Teshome Campbell v. Dan Reardon

In the United States Court of Appeals For the Seventh Circuit ____________________ No. 13-2634 TESHOME CAMPBELL, Petitioner-Appellant, v. DAN REARDON, Warden, Respondent-Appellee. ____________________ Appeal from the United States District Court for the Central District of Illinois, Urbana Division. No. 08-2157—Harold A. Baker, Judge. ____________________ ARGUED NOVEMBER 10, 2014 — DECIDED MARCH 10, 2015 ____________________ Before WOOD, Chief Judge, and ROVNER and HAMILTON, Circuit Judges. HAMILTON, Circuit Judge. Petitioner Teshome Campbell was convicted in an Illinois state court of first-degree murder for participating in a mob-style beating that killed James Shepherd. No physical evidence linked Campbell to the crime. The State’s case against Campbell hinged on the tes- timony of three eyewitnesses. Two of these witnesses, Dami- 2 No. 13-2634 on Johnson and Steven Peete, had been charged with the murder but were granted full immunity in exchange for tes- tifying against Campbell. Both had serious criminal histo- ries, and both denied any involvement in the crime when they testified at Campbell’s trial. The third witness, Rita But- ler, testified that Campbell started the fight with the victim, but she saw the fight from inside a van that was parked around the street corner and facing away from the brawl. In his federal habeas petition, Campbell argues that his trial counsel was ineffective under Strickland v. Washington, 466 U.S. 668 (1984). He contends that a number of errors support his claim. For example, defense counsel failed to impeach Johnson and Peete with the full scope of their crim- inal histories and the details of their plea agreements with the State. Campbell also criticizes his lawyer’s failure to at- tend the trial (or to review trial transcripts) of a co-defendant who was acquitted in a separate trial just weeks before Campbell’s trial. We do not reach these issues, however, but focus instead on the core of Campbell’s petition—that his counsel failed to conduct an adequate pretrial investigation by failing to interview three witnesses who would have said that Campbell played no role in the fatal attack on Shepherd. These witnesses and at least part of the information they could provide were in police reports provided to Campbell’s lawyer before trial. Toni Leonard told police that two other men—not Campbell—started the fight with the victim. This statement flatly contradicted prosecution witness Rita But- ler’s testimony, as well as the State’s entire theory of the case that the fight began when Campbell attacked the victim. Ms. Leonard also told police that prosecution witness Damion Johnson participated in the beating, which contradicted No. 13-2634 3 Johnson’s testimony that he was not involved. Another eye- witness, Leroy Hunter, told police that prosecution witness Steven Peete hit the victim three or four times with “some- thing that looked like a pipe, or a big stick.” This statement contradicted Peete’s testimony that he was also not involved. A third eyewitness, Ieca Hunter, witnessed the beating but told police that she could not identify any of the assailants. According to an affidavit that was presented to the state courts, though, she would have been able to identify Camp- bell as someone present at the scene who did not participate in the beating. According to her affidavit, Campbell was standing directly in front of her house, apart from the group of men who participated in the beating, and he “never touched” the victim. Neither Ms. Leonard, Mr. Hunter, nor Ms. Hunter was called to testify in Campbell’s defense at tri- al. The state courts assumed, as we must on appeal, that Campbell’s lawyer never contacted these eyewitnesses to in- terview them. The state courts rejected Campbell’s ineffec- tive assistance of counsel claim anyway, holding that under Strickland (1) counsel receives wide latitude to make strategic decisions about which witnesses to call and which theory or theories to present to the jury, and (2) Campbell’s lawyer rea- sonably declined to interview Ms. Hunter because the police report implied that such an interview would be fruitless. We hold that the state courts unreasonably applied Strickland when they rejected Campbell’s ineffective assistance of coun- sel claim. That determination does not entitle Campbell to a writ of habeas corpus, at least not yet, but calls for further proceed- ings to determine the actual facts. The state courts and the 4 No. 13-2634 federal district court rejected Campbell’s ineffective assis- tance of counsel claim without ever addressing key factual questions: whether Campbell’s lawyer in fact interviewed these witnesses and whether they would have testified cred- ibly and consistently with their affidavits and statements to the police. On these factual issues, we have no factual find- ings to review and the record is otherwise ambiguous. To be sure, affidavits from two of the three exculpatory witnesses (Mr. Hunter and Ms. Hunter) say that Campbell’s lawyer never contacted them. But these affidavits, which were part of the state court record, have not been tested through the adversarial process at any kind of hearing. There also is no affidavit from the third exculpatory witness (Ms. Leonard) or the lawyer himself, let alone a record of live testimony. Accordingly, we reverse the denial of Campbell’s petition and remand to the district court to resolve these factual questions, as we explain in more detail below. I. Factual and Procedural Background A. The Murder of James Shepherd James Shepherd was beaten to death by a group of men in the early morning hours of Christmas Day in 1997. The beating occurred on a street in Champaign, Illinois. The State initially charged twelve men with first-degree murder, but it eventually dismissed charges against four of them. 1 Three defendants entered plea agreements and were sentenced to 1 The State dismissed charges against Deandrea Donald, Jeffrey Dil- lon, Terron Lyons, and Joe Posey. The charges against Donald, Dillon, and Lyons were dismissed outright. Posey’s charges were dismissed in exchange for him pleading guilty to an unrelated felony. No. 13-2634 5 varying terms of imprisonment. 2 Two other defendants, Damion Johnson and Steven Peete, were granted full im- munity in exchange for testifying against the remaining three defendants. 3 One defendant, Theodis White, was con- victed and sentenced to 45 years’ imprisonment. Another, Bobby Joe Douglas, was tried and acquitted. The remaining defendant was petitioner Teshome Campbell, who was con- victed and sentenced to 55 years in prison. Campbell main- tains his innocence. B. Campbell’s Trial Before trial Campbell asked the court to appoint counsel for him. In January 1998 the court appointed attorney Har- vey Welch to represent him. Although Campbell was facing a first-degree murder charge, his lawyer did not record do- ing any work on Campbell’s case until five months later, in June 1998. The first twelve entries on the lawyer’s billing statement are for the preparation of a trial notebook. No en- tries indicate that the lawyer conducted any pretrial witness interviews. Nor is there any evidence that the lawyer hired a private investigator. Campbell went to trial in October 1998. No physical evi- dence linked Campbell to the crime. The prosecution’s case rested almost entirely on the testimony of three eyewitness- es: Damion Johnson, Steven Peete, and Rita Butler. Johnson and Peete had been charged initially with Shepherd’s mur- der and testified against Campbell in exchange for full im- 2 These co-defendants were Juanchez Booker, Lynntez Holt, and Lamarcus Townsend. 3 Damion Johnson is also referred to as “Damion Holt” throughout the record. 6 No. 13-2634 munity. Both had extensive criminal histories. Butler was with Shepherd before his death on the night of the beating. 1. Rita Butler’s Testimony Rita Butler testified that she was walking home early Christmas morning when Shepherd pulled up beside her in a van and asked her to get in. 4 She agreed. Shepherd asked Butler if she knew where they could buy some crack cocaine. Butler directed him to the 200 block of Bellefontaine Street. When they arrived, Butler saw a man on the street from whom she had bought drugs before. She did not know the man by name but recognized his face, and she later identi- fied Campbell as that man. She took money from Shepherd, approached the man on the street, and bought a $20 bag of what she believed to be crack cocaine and returned to the van. Butler and Shepherd then drove to another location to smoke it. When Shepherd tried to smoke it, he discovered that the substance was fake. Intent on getting his money back, Shep- herd drove back toward Bellefontaine Street and parked the van around a corner, facing away from where Butler had ini- tially bought the fake drugs from the man standing on Belle- fontaine Street. Shepherd got out of the van. Butler stayed in the passenger seat. From inside the van, she looked back to- ward the corner of Bellefontaine Street and watched what happened next. Butler testified that there were “a lot of people standing out” on the street. Butler saw Shepherd approach the man 4 Throughout her testimony, Butler referred to Shepherd as “Jimmy Collins.” No. 13-2634 7 who had sold her the fake cocaine and demand his money back. The man shouted back at Shepherd, and the two men began to fight. She did not see who threw the first blow. As the two men fought, eight or nine other people joined in and began hitting and kicking Shepherd. Shepherd fought back but eventually fell to the ground. The last time Butler saw Shepherd, there was “a crowd” of men beating him as he lay in the middle of the street. She ran away. 2. Damion Johnson’s Testimony The prosecution’s next eyewitness was Damion Johnson. He was arrested in late January 1998. Upon his arrest, he gave police a taped statement about what he saw on the night of the incident. According to this original taped state- ment, Shepherd approached Campbell in the street. Shep- herd yelled at Campbell, demanding his money back. Campbell and Shepherd then began fighting. Bobby Joe Douglas jumped into the fight, but Campbell fought his way out of the melee and walked over to the sidewalk. From there he watched the fight escalate, but he never rejoined it. According to Johnson’s statement, Douglas, Lamarcus Townsend, Juanchez Booker, and Theodis White were the primary participants in the fatal beating. In April 1998, though, Johnson changed his story. In exchange for full immunity on the first-degree murder and related charges, he agreed to testify against Campbell. At trial Johnson testified that on the night of the incident, he was inside a cousin’s residence until he heard gunshots. He walked outside and saw a van driving down Bellefontaine Street. The man driving the van was yelling that he wanted “the real thing or his money back.” The driver parked the van and approached Campbell on foot. A fight broke out. 8 No. 13-2634 Johnson testified that Campbell, Booker, Townsend, White, and Douglas all participated in the beating. According to Johnson, Shepherd did not fight back. At Campbell’s trial, Johnson denied any participation in the fatal beating. He said he witnessed the incident while standing on a sidewalk, away from the fight. He also testi- fied that he did not see Steven Peete kick or punch Shepherd. When Johnson heard police sirens, he fled the scene. 3. Steven Peete’s Testimony The prosecution’s final eyewitness to the murder was Steven Peete. He testified that on the night of the incident he was asleep in his house when his fiancée woke him up, say- ing that she had heard gunfire. He opened the front door and looked outside. He saw a man being beaten directly in front of his house, about 10 feet away. Peete testified that when he first saw the man, he was already on the ground, trying to fight back. Seven or eight people had surrounded him and were kicking and punching him. Peete testified that there was a streetlight near his house, and he could see several of the men’s faces. He identified White, Douglas, Booker, Townsend, and Campbell as parti- cipants in the beating. He testified that Johnson was in front of his bushes outside of his house, and that he was not part of the crowd beating Shepherd. Peete yelled at the crowd to get away from his house, and the men dispersed. After the crowd dispersed, Peete went inside his house and back to sleep. Like Johnson, he denied any participation in the beat- ing. No. 13-2634 9 4. The Defense Trial Strategy Defense counsel did not call any witnesses on Campbell’s behalf. The only evidence he put on was a stipulation that during the investigation, police seized several items of cloth- ing from Douglas and White that had Shepherd’s blood on them. During cross-examinations of Johnson and Peete, defense counsel did not impeach them with the full scope of their criminal histories or the details of their plea agreements with the State. In particular, counsel did not impeach Johnson with a prior juvenile adjudication for felony aggravated bat- tery. Nor did he impeach Peete with a prior juvenile adjudi- cation for felony mob action—a crime that took place less than three years before Shepherd’s beating. 5 Both of these crimes were among the charges against Campbell. 6 Defense counsel also failed to elicit testimony that at the time Johnson agreed to testify against Campbell, he was fac- ing prison time for two reasons separate and apart from the 5 Felony mob action, under Illinois law, consists of “the use of force or violence disturbing the public peace by 2 or more persons acting to- gether and without authority of law.” 720 Ill. Comp. Stat. 5/25-1 (1995). 6 Because defense counsel never cross-examined Peete about his vio- lent criminal history, the prosecution was able to tell the jury in closing argument that Peete was “as clean as they get.” The prosecution also told the jury in closing: “Did you hear any evidence about Steven Peete’s hor- rible criminal record? Did you hear anything about Steven Peete other than that he’s a 19 year old with a child, a girlfriend, a job, living right in the middle of the war zone of Bellefontaine or this lawless area? Did you hear anything else about him?” 10 No. 13-2634 dropped first-degree murder charges: (1) he had been condi- tionally discharged on an adult felony conviction for unlaw- ful use of a weapon in 1997, only one year before Shepherd’s murder, and (2) there was a pending petition to revoke his probation for a domestic battery incident. Both of these facts gave Johnson an even stronger incentive to cooperate with the State. The jury never heard these details. During closing argument, Campbell’s lawyer offered four reasons for finding that the prosecution had not proven its case against Campbell beyond a reasonable doubt. First, he argued that it was too dark on the night of the beating for anyone to identify any suspect reliably. This theory was based on the testimony of a single paramedic, called by the prosecution, who arrived on the scene after the group of men had already fled. The paramedic testified that there was “poor lighting in the area” and called the scene “quite dark.” Second, counsel argued that Shepherd’s injuries were incon- sistent with the testimony of Johnson, Peete, and Butler, who testified that Campbell struck and kicked Shepherd all over his body when, according to the testimony of a medical ex- pert, Shepherd’s injuries were “confined from his neck area up.” Third, counsel observed that no physical evidence linked Campbell to the crime. And fourth, counsel argued that it was reasonable to infer from the evidence that Shep- herd got out of his van with a gun and provoked the group of men into beating him, but that Campbell did not partici- pate. The jury convicted Campbell of first-degree murder. He was sentenced to 55 years in prison. C. Campbell’s Direct Appeal Campbell appealed, arguing among other things that de- fense counsel was constitutionally ineffective for failing to No. 13-2634 11 present the testimony of two exculpatory witnesses, Leroy Hunter and Toni Leonard. According to police reports in the state court record, Mr. Hunter lived on the street where the murder took place and was interviewed on February 3, 1998, after calling police to report that he had witnessed the beating. He told police that he was awakened by gunfire that night. He went to his front window and saw a man lying in the street. He saw several men running away, while two men remained standing over the victim. The two men were Joe Posey and Steven Peete. Mr. Hunter knew Peete before the incident because Peete lived across the street from him. Mr. Hunter told police that Posey struck the victim with a pistol and Peete struck the victim three or four times with “something that looked like a pipe, or a big stick.” He also told police that on the day after the incident, he saw Peete return to the spot where the vic- tim had been lying. There was a bloodstain in the street, and Peete spit on it. Mr. Hunter perceived this gesture as a sign of disrespect toward the victim. During his conversations with police, Mr. Hunter never identified Campbell as a par- ticipant in the beating. Toni Leonard was also interviewed by police. She was first interviewed on January 23, 1998, after she was arrested on an unrelated charge. She told police she witnessed the beating. According to the police report, Ms. Leonard had been walking toward Bellefontaine Street when she heard shouting. She saw Shepherd arguing with Jeffrey Dillon and Lynntez Holt about “dummy bags” of drugs. Dillon struck the victim in the head “very hard” and then Holt “hit him several times.” The victim fought back, and Dillon pulled out a pistol and struck the victim in the head with it twice. A 12 No. 13-2634 crowd soon formed around the altercation. Deandrea Don- ald took the pistol from Dillon and struck the victim in the head with it. Terron Lyons jumped on the victim’s head. Johnson kicked the victim at least once. And Posey kicked the victim at least twice. After her initial interview, the police asked Ms. Leonard to identify suspects in a series of lineups. In total, she posi- tively identified nine men as directly involved in Shepherd’s murder, including Johnson. During this same period, Ms. Leonard viewed an in-person lineup that included Camp- bell. She did not identify him as involved in the murder. She identified another man instead. A divided state appellate panel affirmed Campbell’s con- viction over a strong dissent. People v. Campbell, 773 N.E.2d 776 (Ill. App. 2002). The court held that counsel’s failure to call either Mr. Hunter or Ms. Leonard did not constitute inef- fective assistance of counsel “because it was a matter of trial strategy.” Id. at 785. The court also noted that “none of the testimony which [Campbell] claims Hunter and Leonard would have given would have exonerated” Campbell be- cause all they could do was “further implicate Johnson and Peete as involved in the murder, a fact of which the jury was already aware,” and because testimony that Campbell did not start the beating would not have contradicted the State’s theory that Campbell participated in the beating. Id. Justice Knecht disagreed and would have remanded for a new trial. He concluded that defense counsel’s failure to im- peach Johnson and Peete with the full scope of their criminal histories and the details of their plea agreements with the State, as well as his failure to call Mr. Hunter and Ms. Leon- ard, constituted ineffective assistance of counsel. Id. at 787– No. 13-2634 13 90 (Knecht, J., dissenting). Recognizing that the decision to call a witness is generally considered a matter of trial strate- gy, Justice Knecht called counsel’s failure to call Mr. Hunter and Ms. Leonard “egregious,” explaining: “Defense coun- sel’s failure to call Hunter and Leonard could hardly be called trial strategy. I can fathom no strategy in this case that could overcome the value of their potential testimony.” Id. at 790. Campbell’s petition for leave to appeal to the Supreme Court of Illinois was denied. People v. Campbell, 787 N.E.2d 160 (Ill. 2002) (Table). D. State Post-Conviction Proceedings Campbell then filed a petition for post-conviction relief under 725 Ill. Comp. Stat. 5/122-1 et seq. (2002), raising, as relevant here, the new claim that his trial counsel was also constitutionally ineffective for failing to call Ieca Hunter to testify. Ms. Hunter, the daughter of Mr. Hunter, was first in- terviewed by police on the night of the murder. She lived on the same street as her parents, and police interviewed her on the scene. According to the police report, she had Christmas lights on the outside of her house. The lights were “affect- ing” her vision, but she was nevertheless able to see five black men fighting with the victim. When the victim fell to the ground, she left her window and called 911 for an ambu- lance. The police report said: “Ieca stated that she did not recognize any of the subjects in [the] street and she would not be able to identify any of them. The only thing that she could tell me was that all of them were black males.” In an affidavit attached to Campbell’s post-conviction pe- tition for relief, Ms. Hunter explained that although she ini- tially had “trouble seeing in full view because of the Christ- mas lights” on her door, she then opened the door for a bet- 14 No. 13-2634 ter view and saw “a familiar person standing directly in front of [her] house,” approximately 12 feet away. That per- son was Teshome Campbell. She “had interacted with [him] on numerous occasions and there was no mistaking him even in the dark.” According to her affidavit, Campbell “never touched” Shepherd. She also described her initial in- terview with the police on the night of the incident: “[The police officer] asked me if I knew who did the beating. I told him that I could not see the beating well and did not know the people. He did not ask me who else was present who I knew. If asked by the police, I would have said that I saw [Campbell] present but he was not involved in the beating.” The state trial court dismissed the petition without an ev- identiary hearing. 7 Campbell appealed, and the state appel- late court affirmed. People v. Campbell, No. 4-04-0791 (Ill. App. Jan. 16, 2007) (unpublished order). (That procedural posture has important effects for these federal proceedings because we must take Campbell’s allegations and affidavits at face value.) The court explained: “[Ms.] Hunter told police when they interviewed her that the Christmas lights ob- scured her view of the beating and she did not know any of the men involved. Based on this statement, defendant’s at- torney’s failure to speak with her himself and call her as a witness was not objectively unreasonable as there was not a 7 As detailed in the state appellate court’s order affirming the dis- missal of Campbell’s petition for post-conviction relief, this petition was actually an amended petition. See Supp. App. 85–86. The original pro se petition was dismissed summarily on res judicata grounds, but that dis- missal was reversed by the state appellate court. People v. Campbell, 803 N.E.2d 1047, 1051 (Ill. App. 2004), abrogated by People v. Blair, 831 N.E.2d 604 (Ill. 2005). Campbell then filed the amended petition with the assis- tance of counsel. No. 13-2634 15 sound basis to believe she had any information that would be helpful to the defense.” Supp. App. 96. Campbell’s peti- tion for leave to appeal to the Supreme Court of Illinois was denied. E. Federal Habeas Proceedings Campbell then filed a petition for a writ of habeas corpus in federal district court, raising, among other claims, ineffec- tive assistance of trial counsel. The court denied the petition without an evidentiary hearing but granted a certificate of appealability under 28 U.S.C. § 2253(c) on the ineffective as- sistance of counsel claim. Campbell appeals from the denial of his petition. II. Habeas Corpus Review Under 28 U.S.C. § 2254 We have jurisdiction under 28 U.S.C. § 2253(a), and we review de novo the district court’s denial of habeas corpus re- lief. E.g., Harris v. Thompson, 698 F.3d 609, 622 (7th Cir. 2012). Federal courts have authority to issue writs of habeas corpus for persons in state custody under 28 U.S.C. § 2254, as amended by the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA). Under AEDPA, a federal habeas peti- tion may be granted only if a state court’s ruling on a federal constitutional question “was contrary to, or involved an un- reasonable application of, clearly established Federal law, as determined by the Supreme Court,” or “was based on an un- reasonable determination of the facts in light of the evidence presented in the State court proceeding.” § 2254(d)(1) & (2). The standard under AEDPA is “difficult to meet” and “highly deferential.” Cullen v. Pinholster, 563 U.S. —, 131 S. Ct. 1388, 1398 (2011). Federal courts must avoid “using fed- eral habeas corpus review as a vehicle to second-guess the 16 No. 13-2634 reasonable decisions of state courts.” Parker v. Matthews, 567 U.S. —, 132 S. Ct. 2148, 2149 (2012) (per curiam), quoting Re- nico v. Lett, 559 U.S. 766, 779 (2010). A petitioner cannot pre- vail by showing simply that the state court’s decision was wrong. E.g., White v. Woodall, 572 U.S. —, 134 S. Ct. 1697, 1702 (2014). He instead “must show that the state court’s rul- ing on the claim being presented in federal court was so lacking in justification that there was an error well under- stood and comprehended in existing law beyond any possi- bility for fairminded disagreement.” Harrington v. Richter, 562 U.S. 86, 103 (2011). Our review under § 2254(d) is limited to the record that was before the state court. Pinholster, 131 S. Ct. at 1398. AEDPA’s deferential standard of review applies only to claims that were actually “adjudicated on the merits in State court proceedings.” § 2254(d). Where state courts did not reach a federal constitutional issue, § 2254(d) deference ap- plies “only to those issues the state court explicitly ad- dressed.” Quintana v. Chandler, 723 F.3d 849, 853 (7th Cir. 2013), citing Wiggins v. Smith, 539 U.S. 510, 534 (2003) (“In this case, our 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 … analysis.”). The operative decision under review is that of the last state court to address a given claim on the merits. See Greene v. Fisher, 565 U.S. —, 132 S. Ct. 38, 45 (2011). III. Ineffective Assistance of Counsel The Sixth Amendment provides that “the accused shall enjoy the right to … have the Assistance of Counsel for his defence.” To demonstrate that his right to counsel was vio- lated by ineffective assistance, Campbell must satisfy the No. 13-2634 17 familiar two-prong test set forth in Strickland v. Washington, 466 U.S. 668 (1984). First, he must show that his counsel’s performance was deficient because it “fell below an objective standard of reasonableness.” Id. at 687–88. Second, he must show that the deficient performance prejudiced the defense, which means that “there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceed- ing would have been different.” Id. at 694. A. Standard of Review Under AEDPA, “the bar for establishing that a state court’s application of the Strickland standard was ‘unreason- able’ is a high one.” Allen v. Chandler, 555 F.3d 596, 600 (7th Cir. 2009). “When § 2254(d) applies, the question is not whether counsel’s actions were reasonable. The question is whether there is any reasonable argument that counsel satis- fied Strickland’s deferential standard.” Harrington, 562 U.S. at 105. “However, if a state court does not reach either the issue of performance or prejudice on the merits, then ‘federal re- view of [that prong] … is de novo.’” Sussman v. Jenkins, 636 F.3d 329, 350 (7th Cir. 2011), quoting Toliver v. McCaughtry, 539 F.3d 766, 775 (7th Cir. 2008); see also Porter v. McCollum, 558 U.S. 30, 39 (2009) (per curiam); Wiggins, 539 U.S. at 534. B. Performance To establish deficient performance under Strickland, Campbell must identify acts or omissions by counsel that fell below an objective standard of reasonableness and could not have been the result of professional judgment. Strickland, 466 U.S. at 688, 690. “The question is whether an attorney’s rep- resentation amounted to incompetence under ‘prevailing professional norms,’ not whether it deviated from best prac- 18 No. 13-2634 tices or most common custom.” Harrington, 562 U.S. at 105, quoting Strickland, 466 U.S. at 690. “A fair assessment of at- torney performance requires that every effort be made to eliminate the distorting effects of hindsight, to reconstruct the circumstances of counsel’s challenged conduct, and to evaluate the conduct from counsel’s perspective at the time.” Strickland, 466 U.S. at 689. 1. Leroy Hunter and Toni Leonard We begin with the failure to present testimony by Leroy Hunter and Toni Leonard. On direct appeal, the state appel- late court rejected this claim on the merits, holding that the defense lawyer’s decision not to call Mr. Hunter or Ms. Leonard was not deficient performance. Because this was a merits adjudication, AEDPA deference applies and we ask “whether there is any reasonable argument that counsel sat- isfied Strickland’s deferential standard.” Harrington, 562 U.S. at 105. 8 The state court’s analysis of defense counsel’s perfor- mance was an unreasonable application of Strickland. On the record before the state court, it was unreasonable to con- clude that counsel made a reasonable decision not to call Mr. Hunter and Ms. Leonard as a matter of trial strategy. The state court relied on the fact that counsel chose to argue that it was too dark at the time of the beating to make any reli- able identifications. See People v. Campbell, 773 N.E.2d 776, 8The State agrees that the state appellate court did not reach the is- sue of prejudice and addressed only the issue of performance. See Ap- pellee’s Br. 18 (“Because the state court addressed only one prong of the Strickland standard, this Court defers to the state court’s judgment on that prong, but reviews Strickland’s prejudice prong de novo.”). No. 13-2634 19 785 (Ill. App. 2002) (“Defendant’s counsel chose to argue it was too dark that night to make identifications, for which there was some evidence, rather than mount a defense based upon the conflicting testimony of unreliable witnesses.”). Under the state court’s view, once Campbell’s lawyer select- ed this strategy, it was reasonable for him to decide not to present testimony from any eyewitness claiming that he or she could identify anyone—participant or non-participant— on the street that night. The fundamental problem with the state court’s analy- sis—which made it not just incorrect but unreasonable—is that it ignored counsel’s duty to perform a reasonable pretri- al investigation before committing to a defense strategy. In Strickland itself, the Supreme Court distinguished be- tween “strategic choices made after thorough investigation of law and facts relevant to plausible options,” versus “stra- tegic choices made after less than complete investigation.” 466 U.S. at 690–91. Strategic choices in the first category are “virtually unchallengeable,” but those in the second catego- ry are “reasonable precisely to the extent that reasonable professional judgments support the limitations on investiga- tion. In other words, counsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.” Id. In the second cir- cumstance, the “decision not to investigate must be directly assessed for reasonableness in all the circumstances, apply- ing a heavy measure of deference to counsel’s judgments.” Id. at 691. The Supreme Court has repeatedly emphasized the im- portance of this distinction. In Wiggins v. Smith, for example, counsel failed to investigate the defendant’s background and 20 No. 13-2634 to present mitigating evidence at his sentencing proceedings, settling instead on the argument that the defendant was not directly responsible for the crime. 539 U.S. 510, 515–16 (2003). The state court held that trial counsel had not per- formed deficiently because “counsel had made ‘a deliberate, tactical decision to concentrate their effort at convincing the jury’ that [the petitioner] was not directly responsible for the murder.” Id. at 518, quoting Wiggins v. State, 724 A.2d 1, 15 (Md. App. 1999). Even through the deferential lens of AEDPA, the Su- preme Court held that this was an unreasonable application of Strickland. Where the Strickland claim involves an alleged- ly inadequate investigation, the proper question “is not whether counsel should have presented a mitigation case,” but rather “whether the investigation supporting counsel’s decision not to introduce mitigating evidence of [the peti- tioner’s] background was itself reasonable.” Id. at 522–23. As the Court put it, “Strickland does not establish that a cursory investigation automatically justifies a tactical decision with respect to … strategy. Rather, a reviewing court must con- sider the reasonableness of the investigation said to support that strategy.” Id. at 527. The Court then concluded that the state court’s analysis of Strickland was unreasonable in two respects: (1) the state court merely assumed that counsel’s de- cision not to investigate further was reasonable without ac- tually assessing counsel’s decision to stop investigating when they did, and (2) the state court applied deference to counsel’s strategic decision not to present a mitigation de- fense despite the fact that counsel based this choice on an unreasonably limited investigation. Id. at 527–28. No. 13-2634 21 Here, the state court engaged in a similarly unreasonable application of Strickland. The court merely assumed that the lawyer’s decision not to interview Mr. Hunter and Ms. Leon- ard was reasonable and then gave deference to his “strate- gic” decision to argue that it was too dark for anyone to make reliable identifications. As Wiggins also makes clear, the proper inquiry under Strickland is not whether it was reasonable for counsel to present the too-dark-to-identify theory, but whether the investigation supporting counsel’s decision not to call Mr. Hunter and Ms. Leonard was itself reasonable. If counsel’s decision not to investigate Mr. Hunter or Ms. Leonard was itself unreasonable, then his de- cision not to present their testimony—and to rely on the too- dark-to-identify theory instead—was too ill-informed to be considered reasonable. See Stitts v. Wilson, 713 F.3d 887, 891 (7th Cir. 2013) (“If trial counsel’s investigation of a potential alibi defense was unreasonably limited, then trial counsel’s decision not to present an alibi defense is too ill-informed to be considered reasonable.”); Mosley v. Atchison, 689 F.3d 838, 848 (7th Cir. 2012) (“If … Mosley’s lawyer never found out what [the potential witnesses’] testimony would be, he could not possibly have made a reasonable professional judgment that their testimony would have been cumulative or bol- stered the State’s case and could not have chosen not to call [them] as a matter of strategy.”); Crisp v. Duckworth, 743 F.2d 580, 584 (7th Cir. 1984) (“Though there may be instances when the decision not to contact a potential defense witness is justified, an attorney who fails even to interview a readily available witness whose noncumulative testimony may po- tentially aid the defense should not be allowed automatically to defend his omission simply by raising the shield of ‘trial strategy and tactics.’”). 22 No. 13-2634 Giving defense counsel’s judgments the deference they are due, we see no reason in the record why a decision not to interview Mr. Hunter and Ms. Leonard would have been reasonable under the circumstances. Under the prevailing norms at the time of Campbell’s trial, counsel had an obliga- tion “‘to conduct a prompt investigation of the circumstanc- es of the case and to explore all avenues leading to facts rele- vant to the merits of the case.’” Bobby v. Van Hook, 558 U.S. 4, 7 (2009), quoting the American Bar Association Standards for Criminal Justice in effect in 1985. The police reports that were disclosed to counsel clearly presented two disinterest- ed eyewitnesses who would have (1) described a version of the fatal beating substantially different from that argued by the State, (2) testified that two of the State’s three eyewit- nesses were, despite their denials, directly involved in the beating, and (3) testified they did not see Campbell partici- pate in the murder. Mr. Hunter’s version of the events would have squarely contradicted Peete’s testimony that he was not involved in Shepherd’s beating. In fact, Mr. Hunter identified Peete as one of the attackers. He told police that Peete struck Shep- herd three or four times with something that looked like a pipe or a big stick. He also told police that on the day after the incident, he saw Peete return to the scene of the crime and spit on the bloodstain where Shepherd had been lying. Ms. Leonard’s version of the events would have contra- dicted the testimony of the State’s other two eyewitnesses. Not only would she have contradicted Johnson’s testimony that he was not involved in the murder, she would have squarely contradicted Rita Butler’s testimony about how the fight began. Butler testified at trial that the fight began when No. 13-2634 23 Shepherd approached Campbell and the two men began trading blows. But Ms. Leonard told police that the fight be- gan when Shepherd started arguing with Jeffrey Dillon and Lynntez Holt—not Campbell. According to Ms. Leonard, Dillon struck first, followed by Holt, and then Shepherd at- tempted to fight back. Strickland, of course, “permits counsel to ‘make a reason- able decision that makes particular investigations unneces- sary.’” Harrington, 562 U.S. at 106, quoting Strickland, 466 U.S. at 691. Resources are limited, and trial counsel must eventually shift from pretrial investigation to trial prepara- tion. But here, nothing in the record suggests that counsel made a reasonable decision not to interview Mr. Hunter and Ms. Leonard. The obvious exculpatory value of both eyewit- ness accounts appears on the face of the police reports, which were disclosed to counsel before trial. And nothing in either police report suggests that interviewing either Mr. Hunter or Ms. Leonard would have been fruitless or harm- ful. Cf. Strickland, 466 U.S. at 691 (noting that when counsel has “reason to believe that pursuing certain investigations would be fruitless or even harmful, counsel’s failure to pur- sue those investigations may not later be challenged as un- reasonable”). Because the State’s case hinged on eyewitness testimony, counsel’s “decision that it was unnecessary to look for and contact such eyewitnesses cannot be described as reasonable.” United States ex rel. Hampton v. Leibach, 347 F.3d 219, 251–52 (7th Cir. 2003). The state court gave two other reasons to support its con- clusion, but both were based on an unreasonable characteri- zation of the legal significance of Mr. Hunter’s and Ms. Leonard’s potential testimony. The state court noted that 24 No. 13-2634 “none of the testimony which [Campbell] claims Hunter and Leonard would have given would have exonerated” Camp- bell because all they could do was “further implicate John- son and Peete as involved in the murder, a fact of which the jury was already aware because it knew Johnson and Peete had been charged with the murder.” People v. Campbell, 773 N.E.2d 776, 785 (Ill. App. 2002). The court also mentioned that any “testimony that [Campbell] did not start the beating would not contradict the State’s theory that [Campbell] par- ticipated in the beating.” Id. But the jury heard no evidence that Johnson or Peete par- ticipated in the murder. Live testimony from a disinterested witness is quite different from any inference the jury might have drawn from the bare fact that the two men had been charged initially but then had the charges dropped. And the potential testimony of Mr. Hunter and Ms. Leonard would have done far more than “further implicate” Johnson and Peete. According to their statements and affidavits, Mr. Hunter and Ms. Leonard would have presented a version of the facts markedly different from that presented by the pros- ecution. It is true that under Mr. Hunter’s and Ms. Leonard’s tes- timony, it would not have been physically impossible for Campbell to have participated in the beating. It is at least theoretically possible that Campbell could have participated in the start of the beating but left the scene before Mr. Hunter looked out his window. And because Ms. Leonard could not identify everyone at the scene, it is also theoretical- ly possible that Campbell was simply one of the men she saw but could not identify. But the theoretical possibility that Campbell participated in the beating in spite of these eye- No. 13-2634 25 witness accounts does not undermine the probative value to the defense of their potential testimony. Both witnessed the beating, and neither identified Campbell. Each witness de- scribed a fight that differed dramatically from that described by the prosecution at trial. Viewed separately, then, each witness’s testimony would have been very helpful to the defense. Viewed together, their testimony would have been even more powerful. The only weakness in Mr. Hunter’s testimony identified by the State is that he did not see the beginning of the fight. But Ms. Leon- ard said that she did see the beginning and that Campbell was not involved when the fight started among Dillon, Holt, and Shepherd. Thus, Ms. Leonard’s testimony could have filled in the key gap in Mr. Hunter’s testimony. The version of the events presented by Ms. Leonard and Mr. Hunter, if credited, would be entirely exculpatory to Campbell. Accordingly, as the dissenting justice recognized on di- rect appeal, the testimony of Mr. Hunter and Ms. Leonard would not have simply undermined the credibility of John- son and Peete or “further implicated” them in the crime (though it would have done those things, too). It would have completely contradicted the prosecution’s version of the facts. See Campbell, 773 N.E.2d at 789 (Knecht, J., dissenting) (“[Mr.] Hunter’s testimony would have virtually destroyed Peete’s credibility and contradicted the State’s version of the facts surrounding the beating.”). Neither of these additional reasons could support a reasonable determination that fail- ing to investigate Mr. Hunter or Ms. Leonard was not defi- cient performance. In sum, the state court concluded that counsel’s failure to call two exculpatory witnesses reflected a tactical judgment 26 No. 13-2634 not to present their testimony and to pursue an alternative strategy instead. But it did not address the adequacy of the pretrial investigation, which was clearly established under Strickland as the critical threshold question. Because we must assume that counsel failed to investigate two exculpatory eyewitnesses who were known to him based on police re- ports disclosed before trial, there is no reasonable argument that counsel satisfied even the deferential Strickland stand- ard. 2. Ieca Hunter We next turn to counsel’s failure to present the testimony of Ieca Hunter. Campbell first presented this claim to the state courts during post-conviction proceedings. The state trial court denied Campbell’s petition without an evidentiary hearing. On appeal, the state appellate court affirmed, hold- ing that counsel’s failure to interview Ms. Hunter was not deficient performance because “there was not a sound basis to believe she had any information that would be helpful to the defense.” Supp. App. 96 The state court based this con- clusion on the police report: “Christmas lights obscured her view of the beating and she did not know any of the men in- volved.” Id. This, too, was a merits adjudication, so again we ask “whether there is any reasonable argument that counsel satisfied Strickland’s deferential standard.” Harrington, 562 U.S. at 105. The state court’s analysis of counsel’s performance re- garding Ms. Hunter was also an unreasonable application of Strickland. Under Strickland, again, counsel had “a duty to make reasonable investigations or to make a reasonable de- cision that makes particular investigations unnecessary.” Strickland, 466 U.S. at 691. We see no reasonable argument No. 13-2634 27 that counsel could have reasonably concluded, based solely on the police report, that Ms. Hunter was not worth contact- ing. The police report said: [Ieca Hunter] stated that she looked out her window and saw approximately five black males fighting with one subject. Ieca stated that she was trying to see what was going on but she has Christmas lights on the outside of her house, and it was affecting her vision. Ieca stat- ed that she saw the victim fall to the ground, so she left her window and went and telephoned 911 for an ambulance. Ieca stated that she did not recognize any of the subjects in [the] street and she would not be able to identify any of them. The only thing that she could tell me was that all of them were black males. Supp. App. 143–44 (paragraph break omitted). The fact that Ms. Hunter could not identify any of the men who were involved in the beating does not mean that she would have been unable to identify people who were present but not involved in the beating. She explained this point in an affidavit attached to Campbell’s petition for state post-conviction relief. She initially had trouble seeing the fight because of Christmas lights hanging on her door, but she opened the door for a better view. She then saw Camp- bell standing directly in front of her house, approximately 12 feet away. She called Campbell “a familiar person” and someone she “had interacted with” on “numerous occa- sions.” Although she could not identify the participants in the beating, she could identify Campbell as someone who did not participate: “there was no mistaking him even in the 28 No. 13-2634 dark.” This fact was not addressed in the police report. The police apparently asked her to identify only those who did the beating. Police did not ask whether she could identify other people who were present but not involved. We recognize, of course, that Campbell’s lawyer could not have known by reading the police report that Ms. Hunter could identify Campbell as a non-participant. But he had a duty to perform a reasonable investigation. Contacting one of a few eyewitnesses to the crime falls squarely within the core of that duty. See Stanley v. Bartley, 465 F.3d 810, 813 (7th Cir. 2006) (“The lawyer could not know how complete or accurate a prospective witness’s statement to the police was without talking to the witness.”). From the defense per- spective, witness statements to the police are systemically prone to be incomplete. Police investigations are generally focused on identifying individuals involved in the crime and witnesses helpful to the prosecution’s case. Their focus is not necessarily on clearing bystanders. It is also not uncommon for people to withhold information from the police, or at least not to volunteer everything they know when they first speak to police. That is why Strickland typically demands that counsel go beyond discovery provided by the State and conduct her own pretrial investigation. See Crisp, 743 F.2d at 584 (“We do not agree that police statements can generally serve as an adequate substitute for a personal interview.”); see also, e.g., Washington v. Smith, 219 F.3d 620, 632 (7th Cir. 2000) (coun- sel’s “failure to try to ascertain what exculpatory evidence ‘new’ witnesses might have [was] flagrant example[] of inef- fective assistance”); Anderson v. Johnson, 338 F.3d 382, 391–93 (5th Cir. 2003) (deficient performance where counsel “relied No. 13-2634 29 exclusively on the investigative work of the State and based his own pretrial ‘investigation’ on assumptions divined from a review of the State’s files”). We say “typically” because we recognize “there may be unusual cases when an attorney can make a rational decision that investigation is unnecessary.” Crisp, 743 F.2d at 583. But this is not such a case. The police report identified Ms. Hunter as a disinterested eyewitness to the crime. Although the witness’s vision that night was “affected” by Christmas lights and she could not identify the perpetrators, nothing in the report implies that contacting her would have been fruit- less or harmful. Her statement to the police certainly was not a “blanket denial” that she knew anything about the inci- dent, as the State characterizes it on appeal. Perhaps she could have described the location of the beating in a way that would have cast doubt on the prosecution witnesses’ version of the events. Or perhaps she could have told coun- sel about other eyewitnesses who may have had a better view of the incident than she did. Or perhaps, as we must assume for now, she could have identified non-participants, including Campbell. The point is that counsel could not an- swer those questions by reading the police report, which is why he had a duty to investigate further. See Wiggins, 539 U.S. at 527 (“In assessing the reasonableness of an attorney’s investigation, … a court must consider not only the quantum of evidence already known to counsel, but also whether the known evidence would lead a reasonable attorney to inves- tigate further.”). The State argues that the state court reasonably applied Strickland because the Supreme Court has never clearly es- tablished that defense counsel must interview witnesses 30 No. 13-2634 named in police reports. This argument exaggerates Camp- bell’s position. Neither his position nor our decision is based on a per se requirement that defense counsel must always in- terview witnesses named in police reports. Rather, it is based on the Strickland standard itself: “a particular decision not to investigate must be directly assessed for reasonableness in all the circumstances, applying a heavy measure of defer- ence to counsel’s judgments.” Strickland, 466 U.S. at 691. Un- der the circumstances of this case—where (1) there was no physical evidence linking Campbell to the crime, (2) police reports named two other eyewitnesses (Leroy Hunter and Toni Leonard) who identified two of the State’s main wit- nesses as participants in the crime and did not identify Campbell, and (3) the State agreed to drop murder charges against Peete and Johnson in exchange for their testimony against Campbell—counsel’s decision not to interview Ms. Hunter fell below an objective standard of reasonableness, even after applying deference to counsel’s judgments. The state court’s decision to the contrary was an unreasonable application of Strickland. 9 9Campbell also argues that counsel’s failure to investigate or present the testimony of Minnie Hunter was constitutionally ineffective. The State argues that this claim is procedurally defaulted because Campbell failed to present this claim to the state courts. Campbell’s reply brief ar- gues that he presented the claim by attaching Minnie Hunter’s affidavit to his amended post-conviction petition. That was not enough. Under AEDPA, a habeas petitioner must present both the legal argument and the factual basis for the claim to the state courts to preserve the issue. See Pole v. Randolph, 570 F.3d 922, 939–40 (7th Cir. 2009). Although Campbell presented the factual basis for the claim by attaching the affidavit to the petition, he did not present the legal argument in the petition or in his appellate brief during post-conviction proceedings. Rather, he argued only that counsel’s failure to investigate or present the testimony of Ieca No. 13-2634 31 C. Prejudice Having determined that counsel’s performance was defi- cient, at least based on affidavits and police statements that we must treat as true at this stage of the case, we next ana- lyze whether it prejudiced Campbell. To establish prejudice, Campbell “must demonstrate ‘a reasonable probability that, but for counsel’s unprofessional errors, the result of the pro- ceeding would have been different. A reasonable probability is a probability sufficient to undermine confidence in the outcome.’” Harrington v. Richter, 562 U.S. 86, 104 (2011), quoting Strickland, 466 U.S. at 694. This does not require a showing that counsel’s actions “more likely than not altered the outcome,” but the likelihood of a different result must be “substantial, not just conceivable.” Id. at 111–12. “Making this probability determination requires consid- eration of the ‘totality of the evidence before the judge or ju- ry.’” Harris v. Thompson, 698 F.3d 609, 645 (7th Cir. 2012), quoting Strickland, 466 U.S. at 695. A “verdict or conclusion only weakly supported by the record is more likely to have been affected by errors than one with overwhelming record support.” Strickland, 466 U.S. at 696. Because neither state court addressed the issue of prejudice, we review this prong of Strickland de novo. See Sussman v. Jenkins, 636 F.3d 329, 350 (7th Cir. 2011); see also Porter v. McCollum, 558 U.S. 30, 39 (2009) (per curiam); Wiggins v. Smith, 539 U.S. 510, 534 (2003). The State devotes most of its brief to arguing that the omission of exculpatory testimony from Leroy Hunter, Toni Hunter, Juanchez Booker, and Lynntez Holt was ineffective. According- ly, we hold that Campbell’s argument concerning Minnie Hunter was procedurally defaulted. 32 No. 13-2634 Leonard, and Ieca Hunter was not prejudicial. 10 It approach- es this issue from two angles, arguing that the prosecution’s case against Campbell was strong and that the probative (or persuasive) value of the witnesses’ potential testimony was slight. We disagree with the State in both respects. The prosecution’s case against Campbell was far from overwhelming. No physical evidence linked Campbell to the crime, so the testimony of the State’s eyewitnesses was the most important evidence in the trial. See People v. Campbell, 773 N.E.2d 776, 789 (Ill. App. 2002) (Knecht, J., dissenting) (“The credibility of the State’s witnesses was the critical issue in the jury’s determination of [Campbell’s] guilt or inno- cence.”). The prosecution’s closing argument could not have been more explicit on this point: “The most important evi- dence in this trial is the eyewitness evidence. And you don’t have one eyewitness. You don’t have two. You have three.” Supp. App. 659. Two of these eyewitnesses had been charged with the murder and agreed to testify against Campbell in exchange for full immunity. The third eyewitness observed the fight from the inside of a van that was parked around the corner and facing away from the fatal altercation. This was not a case where the eyewitness testimony presented by the prosecution was particularly strong. Cf. Morales v. Johnson, 10The State also attempts to refute Campbell’s argument that preju- dice is established by virtue of the fact that Leroy Hunter and Toni Leon- ard testified in Bobby Joe Douglas’s trial and Douglas was acquitted. While we agree with Campbell that the circumstances of Douglas’s trial and his ensuing acquittal provide circumstantial evidence of prejudice in Campbell’s case, see United States ex rel. Hampton v. Leibach, 347 F.3d 219, 253 (7th Cir. 2003) (noting that two acquittals in other cases supported finding of prejudice), we do not rely on this comparison here because Campbell has established prejudice independent of that comparison. No. 13-2634 33 659 F.3d 588, 600–02 (7th Cir. 2011) (counsel’s deficient per- formance not prejudicial where prosecution presented two eyewitnesses whose testimony was corroborated by physical evidence). But because defense counsel did not call a single witness or present any other evidence aside from the stipula- tion that Shepherd’s blood had been found on the clothing of two other men, the jury never heard an alternative version of the facts. As a result, the prosecution’s case against Campbell seemed much stronger than it should have been if Campbell had received effective assistance. We must compare the actual trial to a hypothetical trial in which, based on the facts we must assume are true, three neutral eyewitnesses would have testified to a very different account of the fight, one in which Campbell was not a parti- cipant, and in which one of these eyewitnesses would have refuted Rita Butler’s testimony about how the fight began. We conclude that there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceed- ing would have been different. See Stitts v. Wilson, 713 F.3d 887, 894 (7th Cir. 2013) (“And if these witnesses testified, the trial would have been transformed from a one-sided presen- tation of the prosecution’s case into a battle between compet- ing eyewitness testimony, where there would have been a ‘reasonable probability’ that a jury would have reasonable doubt as to Stitts’s guilt and therefore acquit.”). Each wit- ness’s account would have both undermined the credibility of (at least) one of the State’s eyewitnesses and contradicted the State’s version of the facts. All three of these eyewitnesses were disinterested. Two of them were residents of the neigh- borhood who lived across the street from Peete. And none of them were ever implicated in the underlying crime, so they 34 No. 13-2634 did not have the same incentive to lie that two of the prose- cution witnesses did. One clear indicator of the prejudice caused by defense counsel’s failure to present the testimony of Leroy Hunter and Toni Leonard occurred during counsel’s closing argu- ment. He tried to argue that Damion Johnson was more like- ly to blame for Shepherd’s death than Campbell: “Every- thing that [the prosecutor] says makes Mr. Campbell guilty, those same factors could be applied especially to Mr. [John- son]; presence, being named by others—.” Supp. App. 703. At that point the prosecution objected, arguing that defense counsel’s statement was impermissible because there had been no evidence that anybody besides Campbell was named by another person. The court sustained the objection. If counsel had presented the testimony of Mr. Hunter or Ms. Leonard, of course, there would have been evidence that someone other than Campbell—Peete or Johnson, for exam- ple—had been involved in the crime and named by others. The prejudice caused by counsel’s failure to present the testimony of Ms. Hunter is even clearer. She saw Campbell standing directly in front of her house, approximately 12 feet away, separate and apart from the group of men beating Shepherd. According to her affidavit, Campbell “never touched” Shepherd. The State attempts to minimize the significance of this exculpatory testimony in a variety of ways. For example, it argues that the testimony from Toni Leonard would have been “relatively insignificant” because “none of petitioner’s proposed evidence undermines” Butler’s testimony about how the fight began. Yet on the same page of its brief the State concedes: “At best, Leonard might have contradicted But- No. 13-2634 35 ler’s testimony identifying petitioner as the person who be- gan fighting with Shepherd.” Appellee’s Br. 24 (emphasis added). How the fight began and who started it was a major theme in the case against Campbell. In its closing argument, for example, the prosecution told the jury: “Rita [Butler] can tell you [what happened] from the very beginning, from be- fore [it] became violent what was happening. She can tell you about the entire interaction, but she can’t tell you about the ending, because she ran.” Supp. App. 660. If a witness’s testimony contradicts a key aspect of the prosecution’s theo- ry of the crime, it cannot be easily dismissed as “relatively insignificant.” The State also argues that Ms. Leonard’s testimony would have been unreliable because she could not account for everyone at the scene and because she failed to identify Theodis White, who was ultimately convicted of the crime. Neither fact significantly detracts from the reliability of her testimony. None of the State’s eyewitnesses could account for everyone at the scene either, yet the State has found them sufficiently reliable to play a major role in imprisoning Campbell since 1998. And although Ms. Leonard failed to identify White, she ultimately identified nine other partici- pants—all of whom the State charged based in part on her identifications. The State’s characterization of Ms. Leonard’s reliability on appeal is belied by its reliance on her identifica- tions during its own pretrial investigation. The State attempts to minimize the persuasive value of Leroy Hunter’s potential testimony on the ground that when he testified at Bobby Joe Douglas’s trial, he identified Peete as an assailant but did not “feel comfortable swearing” it was Peete. But Mr. Hunter’s qualification at Douglas’s trial 36 No. 13-2634 does not destroy his reliability as a witness or the probative value of his potential testimony. His initial statement to po- lice was unequivocal in identifying Peete as a participant. According to the police report, Mr. Hunter voluntarily con- tacted police because “he had heard that the man had died, and he knew that Steve Peete was still walking around free, and he felt that he had to come forward, and tell what he had seen.” And if Mr. Hunter had been called at Campbell’s trial and refused to cooperate, his original taped statement to police could have been admitted as substantive evidence under Illinois law. See People v. Campbell, 773 N.E.2d 776, 789–90 (Ill. App. 2002) (Knecht, J., dissenting), citing 725 Ill. Comp. Stat. 5/115-10.1(c)(2)(C) (1998). Even if the jury ulti- mately discounted Mr. Hunter’s identification of Peete as a participant, his testimony still would have been valuable be- cause he did not identify Campbell as a participant. At no point did he waver on that critical fact. Because the case against Campbell was far from over- whelming and the omitted exculpatory testimony relatively strong, counsel’s unprofessional errors undermine our con- fidence in the outcome of Campbell’s trial. See United States ex rel. Hampton v. Leibach, 347 F.3d 219, 250 (7th Cir. 2003) (“Opposing testimony from other eyewitnesses to the at- tacks, positing that Hampton was not a participant, would have given the jury a qualitatively different and more pow- erful reason to believe that the State’s witnesses were mis- taken in their identifications of Hampton.”). Accordingly, assuming the witnesses would have testified credibly and consistently with their affidavits and statements to the po- lice, we hold that counsel’s deficient performance prejudiced Campbell. No. 13-2634 37 IV. Remedy Campbell has satisfied § 2254(d) on his claim of ineffec- tive assistance of counsel, but that does not necessarily enti- tle him to the issuance of the writ. “Whether the petitioner is actually entitled to relief—whether under § 2254(a) he is in custody in violation of the Constitution or laws or treaties of the United States—is a separate question.” Mosley v. Atchison, 689 F.3d 838, 853 (7th Cir. 2013), citing 28 U.S.C. § 2254(a). We cannot answer that question on the record be- fore us. Throughout this opinion we have assumed, as the state courts and district court did, that defense counsel did not actually interview these witnesses and that these witnesses would have testified credibly and consistently with their af- fidavits and statements to the police. Having concluded that the affidavits and statements, if true, are sufficient to war- rant habeas relief, we still have no factual findings on these questions to review, and the record is otherwise ambiguous. To be sure, Campbell has steadfastly asserted that de- fense counsel never contacted Leroy Hunter, Toni Leonard, or Ieca Hunter, and the State has not claimed that counsel did in fact contact these witnesses. Nor has the State argued that these witnesses would have testified differently than Campbell has described. But the affidavits on which Camp- bell relies—affidavits from Mr. Hunter and Ms. Hunter— have not been tested through the adversarial process at any kind of hearing. And there is no affidavit from Ms. Leonard or defense counsel in the record. Under these circumstances, an evidentiary hearing is needed to develop the record on (1) the extent of counsel’s 38 No. 13-2634 actual pretrial investigation and (2) what these witnesses would have said if called to testify at trial. See Stitts v. Wil- son, 713 F.3d 887, 895–97 (7th Cir. 2013) (finding § 2254(d) satisfied and remanding for evidentiary hearing); Mosley, 689 F.3d at 852–54 (same); Cullen v. Pinholster, 563 U.S. —, 131 S. Ct. 1388, 1412 (Breyer, J., concurring in part and dis- senting in part) (“For example, if the state-court rejection as- sumed the habeas petitioner’s facts (deciding that, even if those facts were true, federal law was not violated), then (af- ter finding the state court wrong on a [§ 2254(d)] ground), a[] [§ 2254(e)] hearing might be needed to determine whether the facts alleged were indeed true.”). We recognize that this will be a daunting challenge so many years after the fact, but it is still necessary. If the district court finds that defense counsel did not in- vestigate these witnesses (and that no other fact would rea- sonably justify that decision), then Campbell has proven that his counsel performed deficiently under Strickland. If, on the other hand, the district court finds that counsel did contact these witnesses (or, alternatively, that counsel made a rea- sonable decision not to investigate these witnesses based on some other fact), then it must determine de novo whether the entirety of counsel’s pretrial investigation, as well as his de- cision not to present their testimony at trial, was reasonable under Strickland. If the district court finds that counsel per- formed deficiently under Strickland for any reason, it should then determine de novo whether counsel’s errors prejudiced Campbell. If the district court finds that Campbell has satis- fied both prongs of Strickland, it should grant the writ. The state courts’ decision that Campbell was not denied effective assistance of counsel was an unreasonable applica- No. 13-2634 39 tion of clearly established federal law as determined by Su- preme Court precedent. We REVERSE the district court’s de- nial of Campbell’s habeas petition. We REMAND for further proceedings consistent with this opinion.