Justice ALITO delivered the opinion of the Court.

A quarter-century after a California jury convicted Hector Ayala of triple murder and sentenced him to death, the Court of Appeals for the Ninth Circuit granted Ayala's application for a writ of habeas corpus and ordered the State to retry or release him. The Ninth Circuit's decision was based on the procedure used by the trial judge in ruling on Ayala's objections under Batson v. Kentucky, 476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986), to some of the prosecution's peremptory challenges of prospective jurors. The trial judge allowed the prosecutor to explain the basis for those strikes outside the presence of the defense so as not to disclose trial strategy. On direct appeal, the California Supreme Court found that if this procedure violated any federal constitutional right, the error was harmless beyond a reasonable doubt. The Ninth Circuit, however, held that the error was harmful.

The Ninth Circuit's decision was based on the misapplication of basic rules regarding harmless error. Assuming without deciding that a federal constitutional error occurred, the error was harmless under Brecht v. Abrahamson, 507 U.S. 619, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993), and the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), 28 U.S.C. § 2254(d).

I

A

Ayala's conviction resulted from the attempted robbery of an automobile body shop in San Diego, California, in April 1985. The prosecution charged Ayala with three counts of murder, one count of attempted murder, one count of robbery, and three counts of attempted robbery. The prosecution also announced that it would seek the death penalty on the murder counts.

Jury selection lasted more than three months, and during this time the court and the parties interviewed the prospective jurors and then called back a subset for general voir dire. As part of the jury selection process, more than 200 potential jurors completed a 77-question, 17-page questionnaire. Potential jurors were then questioned in court regarding their ability to follow the law. Jurors who were not dismissed for cause were called back in groups for voir dire, and the parties exercised their peremptory challenges.

Each side was allowed 20 peremptories, and the prosecution used 18 of its allotment. It used seven peremptories to strike all of the African-Americans and Hispanics who were available for service. Ayala, who is Hispanic, raised Batson objections to those challenges.

Ayala first objected after the prosecution peremptorily challenged two African-Americans, Olanders D. and Galileo S. The trial judge stated that these two strikes failed to establish a prima facie case of racial discrimination, but he nevertheless required the prosecution to reveal the reasons for the strikes. The prosecutor asked to do this outside the presence of the defense so as not to disclose trial strategy, and over Ayala's objection, the judge granted the request. The prosecution then offered several reasons for striking Olanders D., including uncertainty about his willingness to impose the death penalty. The prosecution stated that it dismissed Galileo S. primarily because he had been arrested numerous times and had not informed the court about all his prior arrests. After hearing and evaluating these explanations, the judge concluded that the prosecution had valid, race-neutral reasons for these strikes.

Ayala again raised Batson objections when the prosecution used peremptory challenges to dismiss two Hispanics, Gerardo O. and Luis M. As before, the judge found that the defense had not made out a prima facie case, but ordered the prosecution to reveal the reasons for the strikes. This was again done ex parte, but this time the defense did not expressly object. The prosecution explained that it had challenged Gerardo O. and Luis M. in part because it was unsure that they could impose the death penalty. The prosecution also emphasized that Gerardo O.'s English proficiency was limited and that Luis M. had independently investigated the case. The trial court concluded a second time that the prosecution had legitimate race-neutral reasons for the strikes.

Ayala raised Batson objections for a third and final time when the prosecution challenged Robert M., who was Hispanic; George S., whose ethnicity was disputed; and Barbara S., who was African-American. At this point, the trial court agreed that Ayala had made a prima facie Batson showing. Ayala's counsel argued that the strikes were in fact based on race. Ayala's counsel contended that the challenged jurors were "not significantly different from the white jurors that the prosecution ha[d] chosen to leave on the jury both in terms of their attitudes on the death penalty, their attitudes on the criminal justice system, and their attitudes on the presumption of innocence." App. 306. Ayala's counsel then reviewed the questionnaire answers and voir dire testimony of Barbara S. and Robert M., as well as the statements made by three of the prospective jurors who had been the subject of the prior Batson objections, Galileo S., Gerardo O., and Luis M. Counsel argued that their answers showed that they could impose the death penalty. The trial court stated that it would hear the prosecution's response outside the presence of the jury, and Ayala once more did not object to that ruling. The prosecution then explained that it had dismissed the prospective jurors in question for several race-neutral reasons, including uncertainty that Robert M., George S., or Barbara S. would be open to imposing the death penalty. The prosecution also emphasized (among other points) that Robert M. had followed a controversial trial, that George S. had been a holdout on a prior jury, and that Barbara S. had given the impression during voir dire that she was under the influence of drugs. The trial court concluded, for a third time, that the prosecution's peremptory challenges were based on race-neutral criteria.

In August 1989, the jury convicted Ayala of all the charges except one of the three attempted robberies. With respect to the three murder convictions, the jury found two special circumstances: Ayala committed multiple murders, and he killed during the course of an attempted robbery. The jury returned a verdict of death on all three murder counts, and the trial court entered judgment consistent with that verdict.

B

Ayala appealed his conviction and sentence, and counsel was appointed to represent him in January 1993. Between 1993 and 1999, Ayala filed 20 applications for an extension of time, 11 of which requested additional time to file his opening brief. After the California Supreme Court eventually ruled that no further extensions would be granted, Ayala filed his opening brief in April 1998, nine years after he was convicted. The State filed its brief in September 1998, and Ayala then asked for four extensions of time to file his reply brief. After the court declared that it would grant him no further extensions, he filed his reply brief in May 1999.

In August 2000, the California Supreme Court affirmed Ayala's conviction and death sentence. People v. Ayala, 24 Cal.4th 243, 99 Cal.Rptr.2d 532, 6 P.3d 193 (2000). In an opinion joined by five justices, the State Supreme Court rejected Ayala's contention that the trial court committed reversible error by excluding the defense from part of the Batson hearing. The court understood Ayala to challenge the peremptory strikes under both Batson and its state-law analogue, People v. Wheeler, 22 Cal.3d 258, 148 Cal.Rptr. 890, 583 P.2d 748 (1978). The court first concluded that the prosecution had not offered matters of trial strategy at the ex parte hearing and that, "as a matter of state law, it was [error]" to bar Ayala's attorney from the hearing. 24 Cal.4th, at 262, 99 Cal.Rptr.2d 532, 6 P.3d, at 203.

Turning to the question of prejudice, the court stated:

The court then reviewed the prosecution's reasons for striking the seven prospective jurors and found that "[o]n this well-developed record, ... we are confident that defense counsel could not have argued anything substantial that would have changed the court's rulings. Accordingly, the error was harmless." Id., at 268, 99 Cal.Rptr.2d 532, 6 P.3d, at 207. The court concluded that the record supported the trial judge's implicit determination that the prosecution's justifications were not fabricated and were instead "grounded in fact." Id., at 267, 99 Cal.Rptr.2d 532, 6 P.3d, at 206. And the court emphasized that the "trial court's rulings in the ex parte hearing indisputably reflect both its familiarity with the record of voir dire of the challenged prospective jurors and its critical assessment of the prosecutor's proffered justifications." Id., at 266-267, 99 Cal.Rptr.2d 532, 6 P.3d, at 206.

The California Supreme Court also rejected Ayala's argument that his conviction should be vacated because most of the questionnaires filled out by prospective jurors who did not serve had been lost at some point during the decade that had passed since the end of the trial. The court wrote that "the record is sufficiently complete for us to be able to conclude that [the prospective jurors who were the subject of the contested peremptories] were not challenged and excused on the basis of forbidden group bias." Id., at 270, 99 Cal.Rptr.2d 532, 6 P.3d, at 208. And even if the loss of the questionnaires was error under federal or state law, the court held, the error was harmless under Chapman and its state-law analogue. Two justices of the State Supreme Court dissented. We then denied certiorari. 532 U.S. 1029, 121 S.Ct. 1978, 149 L.Ed.2d 770 (2001).

C

After the California Supreme Court summarily denied a habeas petition, Ayala turned to federal court. He filed his initial federal habeas petition in 2002, but then went back to state court to exhaust several claims. In December 2004, he filed the operative federal petition and argued, among other things, that the ex parte hearings and loss of the questionnaires violated his rights under the Sixth, Eighth, and Fourteenth Amendments.

In 2006, the District Court denied Ayala relief on those claims. The District Court read the decision of the California Supreme Court to mean that the state court had not decided whether the ex parte proceedings violated federal law, and the District Court expressed doubt "whether the trial court's procedure was constitutionally defective as a matter of clearly established Federal law." App. to Pet. for Cert. 145a. But even if such a violation occurred, the District Court held, the state court's finding of harmlessness was not contrary to or an unreasonable application of clearly established law and thus could not be overturned under AEDPA. The District Court also rejected Ayala's argument about the lost questionnaires, concluding that, even without them, the record was sufficient to resolve Ayala's other claims.

In 2013, a divided panel of the Ninth Circuit granted Ayala federal habeas corpus relief and required California either to release or retry him. Ayala v. Wong, 756 F.3d 656 (2014). Because Ayala's federal petition is subject to the requirements of AEDPA, the panel majority began its analysis by inquiring whether the state court had adjudicated Ayala's claims on the merits. Applying de novo review,1 the panel held that the ex parte proceedings violated the Federal Constitution, and that the loss of the questionnaires violated Ayala's federal due process rights if that loss deprived him of "the ability to meaningfully appeal the denial of his Batson claim." Id., at 671. The panel folded this inquiry into its analysis of the question whether the error regarding the ex parte proceedings was harmless.

Turning to the question of harmlessness, the panel identified the applicable standard of review as that set out in Brecht and added: "We apply the Brecht test without regard for the state court's harmlessness determination." 756 F.3d, at 674 (internal quotation marks omitted).2 The panel used the following complicated formulation to express its understanding of Brecht's application to Ayala's claims: "If we cannot say that the exclusion of defense counsel with or without the loss of the questionnaires likely did not prevent Ayala from prevailing on his Batson claim, then we must grant the writ." 756 F.3d, at 676. Applying this test, the panel majority found that the error was not harmless, at least with respect to three of the seven prospective jurors. The panel asserted that the absence of Ayala and his counsel had interfered with the trial court's ability to evaluate the prosecution's proffered justifications for those strikes and had impeded appellate review, and that the loss of the questionnaires had compounded this impairment.

Judge Callahan dissented. She explained that the California Supreme Court's decision that any federal error was harmless constituted a merits adjudication of Ayala's federal claims. She then reviewed the prosecution's explanations for its contested peremptory challenges and concluded that federal habeas relief was barred because "fairminded jurists can concur in the California Supreme Court's determination of harmless error." Id., at 706.

The Ninth Circuit denied rehearing en banc, but Judge Ikuta wrote a dissent from denial that was joined by seven other judges. Like Judge Callahan, Judge Ikuta concluded that the California Supreme Court adjudicated the merits of Ayala's federal claims. Instead of the panel's "de novo review of the record that piles speculation upon speculation," she would have found that the state court's harmlessness determination was not an unreasonable application of Chapman. 756 F.3d, at 723.

We granted certiorari. 574 U.S. ___, 135 S.Ct. 401, 190 L.Ed.2d 288 (2014).

II

Ayala contends that his federal constitutional rights were violated when the trial court heard the prosecution's justifications for its strikes outside the presence of the defense, but we find it unnecessary to decide that question. We assume for the sake of argument that Ayala's federal rights were violated, but that does not necessarily mean that he is entitled to habeas relief. In the absence of "the rare type of error" that requires automatic reversal, relief is appropriate only if the prosecution cannot demonstrate harmlessness. Glebe v. Frost, 574 U.S. ___, ___, 135 S.Ct. 429, 429, 190 L.Ed.2d 317 (2014) (per curiam). The Ninth Circuit did not hold—and Ayala does not now contend—that the error here falls into that narrow category, and therefore Ayala is entitled to relief only if the error was not harmless.

The test for whether a federal constitutional error was harmless depends on the procedural posture of the case. On direct appeal, the harmlessness standard is the one prescribed in Chapman, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705: "[B]efore a federal constitutional error can be held harmless, the court must be able to declare a belief that it was harmless beyond a reasonable doubt." Id., at 24, 87 S.Ct. 824.

In a collateral proceeding, the test is different. For reasons of finality, comity, and federalism, habeas petitioners "are not entitled to habeas relief based on trial error unless they can establish that it resulted in `actual prejudice.'" Brecht, 507 U.S., at 637, 113 S.Ct. 1710 (quoting United States v. Lane, 474 U.S. 438, 449, 106 S.Ct. 725, 88 L.Ed.2d 814 (1986)). Under this test, relief is proper only if the federal court has "grave doubt about whether a trial error of federal law had `substantial and injurious effect or influence in determining the jury's verdict.'" O'Neal v. McAninch, 513 U.S. 432, 436, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995). There must be more than a "reasonable possibility" that the error was harmful. Brecht, supra, at 637, 113 S.Ct. 1710 (internal quotation marks omitted). The Brecht standard reflects the view that a "State is not to be put to th[e] arduous task [of retrying a defendant] based on mere speculation that the defendant was prejudiced by trial error; the court must find that the defendant was actually prejudiced by the error." Calderon v. Coleman, 525 U.S. 141, 146, 119 S.Ct. 500, 142 L.Ed.2d 521 (1998) (per curiam).

Because Ayala seeks federal habeas corpus relief, he must meet the Brecht standard, but that does not mean, as the Ninth Circuit thought, that a state court's harmlessness determination has no significance under Brecht. In Fry v. Pliler, 551 U.S. 112, 120, 127 S.Ct. 2321, 168 L.Ed.2d 16 (2007), we held that the Brecht standard "subsumes" the requirements that § 2254(d) imposes when a federal habeas petitioner contests a state court's determination that a constitutional error was harmless under Chapman. The Fry Court did not hold—and would have had no possible basis for holding—that Brecht somehow abrogates the limitation on federal habeas relief that § 2254(d) plainly sets out. While a federal habeas court need not "formal[ly]" apply both Brecht and "AEDPA/Chapman," AEDPA nevertheless "sets forth a precondition to the grant of habeas relief." Fry, supra, at 119-120, 127 S.Ct. 2321.

Under AEDPA, 28 U.S.C. § 2254(d):

Section 2254(d) thus demands an inquiry into whether a prisoner's "claim" has been "adjudicated on the merits" in state court; if it has, AEDPA's highly deferential standards kick in. Harrington v. Richter, 562 U.S. 86, 103, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011).

At issue here is Ayala's claim that the ex parte portion of the Batson hearings violated the Federal Constitution. There is no dispute that the California Supreme Court held that any federal error was harmless beyond a reasonable doubt under Chapman, and this decision undoubtedly constitutes an adjudication of Ayala's constitutional claim "on the merits." See, e.g., Mitchell v. Esparza, 540 U.S. 12, 17-18, 124 S.Ct. 7, 157 L.Ed.2d 263 (2003) (per curiam). Accordingly, a federal habeas court cannot grant Ayala relief unless the state court's rejection of his claim (1) was contrary to or involved an unreasonable application of clearly established federal law, or (2) was based on an unreasonable determination of the facts. Because the highly deferential AEDPA standard applies, we may not overturn the California Supreme Court's decision unless that court applied Chapman "in an `objectively unreasonable' manner." Id., at 18, 124 S.Ct. 7 (quoting Lockyer v. Andrade, 538 U.S. 63, 75, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003)). When a Chapman decision is reviewed under AEDPA, "a federal court may not award habeas relief under § 2254 unless the harmlessness determination itself was unreasonable." Fry, supra, at 119, 127 S.Ct. 2321 (emphasis in original). And a state-court decision is not unreasonable if "`fairminded jurists could disagree' on [its] correctness." Richter, supra, at 101, 131 S.Ct. 770 (quoting Yarborough v. Alvarado, 541 U.S. 652, 664, 124 S.Ct. 2140, 158 L.Ed.2d 938 (2004)). Ayala therefore must show that the state court's decision to reject his claim "was so lacking in justification that there was an error well understood and comprehended in existing law beyond any possibility for fairminded disagreement." 562 U.S., at 103, 131 S.Ct. 770.

In sum, a prisoner who seeks federal habeas corpus relief must satisfy Brecht, and if the state court adjudicated his claim on the merits, the Brecht test subsumes the limitations imposed by AEDPA. Fry, supra, at 119-120, 127 S.Ct. 2321.

III

With this background in mind, we turn to the question whether Ayala was harmed by the trial court's decision to receive the prosecution's explanation for its challenged strikes without the defense present. In order for this argument to succeed, Ayala must show that he was actually prejudiced by this procedure, a standard that he necessarily cannot satisfy if a fairminded jurist could agree with the California Supreme Court's decision that this procedure met the Chapman standard of harmlessness. Evaluation of these questions requires consideration of the trial court's grounds for rejecting Ayala's Batson challenges.

A

Batson held that the Equal Protection Clause of the Fourteenth Amendment prohibits prosecutors from exercising peremptory challenges on the basis of race. 476 U.S., at 89, 106 S.Ct. 1712. When adjudicating a Batson claim, trial courts follow a three-step process:

The opponent of the strike bears the burden of persuasion regarding racial motivation, Purkett v. Elem, 514 U.S. 765, 768, 115 S.Ct. 1769, 131 L.Ed.2d 834 (1995) (per curiam), and a trial court finding regarding the credibility of an attorney's explanation of the ground for a peremptory challenge is "entitled to `great deference,'" Felkner v. Jackson, 562 U.S. 594, 598, 131 S.Ct. 1305, 179 L.Ed.2d 374 (2011) (per curiam) (quoting Batson, 476 U.S., at 98, n. 21, 106 S.Ct. 1712). On direct appeal, those findings may be reversed only if the trial judge is shown to have committed clear error. Rice v. Collins, 546 U.S. 333, 338, 126 S.Ct. 969, 163 L.Ed.2d 824 (2006). Under AEDPA, even more must be shown. A federal habeas court must accept a state-court finding unless it was based on "an unreasonable determination of the facts in light of the evidence presented in the State court proceeding." § 2254(d)(2). "State-court factual findings, moreover, are presumed correct; the petitioner has the burden of rebutting the presumption by `clear and convincing evidence.'" Collins, supra, at 338-339, 126 S.Ct. 969 (quoting § 2254(e)(1)).

In this case, Ayala challenged seven of the prosecution's peremptory challenges. As explained above, the Ninth Circuit granted relief based on the dismissal of three potential jurors. The dissent discusses only one, Olanders D. We will devote most of our analysis to the three individuals discussed by the Ninth Circuit, but we hold that any error was harmless with respect to all seven strikes.

B

1

Ayala first contests the prosecution's decision to challenge Olanders D., an African-American man. The prosecution stated that its "primary" reason for striking Olanders D. was uncertainty about whether he could impose the death penalty, and the prosecutor noted that Olanders D. had written on his questionnaire that he did not "believe in the death penalty." 50 Reporter's Tr. on Appeal 6185 (hereinafter Tr.). Providing additional reasons for this strike, the prosecutor first stated that Olanders D.'s responses "did not make a lot of sense," "were not thought out," and "demonstrate[d] a lack of ability to express himself well." App. 283. The prosecutor also voiced doubt that Olanders D. "could actively participate in a meaningful way in deliberations with other jurors" and might have lacked the "ability to fit in with a cohesive group of 12 people." Ibid.

The trial court concluded that the strike was race-neutral. The judge stated: "Certainly with reference to whether or not he would get along with 12 people, it may well be that he would get along very well with 12 people. I think the other observations of counsel are accurate and borne out by the record." 50 Tr. 6186. The California Supreme Court found that the evidence of Olanders D.'s views on the death penalty provided adequate support for the trial judge's finding that the strike exercised against him was not based on race, and the court further found that defense counsel's presence would not have affected the outcome of the Batson hearing. The Ninth Circuit reversed, but its decision rested on a misapplication of the applicable harmless-error standards.

2

As the trial court and the State Supreme Court found, Olanders D.'s voir dire responses amply support the prosecution's concern that he might not have been willing to impose the death penalty. During voir dire, Olanders D. acknowledged that he wrote on his questionnaire, "`I don't believe in the death penalty,'" App. 179, and he agreed that he had at one time "thought that [the death penalty] was completely wrong," id., at 177. Although he stated during the voir dire that he had reconsidered his views, it was reasonable for the prosecution and the trial court to find that he did not clearly or adequately explain the reason or reasons for this change. When asked about this, Olanders D. gave a vague and rambling reply: "Well, I think it's—one thing would be the—the—I mean, examining it more closely, I think, and becoming more familiar with the laws and the—and the behavior, I mean, the change in the people, I think. All of those things contributed to the changes." Id., at 178.

The Ninth Circuit reversed because it speculated that defense counsel, if present when the prosecution explained the basis for this strike, "could have pointed to seated white jurors who had expressed similar or greater hesitancy" in imposing the death penalty. 756 F.3d, at 678. The Ninth Circuit wrote that a seated white juror named Ana L. was "indistinguishable from Olanders D. in this regard" and that she had "made almost precisely the same statement in her questionnaire." Ibid.

The responses of Olanders D. and Ana L., however, were by no means "indistinguishable." Olanders D. initially voiced unequivocal opposition to the death penalty, stating flatly: "I don't believe in the death penalty." He also revealed that he had once thought it was "completely wrong." Ana L., by contrast, wrote on the questionnaire that she "probably would not be able to vote for the death penalty," App. 109 (emphasis added), and she then later said at voir dire that she could vote for a verdict of death.

In a capital case, it is not surprising for prospective jurors to express varying degrees of hesitancy about voting for a death verdict. Few are likely to have experienced a need to make a comparable decision at any prior time in their lives. As a result, both the prosecution and the defense may be required to make fine judgment calls about which jurors are more or less willing to vote for the ultimate punishment. These judgment calls may involve a comparison of responses that differ in only nuanced respects, as well as a sensitive assessment of jurors' demeanor. We have previously recognized that peremptory challenges "are often the subjects of instinct," Miller-El v. Dretke, 545 U.S. 231, 252, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005) (citing Batson, 476 U.S., at 106, 106 S.Ct. 1712 (Marshall, J., concurring)), and that "race-neutral reasons for peremptory challenges often invoke a juror's demeanor," Snyder, 552 U.S., at 477, 128 S.Ct. 1203. A trial court is best situated to evaluate both the words and the demeanor of jurors who are peremptorily challenged, as well as the credibility of the prosecutor who exercised those strikes. As we have said, "these determinations of credibility and demeanor lie peculiarly within a trial judge's province," and "in the absence of exceptional circumstances, we [will] defer to the trial court." Ibid. (alterations and internal quotation marks omitted). "Appellate judges cannot on the basis of a cold record easily second-guess a trial judge's decision about likely motivation." Collins, 546 U.S., at 343, 126 S.Ct. 969 (BREYER, J., concurring).

The upshot is that even if "[r]easonable minds reviewing the record might disagree about the prosecutor's credibility, ... on habeas review that does not suffice to supersede the trial court's credibility determination." Id., at 341-342, 126 S.Ct. 969 (majority opinion). Here, any similarity between the responses of Olanders D. and Ana L. is insufficient to compel an inference of racial discrimination under Brecht or AEDPA.

Ayala contends that the presence of defense counsel might have made a difference because defense counsel might have been able to identify white jurors who were not stricken by the prosecution even though they had "expressed similar or greater hesitancy" about the death penalty. We see no basis for this argument. The questionnaires of all the jurors who sat and all the alternates are in the record, and Ana L., whom we just discussed, is apparently the white juror whose answers come the closest to those of Olanders D. Since neither Ayala nor the Ninth Circuit identified a white juror whose statements better support their argument, there is no reason to think that defense counsel could have pointed to a superior comparator at the ex parte proceeding.

3

In rejecting the argument that the prosecutor peremptorily challenged Olanders D. because of his race, the California Supreme Court appears to have interpreted the prosecutor's explanation of this strike to mean that Olanders D.'s views on the death penalty were alone sufficient to convince him to exercise a strike, see 24 Cal.4th, at 266, 99 Cal.Rptr.2d 532, 6 P.3d, at 206, and this was certainly an interpretation of the record that must be sustained under 28 U.S.C. § 2254(d)(2). As a result, it is not necessary for us to consider the prosecutor's supplementary reason for this strike—the poor quality of Olanders D.'s responses—but in any event, the Ninth Circuit's evaluation of this reason is also flawed.

The Ninth Circuit wrote that its independent "review of the voir dire transcript reveal[ed] nothing that supports the prosecution's claim: Olanders D.'s answers were responsive and complete." 756 F.3d, at 679. The record, however, provides sufficient support for the trial court's determination. Olanders D.'s incoherent explanation during voir dire of the reasons for his change of opinion about the death penalty was quoted above. He also provided a chronology of the evolution of his views on the subject that did not hold together. He stated that he had been "completely against the death sentence" 10 years earlier but seemed to suggest that his views had changed over the course of the intervening decade. See App. 176-177. However, on the questionnaire, which he had completed just a month before the voir dire, he wrote unequivocally: "I don't believe in the death penalty." Id., at 179. And then, at the time of the voir dire, he said that he would be willing to impose the death penalty in some cases. Id., at 180. He explained his answer on the questionnaire as follows: "I answered that kind of fast[.] [N]ormally, I wouldn't answer that question that way, but I mean, I really went through that kind of fast. I should have done better than that." Id., at 179-180. These answers during voir dire provide more than sufficient support for the prosecutor's observation, which the trial court implicitly credited, that Olanders D.'s statements "did not make a lot of sense," "were not thought out," and "demonstrate[d] a lack of ability to express himself well."

In ordering federal habeas relief based on their assessment of the responsiveness and completeness of Olanders D.'s answers, the members of the panel majority misunderstood the role of a federal court in a habeas case. The role of a federal habeas court is to "`guard against extreme malfunctions in the state criminal justice systems,'" Richter, 562 U.S., at 102-103, 131 S.Ct. 770 (quoting Jackson v. Virginia, 443 U.S. 307, 332, n. 5, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979) (Stevens, J., concurring in judgment)), not to apply de novo review of factual findings and to substitute its own opinions for the determination made on the scene by the trial judge.

C

Ayala next challenges the prosecution's use of a peremptory challenge to strike Gerardo O., a Hispanic man. The prosecution offered three reasons for this strike: Gerardo O. had a poor grasp of English; his answers during voir dire and on his questionnaire suggested that he might not be willing to impose the death penalty; and he did not appear to get along with the other prospective jurors. The trial judge accepted this explanation, as did the State Supreme Court.

The Ninth Circuit, however, rejected the state courts' determinations based on speculation that defense counsel, if present at the in camera hearing, "likely could have called into question all of the prosecution's stated reasons for striking Gerardo O." 756 F.3d, at 680. The Ninth Circuit thought that it could grant Ayala relief simply because it "[could not] say that Ayala would not have shown that the trial court would or should have determined that the prosecution's strike of Gerardo O. violated Batson." Id., at 682. But that is not the test. The inquiry under Brecht is not whether the federal habeas court could definitively say that there were no winning arguments that the defense could have made. Instead, the evidence in the record must raise "grave doubt[s]" about whether the trial judge would have ruled differently. O'Neal, 513 U.S., at 436, 115 S.Ct. 992. This requires much more than a "reasonable possibility" that the result of the hearing would have been different. Brecht, 507 U.S., at 637, 113 S.Ct. 1710 (internal quotation marks omitted). And on the record in this case, Ayala cannot establish actual prejudice or that no fairminded jurist could agree with the state court's application of Chapman.

We begin with the prosecution's explanation that it challenged Gerardo O. because of his limited English proficiency. During voir dire, Gerardo O. acknowledged that someone else had written the answers for him on his questionnaire "[b]ecause I couldn't—I cannot read—I cannot spell that well." App. 163. He added that he "didn't get" some of the words on the questionnaire. Ibid. Gerardo O.'s testimony also revealed that he might well have been unable to follow what was said at trial. When asked whether he could understand spoken English, he responded: "It depends if you make long words. If you make—if you go—if you say it straight out, then I might understand. If you beat around the bush, I won't." Id., at 166. At that point, defense counsel and Gerardo O. engaged in a colloquy that suggests that defense counsel recognized that he lacked the ability to understand words not used in basic everyday speech, "legal words," and rapid speech in English:

It is understandable for a prosecutor to strike a potential juror who might have difficulty understanding English.3 The jurors who were ultimately selected heard many days of testimony, and the instructions at both the guilt and the penalty phases included "legal words" and words not common in everyday speech. The prosecution had an obvious reason to worry that service on this jury would have strained Gerardo O.'s linguistic capability.

The Ninth Circuit reached a contrary conclusion by distorting the record and the applicable law. The Ninth Circuit first suggested that Gerardo O.'s English-language deficiencies were limited to reading and writing, 756 F.3d, at 680, but as the portions of the voir dire quoted above make clear, that was not true; the record shows that his ability to understand spoken English was also limited. The Ninth Circuit then suggested that "[t]he prosecution's purported reason for striking Gerardo O.... was directly related to his status as someone who spoke Spanish as his first language," ibid., but the prosecutor voiced no concern about Gerardo O.'s ability to speak Spanish or about the fact that Spanish was his first language. The prosecution's objection concerned Gerardo O.'s limited proficiency in English. The Ninth Circuit quoted the following statement from Hernandez v. New York, 500 U.S. 352, 363, 111 S.Ct. 1859, 114 L.Ed.2d 395 (1991) (plurality opinion): "`[T]he prosecutor's frank admission that his ground for excusing th[is] juror[] related to [his] ability to speak and understand Spanish raised a plausible, though not a necessary, inference that language might be a pretext for what in fact [was a] race-based peremptory challenge[].'" 756 F.3d, at 680 (alterations in original). This statement, however, did not concern a peremptory exercised due to a prospective juror's lack of English proficiency. Instead, it concerned the dismissal of Spanish-speaking members of the venire for fear that, if seated, they might not follow the English translation of testimony given in Spanish. See 500 U.S., at 360, 111 S.Ct. 1859. The Ninth Circuit's decision regarding Gerardo O. was thus based on a misreading of the record and a distortion of our case law. And neither Ayala nor the Ninth Circuit has identified anything that defense counsel might have done at the ex parte hearing to show that the prosecutor's concern about Gerardo O.'s limited English proficiency was pretextual.

The prosecution's second proffered reason for striking Gerardo O. was concern about his willingness to impose the death penalty, and as the trial court found, this observation was also supported by the record. Indeed, when asked in voir dire how he felt about imposing the death penalty, Gerardo O. responded that he was "[k]ind of shaky about it.... I'm not too sure if I can take someone else's life in my hands and say that; say, you know, `death,' or something." App. 168. In response to another question about his thoughts on the death penalty, he replied: "I don't know yet. It's kind of hard, you know, to pick it up like that and say how I feel about the death penalty." 15 Tr. 1052. Answering a question about whether his thoughts on the death penalty would affect how he viewed the evidence presented at trial, he responded, "I don't know, sir, to tell you the truth." App. 165. And when asked if he had "any feeling that [he] would be unable to vote for the death penalty if [he] thought it was a case that called for it," Gerardo O. responded once again, "I don't know." 15 Tr. 1043. While Gerardo O. did say at one point that he might be willing to impose the death penalty, he qualified that statement by adding that he would be comforted by the fact that "there's eleven more other persons on the jury." App. 170.

What we said above regarding jurors who express doubts about their openness to a death verdict applies as well here. The prosecution's reluctance to take a chance that Gerardo O. would ultimately be willing to consider the death penalty in accordance with state law did not compel the trial judge to find that the strike of Gerardo O. was based on race.

Nor is there a basis for finding that the absence of defense counsel affected the trial judge's evaluation of the sincerity of this proffered ground for the strike. Defense counsel had a full opportunity during voir dire to create a record regarding Gerardo O.'s openness to the death penalty. And defense counsel had the opportunity prior to the ex parte proceeding on the Gerardo O. strike to compare the minority jurors dismissed by the prosecution with white jurors who were seated. Counsel argued that the answers on the death penalty given by the minority jurors were "not significantly different from [those of] the white jurors that the prosecution ha[d] chosen to leave on the jury." Id., at 306. The trial judge asked counsel for "particulars," and counsel discussed Gerardo O., albeit briefly. Id., at 307-308. Thus, there is no reason to believe that counsel could have made a more persuasive argument at the ex parte proceeding than he made during this exchange.

The prosecution's final reason for striking Gerardo O. was that he appeared to be "a standoffish type of individual" whose "dress and ... mannerisms ... were not in keeping with the other jurors" and who "did not appear to be socializing or mixing with any of the other jurors." Id., at 298. The trial judge did not dispute that the prosecution's reflections were borne out by the record. The California Supreme Court affirmed and also emphasized that "the trial court's rulings in the ex parte hearing indisputably reflect both its familiarity with the record of voir dire of the challenged prospective jurors and its critical assessment of the prosecutor's proffered justifications." 24 Cal.4th, at 266-267, 99 Cal.Rptr.2d 532, 6 P.3d, at 206.

In light of the strength of the prosecution's first two reasons for striking Gerardo O., it is not at all clear that the prosecution proffered this final reason as an essential factor in its decision to strike, but in any event, there is no support for the suggestion that Ayala's attorney, if allowed to attend the ex parte hearing, would have been able to convince the judge that this reason was pretextual. The Ninth Circuit, however, was content to speculate about what might have been. Mixing guesswork with armchair sociology, the Ninth Circuit mused that "[i]t is likely that Gerardo O.'s dress and mannerisms were distinctly Hispanic. Perhaps in the late 1980's Hispanic males in San Diego County were more likely than members of other racial or ethnic groups in the area to wear a particular style or color of shirt, and Gerardo O. was wearing such a shirt." 756 F.3d, at 680-681. As for the prosecution's observation that Gerardo O. did not socialize with other jurors, the Ninth Circuit posited that, "perhaps, unbeknownst to the trial judge, Gerardo O. did `socializ[e] or mix[]' with a number of other jurors, and had even organized a dinner for some of them at his favorite Mexican restaurant." Id., at 681.

This is not how habeas review is supposed to work. The record provides no basis for the Ninth Circuit's flight of fancy. Brecht requires more than speculation about what extrarecord information defense counsel might have mentioned. And speculation of that type is not enough to show that a State Supreme Court's rejection of the argument regarding Gerardo O. was unreasonable.

D

The final prospective juror specifically discussed in the Ninth Circuit's decision was Robert M., who is Hispanic. The prosecution's primary proffered reason for striking Robert M. was concern that he would not impose the death penalty, though the prosecution added that it was troubled that he had followed the Sagon Penn case, a high-profile prosecution in San Diego in which an alleged murderer was acquitted amid allegations of misconduct by police and prosecutors. In addition, the prosecution also explained to the trial court that Robert M. scored poorly on its 10-point scale for evaluating prospective jurors. The trial court accepted the prosecutor's explanation of the strike.

With respect to the prosecution's concern that Robert M. might not be willing to impose the death penalty, the Ninth Circuit found that defense counsel, if permitted to attend the in camera proceeding, could have compared Robert M.'s statements about the death penalty to those of other jurors and could have reminded the judge that Robert M. had "repeatedly stated during voir dire that he believed in the death penalty and could personally vote to impose it." 756 F.3d, at 682. But as with Olanders D. and Gerardo O., we cannot say that the prosecution had no basis for doubting Robert M.'s willingness to impose the death penalty. For example, when asked at one point whether he could vote for death, Robert M. responded: "Well, I've though[t] about that, but it's a difficult question, and yeah, it is difficult for me to say, you know, one way or the other. I believe in it, but for me to be involved in it is—is hard. It's hard to accept that aspect of it, do you know what I mean?" App. 149-150. In response to another question, he said: "It would be hard, but I think I could, yes. It's—it's hard to say, you know—and I don't care who the person is—to say that they have to put somebody away, you know. It's very hard." Id., at 154. These are hardly answers that would inspire confidence in the minds of prosecutors in a capital case.

While the Ninth Circuit argued that defense counsel's absence at the in camera hearing prejudiced the trial judge's ability to assess this reason for the strike of Robert M., the Ninth Circuit failed to mention that defense counsel specifically addressed this issue during voir dire. At that time, he pointedly reminded the judge that Robert M. had made several statements during voir dire that were favorable to the death penalty. Id., at 307. The trial judge thus heard defense counsel's arguments but nevertheless concluded that the record supplied a basis for a legitimate concern about whether Robert M. could impose the death penalty. That Ayala's attorney did not have the opportunity to repeat this same argument once more at the in camera proceeding does not create grave doubt about whether the trial court would have decided the issue differently.

As for the prosecution's second proffered reason for striking Robert M.—that he had followed the Sagon Penn case4—the Ninth Circuit placed great emphasis on the fact that a seated white juror had followed a different murder trial, that of Robert Alton Harris.5 But the Penn and Harris cases were quite different. Harris was convicted while Penn was acquitted; and since the Harris case was much older, the experience of following it was less likely to have an effect at the time of the trial in this case.

E

Ayala raised a Batson objection about the prosecution's use of peremptory challenges on four additional jurors, George S., Barbara S., Galileo S., and Luis M. The Ninth Circuit did not address these prospective jurors at length, and we need not dwell long on them. With respect to all four of these prospective jurors, we conclude that any constitutional error was harmless.

Of these four additional jurors, Ayala's brief in this Court develops an argument with respect to only two, George S. and Barbara S. And while Ayala's attorney claimed that George S. was Hispanic, the prosecutor said that he thought that George S. was Greek. In any event, the prosecution offered several reasons for striking George S. The prosecutor noted that one of his responses "was essentially, `you probably don't want me to be a juror on this case.'" Id., at 312. The prosecutor was also concerned about whether he would vote for death or even a life sentence and whether he would follow the law as opposed to his personal religious beliefs. In addition, the prosecutor noted that George S. had previously been the sole holdout on a jury and that his prior application to be a police officer had been rejected, for reasons that were not clear. The trial court accepted these explanations.

Ayala contests only two of these justifications. He quibbles that George S. had not been a "`holdout,'" but instead had been the dissenting juror in a civil case on which unanimity was not required. This observation does not render the prosecution's proffered justification "false or pretextual." Brief for Respondent 46. The fact that George S. had been willing to dissent from a jury verdict could reasonably give a prosecutor pause in a capital case since a single holdout juror could prevent a guilty verdict or death sentence. The most that Ayala can establish is that reasonable minds can disagree about whether the prosecution's fears were well founded, but this does not come close to establishing "actual prejudice" under Brecht. Nor does it meet the AEDPA standard. Ayala also points out that a seated white juror, Charles C., had been rejected by a police force, but George S. admitted that he had applied to law enforcement because he was "trying to get out of the Army," App. 222, and the reasons for his rejection were not clear. Charles C., by contrast, had received a qualifying score on a law enforcement exam but was not hired because a position was not available.

As for Barbara S., the prosecution struck her because, during voir dire, she appeared to be "under the influence of drugs" and disconnected from the proceedings. Id., at 314. The prosecution emphasized that she had "an empty look in her eyes, slow responses, a lack of really being totally in tune with what was going on." Ibid. It added that she appeared "somewhat angry," "manifest[ed] a great deal of nervousness," and seemed like someone who would be unlikely to closely follow the trial. Ibid. The trial judge thought that Barbara S. appeared nervous rather than hostile, but he agreed that she gave incomplete answers that were sometimes "non sequiturs." Id., at 315. He concluded, "I certainly cannot quarrel ... with your subjective impression, and the use of your peremptory challenge based upon her individual manifestation, as opposed to her ethnicity." Ibid. Ayala points to the trial court's disagreement with the prosecutor's impression that Barbara S. was hostile, but this ruling illustrates the trial judge's recollection of the demeanor of the prospective jurors and his careful evaluation of each of the prosecutor's proffered reasons for strikes. And the fact that the trial judge's impression of Barbara S.'s demeanor was somewhat different from the prosecutor's hardly shows that the prosecutor's reasons were pretextual. It is not at all unusual for individuals to come to different conclusions in attempting to read another person's attitude or mood.

IV

The pattern of peremptory challenges in this case was sufficient to raise suspicions about the prosecution's motives and to call for the prosecution to explain its strikes. As we have held, the Fourteenth Amendment prohibits a prosecutor from striking potential jurors based on race. Discrimination in the jury selection process undermines our criminal justice system and poisons public confidence in the evenhanded administration of justice.

In Batson, this Court adopted a procedure for ferreting out discrimination in the exercise of peremptory challenges, and this procedure places great responsibility in the hands of the trial judge, who is in the best position to determine whether a peremptory challenge is based on an impermissible factor. This is a difficult determination because of the nature of peremptory challenges: They are often based on subtle impressions and intangible factors. In this case, the conscientious trial judge determined that the strikes at issue were not based on race, and his judgment was entitled to great weight. On appeal, five justices of the California Supreme Court carefully evaluated the record and found no basis to reverse. A Federal District Judge denied federal habeas relief, but a divided panel of the Ninth Circuit reversed the District Court and found that the California Supreme Court had rendered a decision with which no fairminded jurist could agree.

For the reasons explained above, it was the Ninth Circuit that erred. The exclusion of Ayala's attorney from part of the Batson hearing was harmless error. There is no basis for finding that Ayala suffered actual prejudice, and the decision of the California Supreme Court represented an entirely reasonable application of controlling precedent.

* * *

The judgment of the Court of Appeals for the Ninth Circuit is reversed, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

Justice KENNEDY, concurring.

My join in the Court's opinion is unqualified; for, in my view, it is complete and correct in all respects. This separate writing responds only to one factual circumstance, mentioned at oral argument but with no direct bearing on the precise legal questions presented by this case.

In response to a question, respondent's counsel advised the Court that, since being sentenced to death in 1989, Ayala has served the great majority of his more than 25 years in custody in "administrative segregation" or, as it is better known, solitary confinement. Tr. of Oral Arg. 43-44. Counsel for petitioner did not have a clear opportunity to enter the discussion, and the precise details of respondent's conditions of confinement are not established in the record. Yet if his solitary confinement follows the usual pattern, it is likely respondent has been held for all or most of the past 20 years or more in a windowless cell no larger than a typical parking spot for 23 hours a day; and in the one hour when he leaves it, he likely is allowed little or no opportunity for conversation or interaction with anyone. Ibid.; see also Wilkinson v. Austin, 545 U.S. 209, 218, 125 S.Ct. 2384, 162 L.Ed.2d 174 (2005); Amnesty International, Entombed: Isolation in the U.S. Federal Prison System (2014). It is estimated that 25,000 inmates in the United States are currently serving their sentence in whole or substantial part in solitary confinement, many regardless of their conduct in prison. Ibid.

The human toll wrought by extended terms of isolation long has been understood, and questioned, by writers and commentators. Eighteenth-century British prison reformer John Howard wrote "that criminals who had affected an air of boldness during their trial, and appeared quite unconcerned at the pronouncing sentence upon them, were struck with horror, and shed tears when brought to these darksome solitary abodes." The State of the Prisons in England and Wales 152 (1777). In literature, Charles Dickens recounted the toil of Dr. Manette, whose 18 years of isolation in One Hundred and Five, North Tower, caused him, even years after his release, to lapse in and out of a mindless state with almost no awareness or appreciation for time or his surroundings. A Tale of Two Cities (1859). And even Manette, while imprisoned, had a work bench and tools to make shoes, a type of diversion no doubt denied many of today's inmates.

One hundred and twenty-five years ago, this Court recognized that, even for prisoners sentenced to death, solitary confinement bears "a further terror and peculiar mark of infamy." In re Medley, 134 U.S. 160, 170, 10 S.Ct. 384, 33 S.Ct. 835 (1890); see also id., at 168, 10 S.Ct. 384 ("A considerable number of the prisoners fell, after even a short [solitary] confinement, into a semi-fatuous condition ... and others became violently insane; others, still, committed suicide"). The past centuries' experience and consideration of this issue is discussed at length in texts such as The Oxford History of the Prison: The Practice of Punishment in Western Society (1995), a joint disciplinary work edited by law professor Norval Morris and professor of medicine and psychiatry David Rothman that discusses the deprivations attendant to solitary confinement. Id., at 184, 10 S.Ct. 384.

Yet despite scholarly discussion and some commentary from other sources, the condition in which prisoners are kept simply has not been a matter of sufficient public inquiry or interest. To be sure, cases on prison procedures and conditions do reach the courts. See, e.g., Brown v. Plata, 563 U.S. ___, 131 S.Ct. 1910, 179 L.Ed.2d 969 (2011); Hutto v. Finney, 437 U.S. 678, 685, 98 S.Ct. 2565, 57 L.Ed.2d 522 (1978) ("Confinement in a prison or in an isolation cell is a form of punishment subject to scrutiny under the Eighth Amendment"); Weems v. United States, 217 U.S. 349, 365-367, 30 S.Ct. 544, 54 S.Ct. 793 (1910). Sentencing judges, moreover, devote considerable time and thought to their task. There is no accepted mechanism, however, for them to take into account, when sentencing a defendant, whether the time in prison will or should be served in solitary. So in many cases, it is as if a judge had no choice but to say: "In imposing this capital sentence, the court is well aware that during the many years you will serve in prison before your execution, the penal system has a solitary confinement regime that will bring you to the edge of madness, perhaps to madness itself." Even if the law were to condone or permit this added punishment, so stark an outcome ought not to be the result of society's simple unawareness or indifference.

Too often, discussion in the legal academy and among practitioners and policymakers concentrates simply on the adjudication of guilt or innocence. Too easily ignored is the question of what comes next. Prisoners are shut away—out of sight, out of mind. It seems fair to suggest that, in decades past, the public may have assumed lawyers and judges were engaged in a careful assessment of correctional policies, while most lawyers and judges assumed these matters were for the policymakers and correctional experts.

There are indications of a new and growing awareness in the broader public of the subject of corrections and of solitary confinement in particular. See, e.g., Gonnerman, Before the Law, The New Yorker, Oct. 6, 2014, p. 26 (detailing multiyear solitary confinement of Kalief Browder, who was held—but never tried—for stealing a backpack); Schwirtz & Winerip, Man, Held at Rikers for 3 Years Without Trial, Kills Himself, N.Y. Times, June 9, 2015, p. A18. And penology and psychology experts, including scholars in the legal academy, continue to offer essential information and analysis. See, e.g., Simon & Sparks, Punishment and Society: The Emergence of an Academic Field, in The SAGE Handbook of Punishment and Society (2013); see also Venters et al., Solitary Confinement and Risk of Self-Harm Among Jail Inmates, 104 Am. J. Pub. Health 442 (March 2014); Metzner & Fellner, Solitary Confinement and Mental Illness in U.S. Prisons: A Challenge for Medical Ethics, 38 J. Am. Academy Psychiatry and Law 104-108 (2010).

These are but a few examples of the expert scholarship that, along with continued attention from the legal community, no doubt will aid in the consideration of the many issues solitary confinement presents. And consideration of these issues is needed. Of course, prison officials must have discretion to decide that in some instances temporary, solitary confinement is a useful or necessary means to impose discipline and to protect prison employees and other inmates. But research still confirms what this Court suggested over a century ago: Years on end of near-total isolation exact a terrible price. See, e.g., Grassian, Psychiatric Effects of Solitary Confinement, 22 Wash. U.J.L. & Pol'y 325 (2006) (common side-effects of solitary confinement include anxiety, panic, withdrawal, hallucinations, self-mutilation, and suicidal thoughts and behaviors). In a case that presented the issue, the judiciary may be required, within its proper jurisdiction and authority, to determine whether workable alternative systems for long-term confinement exist, and, if so, whether a correctional system should be required to adopt them.

Over 150 years ago, Dostoyevsky wrote, "The degree of civilization in a society can be judged by entering its prisons." The Yale Book of Quotations 210 (F. Shapiro ed. 2006). There is truth to this in our own time.

Justice THOMAS, concurring.

I join the Court's opinion explaining why Ayala is not entitled to a writ of habeas corpus from this or any other federal court. I write separately only to point out, in response to the separate opinion of Justice KENNEDY, that the accommodations in which Ayala is housed are a far sight more spacious than those in which his victims, Ernesto Dominguez Mendez, Marcos Antonio Zamora, and Jose Luis Rositas, now rest. And, given that his victims were all 31 years of age or under, Ayala will soon have had as much or more time to enjoy those accommodations as his victims had time to enjoy this Earth.

Justice SOTOMAYOR, with whom Justice GINSBURG, Justice BREYER, and Justice KAGAN join, dissenting.

At Hector Ayala's trial, the prosecution exercised its peremptory strikes to dismiss all seven of the potential black and Hispanic jurors. In his federal habeas petition, Ayala challenged the state trial court's failure to permit his attorneys to participate in hearings regarding the legitimacy of the prosecution's alleged race-neutral reasons for its strikes. See Batson Batson v. Kentucky, 476 U.S. 79, 97-98, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986). The Court assumes that defense counsel's exclusion from these proceedings violated Ayala's constitutional rights, but concludes that the Ninth Circuit erred in granting habeas relief because there is insufficient reason to believe that counsel could have convinced the trial court to reject the prosecution's proffered reasons. I respectfully dissent. Given the strength of Ayala's prima facie case and the comparative juror analysis his attorneys could have developed if given the opportunity to do so, little doubt exists that counsel's exclusion from Ayala's Batson hearings substantially influenced the outcome.

I

My disagreement with the Court does not stem from its discussion of the applicable standard of review, which simply restates the holding of Fry v. Pliler, 551 U.S. 112, 127 S.Ct. 2321, 168 L.Ed.2d 16 (2007). Fry rejected the argument that the Antiterrorism and Effective Death Penalty Act of 1996, 28 U.S.C. § 2254, compels federal courts to apply any standard other than that set forth in Brecht v. Abrahamson, 507 U.S. 619, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993), when assessing the harmlessness of a constitutional error on habeas review. 551 U.S., at 120, 127 S.Ct. 2321. Brecht, in turn, held that the harmlessness standard federal courts must apply in collateral proceedings is more difficult to meet than the "`beyond a reasonable doubt'" standard applicable on direct review. 507 U.S., at 622-623, 113 S.Ct. 1710 (quoting Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967)). More specifically, under Brecht, a federal court can grant habeas relief only when it concludes that a constitutional error had a "`substantial and injurious effect or influence'" on either a jury verdict or a trial court decision. 507 U.S., at 623, 113 S.Ct. 1710. Later, O'Neal v. McAninch, 513 U.S. 432, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995), clarified that this standard is satisfied when a reviewing judge "is in grave doubt about whether" the error is harmless; that is, when "the matter is so evenly balanced that [a judge] feels himself in virtual equipoise as to the harmlessness of the error." Id., at 435, 115 S.Ct. 992 (emphasis deleted). See also ante, at 2197 - 2198 (quoting O'Neal, 513 U.S., at 436, 115 S.Ct. 992). Put differently, when a federal court is in equipoise as to whether an error was actually prejudicial, it must "treat the error, not as if it were harmless, but as if it affected the verdict (i.e., as if it had a `substantial and injurious effect or influence in determining the jury's verdict')." O'Neal, 513 U.S., at 435, 115 S.Ct. 992.

In addition to confirming the Brecht standard's continued vitality, Fry established its exclusivity. Fry expressly held that federal habeas courts need not first assess whether a state court unreasonably applied Chapman before deciding whether that error was prejudicial under Brecht. Such a requirement would "mak[e] no sense . . . when the latter [standard] obviously subsumes the former." Fry, 551 U.S., at 120, 127 S.Ct. 2321. Nothing in the Court's opinion today calls into question this aspect of Fry's holding. If a trial error is prejudicial under Brecht's standard, a state court's determination that the error was harmless beyond a reasonable doubt is necessarily unreasonable. See ante, at 2198 - 2199.

II

A

To apply Brecht to the facts of this case, it is essential to understand the contours of Ayala's underlying constitutional claim or—perhaps more importantly—to appreciate what his claim is not. Trial judges assess criminal defendants' challenges to prosecutors' use of peremptory strikes using the three-part procedure first announced in Batson. After a defendant makes a "prima facie showing that a peremptory challenge [was] . . . exercised on the basis of race," the prosecution is given an opportunity to "offer a race-neutral basis for striking the juror in question," Miller-El v. Cockrell, 537 U.S. 322, 328, 123 S.Ct. 1029, 154 L.Ed.2d 931 (2003). The court then "decid[es] whether it was more likely than not that the challenge was improperly motivated." Johnson v. California, 545 U.S. 162, 169, 170, 125 S.Ct. 2410, 162 L.Ed.2d 129 (2005). This determination is a factual one, which—as the Court correctly notes—reviewing courts must accord "`great deference.'" See ante, at 2199 - 2200 (quoting Felkner v. Jackson, 562 U.S. 594, 598, 131 S.Ct. 1305, 179 L.Ed.2d 374 (2011) (per curiam)).

Here, Ayala does not claim that the trial court wrongly rejected his Batson challenges based on the record before it. Rather, Ayala's claim centers on the exclusion of his attorneys from the Batson hearings. Ayala contends that there is at least a grave doubt as to whether the trial or appellate court's consideration of his Batson challenges was substantially influenced by the trial court's erroneous refusal to permit his attorneys to appear at the hearings at which those challenges were adjudicated. Ayala's conviction must be vacated if there is grave doubt as to whether even just one of his Batson challenges would have been sustained had the defense been present. Snyder v. Louisiana, 552 U.S. 472, 478, 128 S.Ct. 1203, 170 L.Ed.2d 175 (2008) (reversing a conviction after concluding that use of one peremptory strike was racially motivated).

B

The Court's Brecht application begins and ends with a discussion of particular arguments the Ninth Circuit posited Ayala's lawyers could have raised had they been present at his Batson hearings. This approach fails to account for the basic background principle that must inform the application of Brecht to Ayala's procedural Batson claim: the "[c]ommon sense" insight "that secret decisions based on only one side of the story will prove inaccurate more often than those made after hearing from both sides." Kaley v. United States, 571 U.S. ___, ___, 134 S.Ct. 1090, 1113, 188 L.Ed.2d 46 (2014) (ROBERTS, C.J., dissenting). Our entire criminal justice system was founded on the premise that "[t]ruth . . . is best discovered by powerful statements on both sides of the question." United States v. Cronic, 466 U.S. 648, 655, 104 S.Ct. 2039, 80 L.Ed.2d 657 (1984) (internal quotation marks omitted). There is no reason to believe that Batson hearings are the rare exception to this rule. Instead, defense counsel could have played at least two critical roles had they been present at Ayala's Batson hearings.

First, Ayala's attorneys would have been able to call into question the credibility of the prosecution's asserted race-neutral justifications for the use of its peremptory strikes. Of course, a trial court may identify some pretextual reasons on its own, but Snyder held that when assessing a claimed Batson error, "all of the circumstances that bear upon the issue of racial animosity must be consulted." Snyder, 552 U.S., at 478, 128 S.Ct. 1203. Absent an adversarial presentation, a diligent judge may overlook relevant facts or legal arguments in even a straightforward case. There is also great probative force to a "comparative juror analysis"—an analysis of whether the prosecution's reasons for using its peremptory strikes against non-white jurors apply equally to white jurors whom it would have allowed to serve. Miller-El v. Dretke, 545 U.S. 231, 241, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005). See also Snyder, 552 U.S., at 483, 128 S.Ct. 1203 (emphasizing importance of conducting a comparative juror analysis in the trial court). Trial courts are ill suited to perform this intensive inquiry without defense counsel's assistance.

The risk that important arguments will not be considered rises close to a certainty in a capital case like Ayala's, where jury selection spanned more than three months, involved more than 200 prospective jurors, and generated a record that is massive by any standard. See Ayala v. Wong, 756 F.3d 656, 660, 676 (C.A.9 2014) (case below). It strains credulity to suggest that a court confronted with this mountain of information necessarily considered all of the facts that would have informed its credibility determination without the presence of defense counsel to help bring them to its attention.

Second, not only did the exclusion of defense counsel from Ayala's Batson hearings prevent him from making his strongest arguments before the person best situated to assess their merit, it also impeded his ability to raise these claims on appeal. Because Ayala's lawyers were not afforded any opportunity to respond to the prosecution's race-neutral reasons, we are left to speculate as to whether the trial court actually considered any of the points the defense would have made before it accepted the prosecution's proffered explanations. Moreover, even if we could divine which of the possible considerations the trial judge took into account, our review would still be unduly constrained by a record that lacks whatever material facts the defense would have preserved had it been on notice of the assertions that it needed to challenge. Perhaps some of these facts, such as the jurors' appearance and demeanor, were known to the trial judge, but appellate courts "can only serve [their] function when the record is clear as to the relevant facts" or when they can rely on "defense counsel['s] fail[ure] to point out any such facts after learning of the prosecutor's reasons." United States v. Thompson, 827 F.2d 1254, 1261 (C.A.9 1987). Neither of these conditions is met here.

For the reasons described above, the fact that counsel was wrongfully excluded from Ayala's Batson hearings on its own raises doubt as to whether the outcome of these proceedings—or the appellate courts' review of them—would have been the same had counsel been present.1 This doubt is exacerbated by the loss of the vast majority of the questionnaires that jurors completed at the start of voir dire, including those filled out by the seven black and Hispanic jurors against whom the prosecution exercised its peremptory strikes. The prosecution cited these questionnaires in support of its alleged race-neutral reasons at the ex parte Batson hearings. See e.g., App. 283, 298, 312, 314, 316. Without the underlying documents, however, it is impossible to assess whether the prosecution's characterizations of those prospective jurors' responses were fair and accurate. The loss of the questionnaires has also precluded every court that has reviewed this case from performing a comprehensive comparative juror analysis. The Court today analyzes how the prosecution's statements at the ex parte Batson hearings regarding the black and Hispanic jurors' questionnaires stack up against the actual questionnaires completed by the white seated jurors and alternates. But there is no way to discern how these representations compare with the answers that were given by white jurors whom the prosecution would have permitted to serve but whom the defense ultimately struck. See Miller-El v. Dretke, 545 U.S., at 244-245, 125 S.Ct. 2317 (comparing a juror struck by the prosecution with a juror challenged only by the defense).

C

The above-described consequences of the trial court's procedural error and the fact that the prosecution struck every potential black or Hispanic juror go a long way toward establishing the degree of uncertainty that Brecht requires. Keeping these considerations in mind, the next step is to assess the arguments that Ayala's attorneys may have raised had they been allowed to participate at his Batson hearings. As explained above, Ayala is entitled to habeas relief if a reviewing judge is in "equipoise" as to whether his lawyers' exclusion from the Batson hearings had an "injurious effect" on the trial court's failure to find by a preponderance of the evidence that any of the prosecution's peremptory strikes was racially motivated. With the inquiry so framed, it is easy to see that the Ninth Circuit correctly found that Ayala was actually prejudiced by the trial court's constitutional error. In particular, there is a substantial likelihood that if defense counsel had been present, Ayala could at least have convinced the trial court that the race-neutral reasons the prosecution put forward for dismissing a black juror, Olanders D., were pretextual.2

The prosecution offered three justifications for striking Olanders D.: (1) he might be unable to vote for the death penalty because he had written in his questionnaire that "he does not believe [in] it" and had failed to fully explain a subsequent change in position; (2) his questionnaire answers were poor; and, (3) he might lack the "ability to fit in with a cohesive group of 12 people." App. 283. The trial court rejected the third of these reasons outright, noting that "it may well. . . be that he would get along very well with 12 people." Id., at 283-284. I have grave misgivings as to whether the trial judge would have found it more likely than not that the first two purported bases were legitimate had defense counsel been given an opportunity to respond to them.

Ayala's attorneys could have challenged the prosecution's claim that Olanders D. would hesitate to impose the death penalty by pointing to a seated juror—Ana L.—who made remarkably similar statements concerning capital punishment. Based on his remarks during voir dire, it appears that Olanders D. suggested on his questionnaire that he was or had been opposed to the death penalty.3 Id., at 176, 179. Ana L.'s questionnaire contained numerous comparable statements. When asked to express her "feelings about the death penalty," she wrote: "I don't believe in taking a life." Id., at 108. And, in response to a question regarding whether she "would like to serve as a juror and why?", Ana L. said: "no—If I am selected as a Juror and all Jurors voted for the death penalty I probably would not be able to vote for the death penalty." Id., at 109. Finally, on her questionnaire, Ana L. indicated that she believes the death penalty is imposed "[t]oo often" and that she strongly disagrees with the "adage, `An eye for any eye,'" which she understood to mean, "[a] criminal took a life[,] now [it] is our turn to take his." Id., at 108-109.

A direct comparison of Olanders D.'s and Ana L.'s voir dire answers is equally telling. During voir dire, Olanders D. clarified that he had not intended his questionnaire to reflect that he was categorically opposed to the death penalty, but only that his views on the topic had evolved over the prior decade and that he had come to believe that the death penalty "would be an appropriate sentence under certain circumstances." Id., at 176. To account for this change in his position, Olanders D. cited a number of considerations, including a new understanding of what his religion required, ibid., "more familiar[ity] with the laws," id., at 178, increased violence in our society, ibid., and conversations with his immediate family, id., at 180. Ana L., by contrast, stated at voir dire that she "strongly . . . did not believe in the death penalty" up until she "[f]illed out the questionnaire." Id., at 193. And, only after repeated attempts by both the defense and the prosecution to get her to pinpoint what caused this sudden about face, Ana L. said that she had "listen[ed] to the Bundy evidence that was said and his being put to death, and I started to think; and I said if they were guilty maybe there is a death sentence for these people." Id., at 202.4

Based on this record, it requires little speculation to see that defense counsel could have made a powerful argument that Ana L. was equally or even less likely to impose the death penalty than Olanders D. While both jurors had opposed the death penalty at some point in the past, Olanders D. stated that he had come to believe in capital punishment after a period of sustained deliberation. Ana L., however, purported to change her view due only to one recent execution and the fact that she had been called to serve as a juror on a capital case. Moreover, there is no basis to think that the trial court accounted for the similarities between Ana L. and Olanders D. Approximately two months passed between Olanders D.'s and Ana L.'s voir dire hearings and the date on which the prosecution exercised its peremptory strike against Olanders D. Without the benefit of defense counsel to help jog his recollection, it is absurd to proceed as if the trial judge actually considered one of more than 200 prospective jurors' statements concerning the death penalty when ruling on Ayala's Batson motion. Taken together, it seems highly likely that these arguments—had they been raised—would have convinced the trial judge that the prosecution's first alleged reason for striking Olanders D. was pretextual.

As for the prosecution's second purported justification—that his questionnaire responses "were poor," id., at 283—it is impossible to know what winning arguments the defense could have raised because the questionnaire itself is missing from the record.5 Indeed, for all that is known, counsel may have had a compelling argument that Olanders D.'s answers were cogent and complete. Even if some of them were lacking, however, counsel could still have drawn the trial judge's attention to weak questionnaires completed by several of the seated jurors. For instance, if the prosecution's claim was that Olanders D.'s questionnaire answers were conclusory, Ayala's counsel could have referred the Court to seated juror Charles G.'s questionnaire. In response to a prompt asking prospective jurors to explain why they would or would not like to be empaneled in Ayala's case, Charles G. wrote only "No." Id., at 71. Alternatively, if the prosecution's concern was that Olanders D.'s answer to a particular question demonstrated an inability to clearly express himself, the defense could have directed the court's attention to the questionnaire completed by seated juror Thomas B. When asked to share his "impressions or feelings . . . about gangs based on what [he had] read or s[een]," Thomas B. stated: "I feel the only media coverage they get is bad, however, those whom do constructive events usually seek out positive media coverage." Id., at 30. Finally, it bears noting that if Ayala's lawyers had been able to respond at the Batson hearing, they would have had the questionnaires of many more comparable jurors at their disposal. It is entirely possible that some of the questionnaires completed by prospective jurors who were accepted by the prosecution but dismissed by the defense were weaker than those completed by Charles G. and Thomas B.

In short, it is probable that had Ayala's lawyers been present at the Batson hearing on Olanders D., his strong Batson claim would have turned out to be a winning one. The trial judge rejected one of the reasons advanced by the prosecution on its own and the defense had numerous persuasive arguments that it could have leveled against the remaining two justifications had it been given the opportunity to do so.

III

The Court concludes that Ayala is not entitled to habeas relief because it finds that there is little or no reason to doubt that the trial judge would have accepted both of the above-discussed reasons for striking Olanders D. even if counsel participated at Ayala's Batson hearings. The Court's analysis, however, misunderstands the record and mistakes Ayala's procedural Batson claim for a direct challenge to a trial court's denial of a Batson motion.

In defense of the prosecution's first basis for striking Olanders D.—that he was uncomfortable with the death penalty—the Court begins by asserting that Ana L. was insufficiently similar to Olanders D. to have cast any doubt on the prosecution's position. Olanders D., the Court maintains, "initially voiced unequivocal opposition to the death penalty," whereas Ana L. "wrote on [her] questionnaire that she `probably would not be able to vote for the death penalty.'" Ante, at 2200 - 2201 (emphasis in original). But the Court has plucked one arguably ambiguous statement from Ana L.'s questionnaire while ignoring others (described above) suggesting that she fundamentally opposed capital punishment. More importantly, the Court is not comparing apples with apples. Because Olanders D.'s questionnaire has been lost, there is no way to know the extent to which the views he expressed there were "unequivocal." Consequently, in support of its contention that Olanders D. originally wrote that he was categorically opposed to the death penalty, the Court relies on his response to a question posed by the prosecution during voir dire. To be sure, when asked whether he had stated that he did not "believe in the death penalty" on his questionnaire, Olanders D. responded: "That's correct." App. 179. During voir dire, however, Ana L. described the position she had taken in her questionnaire in identical terms, stating: "I remember saying [on my questionnaire] that I didn't believe in the death penalty." Id., at 201.

Given the difficulty of differentiating between Ana L.'s and Olanders D.'s views toward the death penalty based on the record before us, the Court understandably does not press this factual point further. Instead, it commits a legal error by contending that the trial court's determination is entitled to deference because the judge—unlike this Court—had the benefit of observing both Olanders D.'s and the prosecution's demeanor. Ante, at 2201. Deference may be warranted when reviewing a substantive Batson claim. By suggesting that a trial judge can make a sound credibility determination without the benefit of an adversarial proceeding, however, the Court ignores the procedural nature of the constitutional error whose existence it purports to assume. Courts defer to credibility findings not only because of trial judges' proximity to courtroom events, but also because of the expectations regarding the procedures used in the proceedings that they oversee. A decision to credit a prosecution's race-neutral basis for striking a juror is entitled to great weight if that reason has "survive[d] the crucible of meaningful adversarial testing." Cronic, 466 U.S., at 656, 104 S.Ct. 2039. It warrants substantially less—if any—deference where, as here, it is made in the absence of the "fundamental instrument for judicial judgment: an adversary proceeding in which both parties may participate." Carroll v. President and Comm'rs of Princess Anne, 393 U.S. 175, 183, 89 S.Ct. 347, 21 L.Ed.2d 325 (1968); see also Kaley, 571 U.S., at ___, 134 S.Ct., at 1113 (ROBERTS, C.J., dissenting) ("It takes little imagination to see that . . . ex parte proceedings create a heightened risk of error").6

The Court's analysis of the second reason put forward for striking Olanders D.—that his questionnaire was faulty—fares no better. As a preliminary matter, perhaps because Olanders D.'s questionnaire has been lost, the Court characterizes the prosecution's second proffered reason for dismissing Olanders D. as an objection to all of his "responses" as opposed to simply the responses on his questionnaire. Ante, at 2200. But even if the prosecution had relied on the rationale that the Court now substitutes, there is a real likelihood that the defense would still have been able to undermine its credibility.

The Court asserts that Olanders D.'s "responses" were misleading because he had "unequivocally" stated that he did not believe in the death penalty on his questionnaire, but at voir dire he said that his views on capital punishment had changed over the previous 10 years. Ante, at 2202. The Court's argument thus hinges on the premise that Olanders D.'s questionnaire clearly stated that he was opposed to the death penalty. At least one person, however, did not construe Olanders D.'s questionnaire to express such a categorical view: defense counsel. During voir dire, one of Ayala's lawyers remarked that she thought Olanders D.'s questionnaire "indicated that [he] had had some change in [his] feelings about the death penalty." App. 176. "[M]y understanding," she said, "is that at one time [he] felt one way, and—and then at some point [he] felt differently." Ibid. Thus, if (as the Court now hypothesizes) the trial court was inclined to accept the prosecution's second reason for striking Olanders D. based on apparent tension between his questionnaire and his statements during voir dire (a proposition that is itself uncertain), the defense may have been able to argue persuasively that any claimed inconsistency was illusory.

Batson recognized that it is fundamentally unfair to permit racial considerations to drive the use of peremptory challenges against jurors. When the prosecution strikes every potential black and Hispanic juror, a reviewing court has a responsibility to ensure that the trial court's denial of the defendant's Batson motion was not influenced by constitutional error. But there is neither a factual nor a legal basis for the Court's confidence that the prosecution's race-neutral reasons for striking Olanders D. were unassailable. Because the Court overlooks that Ayala raised a procedural Batson claim, it scours the record for possible support for the trial court's credibility determination without accounting for the flaws in the process that led to it. The proper inquiry is not whether the trial court's determination can be sustained, but whether it may have been different had counsel been present. Given the strength of Ayala's prima facie case and the arguments his counsel would have been able to make based even on the limited existing record, grave doubts exist as to whether counsel's exclusion from Ayala's Batson hearings was harmless. Accordingly, I respectfully dissent.