Opinion by Judge KOZINSKI.

Burks and Celestine were convicted of murdering a fellow prison inmate and sentenced to life without parole. Having exhausted state remedies, they brought a federal habeas petition raising three issues: first, that the State violated Batson v. Kentucky, 476 U.S. 79 , 106 S.Ct. 1712 , 90 L.Ed.2d 69 (1986), by exercising peremptory strikes against three Blacks and two Hispanics; second, that the prosecutor engaged in misconduct during closing argument by suggesting that Burks and Celestine were members of a dangerous prison gang; and third, that the jury’s impartiality was impaired because one juror read a newspaper article relating to the case.

The district court dismissed the petition and we review de novo. Brown v. Borg, 951 F.2d 1011, 1014 (9th Cir.1991).

I

Jury selection in this case lasted over six months; the prosecution exercised 73, and the defense 72, peremptory strikes. Defendants made a timely objection that the prosecution’s exercise of certain peremptories violated Batson as well as People v. Wheeler, 22 Cal.3d 258, 277 , 148 Cal.Rptr. 890 , 583 P.2d 748 (1978) (“[T]he use of peremptory challenges to remove prospective jurors on the sole ground of group bias violates the right to trial by a jury drawn from a representative cross-section of the community under article I, section 13, of the California Constitution.”).

A. The prosecutor conceded that the defense had made a prima facie showing of racial discrimination, State RT 9432, but explained that he struck these prospective jurors because they were squishy on the death penalty, expressed a reluctance to serve and/or lacked certain life experiences — not on account of their race. Defendants argued that race must have been the prosecutor’s real motive because he didn’t strike prospective white jurors who gave answers very similar to those of the Blacks and Hispanics who were struck. State RT 9476-78.

While the defense motions were made during voir dire, the Superior Court didn’t rule on them until after the jury was empaneled. The court was therefore able to consider the defendants’ claims after the prosecution had finished exercising all of its peremptory strikes, and ruled as follows:

State RT 9729. The California Court of Appeal considered the issue, among others, and affirmed in an unpublished disposition.

B. Because the State concedes the defense made out a prima facie case under Batson and the prosecution has come up with race-neutral explanations for its peremptory strikes, the case turns on the third prong of Batson , whether the defense has carried its ultimate burden of showing purposeful discrimination. Hernandez v. New York, 500 U.S. 352, 358-60 , 111 S.Ct. 1859, 1866 , 114 L.Ed.2d 395 (1991) (plurality); Johnson v. Vasquez, 3 F.3d 1327, 1329 (9th Cir.1993). We normally accord relevant findings of the state courts a presumption of correctness on collateral review, Sumner v. Mata, 449 U.S. 539, 547 , 101 S.Ct. 764, 769 , 66 L.Ed.2d 722 (1981), but Celestine and Burks argue no such deference is due here because the findings were made under an erroneous legal standard.

At issue is the method by which the appellate courts should review Batson rulings. In United States v. Chinchilla, 874 F.2d 695 (9th Cir.1989), we held that an appellate court may overturn the finding of the trial court where a comparison between the answers given by prospective jurors who were struck and those who were not fatally undermines the prosecutor’s credibility. The California Supreme Court in People v. Johnson, 47 Cal.3d 1194 , 255 Cal.Rptr. 569 , 767 P.2d 1047 (1989), took a different approach. It held that, for purposes of both Wheeler and Batson , though the Superior Court may compare responses, the appellate court should not conduct such an inquiry on its own, and instead should “give great deference to the trial court’s determination that the use of peremptory challenges was not for an improper or class bias purpose.” Johnson, 47 Cal.3d at 1221 , 255 Cal.Rptr. 569 , 767 P.2d 1047 ; see People v. Montiel, 5 Cal.4th 877, 909 , 21 Cal.Rptr.2d 705 , 855 P.2d 1277 (1993) (“an appellate court will not reassess good faith by conducting its own comparative juror analysis”). The U.S. Supreme Court has not yet ruled on the role of comparative analysis on appellate review, so no one is quite sure whether our circuit or the California Supreme Court is right.

What is the effect of this difference in the methods of appellate review on the presumption of correctness? On the one hand, 28 U.S.C. § 2254 (d) directs us to accord a presumption of correctness to “a determination after a hearing on the merits of a factual issue, made by a State court of competent jurisdiction_” 28 U.S.C. § 2254 (d) (emphasis added). This could be read to apply to a state trial court’s determination, even if there were no meaningful appellate review. Thus, the Supreme Court has noted, in the context of a culpability finding under Enmund v. Florida, 458 U.S. 782 , 102 S.Ct. 3368 , 73 L.Ed.2d 1140 (1982), “[a]t what precise point in its criminal process a State chooses to make the Enmund determination is of little concern from the standpoint of the Constitution.... [T]he court must examine the entire course of the state-court proceeding against the defendant in order to determine whether, at some point in the process, the requisite factual finding ... has been made.” Cabana v. Bullock, 474 U.S. 376, 386-87 , 106 S.Ct. 689, 697 , 88 L.Ed.2d 704 (1986); see also Paradis v. Arave, 20 F.3d 950, 959-60 (9th Cir.1994) (“the requisite culpability finding may be made at any point in state proceedings and is thereafter entitled to a presumption of correctness pursuant to 28 U.S.C. § 2254 (d)”) (emphasis added); Richmond v. Lewis, 948 F.2d 1473, 1490 (9th Cir.1990) (on collateral review, the federal court must only “determine whether, at some point in the process, the requisite factual finding as to the defendant’s culpability has been made”) (emphasis added) (quoting Cabana, 474 U.S. at 386-87 , 106 S.Ct. at 696-97 ), reversed on other grounds, — U.S. —, 113 S.Ct. 528 , 121 L.Ed.2d 411 (1992).

On the other hand, we’ve long recognized “the importance of appellate review to a correct adjudication of guilt or innocence.” Griffin v. Illinois, 351 U.S. 12, 18 , 76 S.Ct. 585, 590 , 100 L.Ed. 891 (1956). Because the presumption of correctness is premised on the assumption that the state courts have provided a full, fair and adequate hearing, see 28 U.S.C. § 2254 (d)(2), (6), the absence of adequate appellate review arguably undermines the presumption. Under this view, state court Batson findings reviewed under Johnson need not be accorded the presumption of correctness because they were reviewed under a legal standard we consider incorrect, even though this standard has a claim to legitimacy as strong as our own.

We need not resolve this intriguing issue, for we find no Batson violation even without the presumption of correctness. The trial court’s Batson ruling turns on an evaluation of the prosecutor’s credibility which is entitled to great deference. Batson, 476 U.S. at 98 n. 21, 106 S.Ct. at 1724 n. 21; Chinchilla, 874 F.2d at 697 . So long as the trial judge made his findings unimpeded by an erroneous reading of the law, later appellate affirmance of those findings under an incorrect legal standard does not diminish the deference we must give to findings properly made by a trial court.

And there’s nothing to suggest that the incorrect legal standard infected the Superi- or Court’s findings. While Johnson limits the scope of appellate review of a trial court’s Batson determination, it does not preclude comparative analysis by the trial court. In any event, Johnson was decided in 1989, long after jurors were selected in this case. At the time the Superior Court ruled on Burks’ and Celestine’s motion, the prevailing law in California was People v. Trevino, 39 Cal.3d 667 , 217 Cal.Rptr. 652 , 704 P.2d 719 (1985), which strongly endorsed comparative analysis. Compare id. at 691-92, 217 Cal.Rptr. 652 , 704 P.2d 719 with id. at 700 , 217 Cal.Rptr. 652 , 704 P.2d 719 (Kaus, J., dissenting). Indeed, defense counsel argued that the prosecution’s reasons for peremptory strikes of Blacks and Hispanies were not credible when “compared and contrasted to ... the non-minority persons with the same problem. ...” See State RT 9478.

Id. at 9476-77. Similarly, the prosecutor often justified his challenges on comparative grounds, explaining that he struck certain minority jurors in part because they would be replaced by jurors whose life experiences he found preferable. State RT at 9461, 9470. Thus, the Batson issue was clearly fought along comparative lines in the trial court. The Superior Court judge is presumed to have performed his factfinding duty under Batson , as we later outlined it in Chinchilla . See Walton v. Arizona, 497 U.S. 639, 653 , 110 S.Ct. 3047, 3057 , 111 L.Ed.2d 511 (1990).

Burks and Celestine argue, however, that the trial court’s findings are belied by the record. They claim a comparative analysis of jurors’ responses conclusively proves that the prosecutor’s explanations were pretextual — despite significant differences in the death penalty views and life experiences of these individuals, and the obvious need for trial counsel to evaluate tone, demeanor and overall credibility.

The matter is not nearly so clear-cut. Chinchilla does not stand for the proposition that Batson is violated whenever prospective jurors of different races provide similar responses and one is excused while the other is not. Nothing in Chinchilla precludes trial counsel from making credibility determinations. In so doing, counsel is entitled to take account of the characteristics of the other prospective jurors against whom perempto-ries might be exercised; to reevaluate the mix of jurors and the weight he gives to various characteristics as he begins to exhaust his peremptories; and to take into account tone, demeanor, facial expression, emphasis — all those factors that make the words uttered by the prospective juror convincing or not. In short, counsel is entitled to exercise his full professional judgment in pursuing his client’s “legitimate interest in using [peremptory] challenges ... to secure a fair and impartial jury.” J.E.B. v. Alabama ex rel. T.B., — U.S. —, —, 114 S.Ct. 1419, 1429 , 128 L.Ed.2d 89 (1994).

The prosecutor in Chinchilla struck the only two Hispanics on objectively verifiable grounds — residential area and age. We found these reasons pretextual because a non-Hispanic who remained on the jury lived in the same area and the record simply didn’t disclose the prospective jurors’ ages. Chinchilla was thus a case where two of the three explanations offered by the prosecutor were objectively and demonstrably false. Where, in such a case, defendant shows the remaining race-neutral reason is not sufficiently “clear and reasonably specific,” he may be deemed to have carried his burden of proving intentional discrimination on the basis of race. See, e.g., Chinchilla, 874 F.2d at 698-99 .

Here, the prosecutor did accept minorities on the jury — a valid, though not necessarily dispositive, consideration in determining whether a prosecutor violated Bat-son. Palmer v. Estelle, 985 F.2d 456, 458 (9th Cir.1993); Chinchilla, 874 F.2d at 698 n. 4. Moreover, the prosecutor’s reasons for striking the prospective jurors were subjective: He didn’t believe they would impose the death penalty. Burks and Celestine certainly were entitled to challenge the bona fides of the prosecutor’s explanation. They did so, but the Superior Court found that the prosecutor’s reasons were not pretextual. We have only a cold transcript to guide us while the trial judge was there to observe the jury selection — day in and day out for six months. “[Evaluation of the prosecutor’s state of mind based on demeanor and credibility lies ‘peculiarly within a trial judge’s province.’ ” Hernandez, 500 U.S. at 365 , 111 S.Ct. at 1869 (quoting Wainwright v. Witt, 469 U.S. 412, 428 , 105 S.Ct. 844, 854 , 83 L.Ed.2d 841 (1985)).

We do not hold that a party’s explanation for the exercise of peremptory challenges will be insulated from appellate review so long as it is couched in vague and subjective terms. While subjective factors may play a legitimate role in the exercise of challenges, reliance on such factors alone cannot overcome strong objective indicia of discrimination such as a clear and sustained pattern of strikes against minority jurors. The stronger the objective evidence of discrimination, the more we will require by way of verifiable facts to sustain a trial court’s finding upholding the exercise of challenges. Here, we have relatively weak evidence of discrimination and the trial court’s decision to accept the prosecutor’s judgment is easily sustainable.

II

Defendants also objected to the prosecutor’s closing argument. The State’s theory of the case was that Brooks was killed because he dropped out of the Black Guerrilla Family prison gang, but the prosecutor was not able to elicit any such testimony, despite calling almost thirty prison inmates to testify. Before closing argument, the trial court warned the prosecutor that, although the court usually allowed “pretty wide leeway” in closing argument, it wouldn’t “on the BGF thing because of the extremely tenuous nature of what’s before this jury.” State RT 12226-27. The prosecutor nonetheless reasserted his claim:

RT 12491-93. Defense counsel moved for a mistrial, and the trial court heard the motion outside the presence of the jury.

While acknowledging the prosecutor stretched the boundaries of legitimate advocacy, the trial court concluded that the closing argument constituted “vigorous, enthusiastic advocacy short of misconduct....” State RT 12498. The judge then exercised his rarely-invoked power under Cal. Const. Art. VI, § 10, to comment on the evidence, repeating at least eight times that, in his view, no evidence in the record supported a connection between the Black Guerrilla Family and the murder:

RT 12503-05 (emphasis added).

In light of this “[pjrompt and effective action by the trial court,” United States v. Simtob, 901 F.2d 799, 806 (9th Cir.1990), Burks and Celestine did not carry their burden of proving that the closing statement had “a substantial and injurious effect or influence” on the jury’s verdict. Brecht v. Abrahamson, — U.S. —, —, 113 S.Ct. 1710, 1722 , 123 L.Ed.2d 353 (1993); Castillo v. Stainer, 997 F.2d 669, 669 (9th Cir.1993), modifying 983 F.2d 145 (9th Cir.1992).

Ill

Finally, Burks argues he suffered prejudice when the trial court failed to excuse a juror who read part of a newspaper article relating to the murder. This art'cle stated that officials believed Brooks was murdered because he had dropped out of the BGF prison gang, that Burks had been identified as the leader of the BGF, that Burks had been sentenced to life for murder, and that while in prison, Burks had been involved in numerous assaults, including another stabbing. The trial court polled the venire to determine whether anyone had read this article, and excused six prospective jurors on the basis of their responses to the questionnaire. Roland Jacques and Clifford Drake, who read portions of the article, were not excused. Burks filed a motion to excuse these two, and after reviewing their questionnaires and conducting an examination, the court excused Jacques, but retained Drake because it found that Jacques had read about and remembered defendants’ prior offenses, while Drake stopped reading as soon as he realized the article was related to this case and before he reached the information about the prior offenses. State RT 11421.

“[T]he extent, if at all, to which the jurors saw or discussed the extrinsic evidence,” Dickson v. Sullivan, 849 F.2d 403, 406 (9th Cir.1988), is a question of historical fact as to which the state court’s findings are entitled to a presumption of correctness. 28 U.S.C. § 2254 (d).

The trial court held an evidentiary hearing, examined Drake and Jacques, and determined that Drake, in contrast to Jacques, was not exposed to any prejudicial information. Drake testified that the only thing he learned from the article was that more than a quarter million dollars was spent on the defense. State RT 11420. When questioned about the listed offenses, Drake recalled that “some things were itemized, numerical,” id., but he didn’t recall specific offenses or how they related to the defendants. State RT 11421. The trial court also found that Drake didn’t discuss the contents of the article with other jurors. State RT 11418.

The Court of Appeal reviewed the record and affirmed the trial court’s denial of Burks’ motion to excuse Drake: “Drake’s testimony established that he did not read the entire newspaper article. He did not read the parts of the article which were potentially prejudicial to the defendants. He assured the court that he learned nothing concerning the defendants from the article and that the article would not influence his performance as a juror.” CR 7, App. B at 39-40. Thus, the Court of Appeal found that the trial court did not abuse its discretion in retaining Drake as a juror. Id. at 40 .

The state courts found that Drake either was not exposed to or had no recollection of the extraneous prejudicial information. We have no basis for overturning this presumptively correct finding.

AFFIRMED.

. Petitioners claim that the trial judge placed himself in the "subjective position" of "sitting as a prosecutor,” Burks' Br. at 23, and that shows the judge was biased in evaluating the prosecutor's reasons for his strikes. Petitioners make too much of this remark. We read it as reflecting the judge's effort to articulate, perhaps somewhat inartfully, his view that the strikes served legitimate prosecutorial purposes. . Such a rule would create an odd discontinuity in the law applicable to habeas petitions raising Batson claims — at least until the U.S. Supreme Court resolves the comparative analysis issue. California appellate courts are, of course, bound by Johnson, which precludes comparative analysis. See Lockhart v. Fretwell, - U.S. -, -, 113 S.Ct. 838, 846 , 122 L.Ed.2d 180 (1993) (Thomas, J., concurring). When the case then comes to federal court in this circuit, however, the issue is governed by Chinchilla. See Wainwright v. Sykes, 433 U.S. 72, 79 , 97 S.Ct. 2497, 2502 , 53 L.Ed.2d 594 (1977) (challenge to state court's resolution of dispositive federal issues “is always fair game on federal habeas”). Ah, the mysteries of federalism! . Petitioners also argue that the prosecutor's reasons must have been pretextual because, if he really objected to the prospective jurors' death penalty views, he would have challenged them for cause under Wainwright v. Witt, 469 U.S. 412 , 105 S.Ct. 844 , 83 L.Ed.2d 841 (1985). But the prosecutor (or defense counsel, for that matter) may have a hunch that a prospective juror is subject to disqualification for cause, yet not have enough by way of articulable facts to support such a challenge. Peremptory strikes are a legitimate means for counsel to act on such hunches and suspicions. J.E.B. v. Alabama ex rel. T.B., — U.S. —, —, 114 S.Ct. 1419, 1429 , 128 L.Ed.2d 89 (1994). . The Brecht majority states that habeas petitioners "are not entitled to habeas relief based on trial error unless they can establish that it resulted in 'actual prejudice.’ " - U.S. at -, 113 S.Ct. at 1722. We have held that Brecht placed the burden of proof to show actual prejudice on the petitioner. Castillo, 997 F.2d at 669 , modifying 983 F.2d 145, 148-49 (9th Cir.1992). Accord Samuels v. Mann, 13 F.3d 522, 526 (2d Cir.1993) (quoting Brecht majority, in dicta); Cumbie v. Singletary, 991 F.2d 715, 724 (11th Cir.1993). But see Brecht, - U.S. at -, 113 S.Ct. at 1723-24 (Stevens, J., concurring) (“unless an error is merely ‘technical,’ the burden of sustaining a verdict by demonstrating that the error was harmless rests on the prosecution. A constitutional violation, of course, would never fall in the 'technical' category”); Libby v. Duval, 19 F.3d 733, 744-45 (1st Cir.1994); Stoner v. Sowders, 997 F.2d 209, 213 (6th Cir.1993). The Supreme Court has recently granted certiorari to clarify this issue. O’Neal v. Morris, 3 F.3d 143 (6th Cir.1993), cert. granted, - U.S. -, 114 S.Ct. 1396 , 128 L.Ed.2d 70 (1994). Even if it turns out that the burden is on the prosecution, we conclude it was carried here. . Celestine does not join the argument challenging failure to excuse this juror because Celes-tine’s trial counsel elected not to join in the motion to excuse him at trial. Celestine’s Opening Br. at 18 n. 7. On direct review, a reversal of Burks’ conviction on this ground would also require us to reverse Celestine’s because our only inquiry would be whether one defendant "suffered the same prejudice from the same fundamental trial error” as another. United States v. Baker, 999 F.2d 412, 417 (9th Cir.1993) (quoting United States v. Olano, 934 F.2d 1425, 1439 (9th Cir.1991), rev'd on other grounds, - U.S. -, 113 S.Ct. 1770 , 123 L.Ed.2d 508 (1993)). However, on collateral review, a habeas petitioner must demonstrate cause in addition to prejudice to overcome a procedural default. Coleman v. Thompson, 501 U.S. 722, 750-51 , 111 S.Ct. 2546, 2565 , 115 L.Ed.2d 640 (1991). As cause obviously requires an individualized showing by each petitioner, we need not consider whether Celes-tine’s constitutional rights were violated by the trial court’s refusal to excuse Drake. . Because we are bound by the state courts’ finding that no prejudicial information was actually introduced, a fortiori, there was no constitutional error. Contrary to Burks' claims, because we need not consider whether the error was prejudicial, Brecht, - U.S. at -, 113 S.Ct. at 1722, Castillo, 997 F.2d at 669 , we have no occasion to engage in de novo review. Cf. Dickson, 849 F.2d at 405 .