsylvania Department of Corrections

834 F.3d 263 | 3rd Cir. | 2016

Lead Opinion

OPINION

RENDELL, Circuit Judge.

James Dennis has spent almost twenty-four years unsuccessfully challenging his conviction for the murder of Chedell Williams. The Pennsylvania Supreme Court repeatedly affirmed Dennis’s first-degree murder conviction and sentence and denied his applications for post-conviction relief. Thereafter, Dennis filed an application under 28 U.S.C. § 2254, and the United States District Court for the Eastern District of Pennsylvania granted Dennis habeas corpus relief, concluding that the Pennsylvania Supreme Court had unreasonably applied Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), with respect to three pieces of evidence suppressed by the Commonwealth. The suppressed Brady material — a receipt corroborating Dennis’s alibi, an inconsistent statement by the Commonwealth’s key eyewitness, and documents indicating that another individual committed the murder — effectively gutted the Commonwealth’s case against Dennis. The withholding of these pieces of evidence denied Dennis a fair trial in state court. We will therefore affirm the District Court’s grant of habeas relief based on his Brady claims.

I. Background

A. Factual Background

On October 22, 1991, Chedell Williams and Zahra Howard, students at Olney High School, climbed the steps of the Fern Rock SEPTA station, located in North Philadelphia. Two men approached the girls and demanded “give me your fucking earrings.” App. 465. The girls fled down the steps; Howard ran to a nearby fruit vendor’s stand and Williams ran into the intersection at Tenth and Nedro Streets. The men followed Williams. The perpetrators tore Williams’s gold earrings from her earlobes. One of the men grabbed her, held a silver handgun to her neck, and shot her. The men then ran up the street to a waiting getaway car and fled the scene. The precise time of injury was 1:54 p.m. Emergency personnel responded within minutes, but Williams was pronounced dead at the hospital less than an hour later.

B. Police Investigation and the Trial

The police undertook an investigation into the Williams murder, primarily aimed at determining the identity of the shooter. Frank Jastrzembski led a team of detectives who pursued the investigation based on rumors that “Jimmy” Dennis from the Abbottsford Homes projects in East Falls1 committed the crime, despite being unable to identify the source of the rumors. Resting on tips by neighbors from *270the projects, police proceeded with Dennis as the primary, if not the sole, suspect.2

Detectives obtained eyewitness reports and identifications, very few of which aligned with Dennis’s appearance. Nearly all of the eyewitnesses who gave height estimates of the shooter described him as between 5'9" and 5'10". He was described as having a dark complexion and weighing about 170 to 180 pounds. The victim, Williams, had a similar build as the shooter; she was 5'10" and weighed 150 pounds. Dennis, on the other hand, is 5'5" tall and weighed between 125 and 132 pounds at the time of trial.

Prior to trial, three eyewitnesses identified Dennis in a photo array, at an in-person lineup, and at a preliminary hearing: Williams’s friend, Zahra Howard; a man working on a garage near the intersection, Thomas Bertha; and a SEPTA employee who was standing in front of the station at the time of the murder, James Cameron.3

Zahra Howard

Photo Array: Howard identified Dennis, saying “this one looks like the guy, but I can’t be sure ... He looks a little like the guy that shot Chedell.” App. 1537. When asked if she could be sure, she replied “No.” Id.
Lineup: Howard indicated that she “thought” Dennis was the shooter. App. 586.4
*271Preliminary Hearing and Trial: Howard testified at trial that she had identified Dennis as the shooter at a preliminary hearing. App. 474-75. She also made an in-court identification during trial. Id.

Thomas Bertha

Photo Array: Bertha initially said that the first photo, which was a photo of Dennis, looked like the man running with the gun and later confirmed his identification.
Lineup: When asked to identify the shooter, Bertha simply stated “three,” which was Dennis. App. 586.
Trial: Bertha identified Dennis as the shooter at trial.

James Cameron

Photo Array: Cameron said that Dennis looked like the shooter, but wavered “I can’t be sure.” App. 1548.
Lineup: Cameron identified Dennis, who was in the third position in the lineup, by simply stating “number three” without reservation. App. 689.
Preliminary Hearing and Trial: At trial, Cameron identified Dennis as the shooter and confirmed that he had identified Dennis at the preliminary hearing.

At trial, the prosecutor introduced testimony from detectives who verified that Howard, Bertha, and Cameron each identified Dennis in the photo array and lineup. No other eyewitness identifications were referenced.

Dennis was arrested on November 22, 1991. His signed statement indicated that he stayed at his father’s house until about 1:30 p.m. on the day in question, when his father drove him to the bus stop and watched him get on the “K” bus toward Abbottsford Homes to attend singing practice that evening. Dennis rode the K bus for approximately thirty minutes to the intersection of Henry and Midvale Avenues. During the trip, Dennis saw Latanya Cason, a woman he knew from Abbotts-ford Homes. In his statement to police, which was read into the record at trial, Dennis asserted that when he and Cason disembarked the bus “[he] waved to her.” App. 710. After getting off the bus, Dennis walked to Abbottsford Homes, where he spent the rest of the day with his friends. Dennis’s father, James Murray, corroborated Dennis’s story. He stated that they spent the morning together, and that he drove Dennis to the bus stop shortly before 2:00 p.m. to catch the K bus to Ab-bottsford Homes:

The Commonwealth’s case rested primarily on eyewitness testimony, which Assistant District Attorney Roger King emphasized in his opening statement to the jury. Though ADA King acknowledged that the Commonwealth had no physical evidence — the silver handgun and the earrings were never recovered — he contended that the eyewitness identifications were sufficient for a conviction. Three eyewitnesses were called to testify at trial: Zahra Howard, Thomas Bertha, and James Cameron.

Zahra Howard, who was present with the victim at the time of the murder, led the Commonwealth’s case. She recounted what had occurred, noting that the shooter was “right in front of’ her and Williams, about one or two feet away, and that she looked the shooter in the face. App. 467-68. About ten seconds passed between the first time she saw the men until she turned around and ran away from the scene; she also saw the shooter for about five to ten seconds while he was grabbing Williams in the street. Howard identified Dennis in a photo array, at an in-person lineup, and at a preliminary hearing. Defense counsel focused his cross-examination on her hesitation in prior identifications. Howard described the shooter as wearing a black hooded sweatshirt and a red sweat suit. In *272her statement, Howard said that the shooter was about same height as Detective Danks, who was 55'9" or 5'10", or taller. Howard testified at trial that she had never seen the shooter or his accomplice before in her life.

Thomas Bertha and his partner, Anthony Overstreet, were installing stones on a garage near Tenth and Nedro Streets on the day in question. After hearing the gunshot, they came down from their ladders and looked down the street from the sidewalk. The two perpetrators ran past them. The shooter passed between three to eight feet in front of Bertha, and Bertha ran after him. Bertha made visual contact with the shooter, who was running toward him, for about three to four seconds. Defense counsel impeached Bertha by recalling that, at the preliminary hearing, Bertha testified that he could not have seen the shooter for longer than about a second. Bertha viewed the photo array and attended the lineup, identifying Dennis at both. He described the shooter as wearing red sweat pants, a red hooded sweatshirt, a black cap, and a leather jacket. Bertha testified at trial that he remembered telling the police that the shooter was 5'9" and 180 pounds.

James Cameron was working as a SEPTA cashier on the day of the murder. He was about eight to ten feet from Williams when she was shot and saw the shooter for a few seconds. Cameron saw the shooter’s face several times but acknowledged that he “didn’t really pay attention.” App. 664. He testified at trial that he saw the shooter for about thirty to forty seconds collectively. This .estimate contradicted Cameron’s prior testimony at the preliminary hearing where he claimed that about twenty seconds passed between when he first saw the shooter and when the shooter ran away. Cameron viewed the array, attended the lineup, and testified at the preliminary hearing, identifying Dennis at each instance, as well as at trial. Cameron stated that Dennis looked like the shooter, “especially from the side.” App. 676. He described the shooter as wearing a red sweat suit and a dark jacket, carrying a small silver revolver. He did not remember giving detectives a specific height and weight description, but remembered telling them that the shooter was “stocky.”5 App. 664.

Aside from eyewitness testimony, the Commonwealth presented testimony from Charles “Pop” Thompson and Latanya Ca-son, who spoke about their interactions with Dennis on October 22, 1991, the day of the murder. Thompson was in Dennis’s singing group, which held rehearsal at Ab-bottsford Homes that day. Thompson did not remember what Dennis was wearing, but told detectives that he saw Dennis with a gun that night. He also identified an illustrative .32 chrome revolver, which had been admitted as a Commonwealth exhibit, as being similar to the one he saw in Dennis’s possession. Thompson had an open drug possession charge at the time of trial, but testified that he was not expecting any help from the Commonwealth with the drug charge in exchange for his testimony. Three years after trial, Thompson attested in a statement that he had never *273seen Dennis with a gun and that his testimony at trial was false.

Latanya Cason, who knew Dennis “by living up [her] way” at Abbottsford Homes, testified that she saw him between 4:00 and.4:30 p.m. at Henry and Midvale Avenues on October 22, 1991. App. 731. Cason’s estimate that she saw Dennis between 4:00 and 4:30 p.m. was “strictly a guess” on her part — she did not know exactly what time she saw Dennis — but there was no question she saw him that day. App. 745. Prior to seeing Dennis, Cason took public transportation to the 3-2 center where she picked up her public assistance check, signing a document to confirm pick up. She then filled her daughter’s prescription, got some fish, ran a few additional errands, and went home via the K bus. Cason testified at trial that she did not see Dennis at 2:00 p.m. that day because she was just leaving work at 2:00 p.m. Although the Commonwealth introduced a schedule of payment and food stamps at trial, which stated that Cason was slated to pick up her public assistance check and food stamps on October 22, 1991, nothing was introduced at trial indi-eating the precise time of day she retrieved her benefits.

Detective Jastrzembski executed a search warrant of Dennis’s father’s home and seized two black jackets, a pair of red pants, and a pair of white sneakers. The police lost the items prior to trial. Detectives and two experts testified at trial about physical aspects of the crime, but the Commonwealth did not introduce any physical evidence at trial.6

Dennis’s defense strategy centered on his alibi, good character, and mistaken identity.7 His defense comprised of testimony by his father, James Murray, Dennis himself, a few members of his singing group, and character witnesses. Dennis did not have evidence to support an “other suspect” defense.

Dennis’s father testified that the two of them were together from the evening of October 21, 1991, until about 1:50 p.m. on October 22, 1991. Murray lives about fifteen to eighteen blocks from the Fern Rock Station, roughly a five-minute drive with traffic. Murray testified that “[he] kn[ew] for a fact that [Dennis] was on [the K bus]” at the time of Williams’s murder *274because he drove Dennis to the stop and watched from his car as Dennis boarded the bus. App. 804. The Commonwealth pointed out that Murray had visited Dennis forty times since his arrest.

Willis Meredith, James Smith, and Marc Nelson, members of Dennis’s singing group who had known Dennis for ten years or more, testified on Dennis’s behalf about rehearsal on the day of the murder. Meredith saw Dennis for about twenty minutes around 2:15 or 2:30 p.m., which aligned with Dennis’s account. Smith testified that Dennis was dressed in dark sweats and a dark hooded shirt at rehearsal that night — he was not wearing any red. Meredith, Smith, and Nelson each testified that Thompson and Dennis frequently got into arguments. Each testified that they had not seen a handgun in Dennis’s possession.8 Other defense witnesses, including Dennis’s brothers, friends, and church leaders, testified to Dennis’s reputation for being honest, truthful, peaceful, and law-abiding.9

Finally, Dennis took the stand. He testified that he had nothing to do. with Williams’s shooting and was not in the area at the time of her murder.10 In line with his father’s testimony, Dennis said he spent the previous night at his father’s house and left at 1:30 or 1:45 p.m. to take the bus to Abbottsford Homes for singing practice. When Dennis left his father’s house, he was wearing a dark blue jeans set; he changed into black sweats at Merri-weather’s house before rehearsal. Dennis testified that he took the K bus, where he “thought” he saw Tammy Cason, to Henry and Midvale Avenues in East Falls, arriving around 2:30 p.m.11 App. 1028. Dennis then went to Willis Meredith’s house for twenty to thirty minutes. Dennis acknowledged getting into frequent arguments with Thompson about Thompson’s desire to be the leader of the singing group.

Counsels’ closings reiterated the trial’s themes — eyewitness identifications and Dennis’s alibi. Defense counsel pointed to eyewitness identifications as the key question in the Commonwealth’s case, but he had no means, of impeaching Howard, the eyewitness with the closest view of the shooter. Defense counsel highlighted Thompson’s motive to lie, but Thompson’s testimony did not directly link Dennis to Williams’s murder. Finally, defense counsel had to backtrack from using Cason to bolster Dennis’s timeline due to the timing discrepancy between her version — that they saw one another between 4:00 and 4:30 — and Dennis’s account that he saw Cason at 2:30. In his closing statement to the jury, counsel urged that Dennis had not, in fact, seen Cason on the bus to detract from the inconsistency.

ADA King similarly saw Howard as the key witness at trial and instructed the jury that “if you believe Zahra Howard, that’s enough to convict James Dennis.” App. 1207. King attacked Dennis’s testimony that he saw Tammy Cason on the K bus as incredible, and undercut Dennis’s father’s testimony by urging that “blood is thicker *275than water,” leaving no disinterested witnesses to support Dennis’s account. App. 1208-09.

The jury found Dennis guilty of first-degree murder, robbery, carrying a firearm without a license, criminal conspiracy, and possession of an instrument of a crime. It found Dennis’s lack of significant criminal history a mitigating factor during the penalty phase, but it also found that the killing was committed in the course of a felony, amounting to an aggravating circumstance. The jury sentenced Dennis to death.

C. Undisclosed Evidence

The prosecution failed to disclose to Dennis’s counsel three pieces of exculpatory and impeachment evidence: (1) a receipt revealing the time that Cason had picked up her welfare benefits, several hours before the time she had testified to at trial, thus corroborating Dennis’s alibi (the “Ca-son receipt”); (2) a police activity sheet memorializing that Howard had given a previous statement inconsistent with her testimony at trial, which provided both invaluable material to discredit the Commonwealth’s key eyewitness and evidence that someone else committed the murder (the “Howard police activity sheet”); and (3) documents regarding a tip from an inmate detailing his conversation with a man other than Dennis who identified himself as the victim’s killer (the “Frazier documents”). .

1. Cason receipt

Detectives interviewed ■ Latanya Cason, the woman identified in Dennis’s initial statement, at Abbottsford Homes a few months after Dennis’s arrest. Cason told detectives that she. thought she remembered seeing Dennis the day of the murder, but her timeline contradicted the one Dennis outlined. She said that she worked until 2:00 p.m., went to the 3-2 center to pick up her public assistance check, picked up a prescription and some fish, boarded the K bus, and got off near Abbottsford Homes. According to Cason, she saw Dennis when she got off the K bus between 4:00 and 4:30. p.m., not between 2:00 and 2:30 p.m. as Dennis indicated. The only discrepancy between Dennis’s testimony and Cason’s was the time of their interaction. Police records indicate that Cason gave detectives a Department of Public Welfare (“DPW”) card marked “Schedule of check payment” at the time of her interview, which was introduced at trial. However, the Commonwealth' possessed another DPW document not disclosed at trial — a receipt bearing the time Cason picked up her check. Cason testified at trial as a witness for the prosecution and her testimony aligned with her initial statement to detectives.

On appeal, Dennis’s new appellate counsel obtained Cason’s time-stamped receipt from the DPW.12 Cason stated in an affidavit that police had a copy of the timestamped receipt when they interviewed her and that she gave police her only copy of the receipt. The receipt indicated that Cason picked up her welfare check at 13:03, or 1:03 p.m. In complete contradiction to her trial testimony, then, Cason could not have been working until 2:00 p.m. that day. Cason attributed her prior incorrect testimony to misunderstanding military time, so that she “may have thought that the 13:03, which is on the receipt, was 3:03 p.m.” App. 1736. Based *276on the discrete time indicated on the receipt, Cason’s affidavit stated she would have seen Dennis “between 2:00 and 2:30 p.m. at the Abbottsford Homes, and hot 4:00 to 4:30 that is in my statement.” Id.

2. Howard police activity sheet

Two days after the murder, detectives interviewed Diane and Mannasett Pugh, Williams’s aunt and uncle. Diane Pugh told detectives that, the night after the murder, Zahra Howard told them that she recognized the assailants from Olney High School, where she and Williams were students. Dennis did not attend Olney High School. Howard’s assertion that she recognized the assailants from school contradicted her prior statements to police that she had never seen the men before and did not recognize them from school. Police recorded in their “THINGS TO DO” list that they planned to interview Howard about her inconsistent statements.

Howard further told the Pughs that two people named “Kim” and “Quinton” had also been present at the murder. The following day, another of Williams’s aunts, Elaine Parker, told police that Howard mentioned Kim and Quinton were at the scene. The Commonwealth disclosed Parker’s statement prior to trial. However, the prosecution did not disclose information about Howard’s inconsistent statement to the Pughs. Mere hours after meeting with Parker and receiving additional information that Howard had omitted or misstated facts in her initial statement to police, two detectives met with Howard, ostensibly to follow up on their “things to do.” Ignoring their recorded intentions, however, the detectives only questioned Howard about a photo array and did not inquire about the inconsistent statements.

S. Frazier documents

Prior to Dennis’s arrest, Philadelphia detectives received a call from Montgomery County police relaying a tip from an inmate at the Montgomery County Correctional Facility, William Frazier. Frazier told Montgomery County detectives that he spoke with the man who may have murdered Williams during a three-way call with Frazier’s friend, Tony Brown, facilitated by Frazier’s aunt. During the call, Brown told Frazier and Frazier’s aunt that he “fucked up” and murdered Williams when the gun went off accidentally during a botched robbery of her earrings. App. 1692. He also said that two other men, Ricky Walker and “Skeet,” aided in committing the crime. Frazier told detectives that Brown had a brown car, that he “like[d] to wear sweat suits,” and that the men knew the victim as “Kev[’s] ... girl.”13 App. 1694-95.

Frazier told police that Brown and the others had hid in Frazier’s empty apartment for two days following the murder. Frazier provided addresses for the men, including their parents’ and girlfriends’ addresses, an address and phone number for his aunt, and an address for the pawn shop Brown frequented. Frazier volunteered to take detectives on a “ride along” to point out the houses and pawn shop.

Following the tip, Detectives Santiago and Jastrzembski interviewed Walker, who admitted to knowing Williams from Olney High School, but denied knowing Brown or Skeet. Walker denied any involvement in the murder, and claimed that his mother could verify that he was sleeping when Williams was murdered. Walker admitted to hanging out around Broad and Olney, the exact area where Overstreet said he had seen the perpetrator before. Detectives never verified Walker’s alibi nor *277showed his photo to any of the eyewitnesses. Detectives never located Brown or Skeet.

Detectives, including Jastrzembski, spoke with Frazier’s landlord, who had no knowledge of anyone entering Frazier’s apartment. Detectives did not interview Frazier’s aunt to obtain her account of the call with Brown.

The Commonwealth suppressed at least six documents relating to the Frazier tip from Dennis’s trial counsel: (1) Frazier’s initial statement to the Montgomery County police (Oct 31, 1991); (2) Frazier’s statement to the Philadelphia police (Nov. 1, 1991); (3) Police Activity Sheet regarding Frazier’s landlord (Nov. 1, 1991); (4) Police Activity Sheet regarding Ricky Walker (Nov. 2, 1991); (5) Frazier’s signed search consent; and (6) Ricky Walker’s statement (Nov. 2, 1991). The Commonwealth concedes that these documents were not disclosed to Dennis until a decade after trial.

D. Review of State Court Conviction

Like many habeas cases, this case has a lengthy procedural history. Only those decisions and arguments relevant to the instant appeal aré summarized below.

On July 22, 1998, the Pennsylvania Supreme Court affirmed Dennis’s conviction and death sentence on direct appeal by a vote of four to three. Commonwealth v. Dennis, 552 Pa. 331, 715 A.2d 404 (1998) (“Dennis /”). Dennis argued on direct appeal that the Commonwealth violated his due process rights by failing to disclose Cason’s time-stamped receipt prior to trial, in opposition to Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963).14

On September 15, 1998, Dennis filed a timely pro se petition pursuant to Pennsylvania’s • Post Conviction Relief Act (“PCRA”), received new counsel, and also received discovery. In December 1999, PCRA counsel was appointed and filed an amended petition, and, subsequently,' a supplemental amended petition and a second supplemental petition on December 1, 2000, and July 10, 2002, respectively. Two pieces of evidence at issue in this appeal were disclosed during PCRA discovery. First, Dennis received the police activity sheet memorializing Howard’s statements to Diane Pugh the night after the murder, which indicated that she recognized the shooter from Olney High School. Second, Dennis received the six documents relating to the Frazier lead that police had abandoned. The PCRA court denied Dennis’s claims that the prosecution violated Brady by failing to disclose the Howard statement and the Frazier documents. Dennis again appealed to the Pennsylvania Supreme Court.

The Pennsylvania Supreme Court affirmed the PCRA court in part and vacated in part, and remanded two claims. Commonwealth v. Dennis, 597 Pa. 159, 950 A.2d 945 (2008) (‘Dennis III”). The court found that the Commonwealth’s failure to disclose the Frazier documents did not violate Brady because the prosecution was not required to disclosé “every fruitless lead” and that “inadmissible evidence cannot be the basis for a Brady violation.” Id. at 968 (internal quotation marks omitted) *278(quoting Commonwealth v. Lambert, 584 Pa. 461, 884 A.2d 848, 857 (2005)).

The Pennsylvania Supreme Court remanded to the PCRA court Dennis’s claim that the Commonwealth violated Brady by suppressing the contents of the police activity sheet memorializing Zahra Howard’s inconsistent statement. After evidentiary hearings on remand, the PCRA court again dismissed Dennis’s petition. Commonwealth v. Dennis, Case No. 92-01-0484, slip op. (Pa. Ct. Com. PL Mar. 17, 2010). The Pennsylvania Supreme Court concluded that it was not relevant that Howard denied her prior inconsistent statement at the evidentiary hearing before the PCRA court. See, e.g., Commonwealth v. Dennis, 609 Pa. 442, 17 A.3d 297, 309 (2011) (“Dennis /V”).

The Pennsylvania Supreme Court affirmed the PCRA denial on appeal. Id. It concluded that the police activity sheet was not material under Brady because “Howard was extensively cross-examined” and because “there were two eyewitnesses other than Howard who observed the shooting ■ at close range ... [and] positively identified [Dennis] as the shooter in a photo array, in a line up, and at trial.” Id.

Following the Pennsylvania Supreme Court ruling, Dennis filed a habeas corpus petition under 28 U.S.C. § 2254 in the United States District Court for the Eastern District of Pennsylvania for review of his conviction and death sentence. The District Court granted Dennis habeas relief based on Dennis’s Brady claims as to the Commonwealth’s failure to disclose the Cason receipt, the Frazier documents, and the police activity sheet containing Howard’s inconsistent statement. Dennis V, 966 F.Supp.2d at 518.

The District Court concluded that the state court’s ruling regarding the Cason receipt involved an unreasonable determination of the facts. The Pennsylvania Supreme Court had concluded that the receipt was not exculpatory because (1) “[Cason’s] testimony would not support Appellant’s alibi”; (2) it would have been cumulative of testimony by another witness; and (3) there was no evidence that the Commonwealth withheld the receipt from the defense. Dennis I, 715 A.2d at 408. The District Court determined that the receipt corroborated Dennis’s alibi, provided direct evidence that Cason’s testimony was false, and would have been strong impeachment evidence. Therefore, the state court’s determination that the receipt was not “exculpatory” was an unreasonable determination of the facts. Dennis V, 966 F.Supp.2d at 508.

The District Court also concluded the Pennsylvania Supreme Court had engaged in a similarly unreasonable determination of facts regarding whether the receipt was actually suppressed by the police. In its opinion, the Pennsylvania Supreme Court stated that the police came into possession of the receipt when interviewing Cason, and that the Commonwealth never claimed to have disclosed the receipt to defense counsel. The District Court relied on Kyles v. Whitley, 514 U.S. 419, 437, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995), for the proposition that favorable evidence in the police’s possession is imputed to the prosecution. Dennis V, 966 F.Supp.2d at 509-10. It also interpreted the three-factor balancing test in United States v. Pelullo, 399 F.3d 197 (3d Cir. 2005), to come out in favor of required disclosure by the Commonwealth. Further, the state court’s conclusion that the receipt was not material was an unreasonable application of clearly established federal law because the “receipt and Ca-son’s accompanying corrected testimony would have provided independent, disinterested corroboration of Dennis’[s] explanation for where he was at the time of Williams’[s] murder,” would have transformed Cason from a government witness into a defense witness who supported Den*279nis’s alibi, and would have provided impeachment evidence to challenge Cason’s testimony that she had worked until 2:00 p.m. that day, which otherwise could not have been challenged. Dennis V, 966 F.Supp.2d at 511.

The District Court also granted habeas relief on the basis of Dennis’s Brady claim regarding the Frazier documents, concluding that the state court had adopted an unreasonably narrow reading of Brady. The Pennsylvania Supreme Court had held that the prosecution did not violate Brady by failing to disclose the Frazier documents because Dennis did not show that the documents were admissible and material. The District Court rejected the assertion that inadmissible evidence cannot be the basis of a Brady claim, reasoning that the United States Supreme Court has never stated such a rule and that most circuit courts, including the Third Circuit, have held to the contrary. Id. at 503. Additionally, that the United States Supreme Court proceeded with the Brady analysis after acknowledging that the polygraph results at issue in Wood v. Bartholomew, 516 U.S. 1, 116 S.Ct. 7, 133 L.Ed.2d 1 (1995), were not admissible indicated to the District Court that there is no admissibility requirement for Brady evidence. Dennis V, 966 F.Supp.2d at 503.

The Pennsylvania Supreme Court had also held that the prosecution need not disclose every “fruitless lead” in order to comply with Brady. The District Court determined that this conclusion was unreasonable under Kyles. The Frazier documents contained “internal markers of credibility,” such as a description of the victim as “Kev[’s] ... girl,” which was accurate, an admission to shooting the victim in the correct location on her body, and a description of the alleged perpetrators that matched other descriptions of the shooter more closely than Dennis did. Id. at 504. The District Court reasoned that the Frazier documents would have led to further investigation that could have proved vital to the defense and could have been used to impeach the police investigation or provide a defense that another person committed the murder. Id. at 505.

Lastly, the District Court granted habe-as relief on the basis of Dennis’s claim that the Commonwealth violated Brady when it withheld the police activity sheet containing Howard’s inconsistent statements. The District Court concluded that the Pennsylvania Supreme Court had unreasonably applied Brady and its progeny in rejecting the Howard Brady claim. First, the Pennsylvania Supreme Court had unreasonably dismissed the impeachment value of the evidence and incorrectly concluded that cross-examination of Howard rendered new impeachment evidence immaterial. The District Court noted that the United States Supreme Court has directly rejected the notion that there can be no Brady claim relating to impeachment evidence where a witness was already impeached with other information. See Banks v. Dretke, 540 U.S. 668, 702, 124 S.Ct. 1256, 157 L.Ed.2d 1166 (2004) (rejecting the state’s argument that no Brady violation occurred because the witness was “heavily impeached at trial,” where the withheld evidence was the only impeachment evidence that the witness was a paid informant).15 The District Court emphasized that, although Howard was cross-examined at trial, she was not impeached. Dennis V, 966 F.Supp.2d at 514-15. Second, the District Court concluded that the Pennsylvania Supreme Court had incorrectly applied a sufficiency of the evidence test in direct contravention of Kyles’ s directive that Brady material be viewed in light of all of the *280evidence. Rather, the state court should have focused on whether the defendant received a fair trial in the absence of the disclosed evidence. Id. at 516. Finally, the District Court found it unreasonable that the state court had failed to consider the effect of the evidence on trial counsel’s investigation, pretrial preparation, decision to interview or call certain witnesses, or the effect of cross-examining detectives on their investigation into Howard. Given that the police themselves thought it was important to follow up with Howard about her possible statements to Pugh, the District Court concluded it was clear that the lead was material from an investigatory point of view. Id.

The District Court also concluded that the Pennsylvania Supreme Court had failed to undertake a cumulative materiality analysis as required by Kyles. Id. at 517-18. It did not rule on Dennis’s remaining claims. Id. at 491, 501 n.19 & 510 n.27. The Commonwealth filed a timely notice of appeal.

A panel of this Court issued an opinion on February 9, 2015. Dennis v. Sec’y, Pa. Dep’t of Corr., 777 F.3d 642 (3d Cir. 2015). This opinion was vacated and rehearing en bane was granted on May 6, 2015.

II. Jurisdiction and Standard of Review

The District Court had jurisdiction under 28 U.S.C. §§ 2241, and 2254 over Dennis’s habeas corpus petition. This Court has appellate jurisdiction under 28 U.S.C §§ 1291 and 2253. The District Court based its decision on a review of the state court record and did not conduct an evidentiary hearing, so our review of its order is plenary and we apply the same standard the District Court applied. Branch v. Sweeney, 758 F.3d 226, 232 (3d Cir. 2014); Brown v. Wenerowicz, 663 F.3d 619, 627 (3d Cir. 2011).

The Antiterrorism and Effective Death Penalty Act (AEDPA) dictates the manner in which we conduct our review. Federal habeas courts cannot grant relief “with respect to any claim that was adjudicated on the merits in State court” unless the adjudication (1) “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States”; or (2) “resulted in a decision that 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).

Under § 2254(d)(1), “clearly established federal law” means “the governing legal principle or principles set forth by the Supreme Court at the time the state court renders its decision.” Lockyer v. Andrade, 538 U.S. 63, 71-72, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003). It “refers to the holdings, as opposed to the dicta, of [the Supreme Court’s] decisions as of the time of the relevant state-court decision.” Williams v. Taylor, 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). AED-PA allows federal courts to grant habeas relief only if the state court decision is contrary to, or an unreasonable application of, clearly established federal law. 28 U.S.C. § 2254(d)(1). .

A state court decision is “contrary to” clearly established federal law if the state court (1) “applies a rule that contradicts the governing law” set forth in Supreme Court precedent or (2) “confronts a set of facts that are materially indistinguishable from a decision of [the Supreme] Court and nevertheless arrives at a result different” from that reached by the Supreme Court. Williams, 529 U.S. at 405-06, 120 S.Ct. 1495. Interpreting Supreme Court precedent in a manner that adds an additional element to the legal standard for proving a constitutional violation is “contrary to” clearly established federal *281law. Id. at 393-94, 397, 120 S.Ct. 1495 (reasoning that the Virginia Supreme Court’s interpretation of Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), which increased the burden on petitioners, was “contrary to” Supreme Court precedent).

A state court decision is an “unreasonable application of federal law” if the state court “identifies the correct governing legal principle,” but “unreasonably applies that principle to the facts of the prisoner’s case.” Id. at 413, 120 S.Ct. 1495. A strong case for habeas relief “does not mean the state court’s contrary conclusion was unreasonable.” Harrington v. Richter, 562 U.S. 86, 102, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011). Habeas relief may not be granted on the basis that the state court applied clearly established law incorrectly; rather, the inquiry is “whether the state court’s application of clearly established federal law was objectively unreasonable.” Williams, 529 U.S. at 409, 120 S.Ct. 1495 (emphasis added). A rule’s unreasonable application corresponds to the specificity of the rule itself: “[t]he more general the rule, the more leeway courts have in reaching outcomes in case-by-case determinations.” Richter, 562 U.S. at 101, 131 S.Ct. 770 (internal quotation marks and citation omitted). “A state court’s determination that a claim lacks merit precludes federal habeas relief so long as fairminded jurists could disagree on the correctness of the state court’s decision.” Id. (internal quotation marks omitted).

Finally, under 28 U.S.C. § 2254(d)(2), a state court decision is based on an “unreasonable determination of the facts” if the state court’s factual findings are “objectively unreasonable in light of the evidence presented in the state-court proceeding,” which requires review of whether there was sufficient evidence to support the state court’s factual findings. See Miller-El v. Cockrell, 537 U.S. 322, 340, 123 S.Ct. 1029, 154 L.Ed.2d 931 (2003). Determinations of factual issues made by state courts are presumed to be correct. 28 U.S.C. § 2254(e)(1); Miller-El, 537 U.S. at 340, 123 S.Ct. 1029. However, “[d]eference does not by definition preclude relief. A federal court can disagree with a state court’s credibility determination and, when guided by AEDPA, conclude the decision was unreasonable or that the factual premise was incorrect by clear and convincing evidence.” Miller-El, 537 U.S. at 340, 123 S.Ct. 1029.

Judges Fisher and Hardiman advance an interpretation of Richter that far exceeds its reach. Further, their approach would have the federal habeas courts “rewrite” state court opinions, as Judge Jordan’s thorough concurrence observes. We recognize that the AEDPA standard is “difficult to meet ... because it was meant to be. As amended by AEDPA, § 2254(d) stops short of imposing a complete bar on federal-court relitigation of claims already rejected in state proceedings.” Richter, 562 U.S. at 102, 131 S.Ct. 770. The highly deferential standard “reflects the view that habeas corpus is a guard against extreme malfunctions in the state criminal justice systems, not a substitute for ordinary error correction through appeal.” Id. at 102-OS, 131 S.Ct. 770 (internal quotation marks omitted). This level of deference stems from deep-rooted concerns about federalism. Williams, 529 U.S. at 406, 120 S.Ct. 1495 (noting that Congress intended to “further the principles of comity, finality, and federalism” in passing AEDPA). That said, Richter and its progeny do not support unchecked speculation by federal ha-beas courts in furtherance of AEDPA’s goals. While we must give state court decisions “the benefit of the doubt,” as Judge Fisher recognizes, federal habeas review does not entail speculating as to what other theories could have supported the state court ruling when reasoning has been pro*282vided, or buttressing a state court’s scant analysis with arguments not fairly presented to it. Make no mistake about it, the Dissents justify the state court ruling based on an argument never presented to it. No case decided by our court or the United States Supreme Court permits this approach. We now write to clarify how we interpret the Supreme Court’s jurisprudence as to when and how federal courts ought to “fill the gaps” in state court opinions on federal habeas review subject to AEDPA.

The United States Supreme Court has clearly laid out the analytical path for federal habeas courts confronted with a state court opinion devoid of reasoning — i.e., a bare ruling. When a state court decision lacks reasoning, the Supreme Court instructed habeas courts to “determine what arguments or theories supported or, as here, could have supported, the state court’s decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of this Court.” Richter, 562 U.S. at 102, 131 S.Ct. 770 (emphasis added). Richter is that case. This is not.

In Richter, the Court faced the question of whether AEDPA deference “applies when a state court’s order is unaccompanied by an opinion explaining the reasons relief has been denied.” Id. at 98; 131 S.Ct. 770. The United States Supreme Court admonished the Ninth Circuit’s de novo review of the California Supreme Court’s one-sentence summary denial of petitioner’s claim under Strickland, and held that state court decisions that are devoid of reasoning, i.e., a bare ruling, constitute adjudications on the merits that trigger AEDPA deference. Richter, 562 U.S. at 98, 131 S.Ct. 770 (“[T]he habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief. This is so whether or not the state court reveals which of the elements in a multipart claim it found insufficient....”). In pther words, state courts need not articulate a statement of reasons to invoke AEDPA deference by federal habeas courts. Id. (“[Djetermining whether a state court’s decision resulted from an unreasonable legal or factual conclusion does not require that there be an opinion from the state court explaining the state court’s reasoning.”). The California Supreme Court had provided no reasoning; accordingly, in order to determine whether the state court had made a decision that was contrary to, or involved an unreasonable application of, clearly established'federal law, or an unreasonable determination of fact, the federal habeas court was required to theorize based on what was presented to the state court.

We suggest that the concept of “gap filling” is fairly limited. It should be reserved for those cases in which the federal court cannot be sure of the precise basis for the state court’s ruling. It permits a federal court to defer while still exploring the possible reasons. It does not permit a federal habeas court, when faced with a reasoned determination of the state court, to fill a non-existent'“gap” by coming up with its own theory or argument, let alone one, as here, never raised to the state court. In Premo v. Moore, 562 U.S. 115, 131 S.Ct. 733, 178 L.Ed.2d 649 (2011), decided on the same day as Richter, the state court had concluded that the petitioner had not received ineffective assistance of counsel under Strickland, but did not specify on which Strickland prong — performance or prejudice — petitioner failed to meet his burden. As in Richter, the Supreme Court instructed the Ninth Circuit to assume “that both findings would have involved an unreasonable application of clearly established federal law.” 562 U.S. at 123, 131 S.Ct. 733. Unsure as to which *283prong formed the basis for the state court’s ruling, the federal court could fill the gap by exploring the two prongs of Strickland.

In contrast, when the state court pens a clear, reasoned opinion, federal habeas courts may not speculate as to theories that “could have supported” the state court’s decision. The Supreme Court established this limitation on Richter “gap filling” in Wetzel v. Lambert, — U.S. -, 132 S.Ct. 1195, 182 L.Ed.2d 35 (2012), where it described the proper analytical path for state court decisions accompanied by reasoning:

Under § 2254(d), a habeas court must determine what arguments or theories supported ... the state court’s decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of this Court.

Id. at 1198 (quoting Richter, 562 U.S. at 102, 131 S.Ct. 770; alterations in original; emphasis added). This is fairly straightforward. As explained above, the Court in Richter included the language “or, as here, could have supported” when it initially instructed courts on gap filling. Courts were tasked with considering what theories “could have supported” the state court decision in cases akin to those “as here,” or, summary denials. Removing the clause “or, as here, could have supported” from the instruction when the state court provides a fully-reasoned decision removed the task of speculative gap-filling from the habeas court’s analysis. Instead, federal habeas courts reviewing reasoned state court opinions are limited to “those arguments or theories” that actually supported, as opposed to “could have supported,” the state court’s decision. The Supreme Court’s intent to limit deference to the state court to those reasons that it articulated in its opinion is further supported by the Supreme Court’s instruction that the court on remand consider whether “each ground supporting the state court decision is examined and found to be unreasonable under AEDPA.” Id. at 1199.

When a state court ruling is based on a reasoned, but erroneous, analysis, federal habeas courts are empowered to engage in an alternate ground analysis— relying on any ground properly presented — but, in such a case, the federal court owes no deference to the state court. In Lafler v. Cooper, — U.S.-, 132 S.Ct. 1376, 182 L.Ed.2d 398 (2012), the state court had “simply found that respondent’s rejection of the plea was knowing and voluntary” in rejecting defendant’s ineffective counsel claim and “failed to apply Strickland, ” despite referencing the performance and prejudice prongs of Strickland in its opinion. Id. at 1390. “By failing to apply Strickland to assess the ineffective-assistance-of-counsel claim respondent raised, the state court’s adjudication was contrary to clearly established federal law” and the Supreme Court analyzed the Strickland claim de novo. Id. at 1390. The Court was not filling a gap in Lafler. Instead, it was employing different analysis that was very much a part of the case, and supplied an alternate ground for concluding, on de novo review, that there was no ineffectiveness of counsel.

Justices of the Supreme Court have indicated in a concurrence from the denial of a petition for certiorari that federal courts are bound to the text of state court opinions. Justice Ginsburg, joined by Justice Kagan, observed

Richter1 s hypothetical inquiry was necessary, however, because no state court opinion explained the reasons relief had been denied. In that circumstance, a federal habeas court can assess whether the state court’s decision involved an unreasonable application of clearly established *284Federal law only by hypothesizing reasons that might have supported it. But Richter makes clear that where the state court’s real reasons can be ascertained, the § 2254(d) analysis can and should be based on the actual arguments or theories that supported the state court’s decision.

Hittson v. Chatman, — U.S. -, 135 S.Ct. 2126, 2127-28, 192 L.Ed.2d 887, reh’g denied, — U.S. -, 136 S.Ct. 15, 192 L.Ed.2d 984 (2015) (mem.) (internal quotation marks, alterations, and citations omitted). Other courts of appeals have similarly limited Richter’s gap-filling instruction to the bare ruling situation. See Johnson v. Sec’y, DOC, 643 F.3d 907, 910 (11th Cir. 2011) (“When faced with an ineffective assistance of counsel claim that was denied on the merits by the state courts, a federal habeas court ‘must determine what arguments or theories supported or, [if none were stated], could have supported, the state court’s decision[.]” (alterations in. original) (quoting Richter, 562 U.S. at 102, 131 S.Ct. 770)); see also Grueninger v. Dir., Va. Dep’t of Corr., 813 F.3d 517, 525-26 (4th Cir. 2016) (“looking through” a state court summary refusal to hear an appeal to the prior reasoned decision and observing that “where there is no indication of the state court’s reasoning, a federal habeas petitioner must show that there was ‘no reasonable basis for the state court to deny relief,’ and a federal habeas court must defer under AEDPA to any reasonable ‘arguments or theories ... [that] could have supported!] the state court’s decision’ ” (quoting Richter, 562 U.S. at 98, 102, 131 S.Ct. 770) (internal citations omitted; alterations in original)); Montgomery v. Bobby, 654 F.3d 668, 700 (6th Cir. 2011) (Clay, J., dissenting) (“If the- state court articulated its reasons, the habeas court must identify and evaluate those reasons under § 2254(d); only if the state court did not articulate its reasons must the habeas court hypothesize as to the state court’s reasoning, and evaluate those hypothetical reasons.”). Federal courts should only gap-fill when presented with a bare ruling or when it is unsure as to the basis of the state court ruling on the issue presented. See Premo, 562 U.S. at 123, 131 S.Ct. 733 (concluding that when the state court neglected to articulate which prong of Strickland was deficient, the federal habeas court ought to evaluate both prongs of Strickland). We will not gap-fill when the state court has articulated its own clear reasoning. Instead, we will evaluate the state court’s analysis and review de novo any properly presented alternative ground(s) supporting its judgment.

Dennis’s claims at issue on appeal stem from the Commonwealth’s violations of Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). Prosecutors have an affirmative duty “to disclose [Brady] evidence ... even though there has been no request [for the evidence] by the accused,” which may include evidence known only to police. Strickler v. Greene, 527 U.S. 263, 280, 119 S.Ct. 1936, 144 L.Ed.2d 286 (1999); Kyles, 514 U.S. at 438, 115 S.Ct. 1555. To comply with Brady, prosecutors must “learn of any favorable evidence known to the others acting on the government’s behalf ..., including the police.” Strickler, 527 U.S. at 281, 119 S.Ct. 1936 (internal quotation marks omitted) (quoting Kyles, 514 U.S. at 437, 115 S.Ct. 1555).

To prove a Brady violation, a defendant must show the evidence at issue meets three critical elements. First, the evidence “must be favorable to the accused, either because it is exculpatory, or because it is impeaching.” Id. at 281-82, 119 S.Ct. 1936; see also United States v. Bagley, 473 U.S. 667, 676, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985) (“Impeachment evidence ..., as well as exculpatory evidence, falls within the Brady rule.”). Second, it *285“must have been suppressed by the State, either willfully or inadvertently.” Strickler, 527 U.S. at 282, 119 S.Ct. 1936. Third, the evidence must have been material such that prejudice resulted from its suppression. Id.; see also Banks, 540 U.S. at 691, 124 S.Ct. 1256. The “touchstone of materiality is a ‘reasonable probability’ of a different result.” Kyles, 514 U.S. at 434, 115 S.Ct. 1555. Materiality “does not require demonstration by a preponderance that disclosure of the suppressed evidence would have resulted ultimately in the defendant’s acquittal ... [Rather], [a] ■ ‘reasonable probability’ of a different result is ... shown when the government’s eviden-tiary suppression undermines confidence in the outcome of the trial.” Id. (internal quotation marks omitted).

III. Discussion

The District Court held that the Pennsylvania Supreme Court had unreasonably applied Brady and its progeny in rejecting Dennis’s claims that the prosecution was required under Brady to disclose the Ca-son receipt, the Frazier documents, and the police activity sheet containing Howard’s inconsistent statements. The Pennsylvania Supreme Court issued a thorough decision on each claim. We conclude, like the District Court, that the Pennsylvania Supreme Court’s decisions regarding Dennis’s Brady claims rested on unreasonable conclusions of fact and unreasonable applications of clearly established law, or were contrary to United States Supreme Court precedent. We will affirm the District Court and grant habeas relief on Dennis’s Brady claims based on the Cason receipt, the Howard police activity sheet, the Frazier documents, and their cumulative prejudice.

A. Cason Receipt

1. Facts

The Commonwealth did not disclose the DPW receipt that was in the police’s possession, provided objective impeachment evidence of a key Commonwealth witness, and bolstered Dennis’s alibi. Cason signed the DPW receipt when she picked up her check on October 22, 1991, the day of Williams’s murder. The receipt’s time stamp shows Cason picked up a $94.00 payment for “public assistance” at “13:03,” or 1:03 p.m. During Dennis’s direct appeal, Cason signed an affidavit detailing her recollection of the interview she had with police prior to Dennis’s trial. According to Cason, detectives brought a copy of the time-stamped receipt to the interview, and she “located and gave the detective [her] pink copy of the same receipt. The detective kept [her] copy of the receipt.” App. 1735.

The Commonwealth called Cason to testify at Dennis’s trial. She testified that she left work around 2:00 p.m., picked up her welfare check, ran errands, and saw Dennis when she got off the K bus “between 4:00 and 4:30.” App. 733. The receipt serves two functions: (1) it negates her testimony that she worked until 2:00 p.m. on October 22; and (2) it demonstrates that, contrary to Cason’s testimony at trial that she retrieved her receipt after 3:00 p.m., Cason actually picked up her check at 1:03 p.m. Cason admits in her affidavit that she “may have thought that the 13:03, which was on the receipt, was 3:03 p.m.” App. 1736. In light of the time-stamped receipt, Cason explained in her affidavit, she “would have seen [James] Dennis between 2:00 and 2:30 p.m. at the Abbotts-ford Homes, and not 4:00 to 4:30 that is in my statement.” Id.

2. State Court Decision

The Pennsylvania Supreme Court rejected Dennis’s Brady claim stemming from the Cason receipt. The Court found, consistent with Cason’s affidavit, that the “police came into possession of a Depart-*286merit of Public Welfare (DPW) receipt showing that Cason cashed her check at 1:03 p.m.” Dennis I, 715 A.2d at 408. In denying Dennis’s ineffective assistance of counsel claim, the Court held that Cason’s new version of events “would not support [Dennis’s] alibi [ ] because the murder occurred at 1:50 p.m., forty minutes earlier than Cason’s earliest estimate” of when she saw Dennis. Id. The Court further held that the corrected testimony “would have been cumulative of testimony of witness Willis Meredith, who testified that he saw [Dennis] at the Abbottsford Homes at approximately 2:15 to 2:30 p.m.” Id. The Court dismissed'the Brady claim because the receipt was “not exculpatory, because it had no bearing on [Dennis’s] alibi, and there [was] no evidence that the Commonwealth withheld the receipt from the defense.” Id.

S. AEDPA Review

The state court ruling was a reasoned ruling that the District Court could understand; no gaps needed to be filled. Dennis was entitled to habeas relief based on the Cason Brady claim only if he could demonstrate that the decision was an unreasonable application of, or contrary to, clearly established law, or an unreasonable determination of the facts. 28 U.S.C. § 2254(d). Addressing the reasoned view of the Pennsylvania Supreme Court, we conclude that it unreasonably applied Brady and its progeny in evaluating the Cason receipt and made unreasonable determinations of fact. The receipt would have served as independent documentary corroboration of a key witness for Dennis’s alibi defense, and suppression by the Commonwealth violated Brady.

a) Favorability

The Cason receipt provided exculpatory and impeachment evidence that would have bolstered Dennis’s alibi defense at trial, so it easily meets Brady’s first prong. Banks, 540 U.S. at 691, 124 S.Ct. 1256 (stating that both impeachment and exculpatory evidence satisfy the first Bradiy prong).

The Pennsylvania Supreme Court erred by failing to recognize the impeachment value of the Cason receipt, which would have provided documentary evidence that Cason testified falsely at trial. The United States Supreme Court has made plain that impeachment evidence may be considered favorable under Brady even if the jury might not afford it significant weight. See Kyles, 514 U.S. at 450-51, 115 S.Ct. 1555 (rejecting the state’s argument that the evidence was “neither impeachment nor exculpatory evidence” because the jury might not have substantially credited it; according to the Court, “[s]uch [an] argument ... confuses the weight of the evidence with its favorable tendency”).16

Dennis’s defense strategy pitted his credibility, and that of his witnesses, against eyewitness credibility, Cason’s testimony, and the testimony of the other prosecution witnesses. No physical evidence was admitted at trial. Evidence that challenged Dennis’s credibility, or that of other defense witnesses dike his father, was therefore particularly crucial to the outcome of the trial. As the District Court aptly noted:.

Armed with the receipt, Dennis’s counsel — at the very least — would have been able to show that Cason was mistaken about the timing of the afternoon, by pointing out that she could not possibly *287have worked until 2 p.m. since she was at the DPW center at 1:03 p.m. ... The time stamped receipt would have directly contradicted [Cason’s testimony that she didn’t get off work until 2:00 p.m.].

Dennis V, 966 F.Supp.2d at 508. Without evidence to challenge the veracity of Ca-son’s testimony, Dennis’s assertion that he saw Cason as he got off the K bus lost significant credibility, as did his father’s corroboration of Dennis’s version of his timeline.

Further, the Pennsylvania Supreme Court erroneously concluded that the receipt was not exculpatory because it did not affect Dennis’s alibi. Dennis I, 715 A.2d at 408. It held that Cason’s revised recollection of the day “would not support [Dennis’s] alibi [ ] because the murder occurred at 1:50 p.m., forty minutes earlier than Cason’s earliest estimate.” Id. This conclusion fails to recognize how Cason’s corrected testimony corroborates testimony provided by Dennis and other witnesses, namely, his father.

The Commonwealth argues that the Pennsylvania Supreme Court reasonably concluded that the receipt did not require disclosure pursuant to Brady because Ca-son’s corrected testimony would not have made it impossible for Dennis to have been at Fern Rock station when Williams was murdered. Cason’s affidavit stated that she saw Dennis at 2:30 p.m. at Abbottsford Homes. The Commonwealth contends that Dennis could have committed the murder at Fern Rock at 1:50 p.m. and returned to Abbottsford Homes by 2:30 p.m. because the shooter entered a waiting getaway car after the murder and it was a thirteen minute drive between the two. This view unreasonably discounts the buttressing effect Cason’s corrected testimony would have on Dennis’s alibi theory. Although Cason’s corrected testimony, assuming it would mirror precisely what she said in her affidavit, would not definitively place Dennis in a location where it was impossible for him to commit the murder, Cason’s testimony would have strengthened Dennis’s and his father’s testimony that Dennis had been with his father that afternoon and was on the bus at the time of the murder.

Validating Dennis’s and his father’s testimony about his alibi on the day in question is sufficient to demonstrate favorability under Brady. Exculpatory evidence need not show defendant’s innocence conclusively. Under Brady, “[e]xculpatory evidence includes material that goes to the heart of the defendant’s guilt or innocence as well as that which may well alter the jury’s judgment of the credibility of a crucial prosecution witness.” United States v. Starusko, 729 F.2d 256, 260 (3d Cir. 1984) (citing Giglio v. United States, 405 U.S. 150, 154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972)). That Cason’s corrected testimony does not wholly undermine the prosecution’s theory of guilt does not sap its exculpatory value. The Commonwealth had an obligation to disclose the receipt under Brady because it would have altered the jury’s judgment about Cason’s credibility. Cason’s evidence is not favorable simply because of where Cason said she saw Dennis as corrected in her affidavit — at Abbottsford Homes. Rather, as Dennis argues, the exculpatory value lies in corroborating testimony of witnesses at trial who otherwise received little objective reinforcement, and whose credibility, as a result of Cason’s mistaken testimony in the absence of the receipt, was seriously undermined.

The only discrepancy between Cason’s testimony and the alibi established by Dennis and his father was the precise time Cason and Dennis saw one another — Ca-son claimed to have seen Dennis around 4:00 or 4:30 p.m., while Dennis said it was around 2:30 p.m. As both parties note, the *288other witnesses that testified on behalf of Dennis were friends and family, who were vulnerable to arguments of bias. To the contrary, Cason offered disinterested testimony that corroborated the government’s theory. Although the Commonwealth indicates that Cason could have been discredited in a similar manner as Dennis’s other witnesses, nothing in the record indicates that Cason shared the type of close relationship with Dennis as other witnesses who testified on his behalf.

The receipt contradicted Cason’s testimony at trial. Her corrected recollection, coupled with a specific documentary basis, would have provided disinterested corroboration of Dennis’s and his father’s testimony. The Pennsylvania Supreme Court made an unreasonable determination of the facts and an unreasonable application of federal law in refusing to acknowledge the receipt’s exculpatory and impeachment value.

b) Suppression of the receipt

The Pennsylvania Supreme Court stated that “the police came into possession of [the] receipt” when interviewing Cason. Dennis I, 715 A.2d at 408. Later, in a section analyzing materiality, it concluded there was “no evidence that the Commonwealth withheld the receipt from the defense.” Id. The Pennsylvania Supreme Court provided no explanation for its latter statement, and we cannot be sure whether the court was assessing the facts or interpreting the law. If it was construing fact, it was clearly unreasonable because the police had the receipt and therefore so did the prosecution.17 See Kyles, 514 U.S. at 437-38, 115 S.Ct. 1555. If it was making a conclusion of law as to the duty to disclose, the conclusion is similarly problematic because the court ignored Kyles. As Judge Jordan observes in his concurrence, “[i]f one follows the instruction of Kyles, those two statements are impossible to harmonize.” J. Jordan Concurring Op. at 352.

Once the Pennsylvania Supreme Court determined that the police detectives had obtained the receipt from Cason, the Commonwealth had constructive possession and was required to disclose the receipt to Dennis prior to trial. In 1995, three years prior to the Pennsylvania Supreme Court’s decision, the United States Supreme Court explained this duty:

[T]he prosecution, which alone can know what is undisclosed, must be assigned the consequent responsibility to gauge the likely net effect of all [favorable] evidence and make disclosure whén the point of “reasonable probability” is reached. This in turn means that the individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government’s behalf in the case, including the police. But whether the prosecutor succeeds or fails in meeting this obligation (whether, that is, a failure to disclose is in good faith or bad faith), the prosecution’s responsibility for failing to disclose known, favorable evidence rising to a material level of importance is inescapable.

Kyles, 514 U.S. at 437-38, 115 S.Ct. 1555 (internal quotation marks and citation *289omitted). In ignoring Kyles’s instruction that prosecutors must disclose evidence obtained by the police, the Pennsylvania Supreme Court unreasonably applied clearly established federal law. The Commonwealth’s argument that the receipt did not appear in the prosecution file does nothing to undercut its duty to disclose under Kyles and, as the District Court correctly notes, borders on bad faith. It explained:

The Commonwealth admits that the entire homicide file — where one may expect a document recovered by the police to exist — went missing in March 1997, before the Commonwealth had submitted its direct appeal briefing. The Commonwealth may not point to a missing file and declare it the petitioner’s burden to prove that the receipt was, at one point, contained inside.

Dennis V, 966 F.Supp.2d at 509 (citation omitted). The Commonwealth has never asserted that it disclosed the receipt to Dennis. We refuse to allow it to evade its duty under Brady based on failure to adequately search or maintain its own files.

The Commonwealth argues that because Dennis’s appellate counsel was able to obtain the receipt from the DPW nearly five years post-trial, the prosecution had no responsibility under Brady to turn it over to defense counsel when the receipt came into its possession. Judge Fisher adopts this approach and excuses the Commonwealth from its Brady responsibility by injecting an argument that was not even mentioned by the Pennsylvania Supreme Court, much less fairly presented before it.

The Commonwealth did not raise a “due diligence” argument, as such, before the state court. Rather, in its Response to Defendant’s Reply Brief, the Commonwealth argued for the first time that there was no Brady violation because the receipt was publicly available. The entirety of the alleged due diligence argument is below.

[A]lthough defendant does not explain how he obtained a copy of [the Cason receipt], he presumably did so from the Department of Public Welfare, thus establishing its public availability. Brady does not require the Commonwealth to produce evidence that was not in its sole possession, but was available, as this document apparently was.

App. 2026. As Judge Jordan observes, Pennsylvania law generally regards arguments raised for the first time in reply briefs as waived. J. Jordan Concurring Op. at 353 n.9.

Further, our review on habeas is limited to the record as presented to the state court. See Cullen v. Pinholster, 563 U.S. 170, 181-82, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011). There was no evidence regarding the availability of the receipt. In fact, the Commonwealth’s assertion that the receipt was publicly available was incorrect, as it runs counter to specific Pennsylvania regulations in effect at the time. As they existed during Dennis’s appeal, the DPW’s privacy regulations protected the vast majority of private information; the only exception was that the Commonwealth may disclose “the address and amount of assistance a person is currently receiving” following a direct request about a specific person. 55 Pa. Code § 105.4(a)(1). Even if the DPW receives a subpoena requesting information about a recipient, it must challenge that demand and “plead, in support of its request to withhold information, that under the Public Welfare Code (62 P.S. §§ 101-1503), the rules of the Department prohibit the disclosure of information in records and files, including the names of clients, except as provided in subsection (a).” Id. § 105.4(b)(3). To the extent that information was publicly available regarding Ca-son’s public assistance payments, it was limited to Cason’s address and her amount *290of assistance, which is irrelevant to her interaction with Dennis on the day of Williams’s murder. Only the Commonwealth held information that would support Dennis’s alibi — the time-stamped receipt Cason provided to the police.

Even if we were to imagine that a diligence argument was presented and considered by the state court, the United States Supreme Court has never recognized an affirmative due diligence duty of defense counsel as part of Brady, let alone an exception to the mandate of Brady as this would t clearly be. The Supreme Court has noted that its precedent “lend[s] no support to the notion that defendants must scavenge for hints of undisclosed Brady material when the prosecution represents that all such material has been disclosed.” Banks, 540 U.S. at 695, 124 S.Ct. 1256. To the contrary, defense counsel is entitled to presume that prosecutors have “discharged their official duties.” Id. at 696, 124 S.Ct. 1256 (quoting Bracy v. Gramley, 520 U.S. 899, 909, 117 S.Ct. 1793, 138 L.Ed.2d 97 (1997)). Further, the duty to disclose under Brady is absolute — it does not depend on defense counsel’s actions. United States v. Agurs, 427 U.S. 97, 107, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976) (“[I]f the evidence is so clearly supportive of a claim of innocence that it gives the prosecution notice of a duty to produce, that duty should equally arise even if no request is made.”). Brady’s mandate and its progeny are entirely focused on prosecutorial disclosure, not de: fense counsel’s diligence.

The emphasis in the United States Supreme Court’s Brady jurisprudence on fairness in criminal trials reflects Brady’s concern with the government’s unquestionable advantage in criminal proceedings, which the Court has explicitly recognized. See, e.g., Strickler, 527 U.S. at 281, 119 S.Ct. 1936 (reasoning that the “special status” of the prosecutor in the American legal system, whose interest “in a criminal prosecution is not that [he] shall win a case, but that justice shall be done ... explains ... the basis fpr the prosecution’s broad duty of disclosure” (quoting Berger v. United States, 295 U.S. 78, 88, 55 S.Ct. 629, 79 L.Ed. 1314 (1935))). Construing Brady in a manner that encourages disclosure reflects the Court’s concern with prosecutorial advantage and prevents shifting the burden onto defense counsel to defend his actions. Requiring an undefined quantum of diligence on the part of defense counsel, however, would enable precisely that result — it would dilute Brady’s equalizing impact on prosecutorial advantage by shifting the burden to satisfy the claim onto defense counsel.

The focus on disclosure by the prosecutor, not diligence by defense, is reiterated in the Supreme Court’s approval of the shift in the traditional adversarial system Brady imposes. In United States v. Bagley, the Court explained that “[b]y requiring the prosecutor to assist the defense in making its case, the Brady rule represents a limited departure from a pure adversary model” because the prosecutor is not tasked simply with winning a case, but ensuring justice. 473 U.S. 667, 675 n.6, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985). Further, the Court placed the burden of obtaining favorable evidence squarely on the prosecutor’s shoulders. See Kyles, 514 U.S. at 437, 115 S.Ct. 1555 (“[T]he individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government’s behalf in the ease.”). That the government may be burdened by the Brady rule does not undercut its need to comply with it. The imposition of an affirmative due diligence requirement on defense counsel would erode the prosecutor’s obligation under, and the basis for, Brady itself.

Indeed, the United States Supreme Court has cautioned against such a rule. It *291has rejected the notion that defense counsel’s diligence is relevant in assessing “cause” for the failure to raise a Brady suppression issue in state court proceedings. In Strickler, it reasoned that because counsel was entitled to rely on the prosecutor fulfilling its Brady obligation, and' had no reason for believing it had failed to comply, the failure to raise the issue earlier in habeas proceedings was justified. See Strickler, 527 U.S. at 286-89, 119 S.Ct. 1936. Similarly here, the prosecutor’s duty is clear. Dennis’s counsel was entitled to rely on the prosecutor’s duty to turn over exculpatory evidence.18 Assessing whether he could or should have discovered the receipt is beside the point.19

In Banks, the Supreme Court explicitly rejected the notion that “the prosecution can lie and conceal and the prisoner still has the burden to ... discover the evidence, so long as the potential existence of a prosecutorial misconduct claim might have been detected.” 540 U.S. at 696, 124 S.Ct. 1256 (internal quotation marks and citations omitted). Banks concluded that “[a] rule ... declaring ‘prosecutor may hide, defendant must seek,’ is not tenable in a system constitutionally bound to accord defendants due process.” Id.; see also United States v. Tavera, 719 F.3d 705, 712 (6th Cir. 2013) (recognizing that “the clear holding in Banks” does away with any belief that Brady imposes a due diligence requirement on defense counsel); Bell v. Bell, 512 F.3d 223, 242 (6th Cir. 2008) (Clay, J., dissenting) (“The rule emerging from Strickler and Banks is clear: Where the prosecution makes an affirmative representation that no Brady material exists, but in fact has Brady material in its possession, the petitioner will not be penalized for failing to discover that material.”).

While we think that the United States Supreme Court has made it clear that Brady requires the prosecution to turn over all material favorable evidence in its possession, we acknowledge that it is not totally frivolous under our Third Circuit jurisprudence for the Commonwealth to have argued, as it did here, that because defense counsel could or should have discovered the Cason receipt with due diligence, the prosecution was not required to disclose it.20 That is because our case law, as we discuss below, is inconsistent and could easily confuse. Thus, we need to clarify our position: the concept of “due diligence” plays no role in the Brady analysis.21 To the contrary, the focus of the *292Supreme Court has been, and it must always be, on whether the government has unfairly “suppressed” the evidence in question in derogation of its duty of disclosure. See Gov’t of the V.I. v. Mills, 821 F.3d 448, 460 n.10 (3d Cir. 2016) (“The critical question in assessing constitutional error is to what extent a defendant’s rights were violated, not the culpability of the prosecutor.” (quoting Marshall v. Hendricks, 307 F.3d 36, 68 (3d Cir. 2002))).

In Brady, the United States Supreme Court held “that the suppression by the prosecution of evidence favorable to an accused upon request violates due process where the evidence is material either to guilt or to punishment, irrespective of the good faith or bad faith of the prosecution.” Brady, 373 U.S. at 87, 83 S.Ct. 1194 (emphasis added). Suppression is “[t]he prosecution’s withholding from the defense of evidence that is favorable to the defendant.” Suppression of Evidence, Black’s Law Dictionary (10th ed. 2014). Inquiries into prosecutorial suppression are, by nature, retrospective as to the actions of the prosecutor — they do not place affirmative duties on defense counsel pre-trial. Agurs, 427 U.S. at 108, 96 S.Ct. 2392 (“[T]he prosecutor will not have violated his constitutional duty of disclosure unless his omission is of sufficient significance to result in the denial of the defendant’s right to a fair trial”).

The government must disclose all favorable evidence. Only when the government is aware that the defense counsel already has the material in its possession should it be held to not have “suppressed” it in not turning it over to the defense. Any other rule presents too slippery a slope. In United States v. Perdomo, 929 F.2d 967, 973 (3d Cir. 1991), and United States v. Starusko, 729 F.2d 256, 262 (3d Cir. 1984), we opened the door to a due diligence exception to Brady. Starusko, 729 F.2d at 262 (“ ‘[T]he government is not obliged under Brady to furnish a defendant with information which he already has or, with any reasonable diligence, he can obtain himself.’ ” (quoting United States v. Campagnuolo, 592 F.2d 852, 861 (5th Cir. 1979))). In Grant v. Lockett, 709 F.3d 224, 230-31 (3d Cir. 2013), we may have widened that opening when we combined our conclusion that defense counsel was constitutionally ineffective in violation of the defendant’s rights with a finding that there was no Brady violation because counsel clearly should have discovered the prosecutor’s key witness’s criminal record and been aware that he was on parole when the shooting occurred and when he testified at trial. We did note in Grant that Grant himself had obtained the witness’s criminal records while in custody, but we did not rest our ruling on that fact.

In Wilson v. Beard, 589 F.3d 651, 663-64 (3d Cir. 2009), we got it right. There we concluded that “[i]f the prosecution has the obligation, pursuant to Perdomo, to notify defense counsel that a government witness has a criminal record even when that witness was represented by someone in defense counsel’s office, the fact that a criminal record is a public document cannot absolve the prosecutor of her responsibility to provide that record to defense counsel.” Id. (emphasis added) (internal quotation marks and citations omitted). Thus, we held that a criminal record, which arguably could have been discovered by defense counsel, is suppressed if not disclosed. Defense counsel in Wilson certainly had the ability to obtain the alleged Brady material — a criminal record — by virtue of his legal training. Yet we required disclosure pursuant to Brady. We also got it right in Pelullo when we rejected defendant’s ar*293gument that certain documents were Brady material and somehow “suppressed” when the government had made the materials available for inspection and they were defendant’s own documents. Pelullo, 399 F.3d at 212 (“[T]he government repeatedly made the warehouse documents available to [the defendant] and his attorneys for inspection and copying.”).

To the extent that we have considered defense counsel’s purported obligation to exercise due diligence to excuse the government’s non-disclosure of material exculpatory evidence, we reject that concept as an unwarranted dilution of Brady’s clear mandate. Subjective speculation as to defense counsel’s knowledge or access may be inaccurate, and it breathes uncertainty into an area that should be certain and sure. See Weisburd, swpra, at 164 (“[P]ros-ecutors ... cannot accurately speculate about what a defendant or defense lawyer could discover through due diligence. Prosecutors are not privy to the investigation plan or the investigative resources of any given defendant or defense lawyer.”). The United States Supreme Court agrees. It has recognized that ample disclosure is “as it should be” because it “tend[s] to preserve the criminal trial, as distinct from the prosecutor’s private deliberations, as the chosen forum for ascertaining the truth about criminal accusations.... The prudence of the careful prosecutor should not therefore be discouraged.” Kyles, 514 U.S. at 439-40, 115 S.Ct. 1555 (internal citations omitted).

All favorable material ought to be disclosed by the prosecution. To hold otherwise would, in essence, add a fourth prong to the inquiry, contrary to the Supreme Court’s directive that we are not to do so. In Williams v. Taylor, the Virginia Supreme Court had interpreted the Supreme Court’s decision in Lockhart v. Fretwell, 506 U.S. 364, 369, 113 S.Ct. 838, 122 L.Ed.2d 180 (1993) “to require a separate inquiry into fundamental fairness even when [petitioner] [was] able to show that his lawyer was ineffective and that his ineffectiveness probably affected the out7 come of the proceeding.” 529 U.S. 362, 393, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). The Court held that the Virginia Supreme Court’s imposition of this additional test was an unreasonable application of, and contrary to, Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Williams, 529 U.S. at 393-94, 120 S.Ct. 1495. Adding due diligence, whether framed as an affirmative requirement of defense counsel or as an exception from the prosecutor’s duty, to the well-established three-pronged Brady inquiry would similarly be an unreasonable application of, and contrary to, Brady and its progeny.

The Pennsylvania Supreme Court’s conclusion that the prosecution did not withhold the Cason receipt was an unreasonable application of law and fact. The receipt was in its possession pursuant to Kyles and, under United States Supreme Court precedent, it is clear that there is no additional prong to Brady and no “hide and seek” exception depending on defense counsel’s knowledge or diligence. See Banks, 540 U.S. at 696, 124 S.Ct. 1256.

c) Materiality

Without a doubt, Dennis suffered prejudice due to the Commonwealth’s failure to disclose the receipt. The defense strategy was rooted in Dennis’s alibi that he was getting on the K bus at the time of the murder. The Commonwealth’s withholding of the receipt transformed a witness who would otherwise have been an alibi witness for Dennis into a witness for the prosecution or, at least, left Dennis powerless to impeach Cason’s false testimony if offered by the prosecution. The state court’s conclusion that Dennis suf*294fered no prejudice is an unreasonable determination of fact and law.

Failure to disclose the Cason receipt made testimony by a key government witness, who provided the sole testimony contradicting Dennis’s alibi, unassailable. The Commonwealth highlighted how weighty Cason’s testimony was at trial. In his opening, referring to Cason as simply a “lady from the neighborhood,” ADA King emphasized the discrepancy between Ca-son’s and Dennis’s testimony: “[Cason] had something very interesting to say. Yeah, I saw him when I was on the bus, but it wasn’t 2:00, it was 4:00.” App. 404. At closing, King reiterated the inconsistencies between Cason’s and Dennis’s testimony, and added that “[the Commonwealth] called her, not the defense. She came in and said, I was at work at 2:00. I saw him somewhere between 4:00 and. 4:30. Try again, Jimmy. That one didn’t work.” App. 1209. Disclosure of the receipt would have given defense counsel evidence to demonstrate that Cason falsely testified when she asserted that she worked until 2:00 p.m. on October 22. Disclosure would have allowed defense counsel to undermine Cason’s credibility or would have caused her to correct her testimony — as she did later in an affidavit — so as to support Dennis’s version of events. Impeachment using the receipt may have caused Cason to explain to the jury that her prior testimony rested on a misunderstanding of military time and allowed Cason to correct her timeline during trial. More likely, the prosecution would not have called Cason at all, and Dennis would have called Cason to corroborate his testimony.22 Finally, ADA King would not have, at closing, been able to point out the inconsistencies between Dennis’s and Cason’s testimonies.

Cason’s uncorreeted testimony left the jury with conflicting stories as to Dennis and Cason’s interactions on the day of the murder. Following Cason’s testimony that she could not have seen Dennis between 2:00 and 2:30 p.m., Dennis qualified his trial testimony and said that he only “thought” he saw Cason. App. 1030. During closing, Dennis’s counsel told the jury, “Remember what [Dennis] told you when he got up there? It’s wrong. He didn’t see [Cason] on the bus. He thought he saw her on the bus, but he didn’t.” App. 1179-80. The District Court thoughtfully explained how Dennis’s uncorrected testimony damaged defense counsel’s strategy:

This scrambled explanation left the jury with two options, equally unhelpful to Dennis: believe that Cason and Dennis had seen each other on the bus, as both testified, but that it happened later than Dennis said — and therefore find no alibi for the time of the crime; or believe counsel’s new story that Dennis was on the earlier bus, and thus could not have committed 'the crime, but never saw Ca-son at all. Cason’s corrected testimony would have transformed Cason from a damaging Commonwealth witness to a uniquely powerful, disinterested defense witness who would provide document-supported corroboration for Dennis’fs] alibi....

Dennis V, 966 F.Supp.2d at 512. The impeachment value the receipt provided would have eliminated the conflicting stories for the jury and, given the weight of Cason’s testimony alleged by the prosecution at trial, could have raised significant doubt about Dennis’s guilt. The state court’s determination that Dennis did not suffer prejudice as a result of Cason’s unchallenged testimony was unreasonable. *295In concluding that the Commonwealth had evidence that its witness’s testimony was false, we need not reach whether the prosecutors here intentionally presented false evidence because the inquiry is solely the impact that the absence of evidence had on the trial. See Brady, 373 U.S. at 87, 83 S.Ct. 1194; Mills, 821 F.3d at 460 n.10.

In Banks, the United States Supreme Court admonished prosecutors for letting statements by an informant, which they believed to be false, stand uncorrected throughout the proceedings. The Court concluded that “prosecutors represented at trial and in. state postconviction proceedings that the State had held nothing back ... It was not incumbent on Banks to prove these representations false; rather, Banks was entitled to treat the prosecutor’s submissions as truthful.” 540 U.S. at 698, 124 S.Ct. 1256. Earlier Brady cases indicate similar concern for allowing false testimony. See, e.g., Agurs, 427 U.S. at 103, 96 S.Ct. 2392 (“[Cjonviction obtained by the knowing use of perjured testimony is fundamentally unfair, and must be set aside if there is any reasonable likelihood that the false testimony could have affected the judgment of the jury.”) (footnotes omitted). Letting Cason’s testimony stand when the Commonwealth had evidence it was false unquestionably violated Brady and entitles Dennis to a new trial.

The state court took an unreasonably narrow view of Brady materiality by focusing on the fact that5 Cason would only have been able to say that she saw him around 2:30 p.m. Cason’s testimony need not fully corroborate Dennis’s alibi in order to show materiality under Brady. Kyles explained that Brady materiality does not turn on a determination of the sufficiency of the evidence, but instead requires the court to consider the constitutional error in light of all the evidence to determine whether it “put[s] the whole case in such a different light as to undermine, confidence in the verdict.” 514 U.S. at 435, 115 S.Ct. 1555. Transforming Cason, a disinterested individual with documentary support, into a defense witness meets the requirements of Brady materiality because it would have necessarily bolstered Dennis’s alibi defense narrative and “put the whole case in ... a different light.” Id.

Dennis testified that his father drove him to the bus stop around 1:50 p.m., where he boarded the K bus. Dennis asserted in his statement to police, which was read into the record at trial, that he waved at Cason when “we got off’ the K bus at Abbottsford Homes, a trip that generally took about thirty minutes. App. 710 (emphasis added). Dennis’s statement implies that they rode the K bus together and, setting aside the difference in time-lines, Cason’s testimony aligns with his account since Cason also took the K bus to Abbottsford Homes and saw Dennis there after she disembarked. Regardless of whether the receipt would have refreshed Cason’s memory enough to cause her to testify that she and Dennis were on the 1:56 p.m. K bus together, it certainly would have empowered defense counsel to elicit testimony from Cason that the loca-' tion in which she saw Dennis was consistent with her exiting the bus at the same time he did and to acknowledge that even if she did not notice him on the bus, she had no reason to disbelieve that he was there.

Cason, unlike the other witnesses Dennis called, did not know him well. Cason testified that she knew Dennis, but when ADA King asked her how long she had known him, Cason replied, “I don’t really, you know, know him, I know him by living up my way” at Abbottsford Homes. App. 731. Because Cason simply knew Dennis from the neighborhood, she served as a significantly less interested witness compared to Dennis’s other testifying witnesses, who were all close friends, family, *296and church leaders. As a result, she was less vulnerable to accusations of bias, and her testimony in support would have carried more weight with the jury. This is particularly important given the nature of her testimony compared to Dennis’s other witnesses. Unlike Dennis’s other witnesses, Cason’s testimony would have been supported by documentary proof of her timeline, the time-stamped receipt, to provide independent credibility to her testimony. In light of the receipt, Cason’s testimony on Dennis’s behalf would have been doubly strong — she was disinterested, and the receipt provided documentary corroboration for her version of the events.23

The Commonwealth criticizes the District Court’s analysis of the Cason receipt Brady claim as a misinterpretation of the record. Primarily, this critique rests on the District Court’s conclusion that the Pennsylvania Supreme Court “overlookfed] the fact that both Cason and Dennis testified that they saw each other on the bus.” Dennis V, 966 F.Supp.2d at 511. While it is true that Cason did not testify at trial that she saw Dennis on the K bus, nor did she deny it, and the Commonwealth’s failure to turn over the receipt deprived defense counsel of the opportunity to refresh Ca-son’s memory with the receipt or at least elicit that she saw Dennis immediately upon exiting the bus, thereby corroborating that they exited at the same location. Given that her unrefreshed testimony put the encounter after 4:00 p.m., defense counsel had no reason to elicit such testimony. But whether Cason testified that she saw Dennis on the bus or disembarking the bus, such testimony would have reinforced Dennis’s own testimony that he was on the bus and placed him in a location that would have made it practically impossible for him to murder Williams. Brady, therefore, required that the Commonwealth disclose the receipt.

At minimum, Cason’s time-stamped receipt would have empowered defense counsel to effectively impeach one of the Commonwealth’s strongest witnesses and mitigated the devastating effect of her testimony on Dennis’s credibility and his father’s. At most, the Commonwealth’s ease would have been short one witness, and Dennis’s alibi defense strategy would have been doubly strong due to (1) Cason’s status as a disinterested defense witness with the documentary corroboration and (2) the resulting increase in Dennis’s and his father’s credibility. The Pennsylvania Supreme Court was therefore unreasonable in concluding that the receipt was not favorable to Dennis when it would have bolstered his alibi. It was unreasonable in concluding that there was “no evidence” that the Commonwealth had suppressed the receipt when the state court found that detectives had the receipt in their possession. And finally, it was unreasonable in concluding that the receipt was not material. Had the Commonwealth disclosed the receipt, the jury may well have credited Dennis’s alibi defense.

B. Howard Police Activity Sheet

1. Facts

A suppressed police activity sheet reveals that two days after Williams’s murder, Zahra Howard, an eyewitness and key witness for the Commonwealth at trial, made a statement to Williams’s aunt and *297uncle, Dianne and Mannasett Pugh, that was inconsistent with an earlier statement she had made to police. Shortly after the murder, Howard told police that she did not recognize the shooter from school. The Pughs told police, however, that Howard told them the day after the murder that she knew the perpetrators from Olney High School, and that “Kim” and “Quinton” were at the scene when the shooting occurred. App. 1506. Quinton was Dianne Pugh’s nephew. The police indicated in their “THINGS TO DO” list that they intended to speak with the Pughs again and “[i]nterview Zahra Howard again” in light of her inconsistent statement to the Pughs. App. 1507. When police met with Howard the following day, however, they did not ask Howard about her conversation with the Pughs.

2. State court decision

The Pennsylvania Supreme Court initially characterized Dennis’s Brady claim regarding Howard’s inconsistent statement as one “with at least arguable merit.” Dennis III, 950 A.2d at 969. But the court was not prepared to rule on the record before it, and it remanded the Howard Brady claim to the PCRA court to address that claim in the first instance. Id.

The PCRA court rejected the Brady claim following an evidentiary hearing. The District Court aptly summarized the PCRA hearing and decision by the Pennsylvania Supreme Court:

Dennis sought to argue the merits of the Brady claim on the papers; he objected to the introduction of evidence from Howard and Diane Pugh because, he argued, their recollections now, a decade after the trial, about who the shooter was or what they told the police had no relevance on the question of whether the Commonwealth had violated Brady by failing to disclose the activity sheet. As Dennis’s PCRA counsel told the court:
The testimony has to be evaluated in its trial context. And all we can do at this point is put on paper for the court what we expect the impeachment to have been, assuming, for example, Zahra Howard denies having made the statement. We have to demonstrate on paper how she could have been impeached, and how that evidence relates to other evidence in the case.... Her testimony today about what she remembers from 16 years ago we can cross-examine, but it doesn’t illuminate the question of materiality in the context of the trial.

NT 12/22/08 at 15. The court allowed the testimony over Dennis’s objections. As expected, both Howard and Pugh denied that Howard had ever suggested that she recognized the assailants. Pugh’s testimony should not carry much weight, however, given that she declared before she was even sworn in, “I don’t remember nothing, nothing at all. It’s been 15, 16 years so I don’t remember. They just subpoenaed me and I’m here.” Id. at 56.

The PCRA court ultimately rejected the Brady claim. It noted that, during the hearing, Howard “testified credibly that she did not know the appellant from Olney High School, nor had she seen him prior to the murder.” Commonwealth v. Dennis, Case No. 92-01-0484, slip op. (Pa.Ct.Com.Pl. Mar. 17, 2010), at 18. Although the question whether Howard recognized James Dennis (“the appellant”) or had seen him before the murder is entirely irrelevant to whether she told Diane Pugh that she had seen the shooter before the murder, this is, in fact, the entirety of the. testimony that the Commonwealth elicited from Howard at the PCRA hearing:

Q: And in that conversation [with Diane Pugh] did you ever say anything about recognizing the defendant before?
*298A: No.
Q: Did you ever see the defendant at Olney High School?
A: No.
Q: Did you ever see him around Olney
High School?
A: No.

NT 12/22/08 at 18. On cross, when Dennis’s lawyer asked her about whether she said she had ever seen the shooter before, or whether she had ever told anyone she recognized the shooter from Olney High School, Howard denied recognizing the shooter or having ever said she did. Id. at 25-27.

Given both trial and PCRA counsel’s thorough cross-examination of Howard, the PCRA court determined that it was “unlikely that any additional impeachment evidence contained in the police activity sheet ... would have created a reasonable probability that the result of the proceeding would have been different had it been disclosed.” Dennis, slip op. at 14. The court further noted that the government’s case at trial “did not rest solely on” Howard’s testimony. Id. Finally, the contents of the activity sheet amounted to inadmissible hearsay, which “cannot be the basis for a Brady violation.” Id. at 15.

The Pennsylvania Supreme Court largely accepted the PCRA court’s determinations, despite its seeming, recognition, in Dennis III, of the investigatory value the activity sheet would have had and its earlier dismissal of the admissibility issue. It agreed that Dennis had failed to prove a reasonable probability of a different result had the activity sheet been disclosed. Commonwealth v. Dennis, 609 Pa. 442, 17 A.3d 297, 309 (2011) (“Dennis TP”). It echoed the PCRA court in noting that “Howard was extensively cross-examined by defense counsel in an attempt to impeach her testimony during trial,” and that “there were two eyewitnesses other than Howard” who identified Dennis; “[t]he disclosure of the activity sheet would have had no impact upon these eyewitnesses’ testimony.” Id. It did not specifically address the question of admissibility.

Dennis V, 966 F.Supp.2d at 513-14.

S. AEDPA Review

There is no question that Howard’s inconsistent statement would have been helpful to the defense but was not revealed to defense counsel until PCRA discovery, ten years after trial. The Pennsylvania Supreme Court denied Dennis’s Brady claim regarding the Howard statement on materiality grounds. Although the court articulated the proper standard for materiality, whether a “reasonable probability” of a different outcome has been established, it applied Kyles in a manner inconsistent with Supreme Court precedent.

First and foremost, defense counsel could have used Howard’s inconsistent statement as an effective means of impeachment during trial. As noted above, impeachment evidence unquestionably falls under Brady’s purview and cannot be suppressed by the prosecution. The Commonwealth notes that evidence is not necessarily material under Brady simply because it may open up avenues for impeachment— the focus of the inquiry is on the “reasonable probability of a different result” under Kyles. Such a probability exists here. The type of impeachment evidence provided by the activity sheet would have undercut the credibility of a key prosecution witness in a manner not duplicated by other challenges the defense was able to level at trial. Consequently, the impeachment material provided by the suppressed activity sheet is material under Brady, and it was unreasonable for the Pennsylvania Supreme Court to hold otherwise.

*299Howard was the Commonwealth’s key eyewitness against Dennis and the Commonwealth accordingly highlighted her testimony. ADA King emphasized the importance of Howard’s testimony in his closing argument: “[I]f you believe Zahra Howard, that’s enough to convict James Dennis.” App. 1207. As Williams’s Mend and the person with the closest view of the shooter, Howard’s testimony carried significant emotional and practical weight with the jury.24

Unlike other testifying eyewitnesses, Howard had views of the perpetrator at numerous stages during the incident. At trial, Howard testified that she saw the shooter for approximately twenty seconds total. This comported with her testimony' at the preliminary hearing. The two other testifying eyewitnesses’ views were much briefer. Bertha testified at the preliminary hearing that he saw the assailant for about a second. At trial, he expanded the amount of time he said he saw the shooter to three or four seconds. Cameron initially testified at the preliminary hearing that he saw the assailant for twenty seconds but upped the amount of time to thirty to forty seconds at trial. Notably, Cameron qualified his testimony by admitting that he “didn’t really pay attention.”25 App. 664. In contrast to Bertha and Cameron’s, Howard’s testimony was consistent, lengthy, and involved numerous views of the assailant— on the subway stairs, during the face-to-face encounter and finally, when Williams was shot. Because of the consistency and emotional weight of Howard’s testimony, defense counsel’s strategy was heavily reliant on impeaching Howard by any means — counsel attempted to “discredit her any ... way [he] could.” App. 1326.

Counsel’s ability to discredit Howard was limited, however. Without evidence that would directly contradict Howard’s testimony at trial, defense counsel sought to discredit Howard by pointing out her initial hesitation in identifying Dennis as the perpetrator during the photo array. Counsel could not challenge Howard’s trial testimony on other grounds. But prosecutors held contradictory statements by Howard about whether she recognized the perpetrators. Howard had initially told police, and later testified at trial, that she had never seen the perpetrators before and had not recognized them from school. According to the Pughs, however, Howard had said she recognized the shooter from Olney High School. The Pughs (along with Parker) also stated that Howard had also identified two other individuals, Kim and Quinton, as being present at the scene.

As noted by the District Court, cross-examination does not equate to actu*300al impeachment. Defense counsel cross-examined Howard, but he could only engage in limited questioning focused on challenging her hesitation identifications of Dennis as the shooter. This is decidedly different from the actual impeachment enabled by the activity sheet. In Banks, a witness was heavily impeached at trial, but the prosecution suppressed evidence that the witness served as a paid informant. 540' U.S. at 702, 124 S.Ct. 1256. Accordingly, none of the impeachment conducted at trial covered his status as an informant; the jury weighed his credibility without knowing this. Id. at 702-03, 124 S.Ct. 1256. The Supreme Court rejected the state’s argument that because the witness was heavily impeached, further impeachment evidence was immaterial. Id. at 702, 124 S.Ct. 1256. We have similarly indicated that additional impeachment evidence helps to substantiate Brady claims in a way that might make them material. In Lambert v. Beard, we stated that “it is patently unreasonable to presume — without explanation — that whenever a witness is impeached in one manner, any other impeachment becomes immaterial.” 633 F.3d 126, 134 (3d Cir. 2011), judgment vacated on other grounds sub nom. Wetzel v. Lambert, — U.S. -, 132 S.Ct. 1195, 182 L.Ed.2d 35 (2012). The mere fact that a witness has been heavily cross-examined or impeached at trial does not preclude a determination that additional impeachment evidence is material under Brady.

Indeed, we have granted habeas relief on the basis of a “significant difference” between the suppressed impeachment and other types of impeachment evidence used at trial. Slutzker v. Johnson, 393 F.3d 373, 387 (3d Cir. 2004). In Slutzker, we held that a police report memorializing a witness’s inconsistent statement was significantly different from the reports used to impeach the witness at trial. In the reports used at trial, the witness failed to identify the defendant, but in the suppressed report, she definitively stated that the man she saw was not the defendant. We concluded that “[t]he latter is much more convincing impeachment evidence, and the failure to disclose it leaves us in doubt that the trial verdict was worthy of confidence.” Id. The police activity sheet memorializing Howard’s statement similarly provides distinct and persuasive impeachment material that discredits Howard’s testimony more thoroughly than the identification challenges defense counsel levelled at trial.

The Commonwealth relies on United States v. Walker, 657 F.3d 160 (3d Cir. 2011), and United States v. Perez, 280 F.3d 318 (3d Cir. 2002), in arguing that the activity sheet does not add anything significant to the record and is consequently immaterial, even if the evidence is unique. However, the activity sheet adds to the record in a distinct and significant way, so Walker and Perez do not compel us to find it immaterial. In Walker, defendants sought a new trial based on the state’s suppression of information, unrelated to the trial itself, about an informant witness. The informant, who testified at defendant’s trial, was found with cocaine and marijuana in his pocket on the day of a controlled buy operation in an unrelated case. We held that suppression of that information did not rise tó the level of a Brady violation. 657 F.3d at 188 (noting that another witness for the prosecution provided direct support). Unlike our case, where Howard’s statement to the Pughs directly undercut the credibility of her eyewitness testimony in Dennis’s case, the alleged Brady evidence in Walker was wholly unrelated to defendant’s case. Further, we reiterated the principle in Walker that “there are some instances where specific impeachment evidence is so important (for issues such as the identity of the culprit) that it is material for Brady purposes even when a witness has already been effectively impeached on other issues.” Id. (emphasis *301added). Thus, Walker supports the view that withholding impeachment material that is germane to a critical aspect of the case — as here, the identity of the perpetrator — violates Brady.

Similarly, Perez does not support the Commonwealth’s contention. The alleged Brady material in Perez was a witness’s later statement inculpating another defendant and exculpating Perez. The initial statement, unlike Howard’s initial statement in this case, was corroborated by documentary evidence and co-defendant testimony at trial. Here, Howard’s eyewitness testimony played a pivotal emotional and practical role that could not be replaced by other evidence. There are material differences in impeachment value as well. In Perez, we concluded that cross-examination on the basis of the later statement would not have induced the co-defendant to admit to committing the crime. Perez, 280 F.3d at 350-51. Here, the type of statement at issue is different — Howard would have been confronted with an inconsistent statement, but not one that would have implicated her in the crime.

Armed with the activity sheet, defense counsel could have impeached Howard in a manner that very well may have led her to admit she recognized the perpetrators from her high school. Regardless of whether she actually recognized the shooter, Howard’s credibility would have been placed counter to that of the victim’s aunt and uncle, the Pughs, who would have undoubtedly been called at trial. Consequently, Howard’s impeachment could have changed the jury’s perception of her credibility.

There are significant, material differences between the type of cross-examination defense counsel engaged in and what he could have done had he known of the police activity sheet. As the District Court noted, “the activity sheet would have shown that [Howard] either lied to Williams’[s] close relatives — only days after the murder and in a manner that implicated Diane Pugh’s own nephew — or she was lying at trial.” Dennis V, 966 F.Supp.2d at 515. Thus, the government’s suppression necessarily undermines confidence in the outcome of Dennis’s trial. Discrediting the prosecution’s central witness, and the eyewitness with the most significant exposure to the shooter, would have had devastating effects on the prosecution’s case at trial. The remaining two eyewitnesses were located farther away from the incident, had only brief glimpses of the perpetrators, or were admittedly paying little attention. Challenging Howard’s identification of the shooter did little to undermine her credibility as a witness; but armed with the inconsistent statement, defense counsel could have undercut Howard’s testimony sufficiently that a jury may not have convicted Dennis. There is a reasonable probability that had the activity sheet been disclosed, the result of the proceeding would have been different.

The Commonwealth argues that Howard did not make the statements attributed to, her in the activity sheet. In support of this assertion, the Commonwealth looks to Howard’s and the Pugh’s testimony during PCRA review — over sixteen years after Dennis’s trial. Her statements during PCRA review carry little weight in how we consider a jury’s credibility determination at trial. In Kyles, the Supreme Court explicitly rejected the contention that post-conviction credibility determinations could replicate the jury’s credibility determinations at trial. Kyles, 514 U.S. at 449 n.19, 115 S.Ct. 1555 (“[N]either observation [during post-conviction proceedings] could possibly have affected the jury’s appraisal of Burns’s credibility at the time of Kyles’s trials.”). The court oriented its analysis around how the jury would have weighed the information, not the credibility of the post-conviction testimony itself. Thus, the *302proper inquiry remains whether use of the activity sheet by defense counsel at trial would have resulted in a different outcome at trial. The jury makes the credibility determination, not the Court sixteen years post-trial.

Although the Supreme Court instructed habeas courts in Wood not to ignore testimony at evidentiary hearings that would undermine the potential usefulness of alleged Brady material, the admissions during a post-conviction hearing in Wood differed significantly from those provided by Howard during PCRA review. In Wood, counsel specifically admitted that “disclosure [of the polygraph results] would not have affected the scope of his cross-examination,” and consequently, he did not bother to obtain admissions during post-conviction review. Wood, 516 U.S. at 7-8, 116 S.Ct. 7. The post-conviction testimony at issue here is markedly different. Dennis’s trial counsel testified that discrediting Howard through inconsistent statements was an integral part of the trial strategy. Interpreting Howard’s statements during PCRA hearings as indicating that she did not, in fact, make the statements to the Pughs contained in the activity sheet would allow the Commonwealth to cure its suppression of material evidence through delay. This we will not do.

The Commonwealth’s argument that the information contained in the activity sheet was double hearsay, so not admissible for impeachment purposes, fairs no better. The Pennsylvania Supreme Court did not rest its decision on an admissibility determination. Rather, it rooted its analysis in a misapplication of the Kyles materiality standard: that “any additional impeachment based on the activity sheet would have created a reasonable probability that the result of the proceeding would have been different.” Dennis IV, 17 A.3d at 309.

Counsel could also have used the information to challenge the adequacy of the police investigation. Defense counsel could have questioned Detectives Jastrzembski and Santiago as to why they did not ask Howard questions about her inconsistent statement when they saw her again only a few hours after indicating that confronting her was part of their “things to do.” Their subsequent meeting with Howard centered on reviewing a photo array. The detectives never asked Howard about admitting to the Pughs that she recognized the assailants from Olney High School. They never asked Howard about Kim and Quinton, despite having recently left a discussion with Parker, who stated that Howard mentioned Kim and Quinton to her as well. There is also no indication that they conducted any further investigation into the Pughs and whether they misheard all of these details or had reason to fabricate Howard’s inconsistent statement. Armed with the statement, defense counsel could have highlighted the investigatory failures for the jury, which could have supported Dennis’s acquittal.

Further, defense counsel could have used the Howard inconsistent statement to mount an “other suspect” defense at trial. According to the Pughs, Howard stated that she recognized the shooter from Ol-ney High School where she and Williams were enrolled. Dennis attended Roxbor-ough High School for his entire high school career. The simple conflict between where Dennis attended school and where Howard stated the assailants went to school would have removed Dennis as a suspect and empowered defense counsel to put forth an “other suspect” defense at trial, which he was otherwise unable to do. Together with the failure to follow up on the statements to the Pughs, defense counsel could have urged that Dennis’s was a case where police arbitrarily put blinders on as to the possibility that someone else committed the crime and pursued the easy lead.

*303Although the Pennsylvania Supreme Court acknowledged that “the omission is to be evaluated in the context of the entire record,” Dennis IV, 17 A.3d at 309, it ultimately applied the Brady materiality standard unreasonably by using sufficiency of the evidence as a touchstone. As pointed out by the District Court, the Supreme Court instructed in Kyles that “[a] defendant need not demonstrate that after discounting the inculpatory evidence in light of the undisclosed evidence, there would not have been enough left to convict.” 514 U.S. at 434-35, 115 S.Ct. 1555. Rather, “the Kyles Court rebuked the dissent for assuming that Kyles must lose on his Brady claim because there would still -have been enough to convict, even if the favorable evidence had been disclosed. ‘The rule is clear, and none of the Brady cases has ever suggested that sufficiency of the evidence (or insufficiency) is the touchstone.’ ” Dennis V, 966 F.Supp.2d at 516 (quoting Kyles, 514 U.S. at 435 n.8, 115 S.Ct. 1555). State courts may not “emphasize[ ] reasons a juror might disregard new evidence while ignoring reasons she might not.” Wearry v. Cain, — U.S.-, 136 S.Ct. 1002, 1007, 194 L.Ed.2d 78 (2016).

The Pennsylvania Supreme Court concluded that “[t]he disclosure of the activity sheet would have had no impact upon [two additional] eyewitnesses’ testimony” and consequently, the activity sheet was not material under Brady. Dennis IV, 17 A.3d at 309. In making its conclusion as to the materiality of the activity sheet, the Pennsylvania Supreme Court tied the materiality of the activity sheet to a requirement that Dennis show that Cameron’s and Bertha’s eyewitness testimony would not be sufficient to support the jury’s finding. This analysis is entirely inconsistent with the Court’s instructions on materiality. The Commonwealth argues, and the Dissent appears to accept, that by citing Commonwealth v. Weiss, 604 Pa. 573, 986 A.2d 808 (2009) — which reiterated the Supreme Court’s admonition of the sufficiency of the evidence test — the Pennsylvania Supreme Court applied the proper standard. However, unreasonable application of federal law under AEDPA occurs when the state court identifies the proper principle, but “unreasonably applies that principle to the facts of the prisoner’s case.” Williams, 529 U.S. at 413, 120 S.Ct. 1495. Indeed, in Lafler, the state court had identified the two Strickland prongs — prejudice and performance — yet the United States Supreme Court concluded that the state court had unreasonably used the “knowing and voluntary” standard and disregarded Strickland. 132 S.Ct. at 1390.

Here, the Commonwealth’s argument that the Pennsylvania Supreme Court knew the proper standard for materiality does little to demonstrate that it actually applied it reasonably. Instead of engaging in a holistic materiality inquiry per Kyles, the Pennsylvania Supreme Court proceeded down an analytical path that hinged the activity sheet’s Brady materiality on the sufficiency of the evidence, namely, the strength of Bertha and Cameron’s eyewitness testimony, in direct contravention of how the Supreme Court has defined materiality.

Judge Fisher’s Dissent relies on the Supreme Court’s decision in Strickler to support the Pennsylvania Supreme Court’s approach to materiality in Dennis IV. Like the activity sheet, the exculpatory materials at issue in Strickler would have cast doubt on the testimony of a key prosecution, Anne Stoltzfus. In Strickler, the Court of Appeals for the Fourth Circuit below had identified the Kyles standard for materiality and had concluded that “without considering Stoltzfus’ testimony, the record contained ample, independent evidence of guilt, as well as evidence sufficient to support the findings of vileness and future dangerousness that warranted *304the imposition of the death penalty.” Strickler, 527 U.S. at 290, 119 S.Ct. 1936. The United States Supreme Court soundly rejected the Fourth Circuit’s approach upon review in Strickler. It instructed that “[t]he standard used by [the Fourth Circuit] was incorrect” and reiterated that “the materiality inquiry is not just a matter of determining whether, after discounting the inculpatory evidence in light of the undisclosed evidence, the remaining evidence is sufficient to support the jury’s conclusions.” Id. (“[T]he question is whether ‘the favorable evidence could reasonably be taken to put the whole case in such a different light as to undermine confidence in the verdict.’ ” (quoting Kyles, 514 U.S. at 435, 115 S.Ct. 1555)). The Pennsylvania Supreme Court did precisely what the Strickler Court rejected — it evaluated whether, after considering Howard’s testimony, the remaining eyewitness testimony was sufficient for Dennis’s conviction. Dennis IV, 17 A.3d at 309 (“[T]here were two eyewitnesses other than Howard who observed the shooting at close range. ... The disclosure of the activity sheet would have had no impact upon these eyewitnesses’ testimony.”).

Further, the materiality of the impeachment evidence in Strickler is distinguishable from the police activity sheet at issue here because the evidence against petitioner in Strickler was far more extensive and varied than the Commonwealth’s case against Dennis. As Judge Fisher recognizes, there was “considerable forensic and other physical evidence” linking the petitioner to the crime in Strickler. 527 U.S. at 293, 119 S.Ct. 1936. The Supreme Court ultimately concluded that “[t]he record provide[d] strong support for the conclusion that petitioner would have been convicted of capital murder and sentenced to death, even if Stoltzfus had been severely impeached.” Id. at 294, 119 S.Ct. 1936. Thus, the Strickler Court held that petitioner had not shown materiality under Brady.

The record laid by the Commonwealth in Dennis’s case pales in comparison to the one mounted by the government in Strick-ler. For instance, the police in Strickler recovered hairs from clothing found with the victim that were microscopically akin to petitioner’s, and petitioner’s fingerprints were found on the inside and outside of the victim’s car. 527 U.S. at 293 n.41, 119 S.Ct. 1936. No similar physical evidence exists on the record in Dennis’s case. The Supreme Court recognized the importance of Stoltzfus’s testimony, as it was the only disinterested narrative account provided at trial, but ultimately concluded in its holistic materiality inquiry that petitioner failed to show that there was “a reasonable probability that his conviction or sentence would have been different had these materials been disclosed.” Id. at 296, 119 S.Ct. 1936. The conclusion that petitioner failed to show materiality against the variety and extensiveness of the evidence against petitioner in Strickler differs from the Pennsylvania Supreme Court indication that two other eyewitness accounts were sufficient for a jury to convict Dennis.

In sum, the Pennsylvania Supreme Court unreasonably applied Brady and its progeny in denying Dennis’s Brady claim based on the Howard inconsistent statement. It unreasonably disregarded the impeachment value of the evidence in discrediting the Commonwealth’s key eyewitness and the adequacy of the investigation. It unreasonably applied a sufficiency of the evidence test by tying the materiality of the activity sheet to the sufficiency of the remaining inculpatory eyewitness testimony. And finally, the Pennsylvania Supreme Court failed to consider that the activity sheet would have enabled defense counsel to raise a defense he was otherwise unable to present — that a student at *305Olney High School committed the murder. There is a reasonable probability that, had the activity sheet been disclosed, the jury would have had a reasonable doubt as to Dennis’s guilt.

C. Frazier Documents

1. Facts

Prior to Dennis’ arrest, Philadelphia police received a lead .from Montgomery County Detectives that someone other than Dennis may have murdered Williams. William Frazier, an initiate at the Montgomery County Correctional Facility called police and told them that Tony Brown “shot ... [a] female in the middle of the street near the Fern Rock station” after the girl resisted his efforts to take her earrings, which Brown sold at a pawn shop for $400. App. 1689-90.

Frazier heard Brown’s confession during a three way call facilitated by his aunt, Angela Frazier. Frazier recounted the conversation in a signed statement given to Philadelphia Police less than two weeks after Williams’s murder. Brown admitted that he — along with Frazier’s cousin, Ricky Walker, and a man called “Skeet”— had “fucked up” and killed Chedell Williams. App 1692. Frazier told police that Brown knew Williams, and identified her as “Kev with the blue pathfinder ... his girl.” App. 1694.

During the call, Brown asked Frazier if he heard about “the incident on the news about the girl that [was] killed over a pair of earrings,” and Brown confessed “that was us.” App. 1692. Frazier reported “[Tony] said that he and Ricky got out of the car and Skeet was driving. They approached the girl, Tony pulled his gun out and told her to give up the earrings she refused. So he put-the gun to her neck ... [and] it accidentally went off.” Id. Walker briefly joined the call and reported that they were scared, and that they left Frazier’s apartment, where they sought cover after the murder, in the middle of the night so that no one would see them. Frazier reported that Brown and Walker sounded “extremely nervous and upset.” App. 1694. Frazier described Tony as 5'7", two inches taller than Dennis, with light brown skin. Like the assailant, Tony “like[d] to wear sweat suits;” he had also committed robberies in the past and owned “a collection of guns.” App. 1693-95.

Frazier gave 'detectives addresses for Brown and Walker, the address where Skeet used to live. Frazier also gave police Angela Frazier’s address and phone number, Brown’s mother’s address, and an address of the pawn shop, along with a description of the proprietor. Frazier agreed to go on a ride along to show detectives the addresses he reported. The Philadelphia police, including Jastrzembski, spoke with Frazier’s landlord, who confirmed that Frazier rented the apartment located at the address he provided. Although the landlord reported that nobody had been in the apartment since Frazier’s arrest, the men used unconventional means to enter Frazier’s apartment the night of the murder — they climbed through Frazier’s right window.26

Detectives interviewed Walker, who told them that he “c[ouldn’t] stand” his cousin, *306Frazier. App. 1703. Walker denied knowing Tony Brown and Skeet and denied any involvement with, or knowledge about, Williams’s murder. He told detectives that he was at his house with his mother on the day of the murder. Police did not conduct an investigation into Walker’s alibi or alert defense counsel to any of the information on Frazier’s tip.

2. State court decision

The Pennsylvania Supreme Court affirmed the PCRA court’s denial of Dennis’s Brady claim as to the Frazier documents on the grounds that Dennis failed to demonstrate that the documents were material and admissible. The Pennsylvania Supreme Court relied on its decision in Commonwealth v. Lambert, 584 Pa. 461, 884 A.2d 848 (2005), in which it emphasized that the prosecution need not “disclose to the defense every fruitless lead followed by investigators of a crime” and asserted that “inadmissible evidence cannot be the basis for a Brady violation.” Lambert, 884 A.2d at 857 (citation omitted). The court concluded: “In the absence of any argument regarding the gravamen of Lambert ... [Dennis] has failed to establish a basis for relief’ regarding the Frazier documents. Dennis III, 950 A.2d at 968. However, as Dennis points out, the Pennsylvania Supreme Court retreated from its decision in Lambert in a later opinion so as to comport with Supreme Court precedent regarding the need for admissibility. Commonwealth v. Willis, 616 Pa. 48, 46 A.3d 648, 670 (2012) (“[W]e hold that admissibility at trial is not a prerequisite to a determination of materiality under Brady.... Therefore, nondisclosed favorable evidence which is not admissible at trial may nonetheless be considered material for Brady purposes[.]”).

3. AEDPA Review

The state court addressed the merits of the Frazier claim and, as a result, Dennis may obtain habeas relief only if he can demonstrate that the decision was an unreasonable application of, or contrary to, clearly established law, or an unreasonable determination of the facts. 28 U.S.C. § 2254(d). It is undisputed that the first two elements of Brady are met. The Frazier documents indicated that someone other than Dennis committed the crime, and were thus exculpatory, and there is no question that the state did not disclose the documents until PCRA discovery. However, the Pennsylvania Supreme Court unreasonably applied Brady and its progeny in concluding that the Frazier documents were immaterial. Also, in appending an admissibility requirement onto Brady, the Pennsylvania Supreme Court acted contrary to clearly established law, as defined by the United States Supreme Court.

The Pennsylvania Supreme Court’s justification that the Frazier documents were a “fruitless lead” was unreasonable. There is no requirement that leads be fruitful to trigger disclosure under Brady, and it cannot be that if the Commonwealth fails to pursue a lead, or deems it fruitless, that it is absolved of its responsibility to turn over to defense counsel Brady material. The rationale behind Brady itself rests on the principle that prosecutors bear an obligation to structure a fair trial for defendants:

Society wins not only when the guilty are convicted but when criminal trials are fair; our system of the administration of justice suffers when any accused is treated unfairly. ... A prosecution that withholds evidence ... which, if made available, would tend to exculpate him or reduce the penalty helps shape a trial that bears heavily on the defendant. That casts the prosecutor in the role of an architect of a proceeding that does not comport with standards of justice, *307even [if] ... his action is not the result of guile[.]

Brady, 373 U.S. at 87-88, 83 S.Ct. 1194 (internal quotation marks and footnote omitted). Structuring a fair trial for defendants demands that prosecutors freely disclose material that is helpful to the defense. Consequently, making Brady disclosure depend on a prosecutor’s own assessment of evidentiary value, as opposed to the benefit to defense counsel, is anathema to the goals of fairness and justice motivating Brady.

The lead was not fruitless, it was simply not rigorously pursued. Detectives did not interview Angela Frazier, who facilitated the three-way call and was on the phone when Brown confessed to the murder. Detectives did not question Walker again— who admitted to having a bias against Frazier — after he stated that he did not know any Brown or Skeet, nor did they attempt to verify Walker’s alibi on the day of the murder. Detectives did not investigate the owner of the pawn shop where Brown purportedly sold Williams’s earrings. Detectives did not obtain the photos of Brown, Skeet, and Walker that were in Frazier’s apartment. Detectives went to an incorrect address seeking information about Skeet and spoke with a woman named Janice Edelen, who said she did not know any man called Skeet. Finally, detectives did not visit the addresses Frazier provided until ten years after the murder. Armed with the Frazier documents, Dennis’s counsel would have been prepared to pursue the lead himself or at least informed the jury of the police’s misguided focus on Dennis and failure to pursue the lead.

The Pennsylvania Supreme Court grafted an admissibility requirement onto the traditional three-prong Brady inquiry when it rejected Dennis’s Brady claim as to the Frazier documents on the ground that he failed to affirmatively show that the documents were admissible. The Pennsylvania Supreme Court’s characterization of admissibility as dispositive under Brady was an unreasonable application of, and contrary to, clearly established law as defined by the United States Supreme Court.

The Commonwealth articulates the Pennsylvania Supreme Court’s decision somewhat differently. It argues that our role on habeas review is determining “whether, under Supreme Court precedent, it was objectively unreasonable for the Pennsylvania Supreme Court to reject Dennis’s claim that he only had to argue or allege that disclosure ‘might’ have affected his investigation or preparation for trial.” Appellants Br. 74. This framing incorrectly states what the Pennsylvania Supreme Court did in Dennis III. It did not simply discount Dennis’s argument that defense counsel could have prepared differently had the documents been disclosed — it appended an admissibility requirement to Brady in contravention of clearly established law.

The Pennsylvania Supreme- Court cited Wood v. Bartholomew, 516 U.S. 1, 116 S.Ct. 7, 133 L.Ed.2d 1 (1995), as attaching an admissibility requirement to Brady. The United States Supreme Court’s holding in Wood compels the opposite conclusion, however. The Supreme Court held in Wood that there was no Brady violation when the prosecution did not disclose the results of two polygraph examinations that were inadmissible at trial. Wood, 516 U.S. at 6, 116 S.Ct. 7. The Wood Court noted that Brady governs “evidence,” and that the polygraph results, since they were inadmissible under state.law, were “not ‘evidence’ at all.” Id. at 5-6, 116 S.Ct. 7. However, under Washington law, poly-graphic examinations cannot be admissible for any purpose at trial, even for impeachment purposes. Id. at 5, 116 S.Ct. 7. At most, the Court’s holding in Wood could *308support the proposition that evidence that cannot be used in any manner at trial under state law may be immaterial under Brady. The holding does not reach so far as to allow state courts to attach a general admissibility requirement onto the Brady inquiry as the Pennsylvania Supreme Court did in Dennis III.

Further, the Wood Court analyzed the effect of suppressing the polygraph results, despite their uncontroverted inadmissibility. After acknowledging their inadmissibility, the Wood Court proceeded to examine whether, if disclosed, the results would have led to the discovery of evidence that would have influenced the course of trial, including pre-trial preparations. See Wood, 516 U.S. at 7, 116 S.Ct. 7 (considering whether trial counsel would have prepared differently given the results, though ultimately concluding that disclosure would not have resulted in a different outcome). The Supreme Court’s decision to continue its inquiry in light of wholly inadmissible alleged Brady material is telling. As the District Court aptly observed, “[i]f inadmissible evidence could never form the basis of a Brady claim, the Court’s examination of the issue would have ended when it noted that the test results were inadmissible.” Dennis V, 966 F.Supp.2d at 503.

The Supreme Court’s choice in Wood to consider the way in which suppression of the polygraph results affected preparation and trial aligns with the way in which materiality is discussed in Kyles. Kyles makes clear that evidence is material under Brady when the defense could have used it to “attack the reliability of the investigation.” 514 U.S. at 446, 115 S.Ct. 1555. As noted by the District Court, in Kyles, defense counsel could have used the information at issue “to throw the reliability of the investigation into doubt and to sully the credibility” of the lead detective. Id. at 447, 115 S.Ct. 1555. The proper inquiry for the Pennsylvania Supreme Court was to consider whether disclosure of the Frazier documents would have impacted the course of trial, which includes investigative activities. Here, disclosure of the Frazier documents would have empowered defense counsel to pursue strategies and preparations he was otherwise unequipped to pursue.

Imposition of an admissibility requirement does not comport with the United States Supreme Court’s longstanding recognition that impeachment evidence may be favorable and material, and if so, is unquestionably subject to Brady disclosure. The Court stated definitively in Strickler that “[o]ur cases make clear that Brady’s, disclosure requirements extend to materials that, whatever their other characteristics, may be used to impeach a witness.” 527 U.S. at 282 n.21, 119 S.Ct. 1936 (emphasis added). As to both the first Brady prong, favorability, and the third Brady prong, materiality, the Supreme Court has held that impeachment evidence falls under Brady’s purview. Id. at 281-82, 119 S.Ct. 1936 (the evidence “must be favorable to the accused, either because it is exculpatory, or because it is impeaching.”); Kyles, 514 U.S. at 445, 115 S.Ct. 1555 (concluding that evidence was material because “the effective impeachment of one eyewitness can call for a new trial even though the attack does not extend directly to others”). Further, nearly all of the cases decided by the United States Supreme Court since Brady have dealt with impeachment evidence. See Wearry v. Cain, — U.S.-, 136 S.Ct. 1002, 194 L.Ed.2d 78 (2016) (per curiam), Wetzel v. Lambert, -U.S.-, 132 S.Ct. 1195, 182 L.Ed.2d 35 (2012); Smith v. Cain, — U.S. -, 132 S.Ct. 627, 181 L.Ed.2d 571 (2012); Cone v. Bell, 556 U.S. 449, 129 S.Ct. 1769, 173 L.Ed.2d 701 (2009); Banks v. Dretke, 540 U.S. 668, 124 S.Ct. 1256, 157 L.Ed.2d 1166 (2004); Strickler v. Greene, 527 U.S. 263, 119 S.Ct. 1936, 144 L.Ed.2d 286 *309(1999); United States v. Bagley, 473 U.S. 667, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985); United States v. Agurs, 427 U.S. 97, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976); Moore v. Illinois, 408 U.S. 786, 92 S.Ct. 2562, 33 L.Ed.2d 706 (1972); Giglio v. United States, 405 U.S. 150, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972); Giles v. Maryland, 386 U.S. 66, 87 S.Ct. 793, 17 L.Ed.2d 737 (1967). It would be difficult to find stronger support for the proposition that admissibility is not a requirement under Brady, and the Supreme Court’s repeated consideration of impeachment material in Brady cases — without any reservation whatsoever — compels us to conclude that it is unreasonable to graft an admissibility requirement onto Brady’s traditional three-pronged inquiry.

Beyond the recognition that impeachment evidence is covered by Brady, the essence of the United States Supreme Court’s Brady jurisprudence focuses on the benefits of disclosure to the defense, not admissibility. This is evidenced by the definition of materiality itself. Kyles provides that evidence is material “if there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” 514 U.S. at 433-34, 115 S.Ct. 1555 (1995) (quoting Bagley, 473 U.S. at 682, 105 S.Ct. 3375 (opinion of Blackmun, J.)) (emphasis added). Quite simply, under Brady, the focus of the inquiry is on whether the information had “been disclosed to the defense,” not whether it was admissible at trial. See id. An admissibility requirement improperly shifts that focus.

The United States Supreme Court’s focus on disclosure is mirrored in the way in which it has applied the “reasonable probability” standard used to assess materiality under Brady. When the Court has reviewed applications of the “reasonable probability” standard, it has weighed the strength of the suppressed evidence against the strength of disclosed evidence to evaluate its. impact, not critiqued the character of the evidence itself. See Strickler, 527 U.S. at 290-94, 119 S.Ct. 1936. In Strickler, the Court denied a Brady claim on materiality grounds because “the record provides strong support for the conclusion that petitioner would have been convicted of capital murder and sentenced to death, even if [an eyewitness] had been severely impeached.” Id. at 294, 119 S.Ct. 1936. Thus, the focus was on disclosure, given the effect of other available material, not the character of the material itself.

The Supreme Court’s later decision in Cone v. Bell similarly affirmed its longstanding focus on disclosure regardless of admissibility at trial. There, the Court considered impeachment evidence including police bulletins, statements contained in official reports, and FBI reports to be Brady material. Cone, 556 U.S. at 470-71, 129 S.Ct. 1769. Neither the Sixth Circuit nor the District Court below fully considered whether the suppressed documents would have persuaded the jury to impose a lesser sentence. Id. at 475, 129 S.Ct. 1769 (“It is possible that the suppressed evidence, viewed cumulatively, may have persuaded the jury that Cone had a far more serious drug problem than the prosecution was prepared to acknowledge, and that Cone’s drug use played a mitigating, though not exculpating, role in the crimes he committed.”). Cone held that the courts below had failed to “thoroughly review the suppressed evidence or consider what its cumulative effect on the jury would have been” regarding Cone’s sentence. Id. at 472,129 S.Ct. 1769. By remanding the case for full consideration of the Brady claim despite the fact that the suppressed evidence was not necessarily admissible, the Court indicated that the admissibility of suppressed evidence ought not to change the materiality inquiry itself, which is understood as “a reasonable probability that, had the evidence been disclosed, the result *310of the proceeding would have been different.” Id. at 470, 476, 129 S.Ct. 1769.

Our recent decision in Johnson v. Folino, 705 F.3d 117 (3d Cir. 2013) further affirms the view that inadmissible evidence is often very material:

[Inadmissible evidence may be material if it could have led to the discovery of admissible evidence. Furthermore ... we think that inadmissible evidence may be material if it could have been used effectively to impeach or corral witnesses during cross-examination. Thus, the admissibility of the evidence itself is not dispositive for Brody purposes. Rather, the inquiry is whether the undisclosed evidence is admissible itself or could have led to the discovery of admissible evidence that could have made a difference in the outcome of the trial sufficient to establish a “reasonable probability” of a different result.

Id. at 130 (citations omitted). Here, however, the Pennsylvania Supreme Court ignored how the United States Supreme Court has evaluated materiality and instead made inadmissibility a determinative factor, indeed, the determinative factor.

The Pennsylvania Supreme Court’s characterization of admissibility as a separate, independent prong of Brady effectively added admissibility as a requirement. This runs afoul of Supreme Court precedent. The Pennsylvania Supreme Court required “evidence sought under Brady be material and admissible.” Dennis III, 950 A.2d at 968 (emphasis added). The Supreme Court has never added a fourth “admissibility” prong to Brady analysis. Like the imposition of a due diligence prong, adding an admissibility prong would alter Brady’s, traditional three-prong inquiry in a manner that the Supreme Court rejected in Williams. See Williams, 529 U.S. at 393, 120 S.Ct. 1495.

Most federal courts have concluded that suppressed evidence may be material for Brady purposes even where it is not admissible. See United States v. Morales, 746 F.3d 310, 314 (7th Cir. 2014) (listing cases). However, the Seventh and Fourth Circuits have indicated that inadmissible evidence cannot be material. Morales, 746 F.3d at 314; see also Jardine v. Dittmann, 658 F.3d 772, 777 (7th Cir. 2011) (“Logically, inadmissible evidence is immaterial under [the Brady] rule”); Hoke v. Netherland, 92 F.3d 1350, 1356 n.3 (4th Cir. 1996). Jar-dine and Hoke involved evidence that was prohibited from being used under state evidence laws and their assertions regarding an admissibility requirement were not determinative to their holdings. Jardine, 658 F.3d at 777 (noting that the undisclosed material was inadmissible under state law and could not be used to impeach, but concluding that no Brady violation occurred only after evaluating other avenues through with the material could be used); Hoke, 92 F.3d at 1355-56 (holding that the undisclosed information about the murder victim’s sexual history would not have been material in light of overwhelming physical and other evidence and resolving the case on grounds other than admissibility). Morales is similarly unpersuasive, as it observed that the Courts of Appeals for the First, Second, Third, Sixth, and Eleventh Circuits have read Brady to include material but inadmissible evidence. 746 F.3d at 314. The Morales court even conceded that “[w]e find the Court’s methodology in Wood to be more consistent with the majority view in the courts of appeals than with a rule that restricts Brady to formally admissible evidence.” Id. at 315.27

*311The Frazier documents were material under Brady. Dennis’s counsel could have used the information contained in the Frazier documents to challenge detectives at trial regarding their paltry investigation of the lead. As we previously noted, the lead was “fruitless” because the Commonwealth failed to take sufficient action to determine if it was fruitful — the Commonwealth essentially abandoned it. The Commonwealth does not dispute that trial counsel could have used the information in the suppressed documents to question the detectives.

Further, had the Commonwealth not suppressed the Frazier documents, Dennis could have presented an “other person” defense at trial, which he was otherwise not able to do. The Frazier documents bring to light that Walker admitted to going to Olney High School — -the school Williams and Howard attended — and he recognized Williams from school. Thus, the documents not only support an alternative shooter theory, but the very same alternative shooter theory that defense counsel could have been prepared to raise had the Howard activity sheet also been disclosed. Alterations in defense preparation and cross-examination at trial are precisely the types of qualities that make evidence material under Brady. Consequently, it was, unreasonable for the Pennsylvania Supreme Court to conclude that the Frazier documents were not material. There is a reasonable probability that had the jury heard an “other person” defense, the result of the proceeding would have been different.

The Pennsylvania Supreme Court unreasonably applied federal law and applied law in a manner contrary to Supreme Court precedent. The Commonwealth’s suppression of the Frazier documents violated Brady as they were favorable to the defense, and could have been used by defense counsel as exculpatory and impeachment evidence. Dennis is entitled to a new trial.

D. Cumulative Materiality

While the suppression of the Ca-son receipt, the Howard police activity sheet, and the Frazier documents support ordering a new trial, the cumulative effect of their suppression commands it. Had the Brady material been disclosed, there is a reasonable probability that the outcome of the trial would have been different, and its *312suppression undermines confidence in the verdict.

The District Court engaged in a cumulative materiality analysis in addition to granting each individual Brady claim. Dennis V, 966 F.Supp.2d at 517-18. This analysis was proper. When the issue ripened in Dennis IV and the Pennsylvania Supreme Court could have assessed the cumulative prejudice of withholding the Cason receipt, Frazier documents, and police activity sheet containing Howard’s statements, it declined to do so explicitly. We are required to presume that the state court considered and rejected Dennis’s cumulative materiality argument. Johnson v. Williams, — U.S. -, 133 S.Ct. 1088, 1097, 185 L.Ed.2d 105 (2013). Just as the Pennsylvania Supreme Court’s rejections of Dennis’s Brady claims constituted unreasonable application of federal law, its rejection of the cumulative materiality of the suppressed evidence, though not done explicitly, was an unreasonable application of Brady and its progeny.

The Supreme Court in Kyles instructed that the materiality of withheld evidence must be “considered collectively, not item by item.” 514 U.S. at 436, 115 S.Ct. 1555. The importance of cumulative prejudice cannot be overstated, as it stems from the inherent power held by the prosecution, which motivated Brady. Id. at 437, 115 S.Ct. 1555 (“[T]he prosecution ... alone can know what is undisclosed[ ] [and] must be assigned the consequent responsibility to gauge the likely net effect of all such evidence and make disclosure when the point of ‘reasonable probability1 is reached.”). The Supreme Court recently reiterated that state courts are required to evaluate the materiality of suppressed evidence cumulatively. Wearry, 136 S.Ct. at 1007 (“[T]he state postconviction court improperly evaluated the materiality of each piece of evidence in isolation rather than cumulatively.”)

As acknowledged by the District Court, the cumulative impeachment value of the suppressed evidence would have undermined the Commonwealth’s case. The Ca-son receipt would have impeached the Commonwealth’s primary response to Dennis’s alibi by providing documentary proof that Cason testified falsely and would have transformed her into a witness for the defense. The inconsistent statement contained in the police activity sheet would have impeached Howard’s credibility, undoubtedly the Commonwealth’s most important eyewitness. Her impeachment by the Pugh statement would challenge her credibility, not simply the reliability of her identification during the photo array and lineup, which was what defense counsel was limited to at trial. Discrediting Cason and Howard may very well have raised sufficient doubt among the jury to acquit Dennis. Moreover, the Frazier documents could have supported the existence of another suspect who attended Howard’s high school, and the significance of this becomes even more pronounced when considered with Howard’s statements to the Pughs that the suspect attended her high school.

Together, the suppressed documents provided ample material to challenge the Commonwealth’s investigation following the murder. As the District Court stated:

Defense would have had a strong case to make that the Commonwealth abandoned promising leads: Police failed to meet with Frazier’s aunt, to verify Walker’s alibi, or to include Walker and Brown in photo arrays or line-ups; police also failed to follow up with Howard about the statement she allegedly made to the Pughs, to take a formal statement from the Pughs, or to interview Quinton. The Commonwealth allowed Cason to testify incorrectly that she worked until 2 p.m., and failed to investigate Den-nises] alibi given the actual timing of *313Cason’s activities. Discrediting the investigation is a crucial corollary to presenting an innocence/alibi defense: If the defense could lead the jury to believe that the Commonwealth conducted a shoddy investigation, the jury would have been more likely to listen to and believe Dennis’fs] alibi.

Dennis V, 966 F.Supp.2d at 518. The withholding of the Brady material would have given defense counsel unique ability to discredit the Commonwealth’s primary witnesses, bolster his alibi defense using objective documentary support from a disinterested party, highlight the shoddiness of the Commonwealth’s investigation, and perhaps point to another perpetrator. The cumulative effect of the suppression of these documents requires habeas relief.

IV. Conclusion

For the foregoing reasons, we will affirm the judgment of the District Court and grant Dennis a conditional writ of habeas corpus. Petitioner shall be released unless the Commonwealth commences a new trial against him within ninety days after issuance of the mandate.

. The Fern Rock SEPTA station is located in North Philadelphia. The Abbottsford projects are located in Northwest Philadelphia.

. Detective Jastrzembski testified at trial that neither the alleged second individual nor the person in the car were ever arrested, although the case was ongoing.

. Chief Judge McKee's masterful concurrence summarizes with great detail the photo array, line up, and the bystanders' identifications. As Chief Judge McKee notes, a majority of the nine eyewitnesses who viewed the photo array were unable to identify Dennis. Anthony Overstreet was installing stone facing on a nearby garage with Bertha at the time of the incident. Overstreet told police that he recognized the shooter from around Broad and Olney Streets in North Philadelphia. Although Overstreet stated that Dennis looked like the shooter when he reviewed the photo array, he identified a different individual as the shooter during a later in-person lineup — not Dennis. George Ritchie, who was across the street from Bertha and Overstreet, was unable to identify anyone as the shooter among photos provided by the police, despite initially asserting that he would be able to identify the perpetrators again. The two fruit vendors Howard ran toward, David Leroy, a hot dog vendor near the station, and Clarence Verdell, a bystander on the SEPTA steps, did not identify Dennis from the photo array. None of these bystanders were called to testify at trial.

. The District Court reasoned that the eyewitnesses' memory may have been supplanted by photos from the array: "That some (but notably not all) of the witnesses went on to identify Mr. Dennis in a life [sic] lineup two months after providing only tentative photo array identification indicates that their memories of the photo array may have 'replaced' their memories of the actual event. Or, more simply, that Mr. Dennis was familiar to them because they had seen his photo previously, and had no prior exposure to the other members of the lineup." Dennis v. Wetzel, 966 F.Supp.2d 489, 492 n.4 (E.D. Pa. 2013) (internal quotation marks and citation omitted), vacated and remanded sub nom. Dennis v. Sec'y, Pa. Dep’t of Corr., 777 F.3d 642 (3d Cir. 2015), reh’g en banc granted, opinion vacated (May 6, 2015) ("Dennis V”).

Chief Judge McKee's concurrence expands on this concern. He observes that “[ajllowing a witness to view a suspect more than once during an investigation can have a powerful corrupting effect on that witness’ memory.” J. McKee Concurring Op. at 328. Research shows "that while fifteen percent of witnesses who mistakenly identify an innocent person during the first viewing of a lineup, that percentage jumps to thirty-seven percent if the witness previously viewed that innocent person's mug shot.” Id. Here, "[t]he witnesses who identified Dennis at trial'were given not two, but three, opportunities to view Dennis. These multiple views could help explain why initially tentative guesses became certain identifications by the time the witnesses took the stand.” Id. at 329

. Detectives Manuel Santiago and William Wynn testified at trial about the eyewitnesses' prior identifications. Detective Santiago supervised the activities at the crime scene on the day of the murder and compiled a photo array to show to Howard, Bertha, and Cameron, which included eight photographs with Dennis’s photo in the first position. Dennis looked different in the photograph than at the time of arrest. Santiago did not ask Howard why she could not be sure that it was the shooter. Detective Wynn, the lineup supervisor for the Philadelphia Police, conducted the in-person lineups for Howard, Bertha, and Cameron. Defense counsel placed Dennis as number three in the lineup. All participants dressed similarly and carried large numbers for identification.

. The Commonwealth's other witnesses did not testify as to Dennis’s connection to the murder. Rather, they spoke to the emergency response to the crime (Fireman Oakes), the scene of the crime (Sergeant Fetscher), Williams's body chart (Detective Brown), and the projectile removed from her body (Detective Reinhold). Williams’s ex-boyfriend recounted a prior incident where Williams had been robbed at gunpoint for the same earrings she wore on the day of the murder. Officer Jachimowicz, a firearms expert, testified as to the type of gun that was likely used in the murder, and although he acknowledged that there were thousands of models of .32 caliber handguns, he asserted with certainty that the nickel finish Harrington Richardson 733 was probably used in the murder. Detective Dominic Mangoni transported Howard and Bertha to the lineup. Detective Thomas Perks participated in Dennis’s arrest. Williams’s mother and father, Barbara and Barry, identified their daughter and testified to her future. Dr. Sekula-Perlman, a medical examiner, ruled Williams’s death a homicide by a shot at close range. Sergeant Fetscher took information from witnesses at the scene, including Howard, Bertha, and Cameron. None of these witnesses testified substantively as to Dennis's alleged involvement in the murder.

. Defense counsel sought to discredit eyewitness testimony put forth by the Commonwealth, primarily that of Zahra Howard. However, counsel's cross-examination was confined to highlighting Howard's prior hesitation in identifying Dennis. Similarly, defense counsel’s cross-examination of Cason focused on shakiness in her " recollection; counsel had nothing to indicate that her time-line was incorrect, or that she was mistaken or testifying falsely.

. Lawrence Merriweather also testified to seeing Dennis on the day in question. Merri-weather testified that he saw Dennis between 3:00 and 3:30 p.m.

. The Commonwealth responded with character witnesses that disputed the testimony of Dennis's character witnesses.

. Dennis testified that Helen Everett, his girlfriend, told him about the rumor that he, Derrick, and Rodney, committed the murder. He testified that Derrick and Rodney spoke with the police about the murder. Neither testified at trial.

.Anthony Sheridan, a SEPTA employee called by the Commonwealth, testified that there was a K bus that left the stop near Dennis's father's house at approximately 1:56 p.m. and that it would take approximately half an hour to arrive at Henry and Midvale.

. It is not clear how counsel would have been able to obtain Cason's receipt on appeal because DPW regulations placed strict limitations on the type of information it would disclose and to whom. See 55 Pa. Code § 105.4(a)(1). Presumably, counsel would have sought permission from Cason, or assistance from Cason herself, in obtaining the receipt.

. Williams, the victim, previously dated a man named Kevin Williams.

. The Pennsylvania Supreme Court's 2004 decision, Commonwealth v. Dennis, 580 Pa. 95, 859 A.2d 1270 (2004) (“Dennis II"), is not relevant to this appeal. On December 12, 2000, Dennis filed a motion for discovery, seeking the prosecutor’s jury selection notes, and the PCRA court granted Dennis’s motion. After granting the Commonwealth’s request for reconsideration of the order, the PCRA court reinstated the discovery order on July 10, 2001. In Dennis II, the Pennsylvania Supreme Court reversed the order granting discovery of the prosecutor’s jury selection notes and remanded the case for completion of PCRA review.

. The parenthetical language here is a direct quote from the parenthetical used by the District Court in its description of Banks. See Dennis V, 966 F.Supp.2d at 514-15.

. This framing of Kyles was taken from Lambert v. Beard, 537 Fed.Appx. 78, 86 (3d Cir. 2013).

. The Commonwealth argues on appeal that the Pennsylvania Supreme Court did not make a factual finding and that the statement that the police had the receipt was merely framing for the later substantive discussion. In Bobby v. Bies, 556 U.S. 825, 129 S.Ct. 2145, 173 L.Ed.2d 1173 (2009), cited in support by the Commonwealth, the Supreme Court held that a state court’s alleged factual finding could not support issue preclusion because there was no evidence that the alleged state court finding was supported by the record at trial or on appeal and further was not necessary to the judgments made by the state court. Bies bears no relation to our case where there is ample evidence in the record that the police took possession of the receipt, as attested by Cason herself.

. Dennis’s trial counsel asserted in an affidavit he "did not specifically request a copy of the welfare check receipt from the Commonwealth, because [he] did not know of its existence,” but he had “[b]y formal motion ... requested] all exculpatory evidence be produced.” App. 1725.

. The Tenth Circuit and the D.C. Circuit agree that defense counsel’s knowledge is not at issue in Brady. Banks v. Reynolds, 54 F.3d 1508, 1517 (10th Cir. 1995) ("[T]he prosecution's obligation to turn over the evidence in the first instance stands independent of the defendant’s knowledge.... The only relevant inquiry is whether the information was exculpatory.” (internal quotation marks omitted)); accord In re Sealed Case, 185 F.3d 887, 896-97 (D.C. Cir. 1999).

. Surprisingly, several courts of appeals have endorsed some form of a due diligence requirement. For a comprehensive overview of common features of the diligence rule and where it emerged, see Kate Weisburd, Prosecutors Hide, Defendants Seek: The Erosion of Brady Through the Defendant Due Diligence Rule, 60 UCLA L. Rev. 138, 141, 147-56 (2012). Common features include that the evidence was equally available to the prosecution and the defense, that the evidence was known by the defendant, and that the relevant facts were accessible by the defendant. Id. at 153-56.

. The Second Circuit also recently recognized in a habeas case that "[t]he [United States] Supreme Court has never required a defendant to exercise due diligence to obtain Brady material.” See Lewis v. Conn. Comm'r *292of Corr., 790 F.3d 109, 121 (2d Cir. 2015). It retained its test for when evidence is not “suppressed” for Brady purposes, however. Id.

. The Commonwealth concedes that if it had the receipt, Cason would have provided little value to the prosecution and they would not have called her. Indeed, Dennis probably would have.

. The Commonwealth argues that Cason’s testimony would be duplicative of Willis Meredith’s non-alibi testimony. Willis Meredith, a friend of Dennis’s, testified that he saw Dennis at Abbottsford Homes around 2:30 p.m. Cason’s testimony is not cumulative for two reasons: (1) Willis, like Dennis’s other witnesses, was a friend and open to accusations of bias from the prosecution; and (2) Cason’s testimony was corroborated by independent documentary evidence. So, even if her testimony simply placed Dennis at Abbottsford Homes around 2:30, it did so with more evi-dentiary weight than Meredith's.

. Howard's testimony undoubtedly bore more emotional weight with the jury than the other eyewitness testimony presented at trial due to Howard’s close friendship with the victim. Because of Howard's personal connection with, and physical proximity to, Williams at the time of her murder, stress may have played a particularly damaging role in the strength of her identification. Chief Judge McKee explains in his concurrence that that stress may impair a witness's identifications. J. McKee Concurring Op. at 329-30. Here, the identification that the Commonwealth so confidently framed as sufficient to support Dennis's conviction may have suffered the greatest from the effect of stress.

. Judge Fisher concedes that Bertha and Cameron may not have been paying attention during the incident, but urges that "the gunshot focused their view and spurred them into action.” J. Fisher Dissent Op. at 366. As Chief Judge McKee's concurrence highlights, however, the presence of a weapon at a crime scene "has a consistently negative impact on both feature recall accuracy and identification accuracy.” J. McKee Concurring Op. at 331. Here, the gunshot may have startled Bertha and Cameron to attention, but research demonstrates that the accuracy of their recollection of the perpetrators would have been reduced, not amplified, by the presence of the silver handgun.

. While this matter was pending before the panel, the government located Frazier in federal prison and interviewed him. During this interview, Frazier admitted the story he told police in 1991 was, in his words, "bullshit,” that the "three-way” phone call with his aunt and "Tony Brown” "never happened,” and that he did not know anyone named “Tony Brown” or "Skeet.” Response to Pet. Rh’g at 17 n.13. Ultimately, Frazier's admission many years post-trial does not change our analysis of whether, given the information the Commonwealth had at the time of the tip, they were required to disclose the lead documents pursuant to Brady.

. Although the United States Supreme Court recently recognized that circuit splits may indicate a possibility of fairminded disagreement under AEDPA, it did so where the circuit split emerged out of an express reservation left by the Supreme Court on the *311precise question decided by the state court. In White v. Woodall, the Kentucky Supreme Court decided that a no-adverse inference instruction, required by the Fifth Amendment to protect a non-testifying defendant at the guilt phase, is not required at the penalty phase. - U.S. -, 134 S.Ct. 1697, 1701, 188 L.Ed.2d 698 (2014) reh’g denied, - U.S. -•, 134 S.Ct. 2835, 189 L.Ed.2d 799 (2014). In so doing, the Kentucky Supreme Court relied on the Supreme Court’s decision in Mitchell v. United States, 526 U.S. 314, 119 S.Ct. 1307, 143 L.Ed.2d 424 (1999), to support its denial. Mitchell included an express reservation on the question the state court decided. See White, 134 S.Ct. at 1703. In the wake of reservation in Mitchell, "[t]he Courts of Appeals ... recognized that Mitchell left [the sentencing question] unresolved; their diverging approaches to the question illustrate the possibility of fairminded disagreement.” White, 134 S.Ct. at 1703 n.3. Thus, the United States Supreme Court opined that the Kentucky Supreme Court’s rejection of respondent's Fifth Amendment claim was not objectively unreasonable because there was an intentional lack of guidance from the Court. The United States Supreme Court has made no such express reservations when it comes to Brady materiality or an admissibility requirement. Consequently, to the extent that language from our sister circuits might be read to recognize a general admissibility requirement in Brady, we respectfully conclude that they have erred. Discrepancies as to the interpretation of Wood ought not to substantiate the Pennsylvania Supreme Court’s erroneous application of the Brady materiality standard in this case.

.Watkins v. Sowders, 449 U.S. 341, 352, 101 S.Ct. 654, 66 L.Ed.2d 549 (1981) (Brennan, J., dissenting) (alterations and emphasis in original) (quoting Elizabeth Loftus, Eyewitness Testimony 19 (1979)).

. Int'l Ass'n of Chiefs of Police, Training Key No. 600: Eyewitness Identification 5 (2006), available at http://www.ripd.org/Documents/ APPENDIX/2/Supporting% 20Materials/IP% 20113% 20IACP% 202006.pdf.






Concurrence Opinion

McKEE, Chief Judge,

concurring.

I. Introduction

More than three decades ago, Justice Brennan cautioned:

[Ejyewitness testimony is likely to be believed by jurors, especially when it is offered with a high level of confidence, even though the accuracy of an eyewitness and the confidence of that witness may not be related to one another at all. All the evidence points rather strikingly to the conclusion that there is almost nothing more convincing than a live human being who takes the stand, points a finger at the defendant, and says ‘That’s the one!’1

James Dennis was sentenced to death because three eyewitnesses appeared at trial and confidently pointed their fingers at him when asked if they saw Chedell Williams’ killer in the courtroom. The prosecution later told the jury that if they believed these witnesses, they should convict James Dennis of first degree murder. And they did.

The Dissent would deny Dennis relief in large part because it believes that “the evidence against Dennis was strong.”2 According to the Dissent, “it is hard to discount the identification testimony of three eyewitnesses.”3 Yet, nearly half a century of scientific research teaches that eyewitness testimony can be one of the greatest causes of erroneous convictions. The jurors in Dennis’ trial, like many juries, were never properly instructed about the dangers of eyewitness identifications. The jury charge given in this ease failed to equip them with the knowledge necessary to accurately assess the reliability of the three eyewitnesses who pointed their fingers at James Dennis and said, “He’s the one.”

I therefore write separately to underscore the problems inherent in eyewitness testimony and the inadequacies of our standard jury instructions relating to that evidence. Jury instructions must educate jurors on the relevant scientific findings regarding eyewitness reliability in order to mitigate the dangers associated with inaccurate eyewitness identifications. The standard instructions, which were used here, are not only insufficient, they are misleading. However, I join the Majority’s *314thoughtful explanation of why Dennis is entitled to relief under AEDPA’s stringent standard of review in its entirety.

In the last thirty years, over 2,000 studies have examined human memory and cognition and their relationship to the reliability of eyewitness identifications.4 This impressive body of scholarship and research has revealed that eyewitness accounts can be entirely untrustworthy. As the International Association of Chiefs of Police has concluded, “[o]f all investigative procedures employed by police in criminal cases, probably none is less reliable than the eyewitness identification.”5

Yet, the law has not caught up to the science. The Innocence Project has documented that, nationwide, eyewitness mis-identifications have been a factor in seventy-five percent of the wrongful convictions that were subsequently overturned by DNA evidence.6 One of the most powerful and prominent examples of such a wrongful conviction is the story of Ronald Cotton and Jennifer Thompson. In July 1984, a man broke into Thompson’s apartment and raped her at knife point.7 When shown a photo array three days later, Thompson tepidly selected Cotton as her attacker.8 “I' think this is the guy,”9 she said, pointing to Cotton’s photo. The lead detective then asked her if she was sure, and she responded, “Positive.”10 But belying her professed certainty, she then asked the detective, “Did I do OK?”11 He reassured her, “You did great.”12 About a month later, Thompson viewed a five lineup, in which Cotton was the only one repeated from the prior photo array.13 When Thompson positively identified Cotton from that lineup, she stated that she was certain he was the one who had attacked her;14 Cotton was then arrested and charged with one count of rape. At his trial, Thompson testified that she was “absolutely sure” that Cotton *315was her rapist.15 There was no corroboration of her identification, and she admitted that she had not been wearing her eyeglasses at the time of the attack.16 Nonetheless, a jury convicted Cotton on the strength of Thompson’s positive identification.17 Cotton was sentenced to life in prison plus fifty-four years.18

The story does not end there. In prison, Cotton learned that a fellow inmate named Bobby Poole had admitted raping Thompson to another inmate. Based on this information, Cotton managed to win a new trial.19 At that retrial, Thompson had an opportunity to view Poole. Her reaction: “I have never seen him in my life.”20 As Thompson later recounted in an interview about the case, when she was asked to look at Poole during Cotton’s second trial, she was angry: “I thought, ‘how dare you. How dare you question me? How dare you try to paint me as someone who could possibly have forgotten what my rapist looked like, I mean, the one person you would never forget. How dare you.’ ”21

Based on Thompson’s unequivocal affirmation of her identification of Cotton, he was once again convicted. He served over a decade in prison before DNA tests finally confirmed that Cotton was innocent and Poole was, in fact, the rapist.22 As one legal commentator described this case, “[t]he fallibility of eyewitness testimony and the malleability of memory could not be clearer, as here a crime victim had seen the scientifically proven perpetrator but instead saw Cotton’s face as that of her assailant.”23

As I will elaborate below when I discuss the even more remarkable story of John White’s erroneous conviction, Cotton’s story cannot readily be dismissed as a fluke. Moreover, problems of erroneous identification remain even where more than one eyewitness identifies the same person as the perpetrator. In thirty-eight percent of misidentification cases documented by the Innocence Project, multiple eyewitnesses misidentified the same innocent person.24 Almost without exception, eyewitnesses who identify the wrong person, express complete confidence that they chose the real perpetrators.25

We should therefore find precious little solace in the fact that three eyewitnesses fingered James Dennis. As I will discuss, the procedures used to elicit the identifications of Dennis and the circumstances surrounding the crime raise serious questions about the accuracy of those identifications. The voluminous studies conducted on the subjects of memory and eyewitness identifications make it painfully clear that many of the identification procedures used in this case were inconsistent with the fundamental concept of neutral inquiry. As a *316result, these identifications lack many of the basic indicia of reliability. Yet, the jury that convicted Dennis was completely unaware of these problems. In addition, the jurors were never even informed that five other eyewitnesses, with similar or better opportunities to observe the shooting, either could not identify Dennis as Chedell Williams’ killer or identified someone else. Accordingly, the three courtroom identifications do little to assuage my concerns about the reliability of the identification testimony that the jury considered. Rather, I cannot help but wonder if an innocent man may have spent more than two decades on death row.

It is as obvious as it is tragic that mistaken identifications have disastrous effects for the unjustly accused. That is particularly true where — as here — the death penalty is imposed. But wrongful convictions are not the only consequence of our continued failure to incorporate the teachings of scientific research into judicial proceedings. Mistaken identifications “also erode public confidence in the criminal justice system as a whole.”26 In addition, when someone is wrongfully convicted, the real perpetrator remains free to victimize again. Thus, this is an issue of far-reaching importance to the defense, prosecutors, police departments, as well as to judges: All have an interest in minimizing the possibility of erroneous identifications. The New Jersey Supreme Court accurately described the situation in its landmark decision discussing eyewitness identifications: “At stake is the very integrity of the criminal justice system and the courts’ ability to conduct fair trials.”27

Before I begin my discussion of the science as applied to this case, I want to emphasize that my point here is not to cast aspersions on the motives of the police or prosecutors involved in this investigation or to insinuate that they intentionally used suggestive procedures to convict Dennis. On the contrary, I have no reason to believe they were motivated by anything other than a sincere desire to bring the killer of Chedell Williams to justice. The science surrounding eyewitness identifications and reliability was simply not as well-understood at the time of Dennis’ investigation and trial as it is today.

II. The Identifications

A. The Crime

As the Majority recounts and the Dissent emphasizes, the shooting at issue here occurred in broad daylight, at the intersection of Tenth Street and Nedro Avenue, in Philadelphia. This intersection is adjacent to the Fern Rock SEPTA station, where steps lead up to a ticketing office. On October 22, 1991, Chedell Williams and her friend Zahra Howard walked up these steps so that Williams could purchase a SEPTA Transpass. As they climbed the steps on opposite sides of a railing that extended up the middle, two men approached them head on. A man with a, red sweat suit — whom witnesses later uniformly described as the shooter — initially approached Howard on her side of the railing and demanded her earrings. The women fled, and Howard managed to hide behind a nearby fruit stand while the man in the red sweat suit pursued Williams into the intersection of Tenth and Nedro. Howard later stated that, up until that point, she had not seen a gun. Howard watched as the man in the red struggled to take *317Williams’ earrings, pulled her close to him, and shot her in the neck with a “silver revolver.”28 Williams fell to the ground, and both men ran north on Tenth Street.

Five other witnesses gave similar accounts of the shooting in police interviews conducted the day of the murder. First, James Cameron, a SEPTA cashier, stated that he was standing at Tenth Street and Nedro Avenue, chatting with another SEPTA employee, when he saw a man grab Williams in the street, pull out a “dull silver gun,” and shoot her.29

As the two .perpetrators fled, they ran past Anthony Overstreet and Thomas Bertha. Overstreet and Bertha were working on a house on North Tenth Street, near the intersection where the shooting occurred. After hearing screaming followed by a gunshot, both men saw Williams fall to the ground as the two perpetrators ran directly toward them. Both Overstreet and Bertha observed the man in the red sweat suit holding a chrome-plated gun in his hand.

Overstreet’s initial interview with police is particularly important because he expressed confidence that he would be able to identify the shooter if he saw him again. Overstreet was about six feet from the perpetrators as they ran past him. In his interview, he recounted that they “both looked right in my face” as they fled.30 Moreover, Overstreet told officers that “he would definitely be able to identify them” because “he ha[d] seen the man with the red hooded jumpsuit who had the gun before.”31 Overstreet then explained that he might have known the shooter from the “area of Broad & Clearview St[reets] where he used to hang.”32 He later clarified that he thought he had seen the shooter at a house where Overstreet used to smoke cocaine, and he gave the police the address of that house.

Another eyewitness who expressed confidence he could identify the shooter was George Ritchie. At the time of the shooting, Ritchie'was repairing a car on Tenth Street. “He heard 2 [black men] hollering and running away from the train station and towards him in the middle of 10th St.”33 Ritchie was about twenty-five feet away from them and “saw them clearly.”34 He told police that “he did get a good look at these two [black males] and can identify them if he sees them again.”35

Another eyewitness, Clarence Verdell, had an opportunity to view the perpetrators immediately prior to the shooting and provided the police with a detailed description of the accomplice’s face. Verdell saw the perpetrators as they initially chased Williams and Howard down the ticketing office steps. A moment later, Verdell heard what sounded like a firecracker. He then turned and saw Williams fall to the ground. Verdell never saw the gun and had never seen either the girls or the males before. He told his interviewer that he would be able to recognize the accomplice, but did not get a good look at the shooter.

Finally, police interviewed David LeRoy, a vendor who sold hot dogs at Tenth and Nedro. He stated that he saw the shooter pull Williams toward him and kill her. He noted that the shooter had on a red hat, pulled down to his eyes.

*318Two weeks after the crime, the police interviewed a fruit vendor and his son, Joseph DiRienzo and Joseph DiRienzo, Jr. They had also been present at the murder scene and echoed the description of the crime provided by the other witnesses.

B. The Photo Arrays

A few days after the shooting, the police heard rumors that James Dennis might have been the shooter, and they decided to show witnesses photo arrays containing his picture. The detectives compiled three arrays of eight photographs each. Dennis’ picture was placed in the first position of the first array, and police used this array to solicit an identification of the shooter (the second array was used to attempt identification of the accomplice, and the third was shown thereafter to offer the witnesses one more opportunity to identify a suspect). At trial, Detective Manuel Santiago explained how he compiled the array: he used the “most recent photo”36 that he could find of Dennis and then “went into [police] files and obtained photos of young black males, which would not be too unlike the photo of Mr. James Dennis.”37 When Detective Santiago showed the witnesses the arrays, he instructed them: “I’m going to show you a photograph spread with eight photos. See if you recognize anyone.”38

Only four of the nine eyewitnesses could make any identification from the arrays: Zahra Howard, Thomas Bertha, Anthony Overstreet, and James Cameron indicated that Dennis “look[ed] familiar.”39 However, none of these witnesses was initially certain about their “identifications.” For example, when Detective Santiago showed Howard the arrays; she pointed to Dennis and stated, “[t]his one looks like the guy, but I can’t be sure.”40 Detective Santiago next showed the same spreads to James Cameron. When asked if he recognized anyone, Cameron stated,'“#1 looks familiar but I can’t be sure.”41 When provided the same arrays, Bertha pointed to Dennis and stated, “[t]hat looks like the one that was running with the gun.”42 Santiago probed further: “Can you be sure that photo #1 is the male that you saw get away from the girl and run at you with the gun after the gunshot?”43 It was then that Bertha replied, “Yes I can.”44 Detective Santiago’s follow-up question and Bertha’s response bear an eerie resemblance to the follow-up question asked of Jennifer Thompson (“Are you sure?”) after her response (“Positive”) following her initial tentative selection of Ronald Cotton from a photo array.

A different detective showed Anthony Overstreet the arrays. After Overstreet had reviewed the first array, the detective asked “[i]s there anyone in these photos that you can identify?”45 Overstreet replied: “Yes, in the first set of photos, #1 looks like the male who shot the girl.”46 The detective then asked Overstreet to repeat his identification: “The male that you identified, is he the male you saw running up the street with the gun?” “Yes he is,” Overstreet confirmed.47 Thus, when *319asked about the male that he had “identified,” Overstreet moved from saying that Dennis’ picture “looked like” the shooter to affirming that Dennis “is” the shooter. This may, at first, appear to be a meaningless distinction that is nothing more than innocuous reply to a simple follow-up question. However, as I will discuss in greater detail below, such subtle, and seemingly innocent, probes can sow seeds that blossom into certain, albeit inaccurate, identifications.48

Significantly, none of the remaining five eyewitnesses selected Dennis from the photo arrays. When a detective showed Verdell the spreads, he stated, “The best I can say is it’s either #1, #5, or #8. I concentrated more on the male that was directly behind Chedell and I believe him to be the accomplice.”49 Verdell returned to the police station a few days later to reexamine the photos. The second time around, he stated “it would be either #1 or #8 who was the [shooter], I lean more towards #1 because of the build of the male but he definitely doesn’t have that cut of hair now. I definitely do not remember him having his hair cut that way.”50 Neither David LeRoy nor either of the DiRienzos identified Dennis from the arrays.

Finally, the Commonwealth denies that police ever showed George Ritchie a photo array. Ritchie vigorously disputes this claim. In 2005, Ritchie testified at Dennis’ Post-Conviction Relief hearing that officers showed him an array during their investigation but became frustrated when Ritchie was unable to identify the shooter from the photos. Assuming arguendo that the Commonwealth’s claim regarding Rit-chie is true, that means that the police and prosecution did not attempt to learn if Ritchie vyould have identified Dennis or someone else as the shooter even though Ritchie had initially expressed confidence in his ability to identify the shooter.

C. The Lineup

On December 19, 1991, about a month and a half after the police showed the witnesses the photo arrays, officers conducted an in-person lineup involving Dennis and five fillers. Fillers are non-suspects who are added to the line-up to provide the witnesses with choices. Although Dennis’ attorney requested that all eyewitnesses be present, only the witnesses who had identified Dennis from the photo array (Howard, Cameron, Bertha, and Over-street) participated.

The police had those four witnesses view the lineup at the same time, in the same room. Accordingly, nothing prevented the witnesses from observing each other’s reactions. As I elaborate below, studies consistently caution against conducting a lineup in this fashion.51 At trial, one of the officers that helped conduct the lineup, Detective William Wynn, testified that the following' instructions were given to the four witnesses:

We’re going to view a lineup of six men. They’ll be numbered from one through six from your left to your right.... I want you to look at each man carefully, see if you can identify any of these men as being involved in your incident. If you can identify any of these men, just remember the number of the man that you can identify, and when we’re through looking at all six men, I’ll order them out of this viewing area or box, as we call it. At that time I will call you outside of the lineup room, one at a time by *320name, and ask you as to whether or not you can make an identification. If you can, just tell me the number of the man that you can identify. If you can’t, simply tell that you cannot. It’s important that while you’re in the lineup room, there will be no pointing, talking, shouting or displaying of emotions so as not to influence one another’s decision. It will be important to you not only this evening but also at a later date.52

After the witnesses viewed each person in the lineup, the police called them out of the room, one by one, and asked if they could make an identification.

Cameron and Bertha identified Dennis. Howard pointed out Dennis, but was less sure, stating only “I think it was [him].”53 Overstreet — the witness who initially expressed the most confidence in his ability to identify the shooter due to his alleged prior exposure to him — identified an entirely different person from the lineup.

D. In-Court Identifications

At Dennis’ trial over a year later, the prosecution called only the three witnesses who had picked him from the photo arrays and lineup. When asked whether Chedell Williams’ killer was in the courtroom, Bertha, Cameron, and Howard each confidently pointed to Dennis, even though all three had expressed doubt in their earlier identifications.

III. The Science of Eyewitness Identifications

As I noted at the outset, we have long known that eyewitness identifications are not-always as reliable as witnesses (and jurors) may believe them to be. In 1927, long before the explosion of research in this area, Justice Felix Frankfurter wrote: “[t]he hazards of [eyewitness identification] testimony are established by a formidable number of instances in the records of English and American trials.”54 In 1932, well before the availability of DNA analysis, Yale Law professor Edwin M. Borchard documented almost seventy cases involving eyewitness errors that caused miscarriages of justice.55 Over thirty years later, the Supreme Court acknowledged this problem in United States v. Wade.56 There, the Court famously proclaimed that “[t]he vagaries of eyewitness identification are well-known; the annals of criminal law are rife with instances of mistaken identification.”57

In the ensuing decades, the scientific community has made significant strides in understanding this phenomenon.58 A combination of basic and applied research on human visual perception and cognition has revealed that the reliability of eyewitness *321identifications is largely contingent on the conditions under which memories are created, stored, and then later recalled. “At its core, eyewitness identification relies on brain systems for visual perception and memory: The witness perceives the face and other aspects of the perpetrator’s physical appearance and bearing, stores that information in memory, and later retrieves the information for comparison with the visual percept of an individual in a lineup.”59 Research has shown that certain variables can impact the processes of these memory functions with serious implications for the reliability of the subsequent memories. These variables generally fall into two basic categories: system variables and estimator variables.

A. System Variables

System variables are the procedures and practices law enforcement use to elicit eyewitness identifications.60 Examples of system variables include the instructions law enforcement officers give to witnesses when they ask them to provide identifications, the comments of police to witnesses during the identification process, and the types of procedures (lineup, photo array, etc.) used to solicit the identification. These factors are important not only because they heavily influence the reliability of identifications, but also because they largely lie within the exclusive control of the criminal justice system. The following section- explores a few critical system variables and their effects on the accuracy of eyewitness identifications.

1. Blinded versus Non-Blinded Procedures

One of the most important system variables that law enforcement can control is the blinding of identification procedures.61 Blinding occurs when the officer administering an identification procedure, such as a photo array, knows who the suspect is but cannot determine when the witness is viewing the suspect’s photo. “In one common ‘blinded’ procedure, the officer places each photo in a separate envelope or folder and then shuffles the envelopes/folders so that only the witness sees the images therein.”62 This blinding can also be doubled: for example, when an officer who neither knows the suspect’s identity nor position in the photo array shows the array to an eyewitness. Such blinding is used to prevent the officer from giving the witness conscious or unconscious cues that can affect the witness’ identification.63

Common sense suggests that idéntification procedures administered without some degree of blinding are inherently untrustworthy, and research confirms this.64 Typically, the greater the level of blinding, the more reliable the procedure. One of the foremost experts on eyewitness identifications has concluded that blind lineup administration is “the single most important characteristic that should apply to eyewitness identification.”65 Social psychologists believe this is crucial to avoiding the “expectancy effect”: “the tendency for experimenters to obtain results they expect ... because they have helped to shape that *322response through their expectations.”66 In a seminal meta-analysis of 345 studies across eight broad categories of behavioral research, researchers found that “[t]he overall probability that there is no such thing as interpersonal expectancy effects is near. zero.”67 “Even seemingly innocuous words and subtle cues — pauses, gestures, hesitations, or smiles — can influence a witness’ behavior.”68 Moreover, the witness usually remains completely unaware of the signals she has been given or their effect on her identification.

Outside the realm of law enforcement, in scientific experiments for instance, it is standard practice to use blinding. The importance of blind administration is so great that a failure to implement such a policy can affect even seemingly objective processes, such as the analysis of DNA samples. In one experiment, researchers gave seventeen experienced DNA analysts a mixed sample of DNA evidence from an actual crime scene — a gang rape committed in Georgia.69 All seventeen analysts worked at the same accredited government laboratory in the United States.70 Years earlier, prosecutors had relied on this evidence to convict a man named Kerry Robinson.71 In the real investigation, two analysts from the Georgia Bureau of Investigation concluded that Robinson “could not be excluded” as a suspect based on his DNA profile relative to the crime scene sample.72 Nevertheless, of the seventeen analysts involved in the study of this case, only one agreed that Robinson “could not be excluded.”73 Four analysts found that the evidence was inconclusive, and the-other twelve said he could be excluded.74 All seventeen analysts were blinded to contextual information about the case.75 Experts speculated that a failure to blind the DNA testing in the real investigation could explain the inconsistency between the results the Georgia Bureau of Investigation and the seventeen independent analysts obtained. “The difference between you giving them the data and saying ‘what do you make of it?’ and the local district attorney giving them the data and saying ‘We’ve arrested someone, is his profile in here?’ is huge.”76

The Supreme Court has recognized the significance of such cues for decades. In 1967, in United States v. Wade, the Court ruled that a pretrial lineup is a “critical stage” of prosecution at which a defendant had a right to the presence of counsel.77 The Court explained:

*323The fact that the police themselves have, in a given case, little or no doubt that the man put up for identification has committed the offense, and that their chief pre-occupation is with the problem of getting sufficient proof, because he has not “come clean,” involves a [ ] danger that this persuasion may communicate itself even in a doubtful case to the witness in some way.78

The importance of conscious and unconscious police persuasion cannot be overstated in the context of a trial because it negates the effect that strenuous cross-examination may otherwise have on the witness’ confidence in her identification. “[E]ven though cross-examination is a precious safeguard to a fair trial, it cannot be viewed as an absolute assurance of accuracy and reliability.”79 Obviously, if an eyewitness is completely unaware that her identification has been shaped by subliminal cues communicated by investigators, it is incredibly difficult, if not impossible, to dissuade that witness of the accuracy of her identification. As was true for Jennifer Thompson in the rape case discussed earlier, vigorous cross-examination may serve only to reinforce the witness’ certainty of her identification.80 The Supreme Court recognized in Wade that once a pretrial identification is made, the identifying wildness becomes “the sole jury.”81 Thus, “[t]he trial which might determine the accused’s fate may well not be that in the courtroom but that at the pretrial confrontation.”82

None of the identifications in Dennis’ case were obtained through processes that included blinding. The officers who showed the photo arrays and conducted the lineup knew that Dennis was the suspect, and they knew his position in the arrays and in the lineup. As the above studies make clear, it is entirely possible that the officers investigating Williams’ killing gave the witnesses unconscious cues about their suspicions. Dennis’ jurors would have been in a far better position to assess the reliability of the three courtroom identifications had they been informed of the importance of blinding procedures and their absence here.

2. Pre-Identification Instructions

The instructions police give witnesses prior to attempting to elicit an identification constitute a second important system variable. There, is broad consensus that police must instruct witnesses that the suspect may not be in the lineup or array and that the witness should not feel compelled to identify anyone.83 In two meta-analyses, researchers found that providing this information to witnesses in advance significantly increased the reliability of the results in target-absent lineups.84 In one study, the number of people that chose innocent fillers in target-absent lineups increased by forty-five percent when the lineup administrators failed to tell the subjects that they need not choose a suspect.85

*324One hardly needs to engage in a protracted review of the wealth of data on this point to appreciate its implications. Without such instructions, witnesses may misidentify innocent suspects merely because they assume the suspect is present and the person misidentified bears the strongest resemblance to the actual perpetrator. Research confirms this.86 It is therefore critical that courts inform jurors of this system variable where present. Such information enables jurors to consider the impact that the absence of such instructions may have had on witness identifications.

The record in Dennis’ case shows that the investigators failed to give such instructions to the witnesses. Accordingly, there is a real risk that the witnesses identified Dennis because he most closely resembled Williams’ killer. Indeed, that is a fair interpretation of this record. Upon seeing Dennis’ photo, Howard did not say “that’s him,” or “I think this is the shooter.” Instead, she tentatively told officers: “This one looks like the guy, but I can’t be sure.”87 Like Howard, Bertha and Cameron also initially responded to these arrays in a manner that strongly suggests that they selected Dennis because his photograph bore a closer resemblance to the shooter than any of the fillers. They qualified their selection of Dennis by saying: “Number 1 looks familiar but I can’t be sure”88; and “that looks like the one that was running with the gun.”89 It simply cannot be assumed that either statement was the equivalent of proclaiming: “that’s him,” or “he’s the one.”

3. Photo Array and Lineup Construction

Researchers have also found that the way that a photo array or live lineup is constructed can affect the reliability of the resulting identifications. A number of considerations are critical. First, not surprisingly, mistaken identifications are more likely where the suspect stands out in comparison to the fillers.90 Using fillers that are relative look-alikes forces a witness to examine her memory, whereas placing the suspect among a group of individuals that bear little resemblance to him causes him to stand out. “[A] biased lineup may [also] inflate a witness’ confidence in the identification because the selection process seemed easy.”91 As of yet, there is no clear agreement among researchers about whether fillers should more closely resemble a witness’ pre-lineup description of the suspect or the actual suspect.92 However, whether the fillers more closely resemble *325the suspect or the witness’ pre-lineup description, the fillers’ appearances should not make the suspect stand out.

Second, all lineups should include a minimum of five fillers.93 The logic here, which appears to be a matter of general agreement, is again clear: the greater the number of choices, the less the chance of making a lucky guess, and the more the witness is forced to rely on her own memory to identify the suspect.

Third, for similar reasons, lineups should not feature more than one suspect. In its landmark decision on the issue of eyewitness identification, the Supreme Court of New Jersey emphasized that, “if multiple suspects are in the lineup, the reliability of a positive identification is difficult to assess, for the possibility of ‘lucky’ guesses is magnified.”94

The trial judge here noted that the composition of the lineup was somewhat suggestive because Dennis was slightly shorter than the rest of the participants, causing him to stand out. The jurors were therefore able to consider this disparity as they evaluated the reliability of the identifications. However, the court did not provide the jury with an explanation of how this may have affected the witnesses’ identifications of Dennis in that lineup. Nor did it give the jurors information that would allow them to consider the lineup construction in context with all of the other factors that were involved in the identifications of Dennis.

4. Interactions with Witnesses: Witness Feedback

Another critical system variable is whether law enforcement provides a witness with any feedback or other information in the course of her identification. As I touched on in my discussion of blinding procedures, “[t]he nature of law enforcement interactions with the eyewitness before, during, and after the identification plays a role in the accuracy of eyewitness identifications and in the confidence expressed in the accuracy of those identifications by witnesses.”95 Elizabeth Loftus, a pioneering researcher in the field of human memory and cognition, has thoroughly documented the effects of received information on memory accuracy. In one study, she showed college students a video of a car crash on a country road.96 Afterward, she asked them to estimate how fast the car was going. Half the students were asked how fast the car was going when it “passed the barn” along the country road; the other half were simply asked how fast the car was going “along the country road.”97 A week later, she asked the same students whether they had seen a barn in the film. Approximately seventeen percent of the students who were given the “passed the barn” cue recalled seeing the barn in the video.98 In contrast, less than three percent of the non-barn cue group remembered a barn.99 In reality, there was no bam in the video.100 This demonstrates the very subtle — yet extremely powerful— *326effect statements at the time of memory recall can have.

In the eyewitness identification context, such information often comes in the form of pre- or post-identification information that may reinforce an identification. For example, research confirms the intuitive proposition that when investigators give cues that suggest “you got the right guy,” the witness’ confidence in the identification is artificially inflated. A meta-analysis of twenty studies covering 2,400 identifications found that witnesses who received feedback “expressed significantly more retrospective confidence in their decision compared with participants who received no feedback.”101 Such feedback not only causes a witness to misjudge the reliability of her identification, it can also result in the witness embellishing the opportunity she had to observe the perpetrator and the crime. “Those who receive a simple post-identification confirmation regarding the accuracy of their identification significantly inflate their reports to suggest better witnessing conditions at the time of the crime, stronger memory at the time of the lineup, and sharper memory abilities in general.”102 Furthermore, eonfirmational feedback need not be immediate to corrupt a witness’ memory. One study showed that the effects of eonfirmational feedback may be the same even when it occurs two days after an identification.103 Other research further substantiates that these effects can withstand the passage of time.104

The particular perils of witness feedback are evident in many of the documented cases of false identifications. Here again, the story of Ronald Cotton and Jennifer Thompson is illustrative: officer feedback led Thompson to harden her false memory of Cotton as her rapist. In the process, her memory was effectively immunized from any impact cross-examination may otherwise have had on her confidence, which impeded the jury’s ability to properly assess her testimony.

I realize, of course, that law enforcement officials are not completely in control of the feedback witnesses receive. Interactions among witnesses outside the confines of police proceedings, for instance, can affect the reliability of the witnesses’ identifications.105 For example, if one witness *327talks to another, she can alter or reinforce the other’s memory of the same event. “[P]ost-identification feedback does not have to be presented by the experimenter or an authoritative figure (e.g. police officer) in order to affect a witness’ subsequent crime-related judg[ ]ments.”106 In one study, after witnesses made incorrect identifications, they were told either that their co-witness made the same or a different identification.107 Not surprisingly, confidence rose among the witnesses that were told that their co-witness had agreed with them and fell among those told that co-witnesses had disagreed.108

Though law enforcement officials may not be able to completely insulate witnesses from this system variable, police did not even attempt to guard against it here. The witnesses who identified Dennis viewed the lineup in the same room and at the same time. Detective Wynn’s instruction to the witnesses not to react or show emotion during the lineup reduces the risk of feedback, but this instruction did not eliminate it. Therefore, the risk that the witnesses’ reactions may have influenced the results of the lineup cannot be discounted, and the jurors should have been instructed about this possibility.

Furthermore, the record of Bertha’s photo array identification establishes the existence of at least some officer-to-witness feedback. Detective Santiago asked Bertha to affirm his identification: “Can you be sure that photo #1 is the male that you saw get away from the girl and run at you with the gun after the gunshot?”109 Only then did Bertha state he was “sure”110 Dennis was the shooter as opposed to his initial statement that Dennis’ photo merely “look[ed] like”111 the shooter.

I am not suggesting that Detective Santiago’s question ultimately negated Bertha’s ability to make an in-court identification. Nor am I suggesting that Detective Santiago intentionally tried to reinforce Bertha’s confidence in his identification or “prime” him for a subsequent in-court identification. I am, however, suggesting that the jury should have been informed of how Detective Santiago’s response to Bertha’s initial selection of Dennis’ photo may have affected the reliability of Bertha’s lineup identification and, as I next explain, his subsequent in-court identification as well.

5. Multiple viewings

Another crucial system variable — and one that was clearly present here — is the *328opportunity to engage in multiple viewings of a suspect. Allowing a witness to view a suspect more than once during an investigation can have a powerful corrupting effect on that witness’ memory. It creates a risk that the witness will merely identify a suspect based on her past views of him rather than her memory of the relevant event. Meta-analysis has revealed that while fifteen percent of witnesses mistakenly identify an innocent person during the first viewing of a lineup, that percentage jumps to thirty-seven percent if the witness previously viewed that innocent person’s mug shot.112 This phenomenon is known as “mug shot exposure.” Related studies have also shown the existence of “mug shot commitment.” This refers to the fact that once witnesses positively identify an innocent person from a mug shot, “a significant number” then “reaffirm! ] their false identification” in a later photo lineup.113 This is true even when the real suspect is actually present in the lineup.114 Nonetheless, multiple viewings seem to have no impact on the reliability of a lineup identification “when a picture of the suspect was not present in photographs examined earlier”115 by the witness.

The incredible story of John White that I mentioned at the outset serves as a powerful example of the impact that multiple viewings can have on witness identifications. In 1979, John White was accused of breaking into the home of a seventy-four-year-old woman and then beating and raping her.116 After the 'victim picked White out of a photo array, he was placed in a live lineup.117 White was the only person repeated in both the photo array and live lineup. The victim identified White from that live lineup.118 DNA analysis later revealed that the victim’s actual assailant was not White, but a man named James Parham. By the cruelest of ironies, Par-ham had actually been placed in the live lineup with White as a filler when the victim identified White as her assailant. Despite having an opportunity to view her real rapist in the lineup, the victim affirmed her initial selection of White. Her erroneous identification led to a life sentence for White, who served twenty-seven years before the DNA evidence exonerated him.119

A leading researcher offered the following explanation of White’s case:

The witness had already identified John White from a photographic lineup. And, John White was the only person who was in both the photographic lineup and the live lineup. Hence, what we have here, I believe, is a strong example of how a mistaken identification from one procedure (a photo lineup) is repeated in the next procedure (a live lineup) even though the real perpetrator is clearly present in the second procedure. Re*329peating the same mistake can occur for several reasons. One possibility is that the initial mistaken identification changed the memory of the witness; in effect John White’s face “became” her memory of the attacker and the face of Parham no longer existed once she mistakenly identified John White. Another possibility is that she approached the live lineup with one goal in mind — find the man she had identified from the photos. Perhaps she never really looked at Parham because she quickly saw the man she identified from photos and did not need to look further.120

The witnesses who identified Dennis at trial were given not two, but three, opportunities to view Dennis. These multiple views could help explain why initially tentative guesses became certain identifications by the time the witnesses took the stand. The possibility cannot be ignored that the witnesses here, like the victims in White and Cotton’s cases, selected Dennis in the live lineup because they were looking for the man they had already identified from the photo arrays. The jurors should have been informed of the impact of multiple viewings so that they could have considered that effect in determining how much weight to afford the lineup identifications and/or the in-court identifications. Absent that information, the jurors were ill equipped to assess the possibility that Howard, Bertha, and Cameron’s lineup and in-court identifications of Dennis may have been based on prior viewings of his picture rather than their memories of the crime.

These system variables on the accuracy of eyewitness identifications highlight the importance of the procedures law enforcement officials use when soliciting identifications. As the Oregon Supreme Court has explained, “it is incumbent on courts and law enforcement personnel to treat eyewitness memory just as carefully as they would other forms of trace evidence, like DNA, bloodstains, or fingerprints, the evi-dentiary value of which can be impaired or destroyed by contamination. Like those forms of evidence, once contaminated, a witness’ original memory is very difficult to retrieve.”121

B. Estimator Variables

Estimator variables are the conditions present during memory formation or storage. They can also have a substantial impact on the reliability of eyewitness identifications.122 Crucial estimator variables include, but are not limited to, the amount of stress on the observer, the presence of weapons, and visibility conditions. Unlike system variables, estimator variables are beyond the control of the criminal justice system. Nevertheless, asking jurors to consider eyewitness identifications without properly instructing them on the impact' that such estimator variables may have had erects yet another barrier to accurate evaluation of identifications.

1. Stress

First, high levels of stress at the time of memory formation can negatively impact a witness’ ability to accurately identify the perpetrator.123 Stressful conditions impair *330a witness’ ability to identify key characteristics of an individual’s face.124 A meta-analysis of the effect of high stress on eyewitness identifications found that stress hampers both eyewitness recall and identification accuracy.125

A recent study examining the effects of stress on identifications at a U.S. Military mock prisoner-of-war camp illustrates this phenomenon.126 In this study, 509 active-duty military personnel, with an average of 4.2 years in the service, underwent two types of interrogations.127 After twelve hours of confinement, participants experienced either a high-stress interrogation involving real physical confrontation followed by a low-stress interrogation without physical confrontation, or vice versa.128 The interrogations were separated by approximately four hours, and about half the participants received the high-stress interrogation first, while the other half experienced the low-stress interrogation first.129 Both interrogations lasted about forty minutes.130 Twenty-four hours after the interrogations, the participants were asked to identify their interrogators from live lineups, sequential photo arrays, or simultaneous photo arrays.131 Across all identification procedures, subjects had far more difficulty accurately identifying their high-stress interrogators.132 Sixty-two percent of subjects could identify their low-stress interrogators in live lineups, while only thirty percent of subjects could accurately identify their high-stress interrogators from such lineups.133 Furthermore, fifty-six percent of subjects erroneously identified a person who was not their interrogator (false positive) during live lineups, while only thirty-eight percent of subjects did so for their low-stress interrogations.134

This study is particularly stunning when one considers that the subjects all had a prolonged and unobstructed opportunity to view their interrogators, and the interrogators were all within arm’s reach of their subjects. The subjects’ ability to see the faces of their interrogators was therefore exponentially better than the opportunity witnesses to most violent crimes have to see perpetrators. Their views were certainly better than those of Howard, Bertha, and Cameron. As the study’s authors explained,

[cjontrary to the popular conception that most people would never forget the face of a clearly seen individual who had physically confronted them and threatened them for more than 30 min[utes], ... [tjhese data provide robust evidence that eyewitness memory for persons encountered during events that are personally relevant, highly stressful, and realistic in nature may be subject to substantial error.135

*331Notably, this study further found that memories formed during a stressful event are highly susceptible to modifications from misinformation received after the event. That has particular relevance here given the presence of the system variables described above.

Stress almost certainly affected all of the witnesses who saw Chedell Williams gunned down. The shooting undoubtedly caused Howard — the prosecution’s star witness — a significant amount of stress. Not only was she herself chased, but she also watched as the perpetrator grabbed her best friend and shot her at point-blank range. It is not surprising that multiple witnesses recalled hearing Howard screaming. Stress also likely affected Bertha’s ability to later make an accurate identification. He saw the shooter as the shooter rushed him, head on, pistol in hand. Jurors cannot properly assess eyewitness identification testimony where stress was present at memory formation unless this variable is explained to them.

2. Weapon Focus

The presence of weapons is a second, and related, estimator variable. The National Research Council has stated, “[research suggests that the presence of a weapon at the scene of a crime captures the visual attention of the witness and impedes the ability of the witness to attend to other important features of the visual scene, such as the face of the perpetrator.... The ensuing lack of memory of these other key features may impair recognition of a perpetrator in a subsequent lineup.”136 In 1992, an analysis of weapon focus studies concluded that the presence of a weapon significantly reduced witnesses’ ability to recall their perpetrators.137 A more recent study of the pertinent literature confirms that weapon presence has a consistently negative impact on both feature recall accuracy and identification accuracy.138

Here, the jury was never informed that visibility of the perpetrator’s gun may well have hampered the witnesses’ ability to observe and/or form an accurate memory of the assailant's face. Howard, Bertha, and Cameron all provided clear descriptions of the gun, revealing their focus on it. But the jury was never informed of how this powerful estimator variable may have affected them.

3. Memory Decay

The period between memory formation and memory recall is known as the “retention interval” and constitutes another important estimator variable. A meta-analysis of fifty-three facial memory studies found “that memory strength will be weaker at longer retention intervals than at briefer ones.”139 Most of the studies analyzed in this meta-analysis examined retention intervals of less than one month, many of them less than one week. This meta-analysis also found agreement among experts that “the rate of memory loss for *332an event is greatest right after an event and then levels off over time.”140 Furthermore,

[t]he effect of the retention interval also is influenced by the strength and quality of the initial memory that is encoded, which, in turn, may be influenced by other estimator variables associated with witnessing the crime (such as the degree of visual attention) and viewing factors (such as distance, lighting, and exposure duration).141

The in-court identifications of Dennis were made nearly one year after the crime occurred — a very significant retention interval under the relevant studies. Research is hardly necessary to appreciate the difficulty of trying to accurately recall the details of this chaotic and traumatizing event— lasting only a matter of seconds — a year later. The jurors should have been informed of that difficulty and its possible impact on the accuracy of these identifications. They were not.

4. Exposure Duration, Distance, and Lighting

As one would expect, exposure duration, distance, and lighting affect the accuracy of eyewitness identifications.142 The charge that was given here did alert the jurors to the impact of these factors on the accuracy of an identification.143 However, as I explain in the following section, it did not adequately convey the impact these factors can have on in-court identifications.

C. The Dissent’s Dismissal of Estimator Variables

As the Majority recounts, nearly all of the eyewitnesses who mentioned the shooter’s height in their initial police interviews described him as between 5'8" and 5'10".144 The witnesses also described the shooter as having a dark complexion and weighing about 170 to 190 pounds. James Dennis is 5'5" tall and weighed between 125 and 182 pounds at the time of trial.

*333The Dissent dismisses and tries to rationalize away this considerable size discrepancy. In an attempt to reinforce the reliability of the three witnesses, the Dissent relies on research that concludes eyewitnesses tend to underestimate the height and weight of taller and heavier targets and overestimate the height and weight of shorter and lighter targets.145 The Dissent’s use of that research is cruelly ironic. The finding of those studies was not that we should disregard eyewitness inaccuracy, as the Dissent’s citation implies. Those researchers found just the opposite. The studies discovered that eyewitness identifications are frequently unreliable,146 As two of the researchers explained, “[t]he width and range of subjects’ errors for the targets’ height and weight in this study showed clearly that some subjects experience great difficulty in accurately judging another individual’s physical characteristics.”147

The Dissent also focuses on the strength of three estimator variables. The Dissent reminds us that “the visual conditions were excellent,”148 the witnesses saw the shooter at “close range,”149 and none of the identifications were cross-racial.150 This is not only misleading, it also ignores many other system and estimator variables that were at least as important (if not more important) than the ones the Dissent focuses upon.

I agree that the lighting was good. However, the lighting here was likely no better than that in the rooms where the military personnel who failed to recognize the faces of their interrogators were questioned under stressful conditions.151 The witnesses here were in close proximity to the shooter. However, they were not as close as Jennifer Thompson was to Ronald Cotton or John White’s accuser was to him. Moreover, these witnesses only had a matter of seconds to view the perpetrators. Howard saw the shooter as he rushed towards her, Cameron in the seconds the crime occurred, and Bertha as the shooter ran past him. All of the witnesses’ views occurred under highly stressful circumstances and their focus appears to have been as much on the gun in the shooter’s hand as on the shooter’s face. As I will explain in greater detail below, the charge that the jurors received did not focus their attention on any of those considerations.

The lack of blinding, thé presence of officer feedback, the fact that the record suggests that the witnesses thought they had to select someone from the photo arrays, the multiple viewings of Dennis, and the witnesses’ viewing of the live lineup in the same room, all suggest that the identifications may have been corrupted by cues from law enforcement and/or other witnesses.

*334We would be justifiably skeptical of any clinical trial where the researcher knew which sample was a placebo or who received the placebo. Yet, we do not think twice about allowing someone to be convicted of a crime and sentenced to death on the basis of identification procedures where the investigator presenting the photo array or lineup is fully aware of who the suspect is. The witnesses who identified Dennis at trial had not one, but three opportunities to view Dennis. And none of the procedures included any level of blinding. Nothing in this record suggests that anyone other than Dennis was present in both the photo array and lineup. Yet, the jury was not made aware of the potential importance of any of these considerations. That should sound a note of caution in assessing the reliability of these identifications.

Finally, we should not ignore the fact that the majority of the witnesses that police interviewed after the crime were unable to identify Dennis as the shooter. Jurors did not know that Joseph DiRienzo, Joseph DiRienzo, Jr., Clarence Verdell, and David LeRoy all were unable to identify Dennis from the photo array. Although Anthony Overstreet did identify Dennis from this array, he did not think Dennis was the shooter once he had an opportunity to view him in the lineup. Overstreet had expressed the most confidence in his ability to positively identify the shooter during the initial police interviews.152 When the totality of circumstances is viewed in context, the evidence of Dennis’ guilt is not as uncompromising as the Dissent suggests.

Moreover, concerns about the reliability of these identifications should not be assuaged by evidence that was introduced in an attempt to corroborate the identification testimony. As the Majority explains, aside from eyewitness testimony, the Commonwealth presented testimony from Charles Thompson, who told detectives that he saw Dennis with a gun the night of the murder. Thompson identified an illustrative .32 chrome revolver (previously admitted as a Commonwealth exhibit) as being similar to the one he saw in Dennis’s possession. As the Majority notes, Thompson had an open drug-possession charge at the time of trial, but testified that he was not expecting help from the Commonwealth in exchange for his testimony. Years after trial, Thompson recanted his testimony, averring that he had never seen Dennis with a gun and that his testimony at trial was false.

I realize, of course, that it can be argued that Thompson’s recantation is not necessarily relevant to the force of the eyewitness identifications because it happened after trial. However, his testimony clearly corroborated the identification evidence, and it underscores the dangers of the inadequate identification instructions. The fact that the jurors were not given a sufficient basis to assess the identifications of Dennis severely undermined the potential force of Dennis’ alibi testimony. Why would jurors believe such testimony (especially since it was offered by his father) when three neutral witnesses identified Dennis as the shooter? Had the jurors been able to assess the identifications with an appropriate understanding of the variables I have discussed, Dennis’s alibi testimony may well *335have had much greater force, and jurors would have been in a better position to weigh Dennis’ alibi against Thompson’s testimony that appeared to corroborate the three eyewitnesses. That is particularly true when we factor in the evidence of the Cason receipt that the Majority explains.153 The Cason receipt could have further bolstered Dennis’ alibi testimony and raised a reasonable doubt about the accuracy of the eyewitness identifications.

IV. Manson v. Brathwaite and its Progeny

In 1977, the Supreme Court established a basic framework for determining whether admission of a particular identification violates a defendant’s Fourteenth Amendment right to due process in Manson v. Brathwaite.154 Under the Manson test, a court must first assess whether the eyewitness identification procedure at issue was, under the “totality of the circumstances,” unnecessarily suggestive.155 If the identification procedure was not unnecessarily suggestive, the inquiry ends. However, if it was unduly suggestive, a court must considers five factors to determine whether the resulting identification is nonetheless reliable. Those factors, drawn from the Supreme Court’s prior decision in Neil v. Biggers,156 are: (1) “the opportunity of the witness to view the criminal at the time of the crime,” (2) “the witness’ degree of attention,” (3) “the accuracy of the witness’ prior description of the criminal,” (4) “the level of certainty demonstrated by the witness at the confrontation,” and (5) “the length of time between the crime and the confrontation.”157 These factors are weighed against “the corrupting effect of the suggestive identification itself.”158 Manson emphasizes that “reliability is the linchpin in determining the admissibility of identification testimony.”159

Since Manson, more than 2,000 scientific studies have been conducted on the reliability of eyewitness identifications.160 As I have explained, we now understand that even seemingly neutral identification procedures can lead to unreliable results due to a myriad of subtle variables. We also now know that a witness’ subjective confidence in the accuracy of her identification has limited correlation to the reliability of her identification. As the National Research Council emphasized in its recent report on eyewitness identifications, the Manson test “treats factors such as the confidence of a witness as independent markers of reliability when, in fact, it is now well established that confidence judgments may vary over time and can be powerfully swayed by many factors.”161

The Supreme Court recently reaffirmed the approach laid out in Manson in Perry v. New Hampshire.162 There, an eyewitness saw a man break into a car, called the police, and then told the responding officer *336that a man standing in the building’s parking lot was the perpetrator.163 That man was then arrested and convicted in state court. On appeal to the Supreme Court, he argued that the highly suggestive nature of the identification process entitled him to a suppression hearing prior to trial in order to determine the admissibility of the identification.164 The Supreme Court rejected this argument. It held that the Due Process Clause of the Fourteenth Amendment only requires such a hearing when law enforcement arranged the unnecessarily suggestive circumstances under which the identification was obtained.165 The Court “linked the due process check, not to suspicion of eyewitness testimony generally, but only to improper police arrangement of the circumstances surrounding an identification.”166

In reaching this conclusion, the Court acknowledged the scientific research on eyewitness reliability.167 It recognized the importance of this body of science and urged more robust jury instructions. As the Court explained, “[ejyewitness-specific jury instructions, which many federal and state courts have adopted, [] warn the jury to take care in appraising identification evidence.”168 The Court also stressed the importance of evidentiary rules “to exclude relevant evidence if its probative value is substantially outweighed by its prejudicial impact or potential for misleading the jury.”169 Thus, instead of considering the relevant system and estimator variables “under the banner of due process,”170 the Supreme Court advocated that courts incorporate the relevant scientific findings through other avenues, such as jury instructions and evidentiary rules.

Some state- courts have heeded Perry’s call and created new procedures and evi-dentiary frameworks that minimize the risks associated with erroneous eyewitness identifications. Most notably, in a unanimous decision, the Supreme Court of New Jersey re-wrote the state’s rules governing the admission of eyewitness identifications in State v. Henderson.171 Prior to that decision, New Jersey courts relied on the Manson test to determine whether certain identifications were admissible.172 Henderson, however, held that the Manson test did “not offer an adequate measure for reliability or sufficiently deter inappropriate police conduct.” The court also concluded that Manson “overstates the jury’s inherent ability to evaluate evidence *337offered by eyewitnesses who honestly believe their testimony is accurate.”173

To remedy these problems, the court pioneered a two-part revision to the judicial procedures related to eyewitness identifications. First, the court changed the requirements related to pre-trial hearings on the admissibility of eyewitness identifications. After Henderson, a defendant can now obtain a pre-trial hearing if she can show “some evidence of suggestiveness that could lead to a mistaken identification.”174 The court specified that this “evidence, in general, must be tied to a system — and not an estimator — variable.”175 The trial court can end this hearing at any time “if it finds from the testimony that defendant’s threshold allegation of suggestiveness is groundless.”176 But if the defendant’s claim is meritorious, the trial judge must weigh both system and estimator variables177 to decide whether, under the “totality of the circumstances,” the defen*338dant has “demonstrated a very substantial likelihood of irreparable misidentification.”178 If the trial court concludes that the defendant has met this burden, the court must suppress the identification evidence.179

Second, the New Jersey Supreme Court directed the state judicial system to develop “enhanced jury charges on eyewitness identification for trial judges to use.”180 As the court explained, “[w]e anticipate that identification evidence will continue to be admitted in the vast majority of cases. To help jurors weigh that evidence, they must be told about relevant factors and their effect on reliability.”181

Henderson also emphasized that the “factors that both judges and juries will consider are not etched in stone.”182 Rather, “the scientific research underlying them will continue to evolve, as it has in the more than thirty years since Manson.”183 Accordingly, the court clarified that its decision does not “limit trial courts from reviewing evolving, substantial, and generally accepted scientific research.”184

Finally, the New Jersey Supreme Court suggested that, where appropriate, trial courts consider giving instructions during the trial before eyewitness identification testimony is elicited. Such instructions would help inform juries, up front, of the problems that can arise from seemingly unequivocal courtroom identifications.185

After Henderson, in July 2012,186 the New Jersey Supreme Court released its expanded set of jury instructions governing evaluation of identifications. These instructions explain that scientific research has shown eyewitness identifications can be unreliable, and they emphasize that eyewitness evidence “must be scrutinized carefully.”187 To this end, the instructions identify a specific set of factors that jurors should consider when deciding whether eyewitness identification evidence is reliable, including estimator and system variables.188 These instructions are consistent with the Supreme Court’s analysis in Perry and will better equip jurors to evaluate the reliability of eyewitness identifications.189

*339The Supreme Court of Oregon has likewise reformed the state judicial system’s approach to eyewitness identifications. However, Oregon has taken a slightly different approach. In State v. Lawson,190 the court addressed the reliability issue from an evidentiary standpoint as opposed to a due process one. Prior to Lawson, Oregon courts adhered to a rule under which trial courts could not consider whether an identification was unreliable until some evidence of suggestiveness was first introduced.191 In rejecting that approach, the Oregon Supreme Court explained:

Such a requirement [ ] conflates eviden-tiary principles with due process concerns. A constitutional due process analysis might properly consider suggestiveness as a separate prerequisite to further inquiry because the Due Process Clause is not implicated absent some form of state action, such as the state’s use of a suggestive identification procedure. As a matter of state evidence law, however, there is no reason to hinder the analysis of eyewitness reliability with purposeless distinctions between suggestiveness and other sources of unreliability.... A trial court tasked with determining a constitutional claim must necessarily assume that the evidence is otherwise admissible; were it inadmissible on evidentiary grounds, the court would never reach the constitutional question. However, a trial court tasked with considering a question of evidentiary admissibility clearly cannot begin by assuming admissibility.192

Lawson then fashioned a new approach to examining eyewitness identifications from existing rules of evidence. Under this revised test, “when a criminal defendant files a pretrial motion to exclude eyewitness identification evidence, the state as the proponent of the eyewitness identification must establish all preliminary facts necessary to establish admissibility of the eyewitness evidence.”193 If the challenged eyewitness evidence implicates the Oregon equivalents of Federal Rules of Evidence 602194 and 701,195 the state must prove that the eyewitness has personal knowledge of the matter on which she will testify, and *340her identification “is both rationally based on [her] first-hand perceptions and helpful to the trier of fact.”196 This flips the burdens in due process cases such as Manson and Henderson. Rather than the defendant proving that the identification at issue is unreliable, the state must first prove that the identification meets the evidentiary requirements of Rules 602 and 701.

If the state successfully shows that the identification evidence is admissible, the burden then shifts to the defendant to establish that “the probative value of the evidence is substantially outweighed by the danger of unfair prejudice, confusion of the issues, misleading the jury, or by considerations of undue delay or needless presentation of cumulative evidence.”197 Thus, Oregon courts now rely on the state equivalent of Federal Rule of Evidence 403198 to exclude unreliable eyewitness identifications that are otherwise admissible. If a trial court concludes that the defendant has made such a showing, “the trial court can either-exclude the identification, or fashion an appropriate intermediate remedy short of exclusion to cure the unfair prejudice or other dangers attending the use of that evidence.”199

State courts are not alone in their responses to the scientific research. Federal circuit courts of appeals have also acknowledged the unreliability of certain eyewitness testimony.200 In United States v. Brownlee,201 we recognized the importance of expert testimony in safeguarding against unreliable eyewitness identifications. There, we held that a district court properly admitted expert testimony concerning the effects of race, hair covering, weapons focus, and exposure on the identification accuracy of multiple witnesses.202 We further held that the district court improperly excluded expert testimony comparing the show-up procedure used in that case (a procedure where law enforcement presents a single individual arguably fitting a witness’ description to that witness for identification) and other identification procedures and analyzing the suggestiveness of the show-up and its potential effect on the identifications. We also held that the district court improperly excluded expert testimony on confidence malleability, post-event suggestiveness, and confidence of accuracy.203 In doing so, we joined the growing chorus in acknowledging that

The recent availability of post-conviction DNA tests demonstrate that there have been an overwhelming number of false convictions stemming from uninformed reliance on eyewitness misidentifica-tions.... In fact, mistaken eyewitness identifications are responsible for more wrongful convictions than all other causes combined. Eyewitness evidence presented from well-meaning and confident citizens is highly persuasive but, at the same time, is among the least reliable forms of evidence.204

*341We then explained that expert testimony can play a crucial role in counteracting the falsely persuasive effect of unreliable eyewitness testimony.205 As the Nationál Research Council has recognized, expert testimony on eyewitness identifications may hold certain advantages over jury instructions as a method to explain the relevant science to juries.206 Expert witnesses: (1) “can explain scientific research in a more flexible manner, by presenting only the relevant research to the jury”; (2) are more “familiar with the research and can describe it in detail”; (3) “can convey the state of the research at the time of the trial”; (4) “can be cross-examined by the other side”; and (5) “can more clearly describe the limitations of the research.”207 Therefore, expert testimony on eyewitness accuracy is a crucial tool for educating juries on the science surrounding identifications.

It is against this backdrop that we must assess the jurors’ acceptance of the three eyewitness identifications of Dennis and the adequacy of the charge that guided their deliberations.

V. The Jury Charge

In Watkins v. Sowders, Justice Brennan wrote: “Surely jury instructions can ordinarily no more cure the erroneous admission of powerful identification evidence than they can cure the erroneous admission of a confession.”208 Although Justice Brennan was referring to the admissibility of certain eyewitness identifications rather than their reliability, his caution underscores the limited utility of a bare bones jury instruction that does not properly inform jurors about the many factors that can undermine courtroom identifications. This is particularly so given the powerful countervailing effect of jurors’ predisposition to believe eyewitness testimony.

Studies have documented that jurors tend to misunderstand how memory works and often believe it to be much more reliable and less susceptible to outside influence than it actually is.209 One survey of 1,000 potential jurors in Washington, D.C. found that almost two-thirds of the respondents thought the statement “I never forget a face” applied “very well” or “fairly well” to them210 Another thirty-seven percent thought the presence of a weapon would enhance the witness’ reliability, while thirty-three percent either believed that the weapon would have no effect or were unsure what effect the weapon would have.211 Finally, thirty-nine percent of respondents believed that when an event is violent, it makes a witness’ memory for details more reliable, while thirty-three percent responded either that this would have no effect or that they were unsure of the effect violence during the commission of the crime would have.212 The studies I have discussed show how wrong these beliefs are. There is no reason to believe the jurors who convicted Dennis were any more enlightened about memory formation and recall than the respondents in these studies.

*342Yet, the jurors who convicted James Dennis were only provided with a “plain vanilla” instruction. They had no knowledge of the potential distortion that can be caused by the factors discussed here. The trial court’s entire jury instruction regarding how the jurors should evaluate the eyewitness identifications was as follows:

There have been several Commonwealth identification witnesses.... However, a mistake can be made in identifying a person even by a witness attempting to be truthful.
Where the opportunity for positive identification is good and the witness is positive in his or her identification and his or her identification is not weakened by prior failure to identify but remains, even after cross-examination, positive and unqualified, the testimony as to identification need not be received with caution and can be treated as a statement of fact.
On the other hand, where a witness is not in a position to clearly observe the assailant or is not positive, as to identify, or his or her positive statements as to identity are weakened by qualification or by inconsistencies or by failure to identify the defendant on one or more prior occasions, then the testimony as to identification must be received with caution. You have heard the testimony in this case to the effect, and I leave it to your judgment and for your determination, but my recollection is that there were some prior identifications that were less than unqualified or positive. I think that’s been gone over at length by counsel. Under those circumstances, you should receive the testimony with caution. But it’s for you to determine whether or not this is so, you decide whether the testimony was weakened and what the evidence was.
If, according to these rules, you decide that caution is required in determining whether or not to accept the testimony of the identifying witnesses, then you must take into consideration the following matters: A, whether the testimony of the identification witness is generally believable; B, whether his or her opportunity to observe was sufficient to allow him or her to make an accurate identification; C, how the identification was arrived at; D, all of the circumstances indicating whether or not the identification was accurate; and E, whether the identification testimony is supported by other evidence. And you must conclude that it is so supported before you can accept it as being accurate.
My advice to you is this. In this case, my recollection, that’s why I’m not being so emphatic, my recollection is that one of the witnesses said, “I think[,]”[] another witness, for example, said, at a certain time, “I can’t be sure.” Witnesses who testified that way, their testimony as to identification should be received with 'caution and you should follow the rules that I’ve given you.213

Absent from this instruction is any explanation of the relevant system or estimator variables that so crucially impact the reliability of witness identifications. The caution the trial court urged is of precious little help given that omission. Jurors need to be informed of the applicable variables before they will be in a position to exercise the caution that this instruction urged. Without those detailed instructions, jurors simply are in no position to fully appreciate that “[t]he witness’ recollection of [a] stranger can be distorted easily by the circumstances or by later actions of the police.”

*343Moreover, as should be evident from my discussion, the italicized text instructing the jurors that they need not be cautious about accepting the identification of a witness who appears certain of her identification and had a good opportunity to observe the crime is extraordinarily dangerous. Contrary to the court’s instruction, that testimony cannot be accepted as fact. Social science aside, one need only consider the professed certainty of the accusers of Ronald Cotton and John White to understand just how problematic such a charge is. We again face a familiar and problematic reality: How ill-equipped these jurors were to assess the accuracy of the three eyewitnesses who pointed to Dennis and said “that’s the one.”

VI. Conclusion: Un-Ringing the Bell

In 1977, Justice Marshall emphasized that “ ‘the vagaries of eyewitness identification are well-known; the annals of criminal law are rife with instances of mistaken identification.’ ”214 They are known far better today. As Justice Marshall continued: “It is, of course, impossible to control one source of such errors[ — ]the faulty perceptions and unreliable memories of witnesses[ — jexcept through vigorously contested trials conducted by diligent counsel and judges.”215 Given the quantity and quality of research that has been conducted since Justice Marshall wrote those words, we judges must do a better job of educating ourselves and jurors about the dynamics of eyewitness identifications. Although no system so dependent on the limits of human abilities will ever be able to totally eliminate the problems endemic in eyewitness testimony, the integrity of the criminal justice system demands that we do better.

“[Jjurors seldom enter a courtroom with the knowledge that eyewitness identifications are unreliable. Thus, while science has firmly established the inherent unreliability of human perception and memory, this reality is outside the jury’s common knowledge and often contradicts jurors’ ‘commonsense’ understandings.”216 Therefore, thorough and appropriately focused jury instructions that reflect the scientific findings are critical to allowing jurors to discharge their solemn obligation to assess evidence.217 Such instructions will also encourage police to use more neutral procedures in investigating crimes. If law enforcement officials know that juries will be informed about best practices for obtaining identifications, police will have a very strong incentive to adopt protocols consistent with those best practices. As the National Research Council has explained, *344such instructions therefore “create an incentive for agencies to adopt written eyewitness identification procedures and to document the identifications themselves.”218

It is difficult to un-ring the bell that an unreliable eyewitness identification tolls. Therefore, in the first instance, it is law enforcement — not the courts — that can best ensure against an undue risk of convicting the innocent. However, robust jury instructions can minimize the dangers associated with inaccurate eyewitness identifications. In this case, had the jury been appropriately informed of the problems associated with the procedures used to solicit the identifications, as well as the numerous estimator variables that could have affected them, the jurors may well have concluded that James Dennis was not the one who shot Chedell Williams.

. Dissent at 357 (Fisher, J.).

. Id.

. State v. Henderson, 208 N.J. 208, 27 A.3d 872, 892 (2011), holding modified by State v. Chen, 208 N.J. 307, 27 A.3d 930 (2011); Charles A. Morgan III et al., Accuracy of Eyewitness Memory for Persons Encountered During Exposure to Highly Intense Stress, 27 Int’l J.L. & Psychiatry 265, 265 (2004).

. The Innocence Project, Reevaluating Lineups: Why Witnesses Make Mistakes and How to Reduce the Chance of a Misidentification 3 (2009), available at http://www.innocence project.org/wp-content/uploads/2016/05/ eyewitness_id_report-5.pdf; see also Brandon L. Garrett, Convicting the Innocent: Where Criminal Prosecutions Go Wrong 8-9, 279 (2011) (finding same in 190 of 250 DNA exoneration cases); Brief for Am. Psychol. Ass'n as Amicus Curiae Supporting Petitioner at 14-15, Perry v. New Hampshire, 132 S.Ct. 716 (2012) ("[Sjtudies have consistently found that the rate of inaccurate identifications is roughly 33 percent.”)..

. 60 • Minutes, Eyewitness: How Accurate is Visual Memory?, CBS News, Mar. 6, 2009, http://www.cbsnews.com/news/eyewitness-how-accurate-is-visual-memory.

. Id.

. Committee on Scientific Approaches to Understanding and Maximizing the Validity and Reliability of Eyewitness Identification in Law Enforcement and the Courts, Committee on Science, Technology, and Law, Committee on Law and Justice, Division of Behavioral and Social Sciences and Education, National Research Council, Identifying the Culprit: Assessing Eyewitness Identification 10 (2014).

. Id.

. Id.

. Id.

. Id.-, 60 Minutes, supra.

. National Research Council, Identifying the Culprit, supra, at 10.

.' Id.

. Jules Epstein, Eyewitnesses and Erroneous Convictions: An American Conundrum, in Controversies in Innocence Cases in America 41, 43 (Sarah Lucy Cooper ed., 2014).

. National Research Council, Identifying the Culprit, supra, at 10.

. Id.

. Epstein, supra, at 43.

. National Research Council, Identifying the Culprit, supra, at 10.

. 60 Minutes, supra.

. National Research Council, Identifying the Culprit, supra, at 10.

. Epstein, supra, at 43 (citation omitted).

. The Innocence Project, Reevaluating Lineups, supra, at 3.

. National Research Council, Identifying the Culprit, supra, at 11.

. National Research Council, Identifying the Culprit, supra, at 22 (citing Int’l Ass’n of Chiefs of Police, National Summit on Wrongful Convictions: Building a Systemic Approach to Prevent Wrongful Convictions (2013)).

. State v. Henderson, 208 N.J. 208, 27 A.3d 872, 879 (2011), holding modified by State v. Chen, 208 N.J. 307, 27 A.3d 930 (2011).

. J.A. 1495.

. J.A. 1496.

. J.A. 1494.

. Id.

. Id.

. J.A. 1493.

. Id.

. Id.

. J.A. 165.

. Id.

. J.A. 161.

. J.A. 1548.

. J.A. 1537.

. J.A. 1548.

. J.A. 1555.

. J.A. 1556.

. Id.

. J.A. 1565.

. Id.

. J.A. 1566.

. See infra Part III.A.4.

. J.A. 1576.

. J.A. 1581.

. See infra Part III.A.4.

. J.A. 226-27.

. J.A. 228-29.

. Felix Frankfurter, The Case of Sacco and Vanzetti: A Critical Analysis for Lawyers and Laymen 30 (Universal Library ed., 1962).

. Edwin M. Borchard, Convicting the Innocent; Sixty-Five Actual Errors of Criminal Justice (1932).

. 388 U.S. 218, 87 S.Ct. 1926, 18 L.Ed.2d 1149 (1967).

. Id. at 228, 87 S.Ct. 1926.

.See, e.g., Gary L. Wells, Nancy K. Steblay, & Jennifer E. Dysart, Double-Blind Photo Lineups Using Actual Eyewitnesses: An Experimental Test of a Sequential Versus Simultaneous Lineup Procedure, 39 L. & Hum. Behav. 1, 1 (2015); Laura Smalarz & Gary L. Wells, Contamination of Eyewitness Self-Reports and the Mistaken-Identification Problem, 24 Current Directions Psychol. 120, 120 (2015); Brian L. Cutler & Steven D. Penrod, Mistaken Identification: The Eyewitness, Psychology, and the Law (1995); Eyewitness Testimony: Psychological Perspectives (Gary L. Wells & Elizabeth A. Loftus eds., 1984).

. National Research Council, Identifying the Culprit, supra, at 14-15.

. See id. at 16, 72, 76.

. See State v. Henderson, 208 N.J. 208, 27 A.3d 872, 896-97 (2011), holding modified by State v. Chen, 208 N.J. 307, 27 A.3d 930 (2011); National Research Council, Identifying the Culprit, supra, at 24-25, 26.

. National Research Council, Identifying the Culprit, supra, at 25.

. Id. at 25.

. See Henderson, 27 A.3d at 896-97; National Research Council, Identifying the Culprit, supra, at 24-25, 26.

. Henderson, 27 A.3d at 896 (internal quotation marks omitted).

. Robert Rosenthal & Donald B. Rubin, Interpersonal Expectancy Effects: The First 345 Studies, 3 Behav. & Brain Sci. 377, 377 (1978).

. Id.

. Henderson, 27 A.3d at 896 (citing Ryann M. Haw & Ronald P. Fisher, Effects of Administrator-Witness Contact on Eyewitness Identification Accuracy, 89 J. Applied Psychol. 1106, 1107 (2004) and Steven E. Clark, Tanya E. Marshall, & Robert Rosenthal, Lineup Administrator Influences on Eyewitness Identification Decisions, 15 J. Experimental Psychol.: Applied 63, 66-73 (2009)).

. Linda Geddes, Fallible DNA Evidence Can Mean Prison or Freedom, 2773 The New Scientist: Special Report 1, 5 (2010).

. Id.

. Id.

. Id.

. Id.

. Id.

. Id.

. Id. at 6 (internal quotation marks omitted).

. 388 U.S. 218, 236-37, 87 S.Ct. 1926, 18 L.Ed.2d 1149 (1967).

. Id. at 235, 87 S.Ct. 1926 (internal alterations, quotation marks, and citation omitted).

. Id.

. See 60 Minutes, supra.

. Wade, 388 U.S. at 235, 87 S.Ct. 1926.

. Id.

. State v. Henderson, 208 N.J. 208, 27 A.3d 872, 897 (2011), holding modified by State v. Chen, 208 N.J. 307, 27 A.3d 930 (2011).

. See Steven E. Clark, A Re-examination of the Effects of Biased Lineup Instructions in Eyewitness Identification, 29 Law & Hum. Behav. 395, 418-20 (2005); Nancy M. Steblay, Social Influence in Eyewitness Recall: A Meta-Analytic Review of Lineup Instruction Effects, 21 Law & Hum. Behav. 283, 285-86, 294 (1997).

. See Roy S. Malpass & Patricia G. Devine, Eyewitness Identification: Lineup Instructions and the Absence of the Offender, 66 J. Applied Psychol. 482, 485 (1981).

. See Clark, Effects of Biased Lineup Instructions, supra, at 421; Steblay, Social Influence in Eyewitness Recall, supra, at 284.

. J.A. 1537 (emphasis added).

. J.A. 1548 (emphasis added).

. J.A. 1555 (emphasis added).

. See Roy S. Malpass, Colin G. Tredoux, & Dawn McQuiston-Surrett, Lineup Construction and Lineup Fairness, in 2 The Handbook of Eyewitnesses Psychology 155, 156-58 (2007).

. State v. Henderson, 208 N.J. 208, 27 A.3d 872, 898 (2011), holding modified by State v. Chen, 208 N.J. 307, 27 A.3d 930 (2011) (citing David F. Ross et al., When Accurate and Inaccurate Eyewitnesses Look the Same: A Limitation of the ‘Pop-Out’ Effect and the 10-to 12-Second Rule, 21 Applied Cognitive Psychol. 677, 687 (2007) and Gary L. Wells & Amy L. Bradfield, Measuring the Goodness of Lineups: Parameter Estimation, Question Effects, and Limits to the Mock Witness Paradigm, 13 Applied Cognitive Psychol. S27, S30 (1999)).

.Compare Steven E. Clark & Jennifer L. Tunnicliff, Selecting Lineup Foils in Eyewitness Identification Experiments: Experimental Control and Real-World Simulation, 25 L. & Hum. Behav. 199, 212 (2001), and Gary L. Wells, Sheila M. Rydell, & Eric P. Seelau, The Selection of Distractors for Eyewitness Lineups, 78 J. Applied Psychol. 835, 842 (1993), with Stephen Darling, Tim Valentine, & Ami-na Memon, Selection of Lineup Foils in Operational Contexts, 22 Applied Cognitive Psychol. 159, 165-67 (2008).

. See Nat’l Inst, of Justice, U.S. Dep't of Justice, Eyewitness Evidence: A Guide for Law Enforcement 29 (1999).

. Henderson, 27 A.3d at 898 (internal quotation marks omitted).

. National Research Council, Identifying the Culprit, supra, at 91 (citing Steven. E. Clark, Tanya E. Marshall, & Robert Rosenthal, Lineup Administrator Influences on Eyewitness Identification Decisions, 15 J. of Experimental Psychol.: Applied 63 (2009)).

. See Elizabeth F. Loftus, Leading Questions and the Eyewitness Report, 7 Cognitive Psy-chol. 560, 566 (1975).

. Id.

. Id.

. Id.

. Id.

. Amy B. Douglass & Nancy M. Steblay, Memory Distortion in Eyewitnesses: A Meta-Analysis of the Post-identification Feedback Effect, 20 Applied Cognitive Psychol. 859, 863 (2006).

. Id. at 864-65; see also Gary L. Wells & Amy L. Bradfield, "Good, You Identified the SuspectFeedback to Eyewitnesses Distorts Their Reports of the Witnessing Experience, 83 J. Applied Psychol. 360 (1998).

. See Gary L. Wells, Elizabeth A. Olson, & Steve D. Charman, Distorted Retrospective Eyewitness Reports as Functions of Feedback and Delay, 9 J. Experimental Psychol.: Applied 42, 49-50 (2003).

. See Jeffrey S. Neuschatz et al., The Effects of Post-Identification Feedback and Age on Retrospective Eyewitness Memory, 19 Applied Cognitive Psychol. 435, 449 (2005).

.See, e.g., Rachel Zajac & Nicola Henderson, Don’t It Make My Brown Eyes Blue: Co-Witness Misinformation About a Target’s Appearance Can Impair Target-Absent Line-up Performance, 17 Memory 266, 275 (2009) ("[Pjarticipants who were [wrongly] told by the [co-witness] that the accomplice had blue eyes were significantly more likely than control participants to provide this information when asked to give a verbal descrip- • tion.”); Lorraine Hope et al., "With a Little Help fi'om My Friends ... ”: The Role of Co-Witness Relationship in Susceptibility to Misinformation, 127 Acta Psychologica 476, 481 (2008) (noting that all participants “were susceptible to misinformation from their co-witness and, as a consequence, produced less accurate recall accounts than participants who did not interact with another witness”); Helen M. Paterson & Richard I. Kemp, Comparing Methods of Encountering Post-Event Information: The Power of Co-Witness Suggestion, 20 Applied Cognitive Psychol. 1083, *3271083 (2006) (“Results suggest that co-witness information had a particularly strong influence on eyewitness memory, whether encountered through co-witness discussion or indirectly through a third party.”); John S. Shaw III, Sena Garven, & James M. Wood, Co-Witness Information Can Have Immediate Effects on Eyewitness Memory Reports, 21 Law & Hum. Behav. 503, 503, 516 (1997) ("[W]hen participants received incorrect information about a co-witness’s response, they were significantly more likely to give that incorrect response than if they received no cohvitness information.”); C.A. Elizabeth Luus & Gary L. Wells, The Malleability of Eyewitness Confidence: Co-Witness and Perseverance Effects, 79 J. Applied Psychol. 714, 717-18 (1994).

. Elin M. Skagerberg, Co-Witness Feedback in Line-ups, 21 Applied Cognitive Psychol. 489, 494 (2007).

. Luus & Wells, The Malleability of Eyewitness Confidence, supra, at 717-18.

. Id.-, see also Skagerberg, supra, at 494-95 (showing similar results).

. J.A. 1556.

. Id.

. J.A. 1555.

. Kenneth A. Deffenbacher, Brian H. Born-stein, & Steven D. Penrod, Mugshot Exposure Effects: Retroactive Interference, Mugshot Commitment, Source Confusion, and Unconscious Transference, 30 L. & Hum. Behav. 287, 299 (2006).

. See Gunter Koehnken, Roy S. Malpass, Michael, S. Wogalter, Forensic Applications of Line-Up Research, in Psychological Issues in Eyewitness Identification 205, 219 (Siegfried L. Sporer, Roy S. Malpass, Gunter Koehnken eds., 1996).

. Id.

. Id. at 218. However, as noted earlier, Dennis’ picture was presented in photo arrays that witnesses saw prior to viewing the lineup.

. The Innocence Project, John Jerome White, http://www.innocenceproject.org/cases/ john-jerome-white/ (last visited July 5, 2016).

. Id.

. Id.

. Id.

. Gary Wells, The Mistaken Identification of John Jerome White, https://public.psych. iastate.edu/glwells/The_Misidentification_of_ John_White.pdf (last visited July 6, 2016).

. State v. Lawson, 352 Or. 724, 291 P.3d 673, 689 (2012).

. See State v. Henderson, 208 N.J. 208, 27 A.3d 872, 895 (2011), holding modified by State v. Chen, 208 N.J. 307, 27 A.3d 930 (2011); National Research Council, Identifying the Culprit, supra, at 1, 72, 92-93.

. See Charles A. Morgan III et al., Accuracy of Eyewitness Identification Is Significantly Associated with Performance on a Standardized Test of Face Recognition, 30 Int’l J.L. & Psychiatry 213 (2007); Kenneth A. Deffen-bacher et al., A Meta-Analytic Review of the *330Effects of High Stress on Eyewitness Memory, 28 L. & Hum. Behav. 687 (2004); Morgan et al., Accuracy of Eyewitness Memory, supra.

. See Charles A. Morgan III et al., Misinformation Can Influence Memory for Recently Experienced, Highly Stressful Events, 36 Int’l J.L. & Psychiatry 11, 15 (2013).

. Deffenbacher et al., Effects of High Stress, supra, at 699.

. Morgan et al., Accuracy of Eyewitness Memory, supra, at 266.

. Id. at 267-68.

. Id. at 268.

. Id.

. Id.

. Id. at 269-70.

. Id. at 272.

. Id.

. Id.

. Id. at 274.

. National Research Council, Identifying the Culprit, supra, at 93.

. Nancy K. Steblay, A Meta-analytic Review of the Weapon Focus Effect, 16 L. & Hum. Behav. 413, 415-17 (1992).

. Jonathan M. Fawcett et ah, Of Guns and Geese: A Meta-Analytic Review of the 'Weapon Focus’ Literature, Psychol., Crime & L. 1, 22 (2011).

. Kenneth A. Deffenbacher et al., Forget- ' ting the Once-Seen Face: Estimating the Strength of an Eyewitness’s Memory Representation, 14 J. Experimental Psychol.: Applied 139, 142 (2008); see also Carol Krafka & Steven Penrod, Reinstatement of Context in a Field Experiment on Eyewitness Identification, 49 J. Personality & Soc. Psychol. 58, 65 (1985) (finding a substantial increase in the misidentification rate in target-absent arrays from two to twenty-four hours after event).

. Deffenbacher et al., Forgetting the Once-Seen Face, supra, at 143.

. National Research Council, Identifying the Culprit, supra, at 99.

. Brian H. Bornstein et al., Effects of Expo- ■ sure Time and Cognitive Operations on Facial Identification Accuracy: A Meta-Analysis of Two Variables Associated with Initial Memory Strength, 18 Psychol., Crime & L. 473 (2012) (meta-analysis of the effect of exposure duration on facial identification accuracy); R.C.L. Lindsay et al., How Variations in Distance Affect Eyewitness Reports and Identification Accuracy, 32 Law & Hum. Behav. 526 (2008) (study of the effect of distance on identification accuracy).

. Race-bias — referring to the relative races of the witness and perpetrator — is another crucial estimator variable. Although this variable does not raise concerns here because the three eyewitnesses and the perpetrator were all Black, it is nevertheless worth noting because it again shows the extent to which circumstances (other than opportunity to observe) can greatly impact the reliability of an eyewitness identification. Research has thoroughly documented a phenomenon known as “own-race bias’’ wherein people more accurately identify faces within their own race as compared to those of members of a different racial group. See National Research Council, Identifying the Culprit, supra, at 96; Roy S. Malpass & Jerome Kravitz, Recognition for Faces of Own and Other Race, 13 J. Personality & Soc. Psychol. 330 (1969). The Innocence Project analyzed 297 DNA exonerations and found that a cross-racial misidentification occurred in forty-two percent of the cases in which an erroneous eyewitness identification was made. Edwin Grimsley, What Wrongful Convictions Teach Us about Racial Inequality, The Innocence Project (Sept. 26, 2012), http:// www.innocenceproject.org/Content/What_ WrongfuLConvictions_Teach_Us_About_ RaciaLInequality .php.

.In fact, one eyewitness — Joseph DiRienzo Jr. — described the shooter's height in terms of his own height: “about my height, about 5'9"." J.A. 1649.

. Dissent at 358 (Fisher, J.) (citing Christian A. Meissner, Siegfried L. Sporer, & Jonathan W. Schooler, Person Descriptions as Eyewitness Evidence, in 2 Handbook of Eyewitness Psychology 3, 8 (Rod C.L. Lindsay et al. eds., 2007) and Rhona H. Flin & John W. Shepherd, Tall Stories: Eyewitnesses’ Ability to Estimate Height and Weight Characteristics, 5 Hum. Learning 29, 34 (1986)).

. Meissner, Sporer, & Schooler, Person Descriptions as Eyewitness Evidence, supra, at 8 (citing the Flin and Shepherd study); Flin & Shepherd, Tall Stories, supra, at 36.

. Flin & Shepherd, Tall Stories, supra, at 36.

. Dissent at 357 (Fisher, J.).

. Id.

. Id. at 358 (citing Arizona v. Youngblood, 488 U.S. 51, 72 n.8, 109 S.Ct. 333, 102 L.Ed.2d 281).

. Morgan et al., Accuracy of Eyewitness Memory, supra, at 268.

. The fact that Overstreet and other non-identifying witnesses could theoretically have been called by defense counsel is no answer. No defense attorney in her right mind would put such witnesses on the stand, knowing that the witnesses had seen photographs of the defendant and would know the person sitting at counsel table was the person the police had arrested for the crime. A criminal justice system seeking fairness and justice should not countenance the creation of such an absurd dilemma.

. See Maj. Op. at 272-73, 274-75, 275-76.

. 432 U.S. 98, 97 S.Ct. 2243, 53 L.Ed.2d 140 (1977).

. Id. at 106, 97 S.Ct. 2243 (internal quotation marks omitted).

. 409 U.S. 188, 93 S.Ct. 375, 34 L.Ed.2d 401 (1972).

. Id. at 199-200, 93 S.Ct. 375; Manson, 432 U.S. at 114, 97 S.Ct. 2243.

. Manson, 432 U.S. at 114, 97 S.Ct. 2243.

. Id.

. State v. Henderson, 208 N.J. 208, 27 A.3d 872, 892 (2011), holding modified by State v. Chen, 208 N.J. 307, 27 A.3d 930 (2011); Morgan et al., Accuracy of Eyewitness Memory, supra, at 265.

. National Research Council, Identifying the Culprit, supra, at 6.

. -U.S.-, 132 S.Ct. 716, 181 L.Ed.2d 694 (2012).

. Id. at 721-22.

. Id. at 722-23.

. Id. at 730.

. Id. at 726 (emphasis added).

. Id. at 727 ("As one of Perry’s amici points out, many other factors bear on "the likelihood of misidentification,” — for example, the passage of time between exposure to and identification of the defendant, whether the witness was under stress when he first encountered the suspect, how much time the witness had to observe the suspect, how far the witness was from the suspect, whether the suspect carried a weapon, and the race of the suspect and the witness.” (internal citation omitted)).

. Id. at 728-29 (internal footnote omitted).

. Id. at 729.

. Id. at 727 ("To embrace Perry's view would thus entail a vast enlargement of the reach of due process as a constraint on the admission of evidence.”).

. 208 N.J. 208, 27 A.3d 872 (2011), holding modified by State v. Chen, 208 N.J. 307, 27 A.3d 930 (2011).

. See id. at 918; State v. Madison, 109 N.J. 223, 536 A.2d 254, 258-59 (1988) holding modified by State v. Henderson, 208 N.J. 208, 27 A.3d 872 (2011).

. Henderson, 27 A.3d at 878.

. Id. at 920.

. Id. The New Jersey Supreme Court instructed courts to consider the following non-exhaustive list of system variables when deciding whether to hold a pre-trial hearing:

1. Blind Administration. Was the lineup procedure performed double-blind? If double-blind testing was impractical, did the police use a technique like the "envelope method” described above, to ensure that the administrator had no knowledge of where the suspect appeared in the photo array or lineup?
2. Pre-identification Instructions. Did the administrator provide neutral, pre-identifi-cation instructions warning that the suspect may not be present in the lineup and that the witness should not feel compelled to make an identification?
3. Lineup Construction. Did the array or lineup contain only one suspect embedded among at least five innocent fillers? Did the suspect stand out from other members of the lineup?
4. Feedback. Did the witness receive any information or feedback, about the suspect or the crime, before, during, or after the identification procedure?
5. Recording Confidence. Did the administrator record the witness' statement of confidence immediately after the identification, before the possibility of any confirmatory feedback?
6. Multiple Viewings. Did the witness view the suspect more than once as part of multiple identification procedures? Did police use the same fillers more than once?
7. Showups. Did the police perform a show-up more than two hours after an event? Did the police warn the witness that the suspect may not be the perpetrator and that the witness should not feel compelled to make an identification?
8. Private Actors. Did law enforcement elicit from the eyewitness whether he or she had spoken with anyone about the identification and, if so, what was discussed?
9. Other Identifications Made. Did the eyewitness initially make no choice or choose a different suspect or filler?

Id. at 920-21.

. Id. at 920.

. The New Jersey Supreme Court told courts to consider the following, non-exhaustive list of estimator variables in assessing the reliability of an eyewitness identification:

1. Stress. Did the event involve a high level of stress?
2. Weapon focus. Was a visible weapon used during a crime of short duration?
3. Duration. How much time did the witness have to observe the event?
4. Distance and Lighting. How close were the witness and perpetrator? What were the lighting conditions at the time?
5. Witness Characteristics. Was the witness under the influence of alcohol or drugs? Was age a relevant factor under the circumstances of the case?
6. Characteristics of Perpetrator. Was the culprit wearing a disguise? Did the suspect have different facial features at the time of the identification?
7. Memory decay. How much time elapsed between the crime and the identification?
8. Race-bias. Does the case involve a cross-racial identification?
9. Opportunity to view the criminal at the time of the crime.
10. Degree of attention.
11. Accuracy of prior description of the criminal.
*33812. Level of certainty demonstrated at the confrontation.
Did the witness express high confidence at the time of the identification before receiving any feedback or other information?
13. The time between the crime and the confrontation. (Encompassed fully by "memory decay” above.)

Id. at 921-22.

. Id. at 920.

. Id.

. Id. at 878.

. Id.

. Id.

. Id.

. Id. at 922.

. Id. at 924.

. These instructions were released a year after the opinion in Henderson.

. Supreme Court of New Jersey, New Jersey Criminal Model Jury Instmctions, Identification: In-Court Identifications Only 2 (2012), http://www.judiciary.state.nj.us/pressrel/2012/ jury_instruction.pdf.

. Id. at 3-9.

. New Jersey is not alone in its response to the vast body of research on the reliability of eyewitness identifications. In 2011, the Justices of the Massachusetts Supreme Judicial Court convened a study group to "offer guidance as to how our courts can most effectively deter unnecessarily suggestive identification procedures and minimize the risk of a wrongful conviction.” Massachusetts Supreme Judicial Court Study Group on Eyewitness Evidence, Report and Recommendations to the Justices 1 (2013) (internal quotation marks omitted). The report made five recommendations aimed at minimizing misidentifications: (1) acknowledge variables affecting identification accuracy; (2) develop a model policy and *339implement best practices for police departments; (3) expand use of pretrial hearings; (4) expand use of improved jury instructions; and (5) offer continuing education to judges and bar leaders. Id. at 2-5. Like Henderson, the Massachusetts report recommended .that, when a defendant contests the reliability of an eyewitness identification, the trial judge should conduct a pretrial hearing to determine whether law enforcement used suggestive identification procedures to elicit that identification. Id. at 109-16. If a suggestive procedure was used, the report recommended that courts assess whether those procedures impacted the reliability of the identification. Id. at 111. The report suggested that courts consider both estimator and system variables in pre-trial hearings. Id.

. 352 Or. 724, 291 P.3d 673 (2012).

. See id. at 688; State v. Classen, 285 Or. 221, 590 P.2d 1198 (1979).

. Lawson, 291 P.3d at 688-89 (citing Perry v. New Hampshire, - U.S. -, 132 S.Ct. ' 716, 730, 181 L.Ed.2d 694 (2012) ("[T]he Due Process Clause does not require a preliminary judicial inquiry into reliability of an eyewitness identification when the identification was not procured under unnecessary suggestive circumstances arranged by law enforcement.”)).

. Id. at 696-97 (emphasis added).

. Fed. R. Evid. 602 ("A witness may testify to a matter only if evidence is introduced sufficient to support a finding that the witness has personal knowledge of the matter. Evidence to prove personal knowledge may consist of the witness’s own testimony.”).

. Fed. R. Evid. 701 ("If a witness is not testifying as an expert, testimony in the form of an opinion is limited to one that is: (a) rationally based on the witness’s perception; (b) helpful to clearly understanding the witness’s testimony or to determining a fact in issue; and (c) not based on scientific, technical, or other specialized knowledge within the scope of Rule 702.”).

. Lawson, 291 P.3d at 697.

. Id.

. Fed. R. Evid. 403 ("The court may exclude relevant evidence if its probative value is substantially outweighed by a danger of one or more of the following: unfair prejudice, confusing the issues, misleading the jury, undue delay, wasting time, or needlessly presenting cumulative evidence.”).

. Lawson, 291 P.3d at 697.

. See, e.g., United States v. Bartlett, 567 F.3d 901, 906 (7th Cir. 2009), cert. denied, 558 U.S. 1147, 130 S.Ct. 1137, 175 L.Ed.2d 971 (2010); United States v. Brownlee, 454 F.3d 131, 141-44 (3d Cir. 2006).

. 454 F.3d 131 (2006).

. Id. at 137.

. Id. at 141.

. Id. at 141-42 (internal quotation marks, citations, and alterations omitted).

. See id. at 144.

. National Research Council, Identifying the Culprit, supra, at 40.

. Id.

. 449 U.S. 341, 350, 101 S.Ct. 654, 66 L.Ed.2d 549 (1981) (Brennan, J. dissenting) (emphasis added).

. Epstein, supra, 46-48; Elizabeth F. Loftus, Timothy P. O'Toole, & Catharine F. Easterly, Juror Understanding of Eyewitness Testimony: A Survey of 1000 Potential Jurors in the District of Columbia 1 (2004).

. Loftus, O’Toole, & Easterly, supra, at 6.

. Id. at 8.

. Id. at 9.

. J.A. 1237-39.

. Manson v. Brathwaite, 432 U.S. 98, 119, 97 S.Ct. 2243, 53 L.Ed.2d 140 (1977) (Marshall, J., dissenting) (internal alteration omitted) (quoting United States v. Wade, 388 U.S. 218, 228, 87 S.Ct. 1926, 18 L.Ed.2d 1149 (1967)).

. Id.

. United States v. Brownlee, 454 F.3d 131, 142 (3d Cir. 2006) (internal quotation marks and citations omitted).

. It is important to note that jury instructions are only one of several promising remedies. As we mentioned in our discussion of Brownlee, expert testimony regarding the reliability of eyewitness identifications can also help jurors accurately assess the reliability of such identifications. The National Research Council has also recommended that, where appropriate, trial judges make basic inquiries into eyewitness identification evidence. National Research Council, Identifying the Culprit, supra, at 109-10. As the National Research Council suggested, “while the contours of such an inquiry would need to be established on a case-by-case basis, at a minimum, the judge could inquire about prior lineups, what information had been given to the eyewitness before the lineup, what instructions had been given to the eyewitness in connection with administering the lineup, and whether the lineup had been administered ‘blindly.’ ” Id. at 110.

. Id. at 110.






Concurrence Opinion

AMBRO, Circuit Judge,

joins in this concurring opinion.

APPENDIX: Eyewitness Identifications

*345Table 1 — Photo Array Identifications

[[Image here]]

*346Table 2 — Lineup Identifications

[[Image here]]






Concurrence Opinion

JORDAN, Circuit Judge,

concurring in part and concurring in the judgment:

To say this case is troubling is a serious understatement. James Dennis was convicted of murder and sentenced to death based almost entirely upon the testimony of three problematic eyewitnesses and despite a dearth of physical evidence. On direct appeal, the Pennsylvania Supreme Court affirmed his conviction and death sentence in an opinion that is no credit to that court’s usual standards. See Dennis I, 552 Pa. 331, 715 A.2d 404 (1998). It rejected in a mere three sentences Dennis’s Brady claim with respect to the Cason receipt, a piece of evidence thoroughly described in today’s Majority opinion. Here is the entirety of the state court’s analysis:

Finally, it is clear that there clearly was no Brady violation. The DPW receipt was not exculpatory, because it had no bearing on Appellant’s alibi, and there is no evidence that the Commonwealth withheld the receipt from the defense. Accordingly, Appellant’s claims of ineffectiveness regarding Cason and the DPW receipt have no arguable merit.

Id. at 408.

Perhaps the most remarkable aspect of that drive-by discussion is the assertion that the Cason receipt was not exculpatory because “it had no bearing on [Dennisj’s alibi.” Id. In reality, the pertinence and importance of the receipt could not be more glaring. It shows exactly what time witness Latanya Cason received her public assistance check, thus shifting the timeline of events that she laid out during her trial testimony so that, instead of contradicting Dennis’s testimony, she almost perfectly corroborated his alibi. The previously-undisclosed receipt thus transforms Cason from a damning prosecution witness into a powerful witness for the defense.

Every judge of our en banc Court has now concluded that the Pennsylvania Supreme Court’s contrary determination was not only wrong, but so obviously wrong that it cannot pass muster even under AEDPA’s highly-deferential standard of review. In other words, it is the unanimous view of this Court that any fairminded jurist must disagree with the Dennis I court’s assessment of the materiality and favorability of the Cason receipt. Yet somehow a majority of the Pennsylvania Supreme Court endorsed Dennis’s conviction and death sentence. The lack of analytical rigor and attention to detail in that decision on direct appeal is all the more painful to contemplate because the proof against Dennis is far from overwhelming. He may be innocent.

*347But the strength of the case against James Dennis need not be the focus of our attention. This case can and should be resolved on a single point: the Brady claim concerning the Cason receipt. That is one reason why I cannot join the more expansive opinion of my colleagues in the Majority. Their correct conclusion that the error in Dennis I regarding the Cason receipt is by itself enough to warrant habeas relief means that we have no call to address the Brady claims with respect to the Howard police activity report and the Frazier documents. And, in fact, I disagree with the' Majority’s analysis of those latter two claims and fully agree with my dissenting colleagues’ rejection of them, which is another reason I cannot join the Majority opinion.

Moreover, I also agree with the Dissent’s position,1 set forth in its discussion of the Cason receipt, that imposing a “reasonable diligence” requirement upon defense counsel does not violate a clearly established holding of the Supreme Court. The “reasonable diligence” requirement is, in effect, a rule that a Brady claim will not lie when the evidence in question was available to the defense by the exercise of reasonable diligence. E.g., Brown v. Cain, 104 F.3d 744, 750 (5th Cir. 1997). We are obligated by AEDPA to uphold a state court’s decision unless it is “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1). Under AEDPA, whether any of us thinks that imposing a reasonable diligence requirement is a good idea or the best interpretation of Brady is irrelevant. What matters is that one can reasonably perceive such a requirement being allowed by Supreme Court jurisprudence.2 We ourselves have applied it repeatedly,3 so we can hardly *348say that it constitutes an unreasonable application of federal law.

Of course, the Pennsylvania Supreme Court never said anything at all in its Dennis I decision about defense counsel’s lack of diligence in locating the Cason receipt. But, under Harrington v. Richter, habeas review requires that we engage in so-called “gap-filling,” and apply AEDPA deference to whatever reasonable “arguments or theories ... could have supported[ ] the state court’s decision,” if that decision does not provide reasoning for its conclusions.4 562 U.S. 86, 102, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011). Thus, despite the fact that the Pennsylvania Supreme Court never itself discussed diligence, Richter might prompt us to apply a reasonable diligence requirement and reject Dennis’s Cason receipt Brady claim — exactly as the Dissent has suggested — if there were a gap in the state-court decision for us to fill. The problem I have with the Dissent is that I see no gap in the state court’s reasoning, at least not in the sense contemplated in Richter. My dissenting colleagues are not filling a gap here; they are rewriting the opinion of the Pennsylvania Supreme Court, adding and then elaborating a theory that was never litigated in state court.

The reality of what happened in Dennis I is more straightforward. The Pennsylvania Supreme Court simply erred. Its opinion stated both that “the police came into possession of’ the Cason receipt and that “there [was] no evidence that the Commonwealth withheld the receipt from the defense.” Dennis I, 715 A.2d at 408. There was, however, no recognition that those statements are fundamentally at odds. Under the Supreme Court’s opinion in Kyles v. Whitley, any evidence in the possession of the police is, for Brady purposes, also in the possession of the prosecution. 514 U.S. 419, 437, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995). If a piece of favorable, material evidence is in the possession of the police but is not turned over to the defense, it is necessarily withheld by the prosecution in violation of Brady. See id. (prosecutors are responsible for “any favorable evidence known to the others acting on the government’s behalf in the case, including the police”).

By entirely failing to apply Kyles, the Pennsylvania Supreme Court acted “contrary to ... clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1).5 In light of the state court’s error, I would review Dennis’s Brady claim with respect to the Cason receipt “unencumbered by the deference AEDPA normally requires,” Panetti v. Quarterman, 551 U.S. 930, 948, 127 S.Ct. 2842, 168 L.Ed.2d 662 (2007), to determine whether Dennis is “in custody in violation of the Constitution or laws ... of the United *349States.” 28 U.S.C. § 2254(a).6 On that de novo review, I would hold that the evidence in question meets all three requirements of Brady — the Cason receipt is material and favorable, and it was suppressed by the Commonwealth — for the reasons set out in Part III.A of the Majority opinion. I therefore concur in the judgment. I also agree with Part II of the Majority opinion and write separately to explain my view of the limits of Richter gap-filling and the proper scope of AEDPA deference.

Recall that in Dennis I, the Pennsylvania Supreme Court said, “there is no evidence that the Commonwealth withheld the [Cason] receipt from the defense.” 715 A.2d at 408. My dissenting colleagues believe “it is not clear what the court meant by [that].” (J. Fisher Dissent Op. at 360.) They then proceed to fill the “gap” they think is created by the ambiguity they perceive, saying, “the Pennsylvania Supreme Court could have meant that the receipt was not withheld because it was available to the defense with reasonable diligence.” (J. Fisher Dissent Op. at 362.)

The precedent that establishes a gap-filling requirement, Richter, dealt with a state court decision that was unsupported by any reasoning. 562 U.S. at 96-97, 131 5.Ct. 770. The state court issued a summary order, with no written opinion, denying a prisoner’s ineffective assistance of counsel claim. Id. The gap in the state court’s reasoning was obvious — there was no reasoning at all. The Supreme Court held that, even in those circumstances, “[w]here a state court’s decision is unaccompanied by an explanation, the habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief.” Id. at 98, 131 S.Ct. 770. Thus federal courts must fill gaps in a state court’s reasoning so that there is something against which to measure a petitioner’s efforts. In short, “a habe-as court must determine what arguments or theories ... could have supported[ ] the state court’s decision” and afford AEDPA deference to those theories. Id. at 102, 131 S.Ct. 770.

Premo v. Moore extended RichtePs gap-filling directive a bit beyond cases devoid of all reasoning. 562 U.S. 115, 131 S.Ct. 733, 178 L.Ed.2d 649 (2011). There, a prisoner claimed ineffective assistance of counsel because his attorney had failed to file a motion to suppress a confession. Id. at 119, 131 S.Ct. 733. In concluding that such a motion “would have been fruitless,” id. the state court’s opinion expressly referenced *350trial counsel’s explanation that “suppression would serve little purpose” because the defendant had made full and admissible confessions to others. Id. at 123, 131 S.Ct. 733. The state court did not, however, specify which of the two prongs of the ineffective assistance of counsel standard from Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) — deficient performance or prejudice — formed the basis of its rejection of the claim. Premo, 562 U.S. at 123, 131 S.Ct. 733. The Supreme Court therefore held that the Court of Appeals for the Ninth Circuit had to fill that gap by assuming “that both findings would have involved an unreasonable application of clearly established federal law.” Id. Critical to the ultimate denial of habeas relief, the Supreme Court believed that the state court’s justification for rejecting the petitioner’s claim was sufficient to address either prong of Strickland.7 Accordingly, the Supreme Court’s decision was not an exercise in speculation but was rooted in the state court’s actual reasoning. Tremo did not require consideration of an entirely new argument that had not already been identified and accepted by the state court. See id. at 124, 131 S.Ct. 733 (“[T]he [state court’s] first and independent explanation — that suppression would have been futile — confirms that [counsel’s] representation was adequate under Strickland.”). The “gap” that the Court filled was thus quite narrow.

The very next year, the Supreme Court put a limit on gap-filling. In Lafler v. Cooper, it upheld a grant of habeas corpus. — U.S. -, 132 S.Ct. 1376, 182 L.Ed.2d 398 (2012). The petitioner, Anthony Cooper, had shot at a woman’s head but missed, instead hitting her in the buttock, hip, and abdomen. Id. at 1383. The prosecution offered Cooper two plea deals, and Cooper expressed interest. Id. He ended up rejecting the offers, though, because (he later alleged) his attorney convinced him that the prosecution would be unable to establish intent to murder because he shot his victim below the waist. Id. After he was convicted on all charges, Cooper claimed ineffective assistance of counsel. Id. The Michigan Court of Appeals rejected his claim, analyzing it as follows:

[T]he record shows that defendant knowingly and intelligently rejected two plea offers and chose to go to trial. The record fails to support defendant’s contentions that defense counsel’s representation was ineffective because he rejected a defense based on [a] claim of self-defense and because he did not obtain a more favorable plea bargain for defendant.

People v. Cooper, No. 250583, 2005 WL 599740, at *1 (Mich. Ct. App. Mar. 15, 2005) (per curiam) (internal citations omitted). After the district court granted Cooper’s petition for habeas relief, the Sixth Circuit affirmed, emphasizing the problem in the state court’s decision with this comment: “it is not clear from the [state] court’s abbreviated discussion (only two sentences of the opinion is even arguably responsive to petitioner’s claim) what the court decided, or even whether the correct *351legal rule was identified.” Cooper v. Lafler, 376 Fed.Appx. 563, 568-69 (6th Cir. 2010), vacated by — U.S.-, 132 S.Ct. 1376, 182 L.Ed.2d 398 (2012).

While it ultimately affirmed the habeas decision, the Supreme Court concluded that the state court’s two-sentence analysis “may not be quite so opaque as the Court of Appeals for the Sixth Circuit thought.... ” Lafler, 132 S.Ct. at 1390. The state court had identified Cooper’s ineffective-assistance-of-counsel claim, but had failed to apply the proper Strickland standard to assess it. Instead, the state court had “simply found that respondent’s rejection of the plea was knowing and voluntary.” Id. Although the Michigan court recited the Strickland standard, the Supreme Court concluded that the state court had mistakenly relied upon an entirely different standard (ie., the “knowing and voluntary” standard), which was contrary to Strickland. By relying upon the wrong standard altogether, “the state court’s adjudication was contrary to clearly 'established federal law.” Id. As a consequence, the Supreme Court declined to apply AEDPA deference to the state court decision and, instead, engaged in de novo review of Cooper’s Strickland claim, concluding that his counsel’s deficient performance and the prejudice therefrom required relief. Id. at 1390-91. The Supreme Court’s analysis in Lafler suggests that we should be hesitant to deem a state court opinion to be so lacking in analysis that it is comparable to an “order ... unaccompanied by an opinion explaining [its] reasons.” Richter, 562 U.S. at 98, 131 S.Ct. 770. In other words, we ought not engage in error correction under the guise of gap-filling.

That holds true here. In Dennis I, the Pennsylvania Supreme Court correctly identified Brady and its requirement that, for relief to be warranted, the evidence in question must be both exculpatory and withheld. Nevertheless, the court applied a standard contrary to Brady and its progeny when it concluded that the prosecution did not withhold evidence that the police had in their possession. Cf. Sears v. Upton, 561 U.S. 945, 952, 130 S.Ct. 3259, 177 L.Ed.2d 1025 (2010) (per curiam) (“Although the court appears to have stated the proper ... standard, it did not correctly conceptualize how that standard applies to the circumstances of this case.”). Kyles is very clear in explaining that, for purposes of a Brady analysis, the prosecution functionally possesses all favorable evidence in the possession of the police. See 514 U.S. at 437, 115 S.Ct. 1555 (“[T]he individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government’s behalf in the case, including the police.”). Just as the Michigan state court in Lafler failed to apply Strickland to assess an ineffective assistance claim, so too the Pennsylvania Supreme Court failed to apply Kyles to assess Dennis’s Brady claim with respect to the Cason receipt. Rather than applying Kyles, the court simply found that there was no evidence that the prosecutor possessed the Cason receipt. Compare Lafler, 132 S.Ct. at 1390 (“Rather than applying Strickland, the state court simply found that respondent’s rejection of the plea was knowing and voluntary. An inquiry into whether the rejection of a plea is knowing and voluntary, however, is not the correct means by which to address a claim of ineffective assistance of counsel.”). Lafler implies a limit on the gap-filling called for by Richter and Premo. As was done in Lafler, we should take the state court’s decision as written, rather than construct our own “not unreasonable” theory to justify that court’s conclusion.

Justice Scalia’s dissent in Lafler further supports the analogy between that case and this one. Indeed, his opinion reads much like the Dissent here. First, he pointed out that the Michigan state court *352had recited the Strickland standard. Lafler, 132 S.Ct. at 1396 (Scalia, J., dissenting). He next read the subsequent paragraph of the state court’s decision as an attempt to apply that standard. Id. He then concluded that the state court did not apply a standard “contrary to” federal law. Instead, by direct analogy to Premo, he argued that his colleagues should have assessed whether the state court opinion constituted an unreasonable application of clearly established federal law, subject to Richter’s gap filling requirement:

Since it is ambiguous whether the state court’s holding was based on a lack of prejudice or rather the court’s factual determination that there had been no deficient performance, to provide relief under AEDPA this Court must conclude that both holdings would have been unreasonable applications of clearly established law.

Id. Justice Scalia’s effort to salvage the state court decision in Lafler provides some support for the Dissent’s approach here. But Justice Scalia was himself writing a dissent. Had the Supreme Court wanted us to save every problematic state court opinion by gap-filling and application of AEDPA deference, Justice Scalia’s opinion would have been the majority position.

I can discern no ambiguity in the Pennsylvania Supreme Court’s Brady analysis regarding the Cason receipt. The Dennis I opinion is clear about it. Very brief and very wrong, but clear. The analysis under the suppression prong of Brady can be distilled from two sentences of the opinion. First, the court says, “During their investigation ... the police came into possession of’ the Cason receipt.8 Dennis I, 715 A.2d at 408. It then says, “there is no evidence that the Commonwealth withheld the receipt from the defense.” Id. If one follows the instruction of Kyles, those two statements are impossible to harmonize. But if one ignores Kyles and assumes there exists some dividing line between the police and the prosecution, the court’s reasoning is plain. To the Pennsylvania Supreme *353Court, the fact that the police had the receipt does not mean that the Common-ivealth had the receipt, and thus the Commonwealth did not suppress what it did not have. There is no hint that “reasonable diligence” was part of the analysis. The Commonwealth did not advance a reasonable diligence argument,9 nor did the court reference a diligence requirement anywhere in its opinion. In failing to apply Kyles, the state court’s opinion was “contrary to” and “involved an unreasonable application of[ ] clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1).

My dissenting colleagues treat the contradictory sentences in Dennis I like a “Magic Eye” image, staring past the obvious error until the illusion of a fillable gap materializes. They do so, I assume, because it is. hard to accept that a court would make such a clear error of law: How could the state court possibly have concluded both that the police possessed the receipt and that the prosecution did not withhold it? That conclusion makes absolutely no sense if one assumes the state court knew of and applied Kyles. See Lopez v. Schriro, 491 F.3d 1029, 1046 (9th Cir. 2007) (Thomas, J., concurring in part and dissenting in part) (noting that we start with the “presumption that state judges know and follow the law”). But state courts, just like us, do sometimes err. And when they do, we are not free to label significant errors as “gaps” to be corrected under Richter and Premo.

Limiting our habeas review to the actual, expressed reasoning of a state court is itself a form of deference. The principles of comity and federalism underlying AED-PA’s highly-deferential standard compel us to acknowledge the state court’s reasoning if we can fairly discern it. See Ylst v. Nunnemaker, 501 U.S. 797, 803, 111 S.Ct. 2590, 115 L.Ed.2d 706 (1991) (describing an “unexplained” state-court order as one from which that court’s rationale is “undis-coverable”).10 We would do real damage to those principles were we to begin re-writing state court opinions to save them. Sometimes what appears to be a fundamental misstep is exactly that. Since the passage of AEDPA, the narrow purpose of federal habeas review has been to address just such missteps. See Richter, 562 U.S. at 103, 131 S.Ct. 770 (“As a condition for obtaining habeas corpus from a federal court, a state prisoner must show that the state court’s ruling on the claim being presented in federal court was so lacking in justification that there was an error well understood and comprehended in existing law beyond any possibility for fairminded disagreement.”).

There is yet another reason to think that Dennis I presents nothing more complicat*354ed than a Kyles error: the Commonwealth advocated it. Before the Pennsylvania Supreme Court, the Commonwealth advanced the incorrect theory that it was not required to turn over favorable evidence in the possession of the police. It emphasized that, “even though Cason claims in her affidavit that [the receipt] was taken by the police,” the failure to produce that document could not constitute a Brady violation because “there [wa]s no reason to believe it was in the Commonwealth’s possession to be produced.” (App. 2026.) That argument presupposes, contrary to Kyles, that there exists a divide between discoverable evidence taken by the police and discoverable evidence in the prosecutor’s case file.

At the time, that argument may have had some basis in Pennsylvania law, although it was already untenable because of Kyles. In 1995, when Kyles was decided, the Pennsylvania rules governing discovery and evidence disclosure were not based on the premise that evidence possessed by the police is possessed by the prosecution. See Pa. R. Crim. Pro. 305B (Repealed) (requiring mandatory disclosure of evidence favorable to the accused only when it “is within the possession or control of the attorney for the Commonwealth”). Even after Kyles was decided, the Pennsylvania Superior Court continued to hew to the outmoded state-law rule. See Commonwealth v. McElroy, 445 Pa.Super. 336, 665 A.2d 813, 819 (1995). The Pennsylvania Supreme Court likewise continued to apply its discovery rules as written. See Commonwealth v. Gribble, 550 Pa. 62, 703 A.2d 426, 435-36 (1997). It did not explicitly abrogate the faulty state rule of discovery until 2001. See Commonwealth v. Burke, 566 Pa. 402, 781 A.2d 1136, 1142 (2001). Dennis I was decided in 1998. Thus,'the court was not leaving agap in its Dennis I opinion. It was accepting the Commonwealth’s unsound argument, and it practically said so.

The wisdom of Richter gap-filling is open to reasonable criticism. A widely respected judge has expressed the view that gap-filling is unfair and incentivizes unreasoned decisions; it is a perspective that my colleague Judge Hardiman evidently shares, as described in his Dissent. See Mann v. Ryan, 774 F.3d 1203, 1225 (9th Cir. 2014) (Kozinski, J., concurring in part and dissenting in part) (Richter “has'the perverse effect of encouraging state courts to deny relief summarily, to insulate their orders from tinkering by the federal courts.”), on reh’g en banc, No. 09-99017, 828 F.3d 1143, 2016 WL 3854234 (9th Cir. July 15, 2016). Given those criticisms, it has been suggested that we should engage in Richter gap-filling, and thus apply AEDPA deference, even when a state court does give a reasoned basis for its conclusions. See id. at 1224 (Kozinski, J., concurring in part and dissenting in part) (“After Richter, it seems clear that we should assess the reasonableness of a state court’s decision, not its reasoning.”). Judge Hardiman would follow that approach here. (See J. Hardiman Dissent Op. at 370 (“I would hold that regardless of the thoroughness — or even the correctness — of the Pennsylvania Supreme Court’s stated reasoning, its judgment may not be upset so long as its decision did not contravene or unreasonably apply clearly established federal law....”).) And, indeed, his approach may have some appeal as a matter of policy — he has identified those policy justifications well — but, as a matter of law, I do not believe we can go so far. Lafler does not accept that logic.

Nor does the Supreme Court’s opinion in Wetzel v. Lambert, a post -Richter decision in which the Court dealt with a fully-reasoned (i.e., gapless) state court opinion. — U.S. -, 132 S.Ct. 1195, 182 L.Ed.2d 35 (2012) (per curiam). Wetzel described the required analytical path as follows:

*355Under § 2254(d), a habeas court must determine what arguments or theories supported ... the state court’s decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of this Court.

Id. at 1198 (quoting Richter, 562 U.S. at 102, 131 S.Ct. 770). The ellipsis in that quotation is significant, as the Court wholly excised the “or, as here, could have supported” language from its quotation of Richter when describing how federal courts review a reasoned state-court decision. Compare supra note 4. Rather than extending Richter, both Lafler and Wetzel suggest that gap-filling is reserved for only those cases where we cannot discern the basis for the state court’s conclusions.11

That is not the case here. Were Dennis in exactly the same position but the Den*356nis I opinion contained one or two fewer sentences, there would perhaps be a gap to fill and I would be joining my dissenting colleagues in applying AEDPA deference, but there is no gap. The Dennis I opinion suffers from erroneous and not opaque reasoning. It may seem odd that so much hinges on so little, with a man’s life depending on the difference between bad reasoning and no reasoning. That, however, is the analytical distinction drawn by Supreme Court precedent, including Richter, Premo, and Lafler.12

Given the magnitude of the Pennsylvania Supreme Court’s error regarding the *357Cason receipt, this case presents the sort of “extreme malfunction[ ] in the state criminal justice system” that demands our intervention. Richter, 562 U.S. at 102, 131 S.Ct. 770 (internal quotation marks omitted). I therefore concur in Part III.A of the Majority’s opinion, insofar as it explains why it is proper to grant Dennis habeas relief on de novo review of the Cason receipt Brady claim, and I concur in Part II of the Majority opinion and in the judgment.

. All references to the "Dissent” refer to Judge Fisher’s dissenting opinion, unless the reference is explicitly made to Judge Hardi-man’s dissent.

. Although the Majority is correct that the “Supreme Court has never recognized an affirmative due diligence duty of defense counsel as part of Brady" (Majority Op. at 290), there is no Supreme Court opinion that forecloses the adoption of that duty. The Supreme Court has emphasized that the Brady rule requires disclosure of evidence that is "unknown to the defense,” United States v. Agurs, 427 U.S. 97, 103, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976), and that the rule is rooted in “the defendant's right to a fair trial,” id. at 108, 96 S.Ct. 2392. Based on that language, several courts of appeals have concluded that information is not unknown to the defense for Brady purposes if it can be obtained by the exercise of reasonable diligence, and that requiring diligence on the part of defense counsel does not implicate the right to a fair trial. See, e.g., Lugo v. Munoz, 682 F.2d 7, 10 (1st Cir. 1982) ("Since the information at issue here- was available to the defense attorney through diligent discovery, we find that the prosecutor’s omission was not of sufficient significance to result in the denial of the defendant’s right to a fair trial” (internal quotation marks omitted).);,United States v. Brown, 628 F.2d 471, 473 (5th Cir. Unit A 1980) ("Truth, justice, and the American way do not ... require the Government to discover and develop the defendant's entire defense.”); United States v. Hedgeman, 564 F.2d 763, 769 (7th Cir. 1977) (establishing a diligence requirement and noting that "the prosecutor will not have violated his constitutional duty of disclosure unless his omission is of sufficient significance to result in the denial of the defendant’s right to a fair trial”). The Dissent has also collected cases to that effect. (See J. Fisher Dissent Op. at 362-63 n.l.) In any event, on AEDPA review it is sufficient for our purposes that there is no Supreme Court decision clearly holding that there is not a reasonable diligence requirement. See Harrington v. Richter, 562 U.S. 86, 103, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011) (noting that a state-court error on habeas review must be one that is "well understood and comprehended in existing law beyond any possibility for fairminded disagreement”).

.See Grant v. Lockett, 709 F.3d 224, 231 (3d Cir. 2013) ("It is therefore clear that trial counsel could have discovered [the otherwise-suppressed evidence] had he exercised rea*348sonable diligence.”); United States v. Pelullo, 399 F.3d 197, 213 (3d Cir. 2005) ("[T]he burden is on the defendant to exercise reasonable diligence.”); United States v. Starusko, 729 F.2d 256, 262 (3d Cir. 1984) (''[T]he government is not obliged under Brady to furnish a defendant with information which he already has or, with any reasonable diligence, he can obtain himself” (internal quotation marks omitted).).

. More specifically, Richter says: "Under § 2254(d), a habeas court must determine what arguments or theories supported or, as here, could have supported, the state court's decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of th[e Supreme] Court.” 562 U.S. at 102, 131 S.Ct. 770.

. At the same time, the court went so far astray in applying Brady that its decision also "involved an unreasonable application of ... clearly established Federal law....” 28 U.S.C. § 2254(d)(1).

. It is important to understand the interplay between §§ 2254(a) and 2254(d). "Section 2254(a) permits a federal court to entertain only those applications alleging that a person is in state custody ‘in violation of the Constitution or laws or treaties of the United States.' " Cullen v. Pinholster, 563 U.S. 170, 181, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011). Section 2254(d) imposes an "additional restriction” on habeas relief in cases where a claim “has been adjudicated on the merits in State court proceedings.” Id. (internal quotation marks omitted). In those circumstances, habeas relief is barred unless the state court’s decision is "contrary to, or involved an unreasonable application of, clearly established Federal law.” 28 U.S.C. § 2254(d)(1). Section 2254(d) thus sets forth a necessary, but not sufficient, prerequisite to habeas relief only for those claims adjudicated on the merits in state court. If that high bar is cleared — i.e., the state court's decision is so unreasonable or contrary to federal law as established by the Supreme Court — we are still restricted to granting habeas relief only if the petitioner has shown he is in custody in violation of federal law under § 2254(a). In that second analysis, we review the petitioner’s claim de novo, without deference to the state court’s legal conclusions. Panetti, 551 U.S. at 953, 127 S.Ct. 2842 (“When a state court’s adjudication of a claim is dependent on an antecedent unreasonable application of federal law, the requirement set forth in § 2254(d)(1) is satisfied. A federal court must then resolve the claim without the deference AEDPA otherwise requires.”).

. See Premo, 562 U.S. at 126-27, 131 S.Ct. 733 (on performance: "It is not clear how the successful exclusion of the confession would have affected counsel’s strategic calculus. The prosecution had at its disposal two witnesses able to relate another confession.... Moore’s counsel made a reasonable choice to opt for a quick plea bargain.”); id. at 129, 131 S.Ct. 733 (on prejudice: "The state court here reasonably could have determined that Moore would have accepted the plea agreement even if his second confession had been ruled inadmissible. By the time, the plea agreement cut short investigation of Moore's crimes, the State’s case was already formidable and included two witnesses to an admissible confession.”).

. The Commonwealth argues that this sentence is not necessarily a factual finding to which we must defer under § 2254(e)(1), but was instead the Pennsylvania Supreme Court’s recapitulation of Dennis’s argument. The Majority rightly rejects that argument. (See Majority Op. at 288 n.17.) The plain language of Dennis I indicates that the statement was a finding of fact. See Paulson v. Newton Corr. Facility, Warden, 703 F.3d 416, 420 (8th Cir. 2013) (interpreting, in a habeas case, a state-court opinion consistent with its "plain language”). When the Pennsylvania Supreme Court was referring to arguments from the parties, it said so: in the very next paragraph of that opinion, every sentence contains some version of the words “appellant argues.” No such language appears in the disputed sentence (or its entire surrounding paragraph, for that matter). Thus, it certainly appears that the Pennsylvania Supreme Court was making a statement of historical fact when it said that "the police came into possession of” the Cason receipt. Dennis I, 715 A.2d at 408.

Without the deference afforded to an express factual finding, it would be an open question whether the police actually possessed the Cason receipt. When Dennis first offered Cason’s affidavit alleging that the police took her receipt, he himself argued that a "remand for an evidentiary hearing” would be "necessary to establish the record” before the Brady issue could be resolved. (App. 2012; see also App. 1891, 2021.) Likewise, the Commonwealth understood Cason’s affidavit to be merely a proffer of her "proposed testimony,” and argued that such testimony would have lacked the support of "competent evidence.” (App. 1923.) Further complicating matters, Cason's 1997 recollection of her interview with the police is in conflict with the police’s contemporaneous record of that encounter in 1992 (which did not enter the court record until after Dennis I, during PCRA proceedings). Were we here on de novo review of that factual finding, we could well question whether the police did, in fact, have the Cason receipt. As it stands, the state court’s factual findings are "presumed to be correct.” 28 U.S.C. § 2254(e)(1).

. In its sur-reply brief before the state court, the Commonwealth mentioned the potential "public availability” of the receipt. (App. 2026.) Under Pennsylvania law, however, arguments raised for the first time in reply briefs are generally regarded as waived. Commonwealth v. Potts, 388 Pa.Super. 593, 566 A.2d 287, 296 (1989).

. In Ylst, the Supreme Court held that when there is one reasoned state judgment rejecting a federal claim, any later unexplained orders upholding that judgment or rejecting the same claim will be presumed to rest upon the same ground. 501 U.S. at 803, 111 S.Ct. 2590. In emphasizing the difficulty of discerning the reasoning behind an unexplained state-court order — or one "whose text or accompanying opinion does not disclose the reason for the judgment,” id. at 802, 111 S.Ct. 2590 — the Court said: "Indeed, sometimes the members of the court issuing an unexplained order will not themselves have agreed upon its rationale, so that the basis of the decision is not merely undiscoverable but nonexistent.” Id. at 803, 111 S.Ct. 2590. Although Ylst predates the passage of AEDPA, the Richter Court cited it favorably, 562 U.S. at 99-100, 131 S.Ct. 770, thus indicating the continued validity of its presumption.

. That reading of Richter has ample support in other circuits. See, e.g., Cannedy v. Adams, 706 F.3d 1148, 1158 (9th Cir. 2013) (“[I]t does not follow from Richter that, when there is a reasoned decision by a lower state court, a federal habeas court may no longer ‘look through' a higher state court’s summary denial to the reasoning of the lower state court."); Johnson v. Secretary, DOC, 643 F.3d 907, 930 n.9 (11th Cir. 2011) (“The Court’s instruction from Harrington does not apply here because the Florida Supreme Court did provide an explanation of its decision....”); Sussman v. Jenkins, 642 F.3d 532, 534 (7th Cir. 2011) (distinguishing Richter because that case "addresses the situation in which a state-court decision 'is unaccompanied by an explanation,’ ” whereas in the instant case "the state appellate court issued an opinion”).

To read Richter to apply to a state court's ultimate decisions, irrespective of stated reasoning, also requires that we assume the Richter Court intended to overrule some precedents sub silentio. In particular, Ylst established a presumption that “[w]here there has been one reasoned state judgment rejecting a federal claim, later unexplained orders upholding that judgment or rejecting the same claim rest upon the same ground.” 501 U.S. at 803, 111 S.Ct. 2590. Judge Hardiman endeavors to narrow the Ylst presumption to only apply when we are uncertain as to whether the state court decided a claim “on the merits.” (J. Hardiman Dissent Op. at 368 n.7.) So, in his view, we look through to the last reasoned state court decision to determine whether the case was decided on the merits, and then, having answered that question, take no account of the reasoning in that state court decision. But, in applying the Ylst presumption, the Supreme Court has analyzed and discussed the expressed reasoning of lower state courts. See Johnson v. Williams, -U.S. -, 133 S.Ct. 1088, 1097-99, 185

L.Ed.2d 105 (2013); see also Hittson v. Chatman, -U.S.-, 135 S.Ct. 2126, 2128, 192 L.Ed.2d 887 (2015) (Ginsburg, J., concurring in the denial of certiorari) ("There is no reason not to 'look through’ ... to determine the particular reasons why the state court rejected the claim on the merits.”). The proper application of the Ylst presumption raises all of the same policy problems Judge Hardiman has noted' — just one step lower in the state review process. If we “look through” an unreasoned state court decision, Ylst presumably requires that we then review the reasoning given in the lower state court. If not, then why bother "looking through” at all? If we truly read Richter in the way Judge Hardiman proposes — and took his reasoning to its logical conclusion — it would require that we void the Ylst presumption, because we need not "look through” unreasoned judgments when we actually review only decisions and not their reasoning. But, in the words of the Supreme Court, "a presumption which gives [unreasoned orders] no effect — which simply 'looks through' them to the last reasoned decision — most nearly reflects the role they are ordinarily intended to play.” Ylst, 501 U.S. at 804, 111 S.Ct. 2590 (emphasis in original). It is hard to accept that the Richter Court intended to implicitly overrule Ylst, particularly because the Court cited Ylst favorably. See Richter, 562 U.S. at 99-100, 131 S.Ct. 770. The Court also applied the Ylst presumption just this past term, thus confirming its continued viability. See Kernan v. Hinojosa,-U.S. -, 136 S.Ct. 1603, 1605-06, 194 L.Ed.2d 701 (2016) (per curiam).

And, under Judge Hardiman's approach, Ylst is not the only precedent that would have to fall. Compare J. Hardiman Dissent Op. at 375 ("Where the state court denies relief but addresses only certain prongs of a test,or components of a claim, the reviewing federal court should likewise consider what reasons *356regarding an unaddressed prong or component could have supported the decision.”), with Wiggins v. Smith, 539 U.S. 510, 534, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003) (reviewing a Strickland claim, and concluding that its "review is not circumscribed by a state court conclusion with respect to prejudice, as neither of the state courts below reached this prong of the Strickland analysis”), and Palmer v. Hendricks, 592 F.3d 386, 400 (3d Cir. 2010) (citing Wiggins for the proposition "that because the state courts did not decide the prejudice issue on the merits, AEDPA's deferential standards do not apply to our resolution of the prejudice question”). In Wiggins, the Supreme Court did not defer to the state court’s order in assessing the second prong of the petitioner’s Strickland claim because "neither of the state courts below reached this prong of the Strickland analysis.” 539 U.S. at 534, 123 S.Ct. 2527. The Court thus acted contrary to Judge Hardiman's proposed holding here — it engaged in de novo review of the second prong even though "the state court denie[d] relief but addresse[d] only certain prongs of a test or components of a claim....” (J. Hardiman Dissent Op. at 375.) Judge Hardiman forthrightly acknowledges that his proposed holding is in tension with Wiggins, but then suggests that Richter (as the later of the two cases) undermines Wiggins. I do not believe that Richter intended that result, especially because the two cases can be reconciled.

. Again, if we determine that a state court’s reasoning is contrary to clearly established federal law, we then engage in de novo review of the claim in question. See supra note 6; Panetti, 551 U.S. at 948-54, 127 S.Ct. 2842. In his dissent, Judge Hardiman posits a hypothetical in which our decision to grant habeas relief could turn on the state court's method of drafting its decision. If the state court issues a summary order, we would apply Richter and deny habeas relief by application of AEDPA deference. If, however, it issues a reasoned decision, and that reasoning is contrary to clearly-established federal law, we would grant habeas relief — to the very same claimant — after de novo review of the underlying claim. My colleague thinks that outcome absurd, but, whether we like it or not, that is what the Supreme Court directs us to do. Under AEDPA, we must defer (1) to the reasoning actually elaborated in a state court decision, and (2) to any basis that can reasonably support a state court’s decision, but only if its own reasoning cannot be fairly discerned. The latter is the import of Richter. If the Supreme Court wanted us to afford AED-PA deference to all state court decisions regardless of the extent of their reasoning, that would be a rule of considerable consequence for habeas petitioners. Presumably the Supreme Court would have said (or at least suggested) as much in Richter, Premo, Lafler, Wetzel, or any of the other numerous habeas appeals it has considered in recent years and that Judge Hardiman has collected in his dissent. If anything, though, the Court has said the contrary. See Panetti, 551 U.S. at 954, 127 S.Ct. 2842 ("§ 2254 does not preclude relief if either the reasoning or the result of the state-court decision contradicts” clearly-established Supreme Court precedent (internal quotation marks and alteration omitted, emphasis added).).

A petitioner does not get any windfall under the approach I have outlined based on Supreme Court precedent. If his claim does not have merit, it will fail even under de novo review. Under Judge Hardiman’s approach, by' contrast, state prosecution teams do get a windfall. They would prevail unless every conceivable route to victory is "contrary to ... clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1). In other words, the prosecution wins even if it never argued a sensible position and the state court gave only a completely erroneous basis for its decision. I do not believe we can or should read Richter as going that far.






Dissenting Opinion

FISHER, Circuit Judge,

dissenting,

with whom SMITH, CHAGARES and HARDIMAN, Circuit Judges, join.

A Philadelphia jury convicted James Dennis of murder and sentenced him to death. The Pennsylvania Supreme Court affirmed his conviction and sentence. His petition for postconviction relief was denied, and, after several intervening decisions, this denial was affirmed by the Pennsylvania Supreme Court. The Majority overturns these state-court decisions by concluding that the prosecution failed to disclose to Dennis exculpatory material in violation of Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). The Majority is particularly concerned about the reliability of eyewitness testimony and about a “shodd[y]” investigation by the Philadelphia police. Maj. Op. 307. By taking this approach, the Majority goes off course for two reasons. First, the evidence against Dennis was strong — it is hard to discount the identification testimony of three eyewitnesses. Second, and more importantly, the Majority fails to adhere to the narrowly circumscribed scope of habe-as review. Congress has decreed that we may not grant a writ of habeas corpus unless the judgment of the state court was clearly unreasonable, not merely incorrect. Applying this standard of review to a case such as this one is difficult, but the Supreme Court has repeatedly reversed those courts of appeals that have not faithfully followed this mandate. The Pennsylvania Supreme Court did not unreasonably apply clearly established federal law, and for that reason I dissent.

I

On a sunny fall afternoon in 1991, Che-dell Williams and her friend Zahra Howard got off the bus that had brought them from their high school and climbed the steps of the Fern Rock SEPTA station in Philadelphia. Two men accosted them and demanded their earrings. Williams ran into the street to escape. One of the men chased her, grabbed her, and ripped her earrings out of her ears. He raised a silver revolver and fired one shot into her neck from less than an inch away. Williams collapsed and died. The shooter fled. Three eyewitnesses, including Howard, observed the shooter at close range. They each identified the shooter in a photo array, in a lineup, and at trial: the shooter was James Dennis.

The Majority discusses in detail the testimony of the three eyewitnesses who testified at trial that Dennis shot Williams: Zahra Howard, Thomas Bertha, and James Cameron. The Majority calls out discrepancies between the eyewitnesses’ descriptions of the shooter’s height and weight (said to be 5'9" or 5'10" and 170 to 180 pounds) and Dennis’s actual size (5'5" and 125 to 135 pounds). The reliability of the eyewitness identifications is irrelevant to the legal question we must decide— which is whether the Pennsylvania Supreme Court unreasonably applied Brady and its progeny. Nevertheless, a few points about the identifications are worth mentioning. First, the visual conditions were excellent. The murder occurred in the afternoon and the weather was clear. Second, the witnesses saw the shooter at close range and had unobstructed views of his face. Howard was one to two feet away *358from the shooter and looked him in the face. Bertha and the shooter made eye contact from less than eight feet away, and Bertha was able to observe the expression on the shooter’s face. Cameron saw the face of the shooter from eight to ten feet away. Third, none of the identifications was cross-racial. See Arizona v. Youngblood, 488 U.S. 51, 72 n.8, 109 S.Ct. 333, 102 L.Ed.2d 281 (1988) (Blackmun, J., dissenting) (noting studies showing that cross-racial identifications are less accurate than same-race identifications). And fourth, witnesses generally overestimate the height and weight of men who are below population averages, as Dennis was. Christian A. Meissner et al., Person Descriptions as Eyewitness Evidence 3, 8, in 2 Handbook of Eyewitness Psychology (Rod C.L. Lindsay et al., eds. 2007) (noting a tendency for witnesses to underestimate the height of taller targets and overestimate the height of shorter targets); Rhona H. Flin & John W. Shepherd, Tall Stories: Eyewitnesses’ Ability to Estimate Height and Weight Characteristics, 5 Human Learning 29, 34 (1986) (noting the same effect for both height and weight); see id. at 36 (citing a study finding that “witnesses tend to overestimate the height of criminals”).

The defense vigorously cross-examined these witnesses and elicited some discrepancies between their testimony and prior statements and between estimates of the shooter’s height and weight and Dennis’s. Nevertheless, the jury found the eyewitnesses’ testimony credible. In addition to that testimony, the prosecution called Charles Thompson, a member of Dennis’s singing group, who testified that he saw Dennis with a small silver handgun several hours after the murder. Whatever one might feel about the testimony of these witnesses or the testimony of eyewitnesses in general, the evidence that convinced the jury to convict Dennis was not, as the district court described it, “scant evidence at best.” Dennis v. Wetzel, 966 F.Supp.2d 489, 491 (E.D. Pa. 2013).

Dennis’s Brady claims concern three documents that he asserts the prosecution should have turned over to him before trial: a receipt from the Department of Public Welfare (DPW), a police activity sheet reporting a conversation with Williams’s aunt and uncle, and police records describing the investigation of a jailhouse tip. The receipt relates to a possible alibi witness, Latanya Cason. Dennis told police that he was riding a bus at the time of the murder — shortly before 2:00 p.m.— and that he saw Cason and waved to her as he left the bus. Cason testified at trial that she saw Dennis at 4:00 or 4:30 p.m., which did not support his alibi. Cason visited the DPW before seeing Dennis that day. Dennis asserts that the police had a time-stamped receipt from Cason’s DPW visit and that, had the receipt been turned over to the defense, Cason would have testified that she saw Dennis at 2:00 or 2:30 p.m. The subject of Dennis’s second claim is a police activity sheet containing detectives’ notes of an interview with Williams’s aunt and uncle, Diane and Man-nasett Pugh. According to the notes, the Pughs told detectives that Zahra Howard told them she recognized the shooter from her high school. This conflicts with Howard’s statements to police and testimony at trial that she had never seen the shooter before. The third Brady claim concerns police records of an investigation of a tip by an inmate, William Frazier, who told police that his friend, Tony Brown, admitted to Frazier that Brown shot Williams. Police never located Tony Brown, and Frazier later admitted that he made up the entire story.

The district court concluded that the prosecution violated Brady by suppressing each of these three items and found that the Pennsylvania Supreme Court’s deter*359minations to the contrary unreasonably applied clearly established Supreme Court precedent. I disagree with the Majority’s affirmance of the district court and will explain my reasons in detail.

II

The source of my disagreement with the Majority is its failure to apply the deferential standard of review prescribed by the Antiterrorism and Effective Death Penalty Act (AEDPA). When a state prisoner applies for a writ of habeas corpus on a claim that was adjudicated on the merits in state court, a federal court may not grant the application unless 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 “resulted in a decision that 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). A state court’s application of the law or determination of the facts is not unreasonable merely because it is — in the eyes of the reviewing federal court — wrong. The decision must be “so lacking in justification that there was an error well understood and comprehended in existing law beyond any possibility for fair-minded disagreement.” Harrington v. Richter, 562 U.S. 86, 103, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011).

We must give state-court decisions “the benefit of the doubt." Cullen v. Pinholster, 563 U.S. 170, 181, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011). This duty to give state-court decisions deference applies even when a state court does not give a reasoned explanation of its decision. ‘Where a state court’s decision is unaccompanied by an explanation, the habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief.” Richter, 562 U.S. at 98, 131 S.Ct. 770. In such a situation, the reviewing federal court must consider arguments and theories that “could have supported” the decision. Id. at 102, 131 S.Ct. 770.

The AEDPA standard is intentionally difficult to meet. The standard reflects state courts’ competence to resolve federal constitutional questions and states’ strong interest in controlling their criminal justice systems. Federal habeas corpus is designed to “ ‘guard against extreme malfunctions in the state criminal justice systems,’ not [to] substitute for ordinary error correction through appeal.” Id. at 102-03, 131 S.Ct. 770 (quoting Jackson v. Virginia, 443 U.S. 307, 332 n.5, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979) (Stevens, J., concurring)). Among the courts of appeals, however, there has been some reluctance to adhere to the AEDPA standard as defined by the Supreme Court. In recent terms, the Court has issued a string of reversals, many as summary per curiam opinions, for failure to apply the correct standard of review under AEDPA. See, e.g., Woods v. Etherton, — U.S. -, 136 S.Ct. 1149, 194 L.Ed.2d 333 (2016) (per curiam); Woods v. Donald, — U.S.-, 135 S.Ct. 1372, 191 L.Ed.2d 464 (2015) (per curiam); Glebe v. Frost, — U.S. -, 135 S.Ct. 429, 190 L.Ed.2d 317 (2014); Lopez v. Smith, — U.S. -, 135 S.Ct. 1, 190 L.Ed.2d 1 (2014) (per curiam); Marshall v. Rodgers, — U.S.-, 133 S.Ct. 1446, 185 L.Ed.2d 540 (2013) (per curiam); Parker v. Matthews, — U.S. -, 132 S.Ct. 2148, 183 L.Ed.2d 32 (2012) (per curiam); Renico v. Lett, 559 U.S. 766, 130 S.Ct. 1855, 176 L.Ed.2d 678 (2010). There are many more. I fear that this case may join that list.

The Majority holds that the Pennsylvania Supreme Court unreasonably applied the United States Supreme Court’s decisions in the line of cases discussing prosecutors’ duty to turn over favorable evidence to the defense. In Brady v. Maryland, the Court held that “the sup*360pression by the prosecution of evidence favorable to an accused upon request violates due process where the evidence is material either to guilt or to punishment, irrespective of the good faith or bad faith of the prosecution.” 373 U.S. at 87, 83 S.Ct. 1194. The Court later ruled that the duty to disclose exculpatory evidence applies whether a defendant requests it or not. United States v. Agurs, 427 U.S. 97, 107, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976). The Court explained that a Bra--dy violation has three components: evidence that is (1) favorable to the defendant, (2) suppressed by the prosecution, and (3) material. Strickler v. Greene, 527 U.S. 263, 281-82, 119 S.Ct. 1936, 144 L.Ed.2d 286 (1999). Favorable evidence includes both exculpatory evidence and evidence that could be used to impeach prosecution witnesses. United States v. Bagley, 473 U.S. 667, 676, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985). Evidence can be suppressed even if it is only known to the police and not to the prosecutor — “the individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government’s behalf in the ease, including the police.” Kyles v. Whitley, 514 U.S. 419, 437, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995). Evidence is material if “there is a reasonable probability that the suppressed evidence would have produced a different verdict.” Strickler, 527 U.S. at 281, 119 S.Ct. 1936. The materiality of suppressed evidence must be assessed cumulatively, “not item by item.” Kyles, 514 U.S. at 436, 115 S.Ct. 1555.

Ill

A

The Pennsylvania Supreme Court addressed the Brady claim based on Latanya Cason’s DPW receipt (and an ineffective assistance of counsel claim based on his counsel’s failure to investigate Cason) without providing much reasoning or detail. The court noted that Cason testified that she saw Dennis at around 4:00 or 4:30 p.m. the day of the murder based on her recollection that she had left work to cash her welfare check at about 2:00 p.m. “During their investigation, however, the police came into possession of a Department of Public Welfare (DPW) receipt showing that Cason cashed her check at 1:03 p.m.” Commonwealth v. Dennis, 552 Pa. 331, 715 A.2d 404, 408 (1998) (“Dennis I”). The court found that the receipt was not material because even if the defense knew of the receipt, Cason’s corrected testimony “would not support [Dennis’s] alibi ... because the murder occurred at 1:50 p.m., forty minutes earlier than Cason’s earliest estimate.” Id. The court concluded: “Finally, it is clear that there clearly was no Brady violation. The DPW receipt was not exculpatory, because it had no bearing on [Dennis’s] alibi, and there is no evidence that the Commonwealth withheld the receipt from the defense.” Id.

I agree with the Majority that the Ca-son receipt was favorable to Dennis and was material, but I disagree with the Majority’s conclusion that the receipt was suppressed. Despite the Pennsylvania Supreme Court’s representations about clarity, it is not clear what the court meant by “there is no evidence that the Commonwealth withheld the receipt from the defense.” The Majority acknowledges that-the Pennsylvania Supreme Court “provided no explanation.” Maj. Op. 288. Yet the Majority assumes that the Pennsylvania Supreme Court made an unreasonable finding of fact or conclusion of law that the prosecution had no duty to disclose the receipt because it was in possession of the police — a finding clearly foreclosed by Kyles, 514 U.S. at 437-38, 115 S.Ct. 1555.

When a state court does not give a reasoned explanation, we are not permitted to assume or guess what the most *361likely explanation is. “Where a state court’s decision is unaccompanied by an explanation, the habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief.” Richter, 562 U.S. at 98, 181 S.Ct. 770 (emphasis added). In other words, when there is an analytical gap in a state court’s reasoning, we must consider “what arguments or theories ... could have supported ... the state court’s decision.” Id. at 102, 131 S.Ct. 770.

Although the state-court decision at issue in Richter was a summary disposition, the Supreme Court’s instruction to consider arguments that could have supported the state court’s decision is not limited to summary dispositions. In Premo v. Moore, 562 U.S. 115, 131 S.Ct. 733, 178 L.Ed.2d 649 (2011), which was decided the same day as Richter, the Supreme Court considered theories that could have supported a reasoned, written decision with an analytical gap. In state postconviction relief proceedings, Moore argued that his counsel had been unconstitutionally ineffective under Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). The state court rejected his Strickland argument, but, as the Supreme Court noted, the “state court did not specify” whether the ineffectiveness claim failed “because there was no deficient performance under Strickland or because Moore suffered no Strickland prejudice, or both.” Moore, 562 U.S. at 123, 131 S.Ct. 733. In order for a federal court to grant habeas relief, both prongs would need to have involved an unreasonable application of clearly established federal law. Id. The Supreme Court found that the state court “reasonably could have concluded that Moore was not prejudiced by [his] counsel’s actions. Under AEDPA, that finding ends federal review.” Id. at 131, 131 S.Ct. 733.

Because the Pennsylvania Supreme Court provided no explanation for why it found that the receipt was not withheld from the defense, there is an analytical gap. This gap is more open-ended than the two possibilities the state court could have considered in Moore and narrower than a summary disposition, such as Richter, where the universe of possible theories is broad. But our obligation to consider what theories could have supported the Pennsylvania Supreme Court’s decision is no less than in Richter and Moore.

Judge Jordan, in his opinion concurring in part and concurring in the judgment, takes the position that there is no gap to be filled under Richter and Moore. He believes that the only way to explain the Pennsylvania Supreme Court’s statements that the police had the receipt but that the Commonwealth did not withhold the receipt is that the court failed to apply Kyles. Judge Jordan concludes that the Pennsylvania Supreme Court “simply found that there was no evidence that the prosecutor possessed the Cason receipt.” Concurring Op. 351. This is a reasonable explanation, but it is not the only explanation. The Pennsylvania Supreme Court’s opinion lacks sufficient analysis to tell what it meant by “there is no evidence the Commonwealth withheld the receipt from the defense.” Dennis I, 715 A.2d at 408. If we “take the state court’s decision as written,” Concurring Op. 351, rather than assuming that the state court made a mistake, there is an analytical gap.

The Majority also takes the position that the Pennsylvania Supreme Court violated Kyles. The Majority notes, however, that “[t]he Pennsylvania Supreme Court provided no explanation for its ... statement [that there was ‘no evidence that the Commonwealth withheld the receipt from the defense’], and we cannot be sure whether the court was assessing the facts or interpreting the law.” Maj. Op. 288. Despite this lack of clarity, the Majority is evident*362ly certain that it knows “the precise basis for the state court’s ruling.” Id. at 282. Unlike the Majority, I am unable to discern the precise basis for the state court’s ruling, and, for that reason, this is one of those cases in which consideration of theories that could have supported the state court’s decision is required.

This required consideration leads to the conclusion that there is a viable gap-filling theory here:' the Pennsylvania Supreme Court could have meant that the receipt was not withheld because it was available to the defense with reasonable diligence. The reasonable diligence “branch of the Brady doctrine” is evident, albeit inconsistent, in our own precedents. See United States v. Perdomo, 929 F.2d 967, 973 (3d Cir. 1991) (“Evidence is not considered to be suppressed if the defendant either knew or should have known of the essential facts permitting him to take advantage of any exculpatory evidence.” (emphasis added)); United States v. Starusko, 729 F.2d 256, 262 (3d Cir. 1984) (“[T]he government is not obliged under Brady to furnish a defendant with information which he already has or, with any reasonable diligence,' he can obtain himself.” (quoting United States v. Campagnuolo, 592 F.2d 852, 861 (5th Cir. 1979))). But see Wilson v. Beard, 589 F.3d 651, 664 (3d Cir. 2009) (“[T]he fact that a criminal record is a public document cannot absolve the prosecutor of her responsibility to provide that record to defense counsel.” (internal quotation mark omitted)).

Despite this inconsistency, we reinforced the conclusion that Brady has a reasonable diligence component in Grant v. Lockett, 709 F.3d 224, 231 (3d Cir. 2013). In Grant, the prosecution failed to disclose that its key witness — the only person who testified that Grant was the shooter — was on parole at the time of the shooting. Grant’s post-conviction relief counsel was able to discover that the witness was on parole, and his trial counsel could have looked up the witness’s criminal history in records kept by the clerk of court. We concluded that Grant’s Brady claim “lacked merit” because “trial counsel could have discovered [the witness’s] parole status had he exercised reasonable diligence.” Id. at 230, 231.

The Majority correctly notes that our case law on Brady reasonable diligence “is inconsistent and could easily confüse” and clarifies that 'reasonable diligence “plays no role in the Brady analysis.” Maj. Op. 291. This clarification to our case law is helpful, and were we reviewing this case on direct appeal it would be entirely appropriate. The “no reasonable diligence” rule may indeed represent the best interpretation of the Supreme Court’s Brady case law. But this rule is nonetheless an interpretation of Supreme Court precedent. It does not represent a clearly established holding of the Court, and it does not mean that any other interpretation is unreasonable.

The reasonableness of interpreting Brady to have a reasonable diligence component is supported by the decisions of other courts of appeals. The Majority notes with surprise that “several Courts of Appeals have endorsed some form of a due diligence requirement.” Maj. Op. 291 n.20. “Several” understates the matter. A majority of the courts of appeals have applied a reasonable diligence requirement at one time or another.1 The number of courts *363(including our court, ten out of the twelve regional courts of appeals) and decisions applying a reasonable diligence requirement hardly evince a clearly established Supreme Court rule that reasonable diligence plays no role in the Brady analysis. Even if the Majority is correct and all these decisions erroneously applied Brady, it is hard to conclude that the error is “well understood and comprehended in existing law” and “beyond any possibility for fairminded disagreement.” Richter, 562 U.S. at 103, 131 S.Ct. 770. Surely, given the number of federal circuit judges who have concluded that reasonable diligence is a consideration in the analysis of a Brady claim, “it is possible fairminded jurists could disagree” that reasonable diligence is inconsistent with Supreme Court precedent. Id. at 102, 131 S.Ct. 770.

Under the specific facts of this case, the Pennsylvania Supreme Court easily could have concluded that Latanya Cason’s DPW receipt was available to Dennis’s counsel had his counsel exercised reasonable diligence. Dennis was aware of Ca-son — the police only interviewed her after Dennis told them she had seen him. Dennis’s appellate counsel obtained the receipt from the DPW. And Dennis argued that his trial counsel would have located the receipt with “minimal investigation.”2 (App. 1800.) It was not an unreasonable application of clearly established Supreme *364Court precedent for the Pennsylvania Supreme Court to conclude that there was no Brady violation where trial counsel could have discovered the evidence by exercising reasonable diligence and investigating his own client’s alibi witness. See United States v. Senn, 129 F.3d 886, 893 (7th Cir. 1997) (“[T]he defendants are hoisted by their own petard: without having obtained the Broward County file they would not have a Brady argument, but the ease with which they obtained their file defeats their claim.”).

The Majority contends the Supreme Court did “away with any belief that Brady imposes a due diligence requirement” in Banks v. Dretke, 540 U.S. 668, 124 S.Ct. 1256, 157 L.Ed.2d 1166 (2004). Maj. Op. 291. But Banks, which was decided after the Pennsylvania Supreme Court’s decision in Dennis I,3 is distinguishable. In Banks, the prosecution withheld evidence that one prosecution witness had been “intensively coached” by prosecutors before his testimony and another witness was a paid police informant. 540 U.S. at 677-78, 124 S.Ct. 1256. The prosecution failed to correct these witnesses’ testimony when the witnesses denied talking to anyone about their testimony or receiving payments from police. Id. The Supreme Court refused to adopt a rule allowing the prosecution to “lie and conceal” evidence so long as the prisoner might have been able to detect the “potential existence” of prosecu-torial misconduct. Id. at 696, 124 S.Ct. 1256. Unlike the DPW receipt at issue in the present case, the evidence in Banks of the witness coaching and police payments was solely in the hands of the prosecution. No amount of diligent investigation would have uncovered that evidence. Banks is not directly applicable to evidence that could have been discovered after “minimal investigation.” See Bell v. Bell, 512 F.3d 223, 235 (6th Cir. 2008) (en banc) (concluding that Banks did not call into question precedents applying a reasonable diligence requirement).

Under these circumstances, I cannot conclude that the Pennsylvania Supreme Court’s denial of Dennis’s Brady claim based on the receipt was an unreasonable application of clearly established Supreme Court precedent.

B

The Pennsylvania Supreme Court reasonably determined that the Pugh statement was immaterial under Brady. The statement was found in a police activity sheet that showed that Chedell Williams’s aunt and uncle, Diane and Mannasett Pugh, told police that Zahra Howard told them that she recognized the shooter from school. This alleged statement is contrary to what Howard repeatedly told police and testified about at trial — that she had never seen the shooter before he accosted Williams and her at the SEPTA station.

The postconviction relief court held an evidentiary hearing about this Brady claim. Howard testified that she never told Williams’s family that she had seen the shooter before. When confronted by the purported statement in the police activity sheet, she denied ever having made it. Diane Pugh testified that, as far as she could remember, Howard never said she recognized the shooter before the murder.

The Pennsylvania Supreme Court concluded that the police activity sheet showing the Pugh statement was not material under Brady because there was no reason*365able probability of a different result had the sheet been turned over. Commonwealth v. Dennis, 609 Pa. 442, 17 A.3d 297, 308-09 (2011) (“Dennis IV’). The Pennsylvania Supreme Court noted that Howard “was extensively cross-examined at trial” about her identification of Dennis, including about whether she had ever seen the shooter before, and she steadfastly testified that Dennis was the shooter and that she had never seen him before. Id. at 309. Two eyewitnesses other than Howard identified Dennis in a photo array, in a line up, and at trial, and these witnesses would not have been affected by any impeachment of Howard. Id. For these reasons, the Pennsylvania Supreme Court held that Dennis “still received a fair trial resulting in a verdict worthy of confidence.” Id. This conclusion was not an unreasonable interpretation of Supreme Court precedent.

The Majority correctly notes that heavy impeachment of a witness does not render further impeachment immaterial. See Banks, 540 U.S. at 702, 124 S.Ct. 1256. In Banks, the prosecution suppressed information that a key witness was a government informant, and the government argued this information was “merely cumulative” because the witness was heavily impeached at trial. Id. None of the testimony at trial concerned the ■witness’s status as an informant, however. The Court concluded this missing information was material because the jury was ignorant of the witness’s “true role” in the case. Id.

The impeachment value of the activity sheet in this case was minor. Howard’s identification of Dennis was cross-examined at trial.' She credibly testified in the postconviction relief hearing that she never made the statements attributed to her in the activity sheet. The activity sheet’s double hearsay makes it inherently weak. This is not the kind of evidence considered material in Banks.

The Majority asserts that had the activity sheet been disclosed, “defense counsel could have impeached Howard'in a manner that very well may have led her to admit she recognized the perpetrators from her high school.” Maj. Op. 301. There is no basis in the record for this speculation, which is undercut by Howard’s consistency in all her sworn testimony at trial and during the postconviction relief hearing. Such a dramatic courtroom reversal is more likely in a Matlock or Perry Mason script than in reality. The unlikely nature of this speculation does not create a reasonable probability of a different result or “undermine confidence in the outcome,” as required for Brady materiality.4 Bagley, 473 U.S. at 682, 105 S.Ct. 3375.

The Pennsylvania Supreme Court’s consideration of the strength of the other evidence against Dennis was also not unreasonable. The materiality of the activity sheet “must be evaluated in the context of the entire record.” Agurs, 427 U.S. at 112, 96 S.Ct. 2392. And “evidence impeaching an eyewitness may not be material if the State’s other evidence is strong enough to sustain confidence in the verdict.” Smith v. Cain, — U.S. -, 132 S.Ct. 627, 630, 181 L.Ed.2d 571 (2012). The Pennsylvania Supreme Court could reasonably have concluded, in the context of the entire record, that any impeachment value of the activity sheet would not undermine confidence in the verdict. Bertha and Cameron also identified Dennis in a photo array, in a line up, and at trial. Impeachment of Howard *366would not have affected the weight of their testimony.

The Majority emphasizes the importance of Howard as “the eyewitness with the most significant exposure to the shooter” and minimizes Bertha and Cameron as “located farther away” with “only brief glimpses of the perpetrators” or “paying little attention.” Maj. Op. 301. But in this case, “farther away” was only eight feet from the shooter for Bertha and ten feet from the shooter for Cameron, and each had an unobstructed view of the shooter’s face. To the extent Bertha and Cameron had not been paying attention to the commotion, the gunshot focused their view and spurred them into action. Bertha stepped into the street as the shooter ran past, stopped as the shooter raised his gun, and then followed behind him. Cameron and the shooter made eye contact. When the shooter fled, Cameron ran to aid Williams. The eyewitness testimony of Bertha and Cameron was powerful evidence of guilt.

The Majority criticizes the Pennsylvania Supreme Court for applying a “sufficiency of the evidence” standard in lieu of the appropriate Brady materiality standard. Nowhere, however, did the Pennsylvania Supreme Court articulate the wrong standard. The Pennsylvania Supreme Court recognized that Brady materiality is not a question of sufficiency of evidence in Commonwealth v. Weiss, 604 Pa. 573, 986 A.2d 808 (2009), and it cited Weiss in Dennis IV.5

The Majority nevertheless concludes that, even if the Pennsylvania Supreme Court knew the correct standard, it unreasonably applied that standard to the facts of this case. The Majority focuses on the Pennsylvania Supreme Court’s statement that “disclosure of the activity sheet would have had no impact upon [Bertha’s and Cameron’s eyewitness] testimony.” Dennis IV, 17 A.3d at 309. According to the Majority, this is evidence that the Pennsylvania Supreme Court was proceeding down a wrong “analytical path.” Maj. Op. 303. But there is nothing inherently wrong with this analytical path. The United States Supreme Court has, at times, made similar statements.

For instance, in Strickler v. Greene, the prosecution withheld exculpatory materials that would have been “devastating ammunition for impeaching” the prosecution’s key witness, Anne Stoltzfus. 527 U.S. 263, 296,119 S.Ct. 1936,144 L.Ed.2d 286 (1999) (Souter, J., dissenting). At the petitioner’s capital murder trial, Stoltzfus testified “in vivid detail” about the abduction of the murder victim. Id. at 266, 119 S.Ct. 1936 (majority opinion). Stoltzfus was the only disinterested eyewitness who testified. The exculpatory materials were police notes of interviews with Stoltzfus and letters Stoltzfus wrote to the police that cast serious doubt on her testimony. The Court found all the elements of Brady met except for materiality. Although the Court recognized the importance of Stoltzfus’s eyewitness testimony, that was not the only evidence before the jury. Other eyewitnesses placed the petitioner at the shopping mall where the abduction occurred, and “considerable forensic and other physical evidence” linked the petitioner to the crime. Id. at 293, 119 S.Ct. 1936. The Court concluded that “[t]he record provides strong support for the conclusion *367that petitioner would have been convicted of capital murder and sentenced to death, even if Stoltzfus had been severely impeached.” Id. at 294, 119 S.Ct. 1936. Thus, the .petitioner did not convince the Court that there was “a reasonable probability that the jury would have returned a different verdict if her testimony had been either severely impeached or excluded entirely.” Id. at 296, 119 S.Ct. 1936. The Pennsylvania Supreme Court’s reasoning in this case is not appreciably different from the reasoning in Strickler.

The Majority’s remaining reason for concluding that the Pennsylvania Supreme Court unreasonably applied the facts is that the Majority considered the same facts and reached a different conclusion. This is not a proper basis for granting habeas relief. There is a reasonable possibility that impeachment of Howard might have produced a different result, but the Pennsylvania Supreme Court did not unreasonably apply the facts or law in concluding that Dennis did not establish a reasonable probability of a different result. See id. at 291, 119 S.Ct. 1936. I would not grant habeas relief on this claim.

C

Dennis’s final Brady claim concerns documents about the police investigation of a lead from William Frazier. Frazier, an inmate at the Montgomery County Correctional Facility, contacted police and informed them that he knew who shot Che-dell Williams. He told a story about a three-way call he received in jail with his aunt and a friend named Tony Brown. During the call, Tony Brown admitted that he accidentally shot Williams while robbing her. Tony Brown told Frazier that he was accompanied by his Mend Ricky Walker, who was Frazier’s cousin, and another man, “Skeet,” who drove the car.

Despite Frazier’s being a jailhouse informant who obviously wanted to parlay information for something in return (even if only a day out of jail), the police investigated his tip. They took Frazier on a ride-along to Tony Brown’s house, Ricky Walker’s house, the pawnshop where Tony Brown allegedly sold Williams’s earrings, Skeet’s house, and Frazier’s girlfriend’s house. Police interviewed Frazier’s landlord and Walker. Walker told police that he never heard of anyone named Tony Brown or “Skeet.” He explained that he “can’t stand” Frazier, who racked up $1,000 in charges on a phone calling card Walker had lent to him. Despite this investigation, police found no trace of a Tony Brown. This is not surprising. Frazier later admitted that he concocted the entire story.6

The Pennsylvania Supreme Court rejected Dennis’s Brady claim about the Frazier lead documents because the documents were inadmissible hearsay. Commonwealth v. Dennis, 597 Pa. 159, 950 A.2d 945, 968 (2008) (“Dennis IIP). This conclusion is not an unreasonable application of clearly established Supreme Court precedent.

Authority is split about whether inadmissible evidence can be the basis for a Brady violation. Our court, along with the First, Second, Sixth, and Eleventh Circuit Courts of Appeals, has concluded that ad*368missibility is not a prerequisite for a Brady claim. See, e.g., Johnson v. Folino, 705 F.3d 117, 130 (3d Cir. 2013) (“[IJnadmissi-ble evidence may be material if it could have led to the discovery of admissible evidence.”). The Fourth and Seventh Circuits have concluded otherwise. Jardine v. Dittmann, 658 F.3d 772, 777 (7th Cir. 2011) (“Logically, inadmissible evidence is immaterial under this rule.”); Hoke v. Netherland, 92 F.3d 1350, 1356 n.3 (4th Cir. 1996) (“[TJhese statements may well have been inadmissible at trial ... and therefore, as a matter of law, ‘immaterial’ for Brady purposes.”).

The Majority recognizes the contrary decisions of the Fourth and Seventh Circuits and “respectfully conclude^] that they have erred.” Maj. Op. 311 n.27. But in order to grant habeas relief, the Majority must conclude that these courts did more than err — the decisions must be so clearly wrong that they are objectively unreasonable. Does the Majority really believe that our fair-minded colleagues on the Fourth and Seventh Circuits are that wrong? As the Supreme Court has noted, the courts of appeals’ “diverging approaches to [a] question illustrate the possibility of fair-minded disagreement.” White v. Woodall, -U.S.-, 134 S.Ct. 1697, 1703 n.3, 188 L.Ed.2d 698 (2014). Circuit precedent cannot create or refíne clearly established Supreme Court law, and lower federal courts “may not canvass circuit decisions to determine whether a particular rule of law is so widely accepted among the Federal Circuits that it would, if presented to [the Supreme] Court, be accepted as correct.” Marshall v. Rodgers, — U.S. -, 133 S.Ct. 1446, 1451, 185 L.Ed.2d 540 (2013) (per curiam). Although “[mjost federal courts have concluded that suppressed evidence may be material for Brady purposes even where it is not admissible,” Maj. Op. 310, that does not transform such a rule into clearly established Supreme Court precedent.

The Majority does not cite any direct holding of the Supreme Court establishing a rule that admissibility is irrelevant under Brady. The Majority instead relies on “the Supreme Court’s repeated consideration of impeachment material in Brady cases.” Maj. Op. 309. The Supreme Court’s consideration of impeachment material does not compel the broad conclusion that admissibility is irrelevant.

Because reasonable judges could — and indeed do — disagree about whether Brady material must be admissible, the Pennsylvania Supreme Court did not unreasonably apply clearly established Supreme Court precedent when it found that the inadmissibility of the Frazier lead documents prevented Dennis’s Brady claim.7

IV

The Majority asserts that the Cason receipt, Pugh statement, and Frazier documents “effectively gutted the Commonwealth’s case against Dennis” and that the failure to turn over these documents denied Dennis a fair trial. Maj. Op. 269. Not true. Dennis’s inability to obtain the Cason receipt before trial was, as Dennis himself argued, due to his trial counsel’s failure to conduct even a minimal investigation. The double hearsay Pugh statement was credibly refuted by Howard. Even if Howard were impeached, based on the eyewitness testimony of Bertha and Cameron, there was not a reasonable probability of a jury’s returning a different verdict. Frazier’s story was fabricated. It was not an unreasonable application of clearly established fed*369eral law to consider the inadmissibility of the Frazier documents. In granting habeas relief for each of these Brady claims, the Majority failed to correctly apply the deferential AEDPA standard. I respectfully dissent.

. First Circuit: United States v. Rodriguez, 162 F.3d 135, 147 (1st Cir. 1998) ("The government has no Brady burden when the necessary facts for impeachment are readily available to a diligent defender....").

Second Circuit: United States v. Payne, 63 F.3d 1200, 1208 (2d Cir. 1995) ("[E]vidence is not considered to have been suppressed within the meaning of the Brady doctrine if the defendant or his attorney either knew, or should have known, of the essential facts *363permitting him to take advantage of that evidence.” (internal quotation marks and alteration omitted)).

Fourth Circuit: United States v. Wilson, 901 F.2d 378, 381 (4th Cir. 1990) ("[W]here the exculpatory information is not only available to the defendant but also lies in a source where a reasonable defendant would have looked, a defendant is not entitled to the benefit of the Brady doctrine.”).

Fifth Circuit: United States v. Dixon, 132 F.3d 192, 199 (5th Cir. 1997) (“Brady does not obligate the government to produce for a defendant evidence or information already known to him, or that he could have obtained from other sources by exercising reasonable diligence.” (internal quotation marks and alteration omitted)); United States v. Prior, 546 F.2d 1254, 1259 (5th Cir. 1977) (“[N]umerous cases have ruled that the government is not obliged under Brady to furnish a defendant with information which he already has or, with any reasonable diligence, he can obtain himself.”).

Sixth Circuit: Matthews v. Ishee, 486 F.3d 883, 891 (6th Cir. 2007) (“Where ... the factual basis for a claim is reasonably available to the petitioner or his counsel from another source, the government is under no duty to supply that information to the defense.” (internal quotation marks omitted)).

Seventh Circuit: Boss v. Pierce, 263 F.3d 734, 740 (7th Cir. 2001) ("Evidence is suppressed for Brady purposes only if ... the evidence was not otherwise available to the defendant through the exercise of reasonable diligence.”).

Eighth Circuit: United States v. Zuazo, 243 F.3d 428, 431 (8th Cir. 2001) (“The government does not suppress evidence in violation of Brady by failing to disclose evidence to which the defendant had access through other channels.”).

Ninth Circuit: Raley v. Ylst, 470 F.3d 792, 804 (9th Cir. 2006) ("[W]here the defendant is aware of the essential facts enabling him to take advantage of any, exculpatory evidence, the Government does not commit a Brady violation by not bringing the evidence to the attention of the defense.” (quoting United States v. Brown, 582 F.2d 197, 200 (2d Cir. 1978))).

Eleventh Circuit: LeCroy v. Sec'y, Fla. Dep’t of Corr., 421 F.3d 1237, 1268 (11th Cir. 2005) ("To establish that he suffered a Brady violation, the defendant must prove that ... the defendant did not possess the evidence and could not have obtained it with reasonable diligence....”).

. The Majority asserts that the 'DPW receipt was not publicly available because DPW regulations prevent disclosure of information about welfare recipients. Maj. Op. 289-90. Dennis did not argue this point below or raise it on appeal, and, to the extent the DPW privacy regulations applied to the receipt, Dennis’s admission that the receipt was available with minimal investigation makes the regulations irrelevant.

. See Williams v. Taylor, 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) ("[T]he phrase ‘clearly established Federal law, as determined by the Supreme Court of the United States’ ... refers to the holdings, as opposed to the dicta, of this Court’s decisions as of the time of the relevant state-court decision.” (emphasis added)).

. The Majority adopts the district court’s conclusion that the activity sheet would have shown that Howard either lied to the Pughs or lied at trial. Maj. Op. 301. Given Howard’s testimony at trial and the postconviction relief hearings, an alternative conclusion is as least as likely: in a crowded and grieving house immediately after the murder, the Pughs misunderstood or later misreported what Howard said.

. See Weiss, 986 A.2d at 816 (remanding to the postconviction relief court to "consider whether disclosure of the impeachment evidence to competent counsel would have made a different result reasonably probable,” which "will necessarily entail a review of all the evidence presented at trial, not for its sufficiency, but for the potential negative effect disclosure of the alleged impeachment evidence would have had thereon”); id. at 815 (“The United States Supreme Court has made clear that Bagley's materiality standard is not a sufficiency of the evidence test.”).

. The Majority asserts that the Frazier "lead was not fruitless, it was simply not rigorously pursued.” Maj. Op. 307. The police did pursue this lead, however, going so far as to take Frazier out of his jail cell and bring him with them on his tour of Philadelphia. The Majority questions why police did not interview more of the people involved in Frazier's tale. Police can always do more investigative work, but they have limited resources. And simply put, this lead coming from a jailhouse snitch was a dead end. The police should not be faulted for deciding not to waste more time on what Frazier himself admitted was "bullshit.” Response to Pet. Rh’g 17 n.13.

. Because the Pennsylvania Supreme Court could reasonably have determined that the Cason receipt was not suppressed and reasonably determined that the Frazier documents were not subject to Brady, materiality was an issue with only the Pugh statement. Accordingly, there is no need to conduct a cumulative materiality analysis.






Dissenting Opinion

HARDIMAN, Circuit Judge,

dissenting,

joined by SMITH and FISHER, Circuit Judges.

At the outset of its analysis of James Dennis’s Brady claims, the Majority notes that the Antiterrorism and Effective Death Penalty Act (AEDPA) “dictates” our review. Majority Op. 280. The opinion describes with precision AEDPA’s strictures. Federal courts are prohibited from granting habeas corpus relief unless the state-court adjudication (1) “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” or (2) “resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” Id. (quoting 28 U.S.C. § 2254(d)). These fetters on our review, the Majority notes, come close to “imposing a complete bar on federal-court relitigation of claims already rejected in state-court proceedings.” Id. at 281 (quoting Harrington v. Richter, 562 U.S. 86, 102, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011)).

It is one thing to recite these demanding limits; it is quite another to abide by them.1 And as Judge Fisher’s dissenting opinion cogently explains, they quickly fall *370by the wayside once the Majority turns to actually reviewing Dennis’s claims. I join Judge Fisher’s opinion in full, but write separately to note that I would reverse the District Court’s judgment even if there were no “analytical gap[s]” in the Pennsylvania Supreme Court’s decision rejecting Dennis’s Brady claims. Fisher Dissent 360-61. Consistent with the text of AED-PA and the precedents of the United States Supreme Court, I would hold that regardless of the thoroughness — or even the correctness — of the Pennsylvania Supreme Court’s stated reasoning, its judgment may not be upset so long as its decision did not contravene or unreasonably apply clearly established federal law and did not rest on an unreasonable determination of the facts. Whatever its flaws, the state court’s decision passes this test.

I

It is a virtue of our judicial system that courts explain their decisions in writing. When an explanation is not good enough— whether due to a legal, logical, factual, or other defect — the decision it supports is often reversed. AEDPA displaces this traditional approach to error review by imposing strict constraints on the writ of habeas corpus designed to stay the hand of federal courts over all but the most glaring of state-court errors. We may issue the writ only “where there is no possibility fairminded jurists could disagree that the state court’s decision conflicts with” the precedents of the Supreme Court. Harrington v. Richter, 562 U.S. 86, 102, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011). For a prisoner in state custody to obtain habeas relief from a federal court, he must demonstrate that the state court’s decision on the claim presented before the federal court “was so lacking in justification that there was an error well understood and comprehended in existing law beyond any possibility for fairminded disagreement.” Id. at 103, 131 S.Ct. 770. “If this standard is *371difficult to meet,” the Supreme Court has explained, “that is because it was meant to be.” Id. at 102,131 S.Ct. 770.

A

By its terms, AEDPA applies to federal review of state-court decisions — not to the specific explanations that support them. See 28 U.S.C. § 2254(d). This distinction might seem technical, but the Supreme Court’s decision in Harrington v. Richter rendered it critical. There,' the Court was faced with the question of AEDPA’s application to a state-court decision that dismissed in a one-sentence summary order a habeas petitioner’s ineffective assistance of counsel claim. 562 U.S. at 96-97, 131 S.Ct. 770. The Court was presented with two issues: whether the state-court decision constituted an “adjudication] on the merits” under AEDPA, and if so, how the Court should go about determining whether the decision was unreasonable under AEDPA given that the opinion provided no reasoning. Id. at 97-102, 131 S.Ct. 770.

The Court’s answer to the first question rested on a straightforward application of AEDPA. Since the text of AEDPA “refers only to a ‘decision’” resulting from an “adjudication” — making no mention of the need for a “statement of reasons” — the Court held that summary decisions unaccompanied by an explanation usually qualify as merits adjudications under AEDPA. Id. at 98,131 S.Ct. 770. Hence, even where the state-court decision under federal review is devoid of reasoning, AEDPA’s deference requirements apply. It followed that “the habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief.” Id. (emphasis added). This rule obtains regardless of “whether or not the state court reveals which of the elements in a multipart claim it found insufficient, for § 2254(d) applies when a ‘claim,’ not a component of one, has been adjudicated.” Id.

The Court’s answer to the second question in Richter — how to assess the reasonableness of a summary state-court decision under AEDPA — is particularly instructive here. The Court held that AEDPA requires federal courts to consider what explanations would nevertheless support the decision under federal law. As the Court explained, “a habeas court must determine what arguments or theories supported or, as here, could have supported, the state court’s decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of this Court.” Id. at 102, 131 S.Ct. 770 (emphasis added). At a minimum, then, when a state-court decision is unaccompanied by an explanation, Richter requires us to ascertain whether it was reasonable.

Circuit co'urts of appeals have divided over whether Richter extends beyond the precise circumstances of that case.2 Those courts that have chosen to cabin Richter can readily point to a limiting principle: single-sentence decisions versus multiple-sentence decisions. That distinction strikes me as unprincipled, however, because neither Richter’s logic nor AEDPA’s text limits the reason-supplying rule to cases in which the state-court “decision” is expressed in just one sentence. A decision is a decision, after all, and AEDPA does not distinguish among one-sentence decisions, one-paragraph decisions, or ten-page decisions; all of them are subject to the same deferential standard. Although the first *372portion of Richter focused on the fact that the state-court decision provided no explanation for the outcome, the reasonableness standard articulated in the rest of the opinion is tied to AEDPA’s general standard itself. “Where a state court’s decision is unaccompanied by an explanation, the habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief.” Id. (emphasis added). In other words, regardless of how extensive or sparse the reasoning of a state-court opinion, the same AEDPA reasonableness test applies to all decisions on the merits.

This approach to AEDPA’s reasonableness standard finds support in Premo v. Moore. There, the petitioner claimed his counsel was ineffective for failing to move for suppression of the petitioner’s confession before advising him regarding a guilty plea. 562 U.S. 115, 119, 131 S.Ct. 733, 178 L.Ed.2d 649 (2011). The state court concluded that the petitioner had not established ineffective assistance of counsel under Strickland v. Washington, reasoning that a “motion to suppress would have been fruitless in light of the other admissible confession by [the petitioner], to which two witnesses could testify.” Id. at 119, 131 S.Ct. 733 (internal quotation marks omitted). Even though the state court “did not specify whether this was because there was no deficient performance under Strickland or because [the petitioner] suffered no Strickland prejudice, or both,” the Supreme Court stated that for a federal habeas court to properly eschew AED-PA deference, it “had to conclude that both findings would have involved an unreasonable application of clearly established federal law.” Id. at 131, 131 S.Ct. 733 (emphasis added).

Athough the state court’s reasoning was quite bare and did not explicitly engage the Strickland prongs, the Court held that its decision was entitled to AEDPA deference because reasons existed that would have supported the decision. Specifically, it highlighted that counsel had explained in state court that his decision to discuss plea bargaining before challenging the petitioner’s confession was based on his rationale that “suppression would serve little purpose in light of [the petitioner’s] other full and admissible confession.” Id. at 123-24, 131 S.Ct. 733. “The state court,” the Supreme Court explained, “would not have been unreasonable to accept this explanation.” Id. at 124, 131 S.Ct.'733 (emphasis added). Indeed, the Court found it unnecessary to consider a second justification that counsel had offered in the underlying proceedings because the first “confirms that his representation was adequate under Strickland, or at least that it would have been reasonable for the state court to reach that conclusion.” Id. (emphasis added). In short, presented with a state-court decision that was not a summary disposition but that provided only some vague reasoning for its decision, the Premo Court looked to the record to posit a rationale that would have supported that decision, finding it not to be an unreasonable application of federal law.3 We should approach Dennis’s case the same way.4

*373B

My understanding of Richter is supported by notions of consistency and coherence as well. If we were to limit Richter to cases involving one-sentence decisions, the outcome of federal review would turn on the state court’s opinion-writing technique. Consider a federal court faced with a state-court decision that rejected a petitioner’s claim that his conviction was invalid because it stemmed from an illegal arrest. Assume the record was unclear with respect to whether the arresting officer had probable cause, but that fairminded jurists could disagree as to whether a Supreme Court precedent demanded the conclusion that there was no probable cause. If the state court re*374jected the petitioner’s claim via summary disposition, Richter requires the reviewing federal court to infer the supportive rationale. Because the record would arguably support probable cause for the arrest, the conviction would be affirmed. But what if the very same claim had been rejected in a partially reasoned state-court opinion with problematic gaps in the logic from which adverse inferences could be drawn or in an opinion that gave incorrect reasons to justify the decision (say, by stating that the arrest was valid because there was “reasonable suspicion”)? Absurdly, appellate courts that circumscribe Richter in the way the Majority has here would require the reviewing federal court to ignore the supportive rationale on de novo review (where a weak case for probable cause wouldn’t be enough) and grant relief.

The asymmetry illustrated by my hypothetical makes a mess of the scheme established by AEDPA. How could a state-court decision be “reasonable” under AEDPA where the state court gives no reasons to explain itself but where we can think of one, yet be “unreasonable” under AEDPA where — although the very same good reason to support the decision exists — the decision is supported by undeveloped or incorrect reasons?5 See Mann, 774 F.3d at 1224-25 (Kozinski, J., concurring and dissenting) (“A habeas petitioner is not entitled to any reasoning at all, so reversing a state court’s reasonable decision on the grounds of incorrect reasoning risks treating defendants inconsistently: Those who are given incorrect reasoning get relief while those who aren’t given any reasoning do not.”). To make AEDPA reasonableness turn on a state court’s drafting decision is inconsistent with AEDPA’s directive that federal courts review the reasonableness of decisions, not opinions. And because it makes AEDPA deference inversely proportional to the amount of information the state court provides, it creates a perverse incentive for state courts to earn the deference of federal courts by saying less.6

II

To sum up, I would hold that when gaps or errors afflict a state court’s habeas adjudication, federal courts may not reverse unless the decision itself is unreasonable. In Dennis’s case, this principle is most pertinent to the Cason receipt. As Judge Fisher explains, the reasons proffered by the Pennsylvania Supreme Court for rejecting Dennis’s Brady claims regarding the Howard police activity report and the Frazier documents are themselves sufficient to pass AEDPA review without any inference from us. The Pennsylvania Supreme Court’s analysis of, the Cason receipt, on the other hand, is incomplete and might ungenerously be read as incorrect. For the reasons explained by Judge Fisher, however, a rationale consistent with Supreme Court precedent supports the decision, and so it must stand. I would simply add that AEDPA would require us to supply this rationale even if the state court’s treatment of the Cason receipt were in fact wrong. After all, “[a] state court could *375write that it rejected a defendant’s claim because Tarot cards dictated that result, but its decision might nonetheless be a sound one.” Brady, 711 F.3d at 827 (Wood, J.).

In my view, AEDPA requires federal courts to take the following approach to habeas review. Where the state court denies relief summarily, Richter requires federal courts to consider what arguments or theories could have supported the state court’s decision such that fairminded jurists could disagree whether those arguments or theories are inconsistent with the holding in a prior decision of the Supreme Court. Where the state court denies relief but addresses only certain prongs of a test or components of a claim, the reviewing federal court should likewise consider what reasons regarding an unaddressed prong or component could have supported the decision. And where, as here, the state court denies relief through vague, ambiguous, incomplete, or even incorrect reasoning, AEDPA still requires the reviewing federal court to consider what theories could have supported the decision, under AEDPA.7

By ignoring these principles, the Majority empowers itself to reweigh evidence that is decades old. Like the District Court, the Majority takes a fresh look at the evidence and concludes, contrary to the consistent testimony of three eyewitnesses, that the alleged Brady violations “effectively gutted the Commonwealth’s case against Dennis.” Majority Op. 269 (emphasis added). AEDPA proscribes such searching review. Because fairminded jurists could disagree as to whether the *376Pennsylvania Supreme Court’s decision was inconsistent with federal law, we owe it our deference. I respectfully dissent from the Majority’s decision to do otherwise.

. The inability of federal courts to follow AEDPA has reached epidemic proportions. As I pointed out in 2012, since 2000

the Supreme Court has granted certiorari in ninety-four cases arising under AEDPA, forty-six of which involved questions of federal court deference to decisions of state courts. Thirty-four of those cases (approximately seventy-four percent) have been reversed because the court of appeals failed to afford sufficient deference to the state court. Remarkably, twenty-two of those cases — almost fifty percent — were reversed without dissent.

Garrus v. Sec’y of Pennsylvania Dep't of Corr., 694 F.3d 394, 412-14 (3d Cir. 2012) (en banc) (Hardiman, J., dissenting) (collecting cases).

In the four short years since we decided Garrus, the errors have continued apace. By my count, of the nineteen cases arising under AEDPA in which the Supreme Court has granted certiorari, fourteen involved questions of federal court deference to state-court decisions. Thirteen of those cases were reversed — ten without dissent. See Kernan v. Hinojosa, -U.S. -, 136 S.Ct. 1603, 1606, 194 L.Ed.2d 701 (2016) (per curiam) (reversing the Ninth Circuit's treatment of a summary decision as a non-merits adjudication and noting that "the Ninth Circuit has already held that state-court denials of claims identical to [the petitioner's] are not contrary to clearly established federal law”); Woods v. Etherton,-U.S. -, 136 S.Ct. 1149, 1153, 194 L.Ed.2d 333 (2016) (unanimously reversing the Sixth Circuit because "a fairminded jurist — applying the deference due the state court under AEDPA — could certainly conclude that the court was not objectively unreasonable in deciding that appellate counsel was not incompetent under Strickland, when she determined that trial counsel was not incompetent under Strickland”); White v. Wheeler,-U.S.-, 136 S.Ct. 456, 461-62, 193 L.Ed.2d 384 (2015) (unanimously reversing the Sixth Circuit’s grant of habeas relief because it "did not properly apply the deference it was required to accord the state-court ruling”); Davis v. Ayala, - U.S. -, 135 S.Ct. 2187, 2208, 192 L.Ed.2d 323 (2015) (reversing the Ninth Circuit’s grant of the writ on the ground that fairminded jurists could disagree as to whether a state court’s exclusion of a defendant's attorney from part of a Batson hearing was harmless error); Woods v. Donald,-U.S.-, 135 S.Ct. 1372, 1377, 191 L.Ed.2d 464 (2015) (unanimously reversing the Sixth Circuit’s grant of habeas relief because the state court's conclusion that the petitioner's counsel was not per se ineffective "was not contrary to any clearly established holding” of the Court); Glebe v. Frost, - U.S. -, 135 S.Ct. 429, 430, 190 L.Ed.2d 317 (2014) (unanimously reversing the Ninth Circuit's conclusion that the state court "unreasonably applied clearly established federal law by failing to classify the trial court’s re*370striction of closing argument as structural error" because no Supreme Court precedent clearly established that such mistakes rank as structural error); Lopez v. Smith, - U.S. -, 135 S.Ct. 1, 5, 190 L.Ed.2d 1 (2014) (unanimously reversing the Ninth Circuit where it “had no basis to reject the state court’s assessment that [the petitioner] was adequately apprised of the possibility of conviction on an aiding-and-abetting theory”); White v. Woodall, - U.S. -, 134 S.Ct. 1697, 1702-04, 188 L.Ed.2d 698 (2014) (reversing the Sixth Circuit’s grant of habeas relief because the state court’s determination that the trial court’s jury instructions did not violate clearly established federal law was not "objectively unreasonable”); Burt v. Titlow, - U.S. -, 134 S.Ct. 10, 17-18, 187 L.Ed.2d 348 (2013) (reversing without dissent the Sixth Circuit’s judgment that the state court's conclusion that counsel’s performance was ineffective was unreasonable); Nevada v. Jackson,-U.S.-, 133 S.Ct. 1990, 1994, 186 L.Ed.2d 62 (2013) (per curiam) (unanimously reversing the Ninth Circuit's grant of habeas relief where the state court reasonably applied federal law in determining that the petitioner had not been denied the right to present a complete defense when he was not allowed to present certain, extrinsic evidence); Metrish v. Lancaster, — U.S.-, 133 S.Ct. 1781, 1792, 185 L.Ed.2d 988 (2013) (unanimously reversing the Sixth Circuit’s grant of the writ where Supreme Court had "never found a due process violation in circumstances remotely resembling [the petitioner’s] case”); Johnson v. Williams, - U.S. -, 133 S.Ct. 1088, 1097, 185 L.Ed.2d 105 (2013) (reversing without dissent the Ninth Circuit’s grant of relief based on the faulty conclusion that the state court had overlooked a meritorious Sixth Amendment claim); Ryan v. Gonzales, -U.S. -, 133 S.Ct. 696, 700, 184 L.Ed.2d 528 (2013) (killing two birds with one stone in unanimously reversing both the Sixth Circuit’s and Ninth Circuit's grants of relief where the courts wrongly concluded that federal law provides a right to incompetent prisoners to suspend their federal habeas proceedings); but see Brumfield v. Cain, — U.S.-, 135 S.Ct. 2269, 2281, 192 L.Ed.2d 356 (2015) (finding the state court's determination of the facts regarding a defendant with an IQ of 75 unreasonable).

. See Noam Biale, Beyond A Reasonable Disagreement: Judging Habeas Corpüs, 83 U. Cin. L. Rev. 1337, 1391 (2015) ("Since Richter ... the circuits have split on whether the opinion’s ‘could have supported’ language for decisions unaccompanied by a reasoned opinion applies to decisions that do include a reasoned opinion.”).

. The Majority and Judge Jordan conclude that the Supreme Court's decision in Lafler v. Cooper “implies a limit” to the reason-supplying rule announced in Richter. Jordan Concurrence 351. I do not read Lafler that way. Significantly, habeas relief in that case rested on the Supreme Court's holding that Michigan Court of Appeals’ application of Strickland was "contrary to" — not an “unreasonable application of” — clearly established federal law. - U.S. --, 132 S.Ct 1376, 1390, 182 L.Ed.2d 398 (2012). Specifically, rather than applying the Strickland ineffective-assistance-of-counsel standard, the state court applied a (completely wrong) "knowing and voluntary” plea rejection rule. Id. Because a decision is categorically “contrary to” clearly established federal law if the state court “applies a rule that contradicts the governing law,” Williams v. Taylor, 529 U.S. 362, 405, 120 S.Ct. 1495, 146 L.Ed.2d 389 *373(2000), AEDPA deference was inappropriate, so de novo review applied. Lafler, 132 S.Ct. at 1390. Consequently, the case was not amenable to Richter's "could have supported” analysis to determine whether the state court decision was an unreasonable application of federal law. For these reasons, I disagree with the Majority and Judge Jordan that La-fler instructs federal courts to "take the state court's decision as written” and apply Richter only in the small subset of cases in which the state court left so-called "gaps” to be filled. Jordan Concurrence 351.

Nor do I read Wetzel v. Lambert to imply any limit on Richter. Although the opinion in that case did not include Richter’s "theories [that] ... could have supported” language in its recitation of AEDPA's general standard, see-U.S. -, 132 S.Ct. 1195, 1198, 182 L.Ed.2d 35 (2012), the Court did not reject that approach by implication. Rather, in Wet-zel the reasons for upholding the state court’s decision under AEDPA were expressed by the state court. The petitioner claimed the prosecution violated Brady by suppressing a police activity sheet consisting of a photo display marked with written notations suggesting that "someone other than or in addition to” the petitioner had committed the crime. Id. at 1196-97. We granted habeas relief, but the Supreme Court vacated and remanded, explaining that we had "overlooked the determination of the state courts that the notations were ... 'not exculpatory or impeaching' but instead 'entirely ambiguous.”' Id. at 1198. The Court criticized us for "focus[ing] solely on the [state court’s] alternative ground that any impeachment value that might have been obtained from the notations would have been cumulative.” Id. The problem was that "[i]f the conclusion in the state courts about the content of the document was reasonable — not necessarily correct, but reasonable — whatever those courts had to say about cumulative impeachment evidence would be beside the point.” Id. Hence, by failing to recognize — as the state courts did' — the " 'ambiguous' nature of the notations” and the “ speculative]’ nature of [the petitioner's] reading of them,” we ran afoul of AEDPA. Id. Far from implying a limitation on Richter, Wetzel merely requires federal habeas courts to review state court opinions in search of a reasonable reading that would support the decision under federal law.

. Some courts have begun to recognize Richter s true reach. See, e.g, Holland v. Rivard, 800 F.3d 224, 235_(6th Cir. 2015) (concluding that although "a state court decision unaccompanied by any explanation differs from a state court decision based on erroneous reasoning ... Richter suggests that this is not a meaningful distinction” and that AEDPA requires a habeas petitioner to show that there was "no reasonable basis for the state court to deny relief ... whether or not the state court reveals [its reasoning]”); Trottie v. Stephens, 720 F.3d 231, 240-41 (5th Cir. 2013) (“We review only the ultimate legal determination by the state court — not every link in its reasoning.”); Brady v. Pfister, 711 F.3d 818, 827 (7th Cir. 2013) (Wood, J.) ("[I]t is clear that a bad reason does not necessarily mean that the ultimate result was an unreasonable application of established doctrine.... If a state court's rationale does not pass muster ... for Section 2254(d)(1) cases, the only consequence is that further inquiry is necessary.”); Mann v. Ryan, 774 F.3d 1203, 1224-25 (9th Cir. 2014) (Kozinski, J., concurring and dissenting) ("I have misgivings about whether, in light .of the Supreme Court's decision in Richter, we are still entitled to reverse a state court’s reasonable decision based on what we consider to be its incorrect reasoning.... After Richter, it seems clear that we should *374assess the reasonableness of a state court’s decision, not its reasoning.”).

. Such arbitrariness is all the more perplexing in light of the fact that AEDPA "does not require citation of [Supreme Court] cases— indeed, it does not even require awareness of [Supreme Court] cases.” Early v. Packer, 537 U.S. 3, 8, 123 S.Ct. 362, 154 L.Ed.2d 263 (2002).

. See Hodges v. Colson, 727 F.3d 517, 537 n.5 (6th Cir. 2013) ("[If Richter is limited to summary dispositions], the more information the state court provides, the less deference we grant it. This is contrary not only to the language of the statute, which speaks of 'claims' not components of claims, but also contrary to the spirit of § 2254(d), which is designed to give more deference to a state court judgment on the merits.”).

. I disagree with Judge Jordan that my understanding of Richter conflicts with Ylst v. Nunnemaker and Wiggins v. Smith. Both of those cases involved the threshold question of whether the petitioners' claim had been decided on the merits. The Ylst Court was faced with an "unexplained” State supreme court order denying the petitioner's habeas petition, wherein it was unclear whether the court rested its denial on a procedural default (the basis of the lower court's holding) or on the merits of his Miranda claim. 501 U.S. 797, 801, 111 S.Ct. 2590, 115 L.Ed.2d 706 (1991). The Supreme Court reversed the Ninth Circuit's conclusion that the decision was on the merits, explaining that, "where, as here, the last reasoned opinion on the claim explicitly imposes a procedural default, we will presume that a later decision rejecting the claim did not silently disregard that bar and consider the merits.” Id. at 803, 111 S.Ct. 2590. To the extent that Ylst requires us to "look through” unreasoned state court opinions to the last reasoned opinion, I have no quarrel with Judge Jordan that we ought to first consider whether the state court's stated explanation is reasonable before deigning to supply reasons of our own under Richter. As for Wiggins, we have explained that the reason the Court declined to apply deference with respect to the prejudice prong of the petitioner's Strickland claim was that the state courts had not decided the Strickland prejudice issue "on the merits.” Palmer v. Hendricks, 592 F.3d 386, 400 (3d Cir. 2010); see also Wiggins, 539 U.S. 510, 534, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003) (“[Ojur review is not circumscribed by a state court conclusion with respect to prejudice, as neither of the state courts below reached this prong of the Strickland analysis.”). Because AEDPA deference only extends to "any claim that was adjudicated on the merits in State court proceedings,” 28 U.S.C.A. § 2254(d), the determination whether the state-court decision under federal review was made on the merits is prior to the consideration, vel non, of whether adequate reasons exist in support of that decision. I do nevertheless agree with Judge Jordan that Wiggins is in some tension with my approach because it engaged in de novo review of the second prong of Strickland even though the state court denied relief but addressed only the first prong. However, Richter — decided after Wiggins — speaks clearly on this point. "[Aj^habeas petitioner's burden still must be met by showing there was no reasonable basis for the state court to deny relief.... whether or not the state court reveals which of the elements in a multipart claim it found insufficient, for § 2254(d) applies when a 'claim,' not a component of one, has been adjudicated.” Richter, 562 U.S. at 98, 131 S.Ct. 770.