Olu Rhodes v. Michael Dittmann

• Court: Court of Appeals for the Seventh Circuit
• Decided: September 10, 2018
• Docket: 17-2223
• Status: Published

Opinion

In the

United States Court of Appeals For the Seventh Circuit ____________________ No. 17‐2223 OLU A. RHODES, Petitioner‐Appellant, v.

MICHAEL A. DITTMANN, Respondent‐Appellee. ____________________

Appeal from the United States District Court for the Eastern District of Wisconsin. No. 13‐CV‐683 — Pamela Pepper, Judge. ____________________

ARGUED JANUARY 18, 2018 — DECIDED SEPTEMBER 10, 2018 ____________________

Before SYKES and HAMILTON, Circuit Judges and LEE, District Judge.* HAMILTON, Circuit Judge. Petitioner‐appellant Olu Rhodes seeks a writ of habeas corpus, arguing that his Sixth Amend‐ ment right to confront witnesses against him was violated. Rhodes was convicted of first‐degree intentional homicide

* Of the Northern District of Illinois, sitting by designation. 2 No. 17‐2223

and first‐degree recklessly endangering safety for shooting two victims and killing one: Robert Davis. The State’s theory at trial was that Rhodes and his brother shot Davis, who was an ex‐boyfriend of their sister, Nari Rhodes (and that the sur‐ viving shooting victim was at the wrong place at the wrong time). Nari had suffered a severe beating the day before Davis was murdered. She was the only connection between Rhodes and the victims. The State called her as a witness. Her direct testimony focused heavily on her injuries from the beating the day before. But when Rhodes tried to cross‐examine Nari to rebut the State’s motive theory, the judge limited the ques‐ tioning on this central issue. In essence, the trial court shut down the defense’s cross‐examination to rebut the prosecu‐ tion’s central theory. Rhodes argues that this violated his rights under the Confrontation Clause of the Sixth Amend‐ ment. The Wisconsin Court of Appeals reversed Rhodes’s con‐ viction, finding that his Confrontation Clause rights were vi‐ olated and that the violation was not harmless. State v. Rhodes, 329 Wis. 2d 268 (Wis. App. 2010). A divided Wisconsin Su‐ preme Court reversed that decision, finding no Confrontation Clause violation and reinstating the conviction. State v. Rhodes, 336 Wis. 2d 64, 799 N.W.2d 850 (Wis. 2011). Rhodes then sought a writ of habeas corpus in federal court, arguing that the Wisconsin Supreme Court’s decision was “contrary to, or involved an unreasonable application of, clearly estab‐ lished Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1). The district court agreed that Rhodes had indeed shown a clear Confrontation Clause violation but found that the violation was harmless. No. 17‐2223 3

Rhodes v. Meisner, 2017 WL 2345671, at *1 (E.D. Wis. May 30, 2017). We agree with the district court that the state courts vio‐ lated clearly established federal law in violating Rhodes’s Confrontation Clause rights. Like the other courts that have reviewed this trial, we recognize that trial judges have consid‐ erable discretion in managing trials and deciding the bound‐ aries of cross‐examination, even by the accused in a criminal case. The reasoning of the state trial court and the Wisconsin Supreme Court looks, superficially, like reasoning that re‐ ceives considerable deference in that area. If the defense had been the party trying to open up the history of domestic vio‐ lence between Nari and the murder victim, perhaps the trial court’s limits might have seemed within bounds of a trial court’s discretion. What happened here was very different, though, as the state appellate court and federal district court explained. The prosecution itself first opened up that history of domestic vi‐ olence. But the prosecution then convinced the trial court that Rhodes’s rebuttal evidence would “confuse” the jury on the subject. As a result, the defense did not have a fair oppor‐ tunity to rebut the prosecution’s central theory about why Rhodes would have murdered the victim. As we explain be‐ low, in affirming Rhodes’s convictions, the state supreme court applied the wrong standard under federal law, and its rationale was not just wrong but unreasonably so. Even under the deferential standard of § 2254(d)(1), Rhodes has shown clear constitutional error. And we must disagree with our col‐ league on the district court, and agree with the state appellate court, on the issue of harmless error. Given the importance of the motive issue and the overall balance of evidence in the 4 No. 17‐2223

trial, the constitutional error was not harmless. Rhodes is en‐ titled to a new trial. I. Factual & Procedural Background On April 4, 2006, two men shot Robert Davis and Jonte Watt as they were standing on Watt’s grandparents’ porch. Davis was shot several times and died at the scene. Watt fled and suffered one gunshot to his leg. The police investigation identified Rhodes and his brother, Jelani Saleem, as the shoot‐ ers. Rhodes and Saleem were tried together. The State’s theory of the motive for the murder was that on April 4, Rhodes and Saleem shot Davis as retribution for the attack on their sister Nari the day before. The prosecutor presented the motive the‐ ory from the start, telling jurors in opening statements that they would hear about Nari’s injuries and how both defend‐ ants “hunted Robert Davis down and shot him dead” in retal‐ iation. Rhodes’s and Saleem’s defense was that they were not involved in the shooting and that the two eyewitnesses who identified them were unreliable. Rebutting the State’s motive theory was central to the defense. A. Nari Rhodes’s Testimony Nari Rhodes’s testimony featured prominently in the prosecution case. The State called Nari as a witness and ques‐ tioned her about what happened on April 3, the day before the shooting. Nari testified that she and Davis got into an ar‐ gument that morning. Davis stole her wallet and telephone and punched her car window so hard that he broke it. Davis was the father of Nari’s child, and at some point, the State asked Nari what her relationship was with Davis at the time of the shooting. Nari responded that they “had had a lot of No. 17‐2223 5

domestic violence problems,” had not been romantically in‐ volved for a year, and “had just been working on being friends.” The afternoon of April 3 took an even more violent turn, which was the focus of the State’s direct examination and its motive theory. Nari testified that she drove past Davis’s girl‐ friend’s house that afternoon and saw Davis, and that he waved her down. She pulled over, thinking he was going to return her telephone and wallet. Instead, Davis’s girlfriend, Nancy Segura, approached, and Nari and Segura argued. Ac‐ cording to Nari’s trial testimony, Davis walked away once the women began arguing. Nari wanted to get out of the car and fight Segura, but before she could, another woman attacked Nari from the passenger side of the car. Segura and the other woman pulled Nari out of the car by her feet. Nari’s head hit the concrete and she lost consciousness. The State questioned Nari about her injuries. Nari testified that she received treatment at a hospital. She had a cut near her eye, a cut on her lip, four displaced teeth, and “a lot of skin missing from the right side” of her face. The State intro‐ duced four photographs of Nari’s injuries and showed them to the jury. This evidence did not concern the fatal shooting that was being tried, but it was detailed and full‐color evi‐ dence of a separate assault to support the State’s motive the‐ ory. The State tried twice to get Nari herself to endorse the mo‐ tive theory on direct examination. She was not as cooperative as the State hoped. She testified that she saw her brothers, Rhodes and Saleem, when she got home from the hospital on April 3. She testified that she told them that the two women, not Davis, were responsible for the beating. She also testified 6 No. 17‐2223

that her brothers were not angry at Davis. Apparently skepti‐ cal, the prosecution asked follow‐up questions: Q: Well, did they become angry? Either of them? A: No. Q: They just took it calmly? A: Yes. Q: That their sister had just been beaten? A: Yes. After introducing the photographs of Nari’s injuries, the State asked Nari again: “And your testimony earlier is that this had no effect on your brothers?” She answered that her brothers were mad only at her—for putting herself “in the predicament to be beaten.” On cross‐examination, the defense tried to respond to the State’s motive theory. The defense asked Nari about an earlier time when Davis himself had beaten her. Rhodes’s attorney asked the following questions before the State objected: Q: You did tell us … on direct examination that there had been domestic violence or violence between yourself and Davis be‐ fore, right? A: Yes. Q: Before that date? A: Yes. Q: In fact, Mr. Davis had attacked you pre‐ vious to April 3, 2006; is that right? No. 17‐2223 7

A: Yes. Q: And you—On direct examination you said something to the effect that there had been a lot of that; is that right? A: Yes. * * * Q: [I]n your conflict with Mr. Davis, have there been other times when you’ve been injured? A: Yes. Q: And what injuries had you received? A: One side—My orbital bone in my eye was broken and it was like really bad. At that point, the State objected and the judge held a side‐ bar conference off the record. In a later proffer on the record, Rhodes’s attorney said that he would have asked Nari more questions to rebut the State’s motive theory. He said: She said her orbital bone had been broken. That’s a fairly serious injury. My next question would have been, well, what’s your orbital bone? The question—She would have described as something around her eye. After that I would have asked her did she make her brothers aware of that injury and who would have inflicted it and she would have said yes. There was no response from the brothers. The State objected that it did not have notice of that particular incident of abuse or that the defense would make “a history of domestic abuse” by Davis an issue. The State also argued 8 No. 17‐2223

that the court had previously ruled that both sides could make general references to domestic abuse, but could not go through individual prior acts, invoking Wis. Stat. § 904.04, the Wisconsin analog to Federal Rule of Evidence 404, which re‐ stricts the use of evidence of a person’s prior bad acts. The court sustained the State’s objection and cut off cross‐ examination on the subject. It agreed with the State’s under‐ standing of its prior evidentiary ruling that the defense was not permitted to make an incident‐by‐incident inquiry into Davis’s prior violent abuse of Nari. The judge also reasoned that the parties should “not get into evidence—extraneous ev‐ idence that would mislead the jury on other issues in a trial within a trial which is the concern.” B. Other Evidence Jonte Watt, who survived his gunshot wound, testified that Rhodes and Saleem were the shooters. Watt’s girlfriend, who said she saw the shooting from a parked car, also identi‐ fied Rhodes and Saleem. A police officer testified that he ar‐ rived at the scene within three minutes and that Watt—bloody and on the ground—told him that Rhodes and Saleem had shot him. Watt also later identified Rhodes and Saleem in pho‐ tograph arrays. Another police officer testified that he questioned Letitia Dotson, the mother of Rhodes’s child, about phone calls Dot‐ son received from Rhodes after the shooting. According to the officer, Dotson told him that Rhodes called and said that he had shot Davis. A second police officer, who had also partici‐ pated in the Dotson interview, corroborated this account. The No. 17‐2223 9

State also offered evidence showing that a cell phone associ‐ ated with Saleem was used in the neighborhood of the shoot‐ ing in the relevant time frame. The defense poked such significant holes in the State’s case that Saleem was found not guilty. The defense focused on in‐ consistencies in Watt’s and Walker’s testimony, including shifting stories about what the shooters were wearing and the witnesses’ failure to notice a significant height difference be‐ tween Saleem and Rhodes. The defense also highlighted the fact that Watt’s grandfather, who testified that he stared one shooter in the eyes for five seconds, could not identify that shooter as either of the defendants. The defense also showed that Rhodes lived around the corner from where the shooting took place, that Watt and Walker knew Rhodes by sight before the crime, and that the cell phone location data was consistent with the brothers making phone calls from Rhodes’s own home. Dotson also testified, and she denied ever telling the police that Rhodes called her about shooting Davis. Rhodes testified in his own defense. He testified that he did not shoot Davis. He claimed that on the day of the shoot‐ ing, he was driving around the neighborhood looking for ma‐ rijuana. He testified that he had known for at least a year be‐ fore the shooting that Davis abused Nari, and that he had a physical confrontation with Davis the first time he found out about an abusive incident. In that confrontation, Rhodes and his cousins threw Davis out of their house. But after that inci‐ dent, Rhodes testified, he “left it alone” because Nari kept go‐ ing back to Davis. Rhodes also testified that Nari did not blame Davis for the most recent beating, the day before Davis was shot and killed. 10 No. 17‐2223

C. Closing Arguments The State returned to the motive theory to wrap up its clos‐ ing arguments. After reminding the jury that it did not need to find a motive to convict, the State argued that “in this par‐ ticular case, there is a motive.” The State argued that Davis broke Nari’s car window and that “he set up her beating.” While displaying the color photographs of Nari’s injuries, the State told the jury: “this is what the defendant[s] saw when she came home from the hospital, and this is what they sought to avenge by killing Robert Davis.” Recognizing that Nari’s testimony was not entirely favor‐ able to the State, the prosecution finished its closing argument by challenging both Nari’s and Rhodes’s credibility: [A]nd now Nari Rhodes attempted to mini‐ mize it, and Olu Rhodes says “I had given up,” and that just doesn’t make sense. It doesn’t make sense at all. What makes sense is that they were horrified and angered by the fact that Nari Rhodes took such a bad beating at the hands of Robert Davis’ other girlfriend. That is why they went out. That is why Olu Rhodes followed them on that morn‐ ing and early afternoon, and that is why they shot and killed Robert Davis. I’m asking you to find both defendants guilty of first degree intentional homicide as party to a crime and first degree recklessly en‐ dangering safety as party to a crime. Thank you. No. 17‐2223 11

During the rebuttal closing argument, the State returned to the motive theory, again challenging Nari’s and Rhodes’s tes‐ timony as not credible: When you look at their version, Nari Rhodes comes home beaten like this after a history of abuse according to them at the hands of Robert Davis. They know Robert Davis had broken out her window, that they just shrug their shoulders and said nothing. That doesn’t make sense. It doesn’t make sense to me. Toward the very end of rebuttal, the State again challenged Nari’s credibility: “Nari Rhodes’ testimony. I can only say that she is the sister of the defendants. Place this in terms of how you weigh her credibility.” The jury deliberated for two days. The jury found Saleem not guilty on both charges but convicted Rhodes of both first‐ degree intentional homicide and first‐degree recklessly en‐ dangering safety. The court sentenced Rhodes to life in prison. D. State Post‐Trial Proceedings Rhodes appealed. As relevant here, he argued that the trial court violated his rights under the Confrontation Clause of the Sixth Amendment when it limited his cross‐examination of Nari. The Wisconsin Court of Appeals agreed with Rhodes. That court held that the trial judge had exceeded her “wide latitude” to impose reasonable limitations on cross‐ examination, in violation of the Sixth Amendment. 329 Wis. 2d 268 at ¶10. The Court of Appeals also found that the State had not shown that the error was harmless beyond a reasonable doubt, the harmless‐error standard that applies on 12 No. 17‐2223

direct review of Confrontation Clause violations. The Court of Appeals remanded for a new trial. 329 Wis. 2d 268 at ¶10 & n.1. In a divided opinion, the Wisconsin Supreme Court re‐ versed. The majority summarized United States Supreme Court precedent on the Confrontation Clause but then held that the balancing test under Wis. Stat. § 904.03—the equiva‐ lent of Federal Rule of Evidence 403—applied “to discretion‐ ary decisions” to exclude evidence “when the defendant’s right to cross‐examination under the confrontation clause may be implicated.” 336 Wis. 2d at 81–83, 799 N.W.2d at 859, ¶¶41, 44. The court applied § 904.03 and found that the trial judge did not abuse its discretion by excluding confusing and misleading evidence. 336 Wis. 2d at 90–91, 799 N.W.2d at 863, ¶¶68–70. The Wisconsin Supreme Court did not reach the question of harmless error. Two justices dissented, saying that the “defendant had a fundamental constitutional right to confront this witness and test the probative value of the testimony through cross‐ examination.” 336 Wis. 2d at 92, 799 N.W.2d at 864, ¶75 (Abrahamson, C.J., dissenting). The dissent said the majority had erred by applying the § 904.03 balancing test to Rhodes’s constitutional challenge, urging that the “fundamental constitutional right of confrontation surely affords the defendant more protection and leeway in cross‐examining a witness than the standard analysis used in discretionary evidentiary decisions” when that right is not implicated. 336 Wis. 2d at 93–94, 799 N.W.2d at 865, ¶80. No. 17‐2223 13

E. Federal Habeas Corpus Proceedings Rhodes then filed a petition for a writ of habeas corpus in federal court. Despite having argued in state court that any error was harmless, the State did not raise harmless error in its response to Rhodes’s petition. In his reply brief, Rhodes asserted that the State had waived harmless error review. After briefing, the case was reassigned to Judge Pepper, who acted sua sponte to order briefing on harmless error. Judge Pepper reasoned that she could revive a waived or for‐ feited issue if she gave the parties notice and an opportunity to brief it. In briefing, Rhodes maintained that the State had waived harmless error review and that the Sixth Amendment error was not harmless. The State argued that the district court could raise harmless error sua sponte and that any error was harmless. After the supplemental briefing, the district court denied Rhodes’s petition for a writ of habeas corpus. The court first found that the Wisconsin Supreme Court had unreasonably applied clearly established Supreme Court precedent inter‐ preting the Confrontation Clause. The court also found, how‐ ever, that it could not grant the writ without engaging in harmless error review. The district court went on to hold that the Sixth Amendment error was harmless in light of the evi‐ dence against Rhodes. The court denied Rhodes’s petition but granted a certificate of appealability. II. Analysis We have jurisdiction under 28 U.S.C. § 2253(a), and we re‐ view de novo the district court’s denial of habeas corpus relief. E.g., Makiel v. Butler, 782 F.3d 882, 896 (7th Cir. 2015). Because the Wisconsin Supreme Court rejected Rhodes’s Sixth 14 No. 17‐2223

Amendment claim on the merits, the Antiterrorism and Effec‐ tive Death Penalty Act (AEDPA), 28 U.S.C. § 2254(d), governs our review. A federal court cannot grant a writ of habeas cor‐ pus on Rhodes’s claim unless the state court’s decision “was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Su‐ preme Court of the United States.” 28 U.S.C. § 2254(d)(1). This standard is quite deferential, though federal courts do not and should not merely rubber‐stamp approval of state‐court deci‐ sions that stray too far from federal constitutional require‐ ments. To obtain relief, Rhodes must show not just that the state court’s ruling was wrong, but that it was “so lacking in justification that there was an error well understood and com‐ prehended in existing law beyond any possibility for fair‐ minded disagreement.” Harrington v. Richter, 562 U.S. 86, 103 (2011). This standard is “difficult to meet,” and was “meant to be.” Id. at 102. This case lies close to the boundary between the “contrary to” and “unreasonable application” branches of § 2254(d)(1), depending on how one reads the state supreme court’s opin‐ ion. “A federal habeas court may issue the writ under the ‘contrary to’ clause if the state court applies a rule different from the law set forth in our cases, or if it decides a case dif‐ ferently than we have done on a set of materially indistin‐ guishable facts.” Bell v. Cone, 535 U.S. 685, 694 (2002), citing Williams v. Taylor, 529 U.S. 362, 405–06 (2000) (O’Connor, J., concurring). A state court unreasonably applies clearly estab‐ lished Supreme Court precedent “if the state court identifies the correct governing legal principle” but “unreasonably ap‐ plies that principle to the facts of the petitioner’s case.” Wig‐ gins v. Smith, 539 U.S. 510, 520 (2003) (internal quotations omitted). That is a high bar: “to grant relief, the state court’s No. 17‐2223 15

decision must be objectively unreasonable, not merely incor‐ rect.” Blackmon v. Williams, 823 F.3d 1088, 1099 (7th Cir. 2016), citing Wiggins, 539 U.S. at 520–21. Even under this demanding standard, Rhodes is entitled to federal habeas corpus relief. The Wisconsin Supreme Court decision was actually both contrary to and an unreasonable application of clearly established Supreme Court precedents recognizing the right of the accused to cross‐examine wit‐ nesses on issues central to the case. It was contrary to Olden v. Kentucky, 488 U.S. 227 (1988) (per curiam), and Delaware v. Van Arsdall, 475 U.S. 673 (1986), which held that trial courts cannot routinely apply Rule 403 balancing to prevent defendants from cross‐examining witnesses on issues central to the case. Even if it were not contrary to these two holdings, we must conclude that the rationale of the state court decision is so un‐ reasonable that it is beyond “fairminded disagreement.” Rich‐ ter, 562 U.S. at 103. No fairminded jurist could find that the defense’s cross‐examination of Nari would confuse the issues or mislead the jury when it was the State, not Rhodes, that made past abuse and motive central issues. The restrictions produced an unfair trial in which the State was allowed to put the murder victim on trial to prove its motive theory, but the defense did not have a fair opportunity to rebut that theory, supposedly for fear of jury “confusion.” A. The Confrontation Clause The Confrontation Clause of the Sixth Amendment guarantees the right of the criminal accused “to be confronted with the witnesses against him.” This right applies equally in federal and state prosecutions, Pointer v. Texas, 380 U.S. 400, 403 (1965), and “means more than being allowed to confront the witness physically,” Davis v. Alaska, 415 U.S. 308, 315 16 No. 17‐2223

(1974). The Supreme Court has repeatedly held that the “main and essential purpose of confrontation is to secure for the opponent the opportunity of cross‐examination.” Van Arsdall, 475 U.S. at 678 (1986) (emphasis in original), quoting Davis, 415 U.S. at 315–16; see also Chambers v. Mississippi, 410 U.S. 284, 295 (1973) (referring to the “rights to confront and cross‐ examine witnesses” as “long … recognized as essential to due process”); Smith v. Illinois, 390 U.S. 129, 131 (1968) (referring to denial of cross‐examination as “constitutional error of the first magnitude”), quoting Brookhart v. Janis, 384 U.S. 1, 3 (1966); Douglas v. Alabama, 380 U.S. 415, 418 (1965) (summarizing Supreme Court precedent as holding “that a primary interest secured by” the Sixth Amendment “is the right of cross‐examination”). Cross‐examination is “an essential and fundamental requirement for the kind of fair trial which is this country’s constitutional goal.” Pointer, 380 U.S. at 405. It “is the principal means by which the believability of a witness and the truth of his testimony are tested.” Davis, 415 U.S. at 316. Because AEDPA governs this case, the question is whether the Wisconsin Supreme Court decision was contrary to or an unreasonable application of Supreme Court precedent inter‐ preting the Confrontation Clause. We conclude that it was both. 1. “Contrary to” Controlling Precedent First, the state court applied the wrong legal standard. It is clearly established that the Sixth Amendment limits the trial court’s ordinary discretion to limit cross‐examination. The Supreme Court has held that ordinary rules of evidence must give way when they prevent a defendant from presenting evidence central to the defense, including through cross‐ No. 17‐2223 17

examination. Olden, 488 U.S. at 232; Van Arsdall, 475 U.S. at 679–80; Davis, 415 U.S. at 319 (finding state’s interest in maintaining anonymity of juvenile offenders was “outweighed by petitioner’s right to probe into the influence of possible bias” through cross‐examination); see also Chambers, 410 U.S. at 295–98 (finding that state could not apply common‐law evidentiary rule to limit cross‐ examination of key witness); id. at 302 (finding that hearsay rule cannot be applied mechanistically when it undermines fundamental elements of defense). And most important for our analysis—because we cannot take the Court’s precedents at too high a level of generality, Carey v. Musladin, 549 U.S. 70 (2006)—it is clearly established that a trial court violates the Confrontation Clause when it routinely applies Rule 403 balancing to limit cross‐examination by the accused on issues central to the defense. Olden, 488 U.S. at 232; Van Arsdall, 475 U.S. at 679–80; see also Redmond v. Kingston, 240 F.3d 590, 592 (7th Cir. 2001) (granting habeas relief under § 2254(d)(1) for Confrontation Clause violation when Wisconsin applied § 904.03 to exclude “highly probative, noncumulative, nonconfusing, nonprejudicial evidence” that was “vital to the central issue in the case”). In Van Arsdall, the Supreme Court affirmed a state su‐ preme court decision that found a Confrontation Clause vio‐ lation. The trial court had relied on the state’s Rule 403 to bar cross‐examination of a prosecution witness about possible bias stemming from charges pending against that witness. The Supreme Court agreed with the state supreme court that the Confrontation Clause was violated: [T]rial judges retain wide latitude insofar as the Confrontation Clause is concerned to impose 18 No. 17‐2223

reasonable limits on such cross‐examination based on concerns about, among other things, harassment, prejudice, confusion of the issues, the witness’ safety, or interrogation that is re‐ petitive or only marginally relevant. And as we observed earlier this Term, “the Confrontation Clause guarantees an opportunity for effective cross‐examination, not cross‐examination that is effective in whatever way, and to whatever extent, the defense might wish.” Delaware v. Fen‐ sterer, 474 U.S. 15, 20 (1985) (per curiam) (em‐ phasis in original). In this case, however, the trial court prohib‐ ited all inquiry into the possibility that [witness] Fleetwood would be biased as a result of the State’s dismissal of his pending public drunken‐ ness charge. By thus cutting off all questioning about an event that the State conceded had taken place and that a jury might reasonably have found furnished the witness a motive for favoring the prosecution in his testimony, the court’s ruling violated respondent’s rights se‐ cured by the Confrontation Clause. Van Arsdall, 475 U.S. at 679. The Van Arsdall Court went on to vacate for fresh consideration of whether the violation was harmless. Two years later, the Court treated the holding of Van Arsdall as so clear that it warranted summary reversal of a state court’s reliance on Rule 403. In Olden, the defendant was convicted of rape and related crimes. The defendant asserted a defense of consent. He wanted to attack the complainant’s No. 17‐2223 19

credibility by showing that she lived with her boyfriend and was lying to protect her relationship with her boyfriend. 488 U.S. at 229–30. The state trial court did not allow cross‐ examination along those lines, reasoning that the evidence would be relevant but should be excluded because its probative value was outweighed by its possibility for prejudice. Id. at 230–31. The state appellate court affirmed, but the Supreme Court reversed summarily, rejecting this attempt to apply Rule 403‐ type balancing to such a central issue in the criminal case. The Court explained: Nonetheless, without acknowledging the significance of, or even adverting to, petitioner’s constitutional right to confrontation, the [state] court held that petitioner’s right to effective cross‐examination was outweighed by the dan‐ ger that revealing [the victim’s] interracial rela‐ tionship would prejudice the jury against her. While a trial court may, of course, impose rea‐ sonable limits on defense counsel’s inquiry into the potential bias of a prosecution witness, to take account of such factors as “harassment, prejudice, confusion of the issues, the witness’ safety, or interrogation that [would be] repeti‐ tive or only marginally relevant,” Delaware v. Van Arsdall, [475 U.S.] at 679, the limitation here was beyond reason. Speculation as to the effect of jurors’ racial biases cannot justify exclusion of cross‐examination with such strong potential to demonstrate the falsity of [the victim’s] testi‐ mony. 20 No. 17‐2223

488 U.S. at 232. The Supreme Court found that the state court had “failed to accord proper weight to petitioner’s Sixth Amendment right ‘to be confronted with the witnesses against him.’” Id. at 231. Rhodes’s case presents the same tension between Rule 403 balancing and the Confrontation Clause rights of the accused. The State called Nari Rhodes as a witness to present evidence of its motive theory, which the State introduced as the central narrative from the start of the trial. On direct examination, the State asked Nari about her relationship with Davis and his vi‐ olence towards her. Rhodes tried to cross‐examine Nari on the same issues, to “delve into the witness’ story to test the wit‐ ness’ perceptions and memory,” and to explore her reliability as a witness. Delaware v. Fensterer, 474 U.S. 15, 19 (1985) (per curiam), quoting Davis, 415 U.S. at 316. As in Olden and Van Arsdall, and our decision in Redmond, the cross‐examination would have been relevant and probative to central issues in the trial. The Wisconsin Supreme Court failed to apply the princi‐ ple established in Olden and Van Arsdall: it applied only Rule 403 balancing when deciding whether Rhodes’s rights under the Confrontation Clause were violated. The state supreme court cited Van Arsdall and Davis v. Alaska, see 336 Wis. 2d at 77–79, 799 N.W.2d at 857, ¶¶29–32, but it did not follow the Supreme Court’s holdings in those cases. Instead, the state court then wrote that Wisconsin’s Rule 403 (Wis. Stat. § 904.03) provided the applicable rule, even for limits on a criminal defendant’s right to cross‐examine witnesses: “We apply the same analysis to discretionary decisions when the defendant’s right to cross‐examination under the confronta‐ tion clause may be implicated.” 336 Wis. 2d at 83, 799 N.W.2d No. 17‐2223 21

at 859, ¶44. The state court’s opinion went on to make unmis‐ takably clear that it was applying the ordinary Rule 403 dis‐ cretionary rule. 336 Wis. 2d at 83–84, 799 N.W.2d at 859–60, ¶¶45–48, citing State v. McCall, 202 Wis. 2d 29, 549 N.W.2d 418 (Wis. 1996) (where testimony sought was not relevant, de‐ fendant’s Sixth Amendment right to confront witness was not violated). Turning to Rhodes’s case, the state supreme court actually recognized that the testimony he sought to admit “would have been relevant.” 336 Wis. 2d at 87, 799 N.W.2d at 861–62, ¶58. The court went on to hold, however, that the trial court did not abuse its discretion under § 904.03 by excluding the evidence because of the danger of confusion of the issues and misleading the jury. 336 Wis. 2d at 87, 799 N.W.2d at 862, ¶¶59–60. That analysis was contrary to clearly established Supreme Court precedent holding that the Confrontation Clause is not satisfied merely because the evidence offered by the accused might be excluded properly under Rule 403 or its equivalent. Olden, 488 U.S. at 232 (per curiam); Van Arsdall, 475 U.S. at 679–80. The Confrontation Clause demands careful scrutiny of the purported reason for limiting cross‐examination. See Maryland v. Craig, 497 U.S. 836, 848 (1990) (“Thus, in certain narrow circumstances, ‘competing interests, if “closely exam‐ ined,” may warrant dispensing with confrontation at trial.’”), quoting Chambers, 410 U.S. at 295; Olden, 488 U.S. at 232 (per curiam); Van Arsdall, 475 U.S. at 679–80. In applying ordinary Rule 403 balancing, without giving any special consideration to the defendant’s constitutional right to confront witnesses against him, the state supreme court’s decision was contrary to controlling United States Supreme Court precedent. 22 No. 17‐2223

2. “Unreasonable Application” of Controlling Precedent If there were doubts under the “contrary to” clause in § 2254(d)(1)—that is, even if we assume that Supreme Court precedent on the Sixth Amendment did not clearly require something other than ordinary Rule 403 balancing, or if we assume, despite its express language otherwise, that the state supreme court applied the correct standard—the state court’s balancing was not reasonable in this case. The “denial or sig‐ nificant diminution” of cross‐examination “requires that the competing interest be closely examined” because it “calls into question the ultimate ‘integrity of the fact‐finding process.’” Chambers, 410 U.S. at 295, quoting Berger v. California, 393 U.S. 314, 315 (1969). Here, the trial court found that Rhodes’s cross‐ examination of Nari would confuse the issues, mislead the jury, and prejudice the victim. The state supreme court agreed that these concerns were weighty enough to justify prevent‐ ing Rhodes from questioning Nari on relevant, probative is‐ sues. The court’s rationale simply does not hold up once one looks closely at the trial. Rhodes’s cross‐examination could not have done any of these things because he was not the party who made Davis’s violence toward Nari and the ques‐ tion of motive central issues in the case. The prosecution built its case by raising both issues. This contradiction was so clear as to make the state court’s contrary conclusion not just an ordinary error, which would not support federal habeas relief under § 2254(d)(1), but an error that went beyond fairminded disagreement. The state court’s majority opinion rejected the Confronta‐ tion Clause claim on the rationale that the defense’s intended lines of questioning would have confused or misled the jury or prejudiced the victim. This rationale lost sight of the facts No. 17‐2223 23

of the case and the dynamics of the trial. It applied an analysis that might have made sense if the defense had been trying to open up the issue of the history of Davis’s violence against Nari, but not where the State itself had already introduced the subject to prove motive. Here is the heart of the state court’s explanation: ¶ 62 Judge McMahon was concerned that the jury would be misled into an improper focus on questions about motive and the alleged history of abuse between the victim and Nari. There was also the possibility that if Rhodes were al‐ lowed to emphasize his rebuttal theory he would not have retaliated against Davis for the April 3 incident because he had not retaliated against him before, he would have confused the issues. First, such testimony would, in effect, have put Davis—the deceased victim—on trial for alleged prior incidents of domestic violence. Second, it would have required the jury to spec‐ ulate as to whether a lack of retaliation for Da‐ vis’s prior assaults on Nari magnified Rhodes’ motive in this instance. Both were legitimate concerns. ¶ 63 At the same time, Judge McMahon was clearly mindful of the importance of allowing Rhodes to rebut the State’s theory of motive. Both Nari and Rhodes were allowed to present their side of the story to rebut the State’s theory of motive. The jury could reasonably have viewed their testimony as contradicting the State’s theory. 24 No. 17‐2223

¶ 64 If anything, Nari’s testimony seemed to refute the State’s theory of motive, because she testified that she and Davis had a friendly rela‐ tionship, and that Davis repeatedly warned her to leave the scene when Segura arrived and be‐ came angry. Nothing in her testimony sug‐ gested that she believed Davis orchestrated the beating she received at the hands of Segura and Bell. According to her testimony, Nari told her brothers that the two women were responsible, not Davis, and that the brothers took the news calmly. Counter to the State’s theory that Rhodes and Saleem believed Davis was respon‐ sible, Nari testified that her brothers were upset that she had put herself in the situation because of the bad blood between Nari and Segura, not between Nari and Davis. 336 Wis. 2d at 88–89, 799 N.W.2d at 862 ¶¶62–64. First, we do not see how the state court could reasonably find in Rhodes’s proposed cross‐examination a legitimate concern about an “improper focus on questions about motive and the alleged history of abuse between the victim and Nari.” Questions of “motive and the alleged history of abuse” were already a central part of the trial. The State, not Rhodes, had already put both of those subjects squarely before the jury. The State raised motive in its opening statement. The State called Nari as a witness. The motive theory was the only point of her testimony. The State questioned her about her in‐ juries and introduced graphic photographs of her injuries. The State then featured those photographs prominently in closing arguments, displaying them to the jury and arguing No. 17‐2223 25

that “this is what the defendant[s] saw when [Nari] came home from the hospital.” The State urged the jury: “this is what they sought to avenge by killing Robert Davis.” We do not see how it could be “improper” to allow the defense to respond, and with evidence the state supreme court itself recognized was relevant. In light of the State’s re‐ peated emphasis on Rhodes’s alleged motive, the conclusion that Rhodes’s questioning about the rest of the Nari/Davis story would mislead the jury into an improper focus on mo‐ tive was “so lacking in justification that there was an error well understood and comprehended in existing law beyond any possibility for fairminded disagreement.” Richter, 562 U.S. at 103. It was the State itself that wanted the jury to focus on motive. Second, for essentially the same reasons, we do not see how the state court could reasonably have found a risk that Rhodes’s rebuttal evidence on cross‐examination would have “confused the issues” by “in effect, [putting] Davis—the de‐ ceased victim—on trial for alleged prior incidents of domestic violence,” or by requiring the jury “to speculate as to whether a lack of retaliation for Davis’s prior assaults on Nari magni‐ fied Rhodes’ motive in this instance.” 336 Wis. 2d at 88, 799 N.W.2d at 862, ¶62. Under other circumstances, those might have been “legitimate concerns,” in the court’s words, but not after the State put on its motive evidence. It was the State itself that “in effect put Davis on trial” for the April 3 assaults. It was the State that invited the jury to—call it “speculate” or call it “infer”—from Nari’s injuries on April 3 that Rhodes and Saleem would have been motivated to shoot Davis. Rhodes wanted to respond by asking the jury to infer, from testimony that Nari could have provided, that he was not motivated to 26 No. 17‐2223

kill Davis. We see no qualitative differences between the very recent violence (the day before the murder) and the evidence that Nari could have provided about earlier instances of abuse, which did not provoke her brothers. The fact that past similar provocations by Davis had not driven Rhodes or Saleem to violence was certainly relevant, as the state supreme court acknowledged. In fact, the state court found Rhodes’s intended cross‐examination “confus‐ ing” for the precise reasons that it was relevant and probative: it would have provided facts from which the jury could infer the Rhodes was motivated—or not—to kill Davis. A jury that heard the full story of Davis’s violence against Nari might well have concluded that she was testifying truthfully and that Rhodes had given up trying to protect her from this abu‐ sive relationship. Or perhaps that hypothetical jury might have concluded, as the State hoped, that even if Rhodes was not angry in the past, the beating on April 3 was the final straw for him. The notion that the State could put on its evi‐ dence of motive but that rebuttal evidence would “confuse” the jury misunderstands the role of the jury. It is not to be pro‐ tected from evidence that would directly rebut the State’s the‐ ory. Giving the jury conflicting evidence relevant to that ques‐ tion cannot reasonably be called “confusing” the jury within the meaning of the rules of evidence. At the risk of sounding trite, that’s what we ask juries to do: sort out conflicting evi‐ dence. Finally, the state court’s third rationale for limiting cross‐ examination—the potential for unfair prejudice to Davis—is equally unreasonable. The State had already painted Davis in a negative light. The State’s direct examination focused on a fight earlier on April 3, the day before the shooting, in which No. 17‐2223 27

Davis stole Nari’s keys, wallet, and phone, scared her by an‐ grily trying to pull their baby from the car, and got so angry that he punched and broke her car window. The State’s questions then shifted to the brutal beating of Nari that same afternoon—a beating that the State wanted the jury to believe that Davis caused. And Nari had already testi‐ fied on direct and cross‐examination that Davis had abused her in the past. By the time the court cut off cross‐examination, she had also testified that Davis had beaten her in the past, including breaking her orbital bone. Rhodes’s counsel was prohibited from asking follow‐up questions that would have added little if any prejudice to Davis, specifically: what an or‐ bital bone is; whether her brothers were aware of that previ‐ ous injury; and whether they had reacted to that beating. Rhodes was unable to question Nari about her repeated re‐ turns to Davis, which he claims led him and his brother to give up on their efforts to protect her from Davis. These addi‐ tional questions could not have been unfairly prejudicial to the prosecution. See Redmond, 240 F.3d at 592. They were “preju‐ dicial” only insofar as they presented facts that could lead a jury to reject the State’s theory of a motive and to raise reason‐ able doubts about whether Rhodes shot and killed Davis. In Paragraphs 63 to 66 of its opinion, the state supreme court took a different approach, asserting in effect that the trial judge had struck a reasonable balance between compet‐ ing interests. The court asserted that both Nari and Rhodes “were allowed to present their side of the story to rebut the State’s theory of motive,” that Nari’s direct testimony actually helped Rhodes on the issue of motive, and that more evidence about the history of Davis’s abuse of Nari, and the brothers’ 28 No. 17‐2223

lack of violent response, would not likely have helped the de‐ fense. 336 Wis. 2d at 88–90, 799 N.W.2d at 862–63, ¶¶63–66. In this appeal, the State makes similar arguments. It argues that Rhodes had a sufficient opportunity to cross‐examine Nari and implies that Rhodes’s Sixth Amendment rights were sat‐ isfied because he was able to testify on similar topics himself. Neither argument cures the basic unfairness of the state court decision. The notion that Nari and Rhodes were “allowed to present their side of the story” cannot be squared with the trial record. Yes, they managed to fit in a conclusory, bobtailed version of the rebuttal. But that was matched up against the State’s detailed and full‐color presentation of its motive theory limited to the April 3 beating. The Sixth Amendment guarantees only “an opportunity” to cross‐ examine the witness, Fensterer, 474 U.S. at 20, but the Sixth Amendment is not satisfied when the defendant is permitted to ask only general questions. Davis, 415 U.S. at 318. To make cross‐examination “effective, defense counsel should have been permitted to expose to the jury the facts from which jurors, as the sole triers of fact and credibility, could appropriately draw inferences relating to the reliability of the witness.” Id. And it makes no difference that Rhodes himself testified on the topic, albeit also in a very limited way. The focus of the analysis for “whether the confrontation right has been vio‐ lated must be on the particular witness,” not on whether the defendant could admit similar evidence through other means. Van Arsdall, 475 U.S. at 680. Any suggestion that the accused’s own testimony is a sufficient substitute is unreasonable and not supported by precedent. Every trial lawyer and trial judge understands that the accused’s own testimony poses unique No. 17‐2223 29

credibility problems. He has the greatest incentive to lie or conceal. That’s why corroborating testimony from other wit‐ nesses—even from family members—can be essential for the accused. To the extent the state court was suggesting the re‐ strictions on cross‐examination of Nari were not actually harmful to the defense, that view is not consistent with this trial record or the realities of trials. To start, the state supreme court stated expressly that it was not deciding the issue of harmless error. 336 Wis. 2d at 91, 799 N.W.2d at 863, ¶69 n.8. (Recall that the state appellate court found that the Confron‐ tation Clause violation was not harmless.) Plus, the state su‐ preme court was surely correct in saying: “In the end, the jury was required to make a determination of credibility as to the testimony presented.” 336 Wis. 2d at 89, 799 N.W.2d at 863, ¶66. The problem is that the jury heard a badly lopsided presentation of evidence on the motive theory. Given the un‐ fair skewing of the opposing sides’ opportunities to address motive, we do not see how the state supreme court could rea‐ sonably assert with any confidence that a more detailed ac‐ count of the relevant facts was unlikely to cause the jury to doubt Rhodes’s guilt. The State raises one additional point that warrants atten‐ tion. At oral argument, the State suggested that Rhodes’s cross‐examination would not have addressed Nari’s credibil‐ ity or reliability. The state court suggested something similar, mentioning briefly that if Rhodes had attempted to challenge Nari’s motivation or credibility, “we might have a different case.” 336 Wis. 2d at 84–85, 799 N.W.2d at 860, ¶51. But Rhodes was attempting to cross‐examine Nari about her cred‐ ibility and reliability. The State’s theory was that Rhodes 30 No. 17‐2223

killed Davis to avenge Nari. Once Nari testified that Rhodes was not angry about the beating, the State attacked her credi‐ bility, arguing she was lying on that point to protect her broth‐ ers. The State said as much during closing arguments. The prosecution argued that it “doesn’t make sense at all” that Rhodes was not angry, and that what “makes sense is that they were horrified and angered by the fact that Nari Rhodes took such a bad beating at the hands of Robert Davis’ other girlfriend.” As one of the final statements in its closing rebut‐ tal, the State told the jury: “Nari Rhodes’ testimony. I can only say that she is the sister of the defendants. Place this in terms of how you weigh her credibility.” There can be no genuine dispute whether Nari’s credibility was at issue, or whether Rhodes’s disallowed cross‐examination was relevant to her credibility and reliability. As noted, the state supreme court recognized as much in Paragraph 66: “In the end, the jury was required to make a determination of credibility as to the testimony presented.” 336 Wis. 2d at 89, 799 N.W.2d at 863, ¶66. Rhodes sought to cross‐examine Nari further to bolster the credibility of her testimony that favored him, and that the State was attacking. The “jurors were entitled to have the benefit of the defense theory before them so that they could make an informed judgment as to the weight to place on” Nari’s “testimony which provided a ‘crucial link in the proof … of petitioner’s act.’” Davis, 415 U.S. at 317 (finding Confrontation Clause violation), quoting Douglas, 380 U.S. at 419. To sum up the merits, the state appellate court correctly found a Confrontation Clause violation here. In reversing, the state supreme court cited but then failed to apply the correct standard from federal law. To the extent the state supreme No. 17‐2223 31

court’s opinion might be read as trying to apply the correct standard, it did so in an unreasonable way. It unreasonably gave weight to concerns about misleading or confusing the jury with evidence that rebutted a theory that the State had put at the center of the trial. And it rationalized the result without recognizing the clearly unfair balance between allow‐ ing the State to present its evidence on motive while barring virtually all defense evidence on the subject. In short, the state supreme court lost sight of the central role of motive in the State’s case and the defendants’ need—and right—to rebut it. We recognize that trial courts deal, all the time, with ef‐ forts by guilty defendants to change the subject of the trial— to put on trial the police, or the victim, or society at large. In some cases, defendants might offer circumstantial evidence through cross‐examination that requires implausible or far‐ fetched speculation about tangential issues. Trial judges are entitled to insist that evidence be relevant and to impose rea‐ sonable limits on such efforts to change the subject. But after the State offered its evidence and arguments about motive, it was not reasonable to extend these ordinary rationales to this case, to the defense efforts to rebut the prosecution motive theory by telling the more complete story of Davis’s violent abuse of Nari. B. Harmless Error? Ordinarily, the next question is whether the constitutional violation was harmless error. The only state court to rule on the issue found that the error was not harmless, and we agree with that conclusion. The matter is complicated here proce‐ durally because the State waived the issue by not raising it in the district court, but the district court then acted to relieve the State of the waiver by inviting briefing on the question. 32 No. 17‐2223

The district judge believed that she was required to consider harmless error even though the State had waived that issue in the habeas case. We agree that the State waived or at least forfeited the harmless error defense. The district court was not required to rescue the State from that waiver or forfeiture. Having invited briefing on the question, though, the district court had author‐ ity to decide it. Trial judges often act in the interests of justice to relieve one party or another from a strategic or tactical er‐ ror. But we disagree with the district judge on the bottom line. This Confrontation Clause violation was not harmless. Rhodes is entitled to a new trial. In believing it was obliged to raise the issue of harmless error, the district court relied on Brecht v. Abrahamson, 507 U.S. 619 (1993). The issue in Brecht was which harmless‐error standard applied during habeas proceedings: the standard under Chapman v. California, 386 U.S. 18 (1967), which is more generous to petitioners, or the standard under Kotteakos v. United States, 328 U.S. 750 (1946), which is more generous to the government. 507 U.S. at 622–23. Even though the Brecht Court spoke in terms of what a habeas petitioner “must show” and what a court “must determine,” the Court was not considering whether harmless‐error review could be waived or forfeited. And the Supreme Court has warned against tak‐ ing the holding in Brecht out of context. O’Neal v. McAninch, 513 U.S. 432, 438–39 (1995) (Brecht addressed only the ques‐ tion before the Court). Some circuits had relied on Brecht’s statement that habeas petitioners “are not entitled to habeas relief based on trial error unless they can establish that it re‐ sulted in ‘actual prejudice’” to mean that Brecht placed the burden of showing prejudice on habeas petitioners. Id. at 438 No. 17‐2223 33

(emphasis in original). O’Neal explained that this language was “not determinative” in part because the “issue in Brecht involved a choice of substantive harmless‐error standards.” Id. at 438. Brecht did not address whether states can waive or forfeit harmless error review. We have held before that the State can waive or forfeit the harmless error issue. AEDPA is full of procedural pitfalls that prevent prisoners from challenging potentially unconstitu‐ tional convictions. “Procedural rules apply to the government as well as to defendants.” Wilson v. O’Leary, 895 F.2d 378, 384 (7th Cir. 1990) (affirming grant of habeas relief where state waived harmless error arguments). We enforce the law against habeas petitioners who fail to navigate procedural hurdles. We see no reason, in this context, to hold that a court can never enforce a forfeiture or waiver when a state fails to raise harmless error. Harmless error is a fact‐intensive in‐ quiry. It is not the court’s job to “search the record—without any help from the parties—to determine that the errors we find are prejudicial.” United States v. Giovannetti, 928 F.2d 225, 226 (7th Cir. 1991). Enforcing forfeiture also encourages effi‐ cient briefing and thus efficient use of the court’s time. Id. In Giovannetti, we therefore held that the government can forfeit harmless error review. In Sanders v. Cotton, 398 F.3d 572, 582 (7th Cir. 2005), we applied Giovannetti when a state failed to raise harmless error during a habeas proceeding gov‐ erned by AEDPA. Other circuits agree. See Miller v. Stovall, 608 F.3d 913, 926–27 (6th Cir. 2010) (finding that state waived harmless error review), reversed on other grounds, Stovall v. Miller, 565 U.S. 1031 (2011); Jones v. Cain, 600 F.3d 527, 540–41 (5th Cir. 2010) (same); Cook v. McKune, 323 F.3d 825, 840 & n.9 (10th Cir. 2003) (same). Sanders post‐dates Brecht and controls 34 No. 17‐2223

this case. It was therefore error for the district court to con‐ clude that it had to find actual prejudice before granting Rhodes’s petition. Nevertheless, the district court acted to relieve the state of its forfeiture of harmless error. We view harmless error in a habeas case as an issue where a court has discretion, in the interests of justice, to relieve a party of its mistake or forfei‐ ture. This discretion has its parallels in doctrines that allow courts to overlook procedural mistakes by habeas petitioners when the circumstances are sufficiently compelling. It would be unfortunate to reverse a conviction for a serious crime only because the state’s lawyer failed to argue harmless error. Re‐ versal imposes costs on the state that must retry the petitioner and implicates concerns about finality and federalism. Recog‐ nizing that “reversal may be an excessive sanction for the gov‐ ernment’s having failed to argue harmless error,” we “have discretion to overlook a failure to argue harmlessness.” Gio‐ vannetti, 928 F.2d at 227. We “generally do so only when the ‘harmlessness of the error or errors found is certain’ and ‘a reversal would result in protracted, costly, and ultimately fu‐ tile proceedings.’” Sanders, 398 F.3d at 582, quoting Giovan‐ netti, 928 F.2d at 227. The district court should have applied Giovannetti and Sanders and asked whether the error was cer‐ tainly harmless. Rather than rely now on forfeiture or the difference be‐ tween “harmless” and “certainly harmless,” however, we find that the most straight‐forward approach to decide this appeal is to confront the issue of harmless error, and to do so de novo. Recall that the state appellate court found that the violation No. 17‐2223 35

was not harmless, and the state supreme court did not reach that issue.1 On direct appeal, the harmless‐error standard is whether the error was harmless beyond a reasonable doubt, but the test is different in collateral proceedings like this case. Davis v. Ayala, 135 S. Ct. 2187, 2197 (2015). When reviewing a state‐ court judgment in a habeas corpus proceeding, we ask whether the error “had substantial and injurious effect or in‐ fluence in determining the jury’s verdict.” Fry v. Pliler, 551 U.S. 112, 116 (2007), quoting Brecht, 507 U.S. at 631; see also Jensen v. Clements, 800 F.3d 892, 902–903 (7th Cir. 2015) (ap‐ plying harmless‐error analysis in habeas review of Confron‐ tation Clause violation). This standard applies even when the state court did not address the issue of harmless error. Fry, 551 U.S. at 121–22. Harmless error “is not the same as a review for whether there was sufficient evidence at trial to support a verdict.” Jen‐ sen, 800 F.3d at 902. The “inquiry cannot be merely whether there was enough to support the result, apart from the phase affected by the error.” Id., quoting Kotteakos, 328 U.S. at 764– 65. The question is “rather, even so, whether the error itself had substantial influence.” Id., again quoting Kotteakos, 328

1 The State’s brief in this appeal dedicated a single footnote to Rhodes’s arguments on waiver, arguing only that the question of waiver is not within the scope of the certificate of appealability. The district court granted the certificate of appealability on Rhodes’s Sixth Amendment claim, without limiting it to only certain issues relevant to that claim. If the certificate were as limited as the State proposes, the certificate would have been futile—affirmance would have been inevitable, regardless of our view of the merits of the Confrontation Clause issue. We do not think the district court intended its certificate to be futile. 36 No. 17‐2223

U.S. at 764–65. This requires “more than a ‘reasonable possi‐ bility’ that the error was harmful.” Ayala, 135 S. Ct. at 2198, quoting Brecht, 507 U.S. at 637. It requires the court to “’find that the defendant was actually prejudiced by the error.’” Id., quoting Calderon v. Coleman, 525 U.S. 141, 146 (1998) (per cu‐ riam). To answer that question in a Confrontation Clause case, we consider a variety of factors, such as “the importance of the witness’ testimony in the prosecution’s case, whether the testimony was cumulative, the presence or absence of evi‐ dence corroborating or contradicting the testimony of the wit‐ ness on material points, the extent of cross‐examination oth‐ erwise permitted, and, of course, the overall strength of the prosecution’s case.” Jones v. Basinger, 635 F.3d 1030, 1052 (7th Cir. 2011), quoting Van Arsdall, 475 U.S. at 684. Most of these factors weigh in Rhodes’s favor. A critical point here is that the same jury that convicted Rhodes found his brother Saleem not guilty, as the state appellate court noted. 329 Wis. 2d 268 at ¶10 n.1. The prosecution case was legally sufficient but not overwhelming. Nari’s testimony and the motive theory were central to the prosecution’s case. The prosecution focused on her beating as a motive for the shooting from the start of the case. As noted, the State used graphic photographs of her injuries during closing arguments to argue that Rhodes must have decided to avenge his sister. Nari’s testimony about the history of her abuse by Davis and the initial reactions by her brothers, her repeated returns to Davis, and her brothers’ decision not to react to the beating that broke her orbital‐bone would not have been cumulative. No. 17‐2223 37

No other witness testified about whether Rhodes reacted an‐ grily when Davis broke Nari’s orbital bone—not even Rhodes, since the defense had been told not to review this prior abuse. The defense also was not allowed to offer evi‐ dence as to why Rhodes had not been angry with Davis be‐ cause of the orbital bone injury, for that required even more history to explain fairly. Rhodes was able to cross‐examine Nari, but not on an issue that was key to rebutting the State’s motive theory. In closing arguments, Rhodes’s lawyer could and did argue that he was not angry with Davis for Nari’s beating, but Rhodes was not allowed to offer, through cross‐ examination, key evidence to support that argument, particu‐ larly in the face of the prosecution’s sharp attacks on his and Nari’s credibility. The final factor is the strength of the overall case against Rhodes. There was no physical evidence implicating him. The only direct evidence was the testimony of Watt and Walker. Walker’s description of the shooters changed after she had a chance to talk to Watt. According to one detective’s report, Walker initially said that the shooter she said was Rhodes was wearing a dark t‐shirt and hat. Later, Walker testified that the same man was wearing a red, white, and blue sweater and no hat. And it took Walker two passes through a photo array to identify Rhodes as a shooter. The third eyewitness, Watt’s grandfather, testified that he had stared one shooter in the eyes, but he could not identify that shooter as either Rhodes or Saleem. Yet Rhodes grew up around the corner from where Watt’s grandfather had lived for decades. The grandfather also testified that he saw only one shooter, while Watt and Walker testified there were two. 38 No. 17‐2223

The cell phone location evidence was not probative since Rhodes still lived around the corner from the shooting. It is true that two officers testified that Dotson initially told them that Rhodes called to say he had shot Davis. Their testi‐ mony was corroborated by a contemporaneous written report of Dotson’s interview. The jury may have found their evi‐ dence persuasive, which could explain why the jury con‐ victed Rhodes but acquitted Saleem. But the officers never had Dotson sign a statement. Dotson testified that she never told the officers that Rhodes said he had shot Davis, only that Rhodes said “someone had got shot.” (Recall again that Rhodes lived around the corner from the shooting.) She also testified that she agreed to speak to the officers at all only after they threatened to have a social worker take her three‐week‐ old baby away from her. In the end, we cannot say that the Confrontation Clause violation had no “substantial and injurious effect” on the jury’s deliberations. The State presented other evidence against Rhodes, but none of the evidence was without its problems. Motive was central to the case, and Nari’s testi‐ mony was central to Rhodes’s attempt to rebut the State’s the‐ ory. Without Rhodes’s rebuttal evidence, the jury was left without key facts relevant to Nari’s credibility and Rhodes’s guilt or innocence. The judgment of the district court is therefore REVERSED and this case is REMANDED to the district court with instruc‐ tions to grant the writ of habeas corpus ordering that Rhodes be released or retried promptly.