delivered the opinion of the Court.

Two years ago, we ordered that a certificate of appealability, under 28 U. S. C. § 2253(c), be issued to habeas petitioner Miller-El, affording review of the District Court’s rejection of the claim that prosecutors in his capital murder trial made peremptory strikes of potential jurors based on race. Today we find Miller-El entitled to prevail on that claim and order relief under § 2254.

I

In the course of robbing a Holiday Inn in Dallas, Texas, in late 1985, Miller-El and his accomplices bound and gagged *236two hotel employees, whom Miller-El then shot, killing one and severely injuring the other. During jury selection in Miller-El’s trial for capital murder, prosecutors used peremptory strikes against 10 qualified black venire members. Miller-El objected that the strikes were based on race and could not be presumed legitimate, given a history of excluding black members from criminal juries by the Dallas County District Attorney’s Office. The trial court received evidence of the practice alleged but found no “systematic exclusion of blacks as a matter of policy” by that office, App. 882-883, and therefore no entitlement to relief under Swain v. Alabama, 380 U. S. 202 (1965), the case then defining and marking the limits of relief from racially biased jury selection. The court denied Miller-El’s request to pick a new jury, and the trial ended with his death sentence for capital murder.

While an appeal was pending, this Court decided Batson v. Kentucky, 476 U. S. 79 (1986), which replaced Swain’s threshold requirement to prove systemic discrimination under a Fourteenth Amendment jury claim, with the rule that discrimination by the prosecutor in selecting the defendant’s jury sufficed to establish the constitutional violation. The Texas Court of Criminal Appeals then remanded the matter to the trial court to determine whether Miller-El could show that prosecutors in his case peremptorily struck prospective black jurors because of race. Miller-El v. State, 748 S. W. 2d 459 (1988) (en banc).

The trial court found no such demonstration. After reviewing the voir dire record of the explanations given for some of the challenged strikes, and after hearing one of the prosecutors, Paul Macaluso, give his justification for those previously unexplained, the trial court accepted the stated race-neutral reasons for the strikes, which the judge called “completely credible [and] sufficient” as the grounds for a finding of “no purposeful discrimination.” Findings of Fact and Conclusions of Law Upon Remand from the Court of Criminal Appeals in State v. Miller-El, No. 8668-NL (5th Crim. Dist. Ct., Dallas County, Tex., Jan. 13, 1989), pp. 5-6, *237App. 928-929. The Court of Criminal Appeals affirmed, stating it found “ample support” in the voir dire record for the race-neutral explanations offered by prosecutors for the peremptory strikes. Miller-El v. State, No. 69,677 (Sept. 16, 1992) (per curiam), p. 2, App. 931.

Miller-El then sought habeas relief under 28 U. S. C. §2254, again pressing his Batson claim, among others not now before us. The District Court denied relief, Miller-El v. Johnson, Civil No. 3:96-CV-1992-H (ND Tex., June 5, 2000), App. 987, and the Court of Appeals for the Fifth Circuit precluded appeal by denying a certificate of appealability, Miller-El v. Johnson, 261 F. 3d 445 (2001). We granted certiorari to consider whether Miller-El was entitled to review on the Batson claim, Miller-El v. Cockrell, 534 U. S. 1122 (2002), and reversed the Court of Appeals. After examining the record of Miller-El’s extensive evidence of purposeful discrimination by the Dallas County District Attorney’s Office before and during his trial, we found an appeal was in order, since the merits of the Batson claim were, at the least, debatable by jurists of reason. Miller-El v. Cockrell, 537 U. S. 322 (2003). After granting a certificate of appealability, the Fifth Circuit rejected Miller-El’s Batson claim on the merits. 361 F. 3d 849 (2004). We again granted certiorari, 542 U. S. 936 (2004), and again we reverse.

a

<

It is well known that prejudices often exist against particular classes in the community, which sway the judgment of jurors, and which, therefore, operate in some cases to deny to persons of those classes the full enjoyment of that protection which others enjoy.” Strauder v. West Virginia, 100 U. S. 303, 309 (1880); see also Batson v. Kentucky, supra, at 86. Defendants are harmed, of course, when racial discrimination in jury selection compromises the right of trial by impartial jury, Strauder v. West Virginia, supra, at 308, but racial minorities are harmed more generally, for prosecu*238tors drawing racial lines in picking juries establish “state-sponsored group stereotypes rooted in, and reflective of, historical prejudice,” J. E. B. v. Alabama ex rel. T. B., 511 U. S. 127, 128 (1994).

Nor is the harm confined to minorities. When the government’s choice of jurors is tainted with racial bias, that “overt wrong ... casts doubt over the obligation of the parties, the jury, and indeed the court to adhere to the law throughout the trial . . . .” Powers v. Ohio, 499 U. S. 400, 412 (1991). That is, the very integrity of the courts is jeopardized when a prosecutor’s discrimination “invites cynicism respecting the jury’s neutrality,” ibid., and undermines public confidence in adjudication, Georgia v. McCollum, 505 U. S. 42, 49 (1992); Edmonson v. Leesville Concrete Co., 500 U. S. 614, 628 (1991); Batson v. Kentucky, supra, at 87. So, “[f]or more than a century, this Court consistently and repeatedly has reaffirmed that racial discrimination by the State in jury selection offends the Equal Protection Clause.” Georgia v. McCollum, supra, at 44; see Strauder v. West Virginia, supra, at 308, 310; Norris v. Alabama, 294 U. S. 587, 596 (1935); Swain v. Alabama, supra, at 223-224; Batson v. Kentucky, supra, at 84; Powers v. Ohio, supra, at 404.

The rub has been the practical difficulty of ferreting out discrimination in selections discretionary by nature, and choices subject to myriad legitimate influences, whatever the race of the individuals on the panel from which jurors are selected. In Swain v. Alabama, we tackled the problem of “the quantum of proof necessary” to show purposeful discrimination, 380 U. S., at 205, with an eye to preserving each side’s historical prerogative to make a peremptory strike or challenge, the very‘nature of which is traditionally “without a reason stated,” id., at 220. The Swain Court tried to relate peremptory challenge to equal protection by presuming the legitimacy of prosecutors’ strikes except in the face of a longstanding pattern of discrimination: when “in case after case, whatever the circumstances,” no blacks served on ju-*239ríes, then “giving even the widest leeway to the operation of irrational but trial-related suspicions and antagonisms, it would appear that the purposes of the peremptory challenge [were] being perverted.” Id., at 223-224.

Swain’s demand to make out a continuity of discrimination over time, however, turned out to be difficult to the point of unworkable, and in Batson v. Kentucky, we recognized that this requirement to show an extended pattern imposed a “crippling burden of proof” that left prosecutors’ use of pe-remptories “largely immune from constitutional scrutiny.” 476 U. S., at 92-93. By Batson’s day, the law implementing equal protection elsewhere had evolved into less discouraging standards for assessing a claim of purposeful discrimination, id., at 93-95 (citing, e. g., Washington v. Davis, 426 U. S. 229 (1976), and Arlington Heights v. Metropolitan Housing Development Corp., 429 U. S. 252 (1977)), and we accordingly held that a defendant could make out a prima facie case of discriminatory jury selection by “the totality of the relevant facts” about a prosecutor’s conduct during the defendant’s own trial. Batson v. Kentucky, 476 U. S., at 94, 96. “Once the defendant makes a prima facie showing, the burden shifts to the State to come forward with a neutral explanation for challenging ... jurors” within an arguably targeted class. Id., at 97. Although there may be “any number of bases on which a prosecutor reasonably [might] believe that it is desirable to strike a juror who is not excusable for cause . .., the prosecutor must give a clear and reasonably specific explanation of his legitimate reasons for exercising the challeng[e].” Id., at 98, n. 20 (internal quotation marks omitted). “The trial court then will have the duty to determine if the defendant has established purposeful discrimination.” Id., at 98.

Although the move from Swain to Batson left a defendant free to challenge the prosecution without having to cast Swain’s wide net, the net was not entirely consigned to history, for Batson’s individualized focus came with a weakness *240of its own owing to its very emphasis on the particular reasons a prosecutor might give. If any facially neutral reason sufficed to answer a Batson challenge, then Batson would not amount to much more than Swain. Some stated reasons are false, and although some false reasons are shown up within the four corners of a given case, sometimes a court may not be sure unless it looks beyond the case at hand. Hence Batson’s explanation that a defendant may rely on “all relevant circumstances” to raise an inference of purposeful discrimination. 476 U. S., at 96-97.

B

This case comes to us on review of a denial of habeas relief sought under 28 U. S. C. §2254, following the Texas trial court’s prior determination of fact that the State’s race-neutral explanations were true, see Purkett v. Elem, 514 U. S. 765, 769 (1995) (per curiam); Batson v. Kentucky, supra, at 98, n. 21.

Under the Antiterrorism and Effective Death Penalty Act of 1996, Miller-El may obtain relief only by showing the Texas conclusion to be “an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” 28 U. S. C. § 2254(d)(2). Thus we presume the Texas court’s factual findings to be sound unless Miller-El rebuts the “presumption of correctness by clear and convincing evidence.” § 2254(e)(1). The standard is demanding but not insatiable; as we said the last time this case was here, “[d]eference does not by definition preclude relief.” Miller-El v. Cockrell, 537 U. S., at 340.

r-H J-H H-\

A

The numbers describing the prosecution’s use of perempto-ries are remarkable. Out of 20 black members of the 108-person venire panel for Miller-El’s trial, only 1 served. Although 9 were excused for cause or by agreement, 10 were *241peremptorily struck by the prosecution. Id., at 331. “The prosecutors used their peremptory strikes to exclude 91% of the eligible African-American venire members .... Happenstance is unlikely to produce this disparity.” Id., at 342.

More powerful than these bare statistics, however, are side-by-side comparisons of some black venire panelists who were struck and white panelists allowed to serve. If a prosecutor’s proffered reason for striking a black panelist applies just as well to an otherwise-similar nonblack who is permitted to serve, that is evidence tending to prove purposeful discrimination to be considered at Batson’s third step. Cf. Reeves v. Sanderson Plumbing Products, Inc., 530 U. S. 133, 147 (2000) (in employment discrimination cases, “[p]roof that the defendant’s explanation is unworthy of credence is simply one form of circumstantial evidence that is probative of intentional discrimination, and it may be quite persuasive”). While we did not develop a comparative juror analysis last time, we did note that the prosecution’s reasons for exercising peremptory strikes against some black panel members appeared equally on point as to some white jurors who served. Miller-El v. Cockrell, supra, at 343.1 The details of two panel member comparisons bear this out.2

*242The prosecution used its second peremptory strike to exclude Billy Jean Fields, a black man who expressed unwavering support for the death penalty. On the questionnaire filled out by all panel members before individual examination on the stand, Fields said that he believed in capital punishment, Joint Lodging 14, and during questioning he disclosed his belief that the State acts on God's behalf when it imposes the death penalty. “Therefore, if the State exacts death, then that’s what it should be.” App. 174. He testified that he had no religious or philosophical reservations about the death penalty and that the death penalty deterred crime. Id., at 174-175. He twice averred, without apparent hesitation, that he could sit on Miller-El’s jury and make a decision to impose this penalty. Id., at 176-177.

Although at one point in the questioning, Fields indicated that the possibility of rehabilitation might be relevant to the likelihood that a defendant would commit future acts of violence, id., at 183, he responded to ensuing questions by saying that although he believed anyone could be rehabilitated, this belief would not stand in the way of a decision to impose the death penalty:

“[B]ased on what you [the prosecutor] said as far as the crime goes, there are only two things that could be rendered, death or life in prison. If for some reason the testimony didn’t warrant death, then life imprisonment *243would give an individual an opportunity to rehabilitate. But, you know, you said that the jurors didn’t have the opportunity to make a personal decision in the matter with reference to what I thought or felt, but it was just based on the questions according to the way the law has been handed down.” Id., at 185 (alteration omitted).

Fields also noted on his questionnaire that his brother had a criminal history. Joint Lodging 13. During questioning, the prosecution went into this, too:

“Q Could you tell me a little bit about that?

“A He was arrested and convicted on [a] number of occasions for possession of a controlled substance.

“Q Was that here in Dallas?

“A Yes.

“Q Was he involved in any trials or anything like that? “A I suppose of sorts. I don’t really know too much about it.

“Q Was he ever convicted?

"A Yeah, he served time.

“Q Do you feel that that would in any way interfere with your service on this jury at all?

“A No.” App. 190 (alteration omitted).

Fields was struck peremptorily by the prosecution, with prosecutor James Nelson offering a race-neutral reason:

“[W]e . . . have concern with reference to some of his statements as to the death penalty in that he said that he could only give death if he thought a person could not be rehabilitated and he later made the comment that any person could be rehabilitated if they find God or are introduced to God and the fact that we have a concern that his religious feelings may affect his jury service in this case.” Id., at 197 (alteration omitted).

*244Thus, Nelson simply mischaracterized Fields’s testimony. He represented that Fields said he would not vote for death if rehabilitation was possible, whereas Fields unequivocally stated that he could impose the death penalty regardless of the possibility of rehabilitation. Perhaps Nelson misunderstood, but unless he had an ulterior reason for keeping Fields off the jury we think he would have proceeded differently. In light of Fields’s outspoken support for the death penalty, we expect the prosecutor would have cleared up any misunderstanding by asking further questions before getting to the point of exercising a strike.

If, indeed, Fields’s thoughts on rehabilitation did make the prosecutor uneasy, he should have worried about a number of white panel members he accepted with no evident reservations. Sandra Hearn said that she believed in the death penalty “if a criminal cannot be rehabilitated and continues to commit the same type of crime.” Id., at 429.3 Hearn went so far as to express doubt that at the penalty phase of a capital case she could conclude that a convicted murderer “would probably commit some criminal acts of violence in the future.” Id., at 440. “People change,” she said, making it hard to assess the risk of someone’s future dangerousness. “[T]he evidence would have to be awful strong.” Ibid. But the prosecution did not respond to Hearn the way it did to Fields, and without delving into her views about rehabilitation with any further question, it raised no objection to her serving on the jury. White panelist Mary Witt said she would take the possibility of rehabilitation into account in deciding at the penalty phase of the trial about a defendant’s probability of future dangerousness, 6 Record of Voir Dire 2433 (hereinafter Record), but the prosecutors asked her no further question about her views on reformation, and they *245accepted her as a juror, id., at 2464-2465.4 Latino venireman Fernando Gutierrez, who served on the jury, said that he would consider the death penalty for someone who could not be rehabilitated, App. 777, but the prosecutors did not question him further about this view. In sum, nonblack jurors whose remarks on rehabilitation could well have signaled a limit on their willingness to impose a death sentence were not questioned further and drew no objection, but the prosecution expressed apprehension about a black juror’s belief in the possibility of reformation even though he repeatedly stated his approval of the death penalty and testified that he could impose it according to state legal standards even when the alternative sentence of life imprisonment would give a defendant (like everyone else in the world) the opportunity to reform.5

The unlikelihood that his position on rehabilitation had anything to do with the peremptory strike of Fields is underscored by the prosecution’s response after Miller-El’s lawyer pointed out that the prosecutor had misrepresented Fields’s responses on the subject. A moment earlier the prosecutor *246had finished his misdescription of Fields’s views on potential rehabilitation with the words, “Those are our reasons for exercising our . . . strike at this time.” Id., at 197. When defense counsel called him on his misstatement, he neither defended what he said nor withdrew the strike. Id., at 198. Instead, he suddenly came up with Fields’s brother’s prior conviction as another reason for the strike. Id., at 199.

It would be difficult to credit the State’s new explanation, which reeks of afterthought. While the Court of Appeals tried to bolster it with the observation that no seated juror was in Fields’s position with respect to his brother, 361 F. 3d, at 859-860, the court’s readiness to accept the State’s substitute reason ignores not only its pretextual timing but the other reasons rendering it implausible. Fields’s testimony indicated he was not close to his brother, App. 190 (“I don’t really know too much about it”), and the prosecution asked nothing further about the influence his brother’s history might have had on Fields, as it probably would have done if the family history had actually mattered. See, e.g., Ex parte Travis, 776 So. 2d 874, 881 (Ala. 2000) (“[T]he State’s failure to engage in any meaningful voir dire examination on a subject the State alleges it is concerned about is evidence suggesting that the explanation is a sham and a pretext for discrimination”). There is no good reason to doubt that the State’s afterthought about Fields’s brother was anything but makeweight.

The Court of Appeals’s judgment on the Fields strike is unsupportable for the same reason the State’s first explanation is itself unsupportable. The Appeals Court’s description of Fields’s voir dire testimony mentioned only his statements that everyone could be rehabilitated, failing to note that Fields affirmed that he could give the death penalty if the law and evidence called for it, regardless of the possibility of divine grace. The Court of Appeals made no mention of the fact that the prosecution mischaracterized Fields as *247saying he could not give death if rehabilitation were possible. 361 F. 3d, at 856.

In sum, when we look for nonblack jurors similarly situated to Fields, we find strong similarities as well as some differences.6 But the differences seem far from significant, particularly when we read Fields’s voir dire testimony in its entirety. Upon that reading, Fields should have been an ideal juror in the eyes of a prosecutor seeking a death sentence, and the prosecutors’ explanations for the strike cannot reasonably be accepted. See Miller-El v. Cockrell, 537 U. S., at 339 (the credibility of reasons given can be measured by “how reasonable, or how improbable, the explanations are; and by whether the proffered rationale has some basis in accepted trial strategy”).

The prosecution’s proffered reasons for striking Joe Warren, another black venireman, are comparably unlikely. Warren gave this answer when he was asked what the death penalty accomplished:

“I don’t know. It’s really hard to say because I know sometimes you feel that it might help to deter crime and then you feel that the person is not really suffering. You’re taking the suffering away from him. So it’s like I said, sometimes you have mixed feelings about whether or not this is punishment or, you know, you’re *248relieving personal punishment.” App. 205; 3 Record 1532.

The prosecution said nothing about these remarks when it struck Warren from the panel, but prosecutor Paul Macaluso referred to this answer as the first of his reasons when he testified at the later Batson hearing:

“I thought [Warren’s statements on voir dire] were inconsistent responses. At one point he says, you know, on a case-by-case basis and at another point he said, well, I think — I got the impression, at least, that he suggested that the death penalty was an easy way out, that they should be made to suffer more.” App. 909.

On the face of it, the explanation is reasonable from the State’s point of view, but its plausibility is severely undercut by the prosecution’s failure to object to other panel members who expressed views much like Warren’s. Kevin Duke, who served on the jury, said, “sometimes death would be better to me than — being in prison would be like dying every day and, if you were in prison for life with no hope of parole, I[’d] just as soon have it over with than be in prison for the rest of your life.” Id., at 372. Troy Woods, the one black panelist to serve as juror, said that capital punishment “is too easy. I think that’s a quick relief.... I feel like [hard labor is] more of a punishment than putting them to sleep.” Id., at 408. Sandra Jenkins, whom the State accepted (but who was then struck by the defense) testified that she thought “a harsher treatment is life imprisonment with no parole.” Id., at 542. Leta Girard, accepted by the State (but also struck by the defense) gave her opinion that “living sometimes is a worse — is worse to me than dying would be.” Id., at 624. The fact that Macaluso’s reason also applied to these other panel members, most of them white, none of them struck, is evidence of pretext.

*249The suggestion of pretext is not, moreover, mitigated much by Macaluso’s explanation that Warren was struck when the State had 10 peremptory challenges left and could afford to be liberal in using them. Id., at 908. If that were the explanation for striking Warren and later accepting panel members who thought death would be too easy, the prosecutors should have struck Sandra Jenkins, whom they examined and accepted before Warren. Indeed, the disparate treatment is the more remarkable for the fact that the prosecutors repeatedly questioned Warren on his capacity and willingness to impose a sentence of death and elicited statements of his ability to do so if the evidence supported that result and the answer to each special question was yes, id., at 202.2, 202.3, 205, 207, whereas the record before us discloses no attempt to determine whether Jenkins would be able to vote for death in spite of her view that it was easy on the convict, id., at 541-546. Yet the prosecutors accepted the white panel member Jenkins and struck the black venireman Warren.

Macaluso’s explanation that the prosecutors grew more sparing with peremptory challenges as the jury selection wore on does, however, weaken any suggestion that the State’s acceptance of Woods, the one black juror, shows that race was not in play. Woods was the eighth juror, qualified in the fifth week of jury selection. Joint Lodging 125. When the State accepted him, 11 of its 15 peremptory strikes were gone, 7 of them used to strike black panel members. Id., at 137. The juror questionnaires show that at least three members of the venire panel yet to be questioned on the stand were opposed to capital punishment, Janice Mackey, id., at 79; Paul Bailey, id., at 63; and Anna Keaton, id., at 55.7 With at least three remaining panel members *250highly undesirable to the State, the prosecutors had to exercise prudent restraint in using strikes. This late-stage decision to accept a black panel member willing to impose a death sentence does not, therefore, neutralize the early-stage decision to challenge a comparable venireman, Warren. In fact, if the prosecutors were going to accept any black juror to obscure the otherwise consistent pattern of opposition to seating one, the time to do so was getting late.8

The Court of Appeals pretermitted these difficulties by stating that the prosecution’s reason for striking Warren was a more general ambivalence about the penalty and his ability to impose it, 361 F. 3d, at 856-857 (and the dissent presses that explanation here, post, at 286-289). But this rationalization was erroneous as a matter of fact and as a matter of law.

As to fact, Macaluso said nothing about any general ambivalence. He simply alluded to the possibility that Warren might think the death penalty too easy on some defendants, saying nothing about Warren’s ability to impose the penalty when it appeared to be warranted.9 On the contrary, though *251Warren had indeed questioned the extent to which the death penalty served a purpose in society, App. 205, he explained his position in response to the very next question: it was not any qualm about imposing what society generally deems its harshest punishment, but his concern that the death penalty might not be severe enough, ibid. When Warren was asked whether he could impose the death penalty he said he thought he could; when told that answering yes to the special issue questions would be tantamount to voting for death he said he could give yes answers if the evidence supported them. Id., at 207.10

As for law, the rule in Batson provides an opportunity to the prosecutor to give the reason for striking the juror, and *252it requires the judge to assess the plausibility of that reason in light of all evidence with a bearing on it. 476 U. S., at 96-97; Miller-El v. Cockrell, 537 U. S., at 339. It is true that peremptories are often the subjects of instinct, Batson v. Kentucky, supra, at 106 (Marshall, J., concurring), and it can sometimes be hard to say what the reason is. But when illegitimate grounds like race are in issue, a prosecutor simply has got to state his reasons as best he can and stand or fall on the plausibility of the reasons he gives. A Batson challenge does not call for a mere exercise in thinking up any rational basis. If the stated reason does not hold up, its pretextual significance does not fade because a trial judge, or an appeals court, can imagine a reason that might not have been shown up as false. The Court of Appeals’s and the dissent’s substitution of a reason for eliminating Warren does nothing to satisfy the prosecutors’ burden of stating a racially neutral explanation for their own actions.

The whole of the voir dire testimony subject to consideration casts the prosecution’s reasons for striking Warren in an implausible light. Comparing his strike with the treatment of panel members who expressed similar views supports a conclusion that race was significant in determining who was challenged and who was not.11

*253B

The case for discrimination goes beyond these comparisons to include broader patterns of practice during the jury selection. The prosecution’s shuffling of the venire panel, its en-quiry into views on the death penalty, its questioning about minimum acceptable sentences: all indicate decisions probably based on race. Finally, the appearance of discrimination is confirmed by widely known evidence of the general policy of the Dallas County District Attorney’s Office to exclude black venire members from juries at the time Miller-El’s jury was selected.

The first clue to the prosecutors’ intentions, distinct from the peremptory challenges themselves, is their resort during voir dire to a procedure known in Texas as the jury shuffle. In the State’s criminal practice, either side may literally reshuffle the cards bearing panel members’ names, thus rearranging the order in which members of a venire panel are seated and reached for questioning.12 Once the order is established, the panel members seated at the back are likely to escape voir dire altogether, for those not questioned by the end of the week are dismissed. As we previously explained,

*254“the prosecution’s decision to seek a jury shuffle when a predominant number of African-Americans were seated in the front of the panel, along with its decision to delay a formal objection to the defense’s shuffle until after the new racial composition was revealed, raise a suspicion that the State sought to exclude African-Americans from the jury. Our concerns are amplified by the fact that the state court also had before it, and apparently ignored, testimony demonstrating that the Dallas County District Attorney’s Office had, by its own admission, used this process to manipulate the racial composition of the jury in the past.” Miller-El v. Cockrell, supra, at 346.

In this case, the prosecution and then the defense shuffled the cards at the beginning of the first week of voir dire; the record does not reflect the changes in order. App. 113-114. At the beginning of the second week, when a number of black members were seated at the front of the panel, the prosecution shuffled.13 2 Record 836-837. At the beginning of the third week, the first four panel members were black. The prosecution shuffled, and these black panel members ended up at the back. Then the defense shuffled, and the black panel members again appeared at the front. The prosecution requested another shuffle, but the trial court refused. App. 124-132. Finally, the defense shuffled at the beginning of the fourth and fifth weeks of voir dire; the record does not reflect the panel’s racial composition before or after those shuffles. Id., at 621-622; 9 Record 3585-3586.

The State notes in its brief that there might be racially neutral reasons for shuffling the jury, Brief for Respondent 36-37, and we suppose there might be. But no racially neutral reason has ever been offered in this case, and nothing *255stops the suspicion of discriminatory intent from rising to an inference.14

The next body of evidence that the State was trying to avoid black jurors is the contrasting voir dire questions posed respectively to black and nonblack panel members, on two different subjects. First, there were the prosecutors’ statements preceding questions about a potential juror’s thoughts on capital punishment. Some of these prefatory statements were cast in general terms, but some followed the so-called graphic script, describing the method of execution in rhetorical and clinical detail. It is intended, Miller-El contends, to prompt some expression of hesitation to consider the death penalty and thus to elicit plausibly neutral grounds for a peremptory strike of a potential juror subjected to it, if not a strike for cause. If the graphic script is given to a higher proportion of blacks than whites, this is evidence that prosecutors more often wanted blacks off the jury, absent some neutral and extenuating explanation.

As we pointed out last time, for 94% of white venire panel members, prosecutors gave a bland description of the death penalty before asking about the individual’s feelings on the subject. Miller-El v. Cockrell, 537 U. S., at 332. The abstract account went something like this:

“I feel like it [is] only fair that we tell you our position in this case. The State of Texas ... is actively seeking the death penalty in this case for Thomas Joe Miller-El. We anticipate that we will be able to present to a jury the quantity and type of evidence necessary to convict him of capital murder and the quantity and type of evi*256dence sufficient to allow a jury to answer these three questions over here in the affirmative. A yes answer to each of those questions results in an automatic death penalty from Judge McDowell.” App. 564-565.

Only 6% of white venire panelists, but 53% of those who were black, heard a different description of the death penalty before being asked their feelings about it. This is an example of the graphic script:

“I feel like you have a right to know right up front what our position is. Mr. Kinne, Mr. Macaluso and myself, representing the people of Dallas County and the state of Texas, are actively seeking the death penalty for Thomas Joe Miller-El....

“We do that with the anticipation that, when the death penalty is assessed, at some point Mr. Thomas Joe Miller-El — the man sitting right down there — will be taken to Huntsville and will be put on death row and at some point taken to the death house and placed on a gurney and injected with a lethal substance until he is dead as a result of the proceedings that we have in this court on this case. So that’s basically our position going into this thing.” Id., at 572-573 (alteration omitted).

The State concedes that this disparate questioning did occur but argues that use of the graphic script turned not on a panelist’s race but on expressed ambivalence about the death penalty in the preliminary questionnaire.15 Prosecu*257tors were trying, the argument goes, to weed out noncommittal or uncertain jurors, not black jurors. And while some white venire members expressed opposition to the death penalty on their questionnaires, they were not read the graphic script because their feelings were already clear. The State says that giving the graphic script to these panel members would only have antagonized them. Brief for Respondent 27-32.

This argument, however, first advanced in dissent when the case was last here, Miller-El v. Cockrell, supra, at 364-368 (opinion of Thomas, J.), and later adopted by the State and the Court of Appeals, simply does not fit the facts. Looking at the answers on the questionnaires, and at voir dire testimony expressly discussing answers on the question*258naires,16 we find that black venire members were more likely than nonblacks to receive the graphic script regardless of their expressions of certainty or ambivalence about the death penalty, and the State’s chosen explanation for the graphic script fails in the cases of four out of the eight black panel members who received it.17 Two of them, Janice Mackey and Anna Keaton, clearly stated opposition to the death penalty but they received the graphic script,18 while the black panel members Wayman Kennedy and Jeannette Butler were unambiguously in favor19 but got the graphic *259description anyway.20 The State’s explanation does even worse in the instances of the five nonblacks who received the graphic script, missing the mark four times out of five: Vivian Sztybel and Filemon Zablan received it,21 although each was unambiguously in favor of the death penalty,22 while Dominick Desinise and Clara Evans unambiguously opposed it23 but were given the graphic version.24

The State’s purported rationale fails again if we look only to the treatment of ambivalent panel members, ambivalent black individuals having been more likely to receive the graphic description than ambivalent nonblacks. Three non-black members of the venire indicated ambivalence to the death penalty on their questionnaires;25 only one of them, *260Fernando Gutierrez, received the graphic script.26 But of the four black panel members who expressed ambivalence,27 all got the graphic treatment.28

The State’s attempt at a race-neutral rationalization thus simply fails to explain what the prosecutors did. But if we posit instead that the prosecutors’ first object was to use the graphic script to make a case for excluding black panel members opposed to or ambivalent about the death penalty, there is a much tighter fit of fact and explanation.29 Of the 10 nonblacks whose questionnaires expressed ambivalence or opposition,30 only 30% received the graphic treatment.31 But of the seven blacks who expressed ambivalence or opposition,32 86% heard the graphic script.33 As between the State’s ambivalence explanation and Miller-El’s racial one, race is much the better, and the reasonable inference is that race was the major consideration when the prosecution chose to follow the graphic script.

*261The same is true for another kind of disparate questioning, which might fairly be called trickery. The prosecutors asked members of the panel how low a sentence they would consider imposing for murder. Most potential jurors were first told that Texas law provided for a minimum term of five years, but some members of the panel were not, and if a panel member then insisted on a minimum above five years, the prosecutor would suppress his normal preference for tough jurors and claim cause to strike. Two Terms ago, we described how this disparate questioning was correlated with race:

“Ninety-four percent of whites were informed of the statutory minimum sentence, compared [with] only twelve and a half percent of African-Americans. No explanation is proffered for the statistical disparity. Pierre v. Louisiana, 306 U. S. 354, 361-362 (1939) ( “The fact that the testimony... was not challenged by evidence appropriately direct, cannot be brushed aside.” Had there been evidence obtainable to contradict and disprove the testimony offered by petitioner, it cannot be assumed that the State would have refrained from introducing it’ (quoting Norris v. Alabama, 294 U. S. 587, 594-595 (1935))). Indeed, while petitioner’s appeal was pending before the Texas Court of Criminal Appeals, that court found a Batson violation where this precise line of disparate questioning on mandatory minimums was employed by one of the same prosecutors who tried the instant case. Chambers v. State, 784 S. W. 2d 29, 31 (Tex. Crim. App. 1989).” Miller-El v. Cockrell, 537 U. S., at 345.

The State concedes that the manipulative minimum punishment questioning was used to create cause to strike, Brief for Respondent 33, and n. 26, but now it offers the extenuation that prosecutors omitted the 5-year information not on the basis of race, but on stated opposition to the death pen*262alty, or ambivalence about it, on the questionnaires and in the voir dire testimony, id., at 34-35. On the State’s identification of black panel members opposed or ambivalent, all were asked the trick question.34 But the State’s rationale flatly fails to explain why most white panel members who expressed similar opposition or ambivalence were not subjected to it. It is entirely true, as the State argues, id., at 35, that prosecutors struck a number of nonblack members of the panel (as well as black members) for cause or by agreement before they reached the point in the standard voir dire sequence to question about minimum punishment. But this is no answer; 8 of the 11 nonblack individuals who voiced opposition or ambivalence were asked about the acceptable minimum only after being told what state law required.35 *263Hence, only 27% of nonblaeks questioned on the subject who expressed these views were subjected to the trick question, as against 100% of black members. Once again, the implication of race in the prosecutors’ choice of questioning cannot be explained away.36

There is a final body of evidence that confirms this conclusion. We know that for decades leading up to the time this case was tried prosecutors in the Dallas County office had followed a specific policy of systematically excluding blacks from juries, as we explained the last time the case was here.

*264“Although most of the witnesses [presented at the Swain hearing in 1986] denied the existence of a systematic policy to exclude African-Americans, others disagreed. A Dallas County district judge testified that, when he had served in the District Attorney’s Office from the late-1950’s to early-1960’s, his superior warned him that he would be fired if he permitted any African-Americans to serve on a jury. Similarly, another Dallas County district judge and former assistant district attorney from 1976 to 1978 testified that he believed the office had a systematic policy of excluding African-Americans from juries.

“Of more importance, the defense presented evidence that the District Attorney’s Office had adopted a formal policy to exclude minorities from jury service. ... A manual entitled ‘Jury Selection in a Criminal Case’ [sometimes known as the Sparling Manual] was distributed to prosecutors. It contained an article authored by a former prosecutor (and later a judge) under the direction of his superiors in the District Attorney’s Office, outlining the reasoning for excluding minorities from jury service. Although the manual was written in 1968, it remained in circulation until 1976, if not later, and was available at least to one of the prosecutors in Miller-El’s trial.” Miller-El v. Cockrell, 537 U. S., at 334-335.37

Prosecutors here “marked the race of each prospective juror on their juror cards.” Id., at S47.38

*265The Court of Appeals concluded that Miller-El failed to show by clear and convincing evidence that the state court’s finding of no discrimination was wrong, whether his evidence was viewed collectively or separately. 361 F. 3d, at 862. We find this conclusion as unsupportable as the “dismissive and strained interpretation” of his evidence that we disapproved when we decided Miller-El was entitled to a certificate of appealability. See Miller-El v. Cockrell, supra, at 344. It is true, of course, that at some points the significance of Miller-El’s evidence is open to judgment calls, but when this evidence on the issues raised is viewed cumulatively its direction is too powerful to conclude anything but discrimination.

In the course of drawing a jury to try a black defendant, 10 of the 11 qualified black venire panel members were peremptorily struck. At least two of them, Fields and Warren, were ostensibly acceptable to prosecutors seeking a death verdict, and Fields was ideal. The prosecutors’ chosen race-neutral reasons for the strikes do not hold up and are so far at odds with the evidence that pretext is the fair conclusion, indicating the very discrimination the explanations were meant to deny.

The strikes that drew these incredible explanations occurred in a selection process replete with evidence that the prosecutors were selecting and rejecting potential jurors because of race. At least two of the jury shuffles conducted by the State make no sense except as efforts to delay consideration of black jury panelists to the end of the week, when they might not even be reached. The State has in fact never offered any other explanation. Nor has the State denied that disparate lines of questioning were pursued: 53% of black panelists but only 3% of nonblacks were questioned with a graphic script meant to induce qualms about applying the death penalty (and thus explain a strike), and 100% of blacks but only 27% of nonblacks were subjected to a trick question about the minimum acceptable penalty for murder, *266meant to induce a disqualifying answer. The State’s attempts to explain the prosecutors’ questioning of particular witnesses on nonracial grounds fit the evidence less well than the racially discriminatory hypothesis.

If anything more is needed for an undeniable explanation of what was going on, history supplies it. The prosecutors took their cues from a 20-year-old manual of tips on jury selection, as shown by their notes of the race of each potential juror. By the time a jury was chosen, the State had peremptorily challenged 12% of qualified nonblack panel members, but eliminated 91% of the black ones.

It blinks reality to deny that the State struck Fields and Warren, included in that 91%, because they were black. The strikes correlate with no fact as well as they correlate with race, and they occurred during a selection infected by shuffling and disparate questioning that race explains better than any race-neutral reason advanced by the State. The State’s pretextual positions confirm Miller-El’s claim, and the prosecutors’ own notes proclaim that the Sparling Manual’s emphasis on race was on their minds when they considered every potential juror.

The state court’s conclusion that the prosecutors’ strikes of Fields and Warren were not racially determined is shown up as wrong to a clear and convincing degree; the state court’s conclusion was unreasonable as well as erroneous. The judgment of the Court of Appeals is reversed, and the case is remanded for' entry of judgment for petitioner together with orders of appropriate relief.

It is so ordered.

concurring.

In Batson v. Kentucky, 476 U. S. 79 (1986), the Court adopted a burden-shifting rule designed to ferret out the unconstitutional use of race in jury selection. In his separate opinion, Justice Thurgood Marshall predicted that the Court’s rule would not achieve its goal. The only way to *267“end the racial discrimination that peremptories inject into the jury-selection process,” he concluded, was to “eliminat[e] peremptory challenges entirely.” Id., at 102-103 (concurring opinion). Today’s case reinforces Justice Marshall’s concerns.

I

To begin with, this case illustrates the practical problems of proof that Justice Marshall described. As the Court’s opinion makes clear, Miller-El marshaled extensive evidence of racial bias. But despite the strength of his claim, Miller-El’s challenge has resulted in 17 years of largely unsuccessful and protracted litigation — including 8 different judicial proceedings and 8 different judicial opinions, and involving 23 judges, of whom 6 found the Batson standard violated and 16 the contrary.

The complexity of this process reflects the difficulty of finding a legal test that will objectively measure the inherently subjective reasons that underlie use of a peremptory challenge. Batson seeks to square this circle by (1) requiring defendants to establish a prima facie case of discrimination, (2) asking prosecutors then to offer a race-neutral explanation for their use of the peremptory, and then (3) requiring defendants to prove that the neutral reason offered is pretextual. See ante, at 239. But Batson embodies defects intrinsic to the task.

At Batson’s first step, litigants remain free to misuse peremptory challenges as long as the strikes fall below the prima facie threshold level. See 476 U. S., at 105 (Marshall, J., concurring). At Batson’s second step, prosecutors need only tender a neutral reason, not a “persuasive, or even plausible,” one. Purkett v. Elem, 514 U. S. 765, 768 (1995) (per curiam); see also id., at 766 (“‘mustaches and the beards look suspicious’”). And most importantly, at step three, Batson asks judges to engage in the awkward, sometime hopeless, task of second-guessing a prosecutor’s instinctive judgment — the underlying basis for which may be invisible *268even to the prosecutor exercising the challenge. See 476 U. S., at 106 (Marshall, J., concurring) (noting that the unconscious internalization of racial stereotypes may lead litigants more easily to conclude “that a prospective black juror is ‘sullen/ or ‘distant/” even though that characterization would not have sprung to mind had the prospective juror been white); see also Page, Batson’s Blind-Spot: Unconscious Stereotyping and the Peremptory Challenge, 85 B. U. L. Rev. 155, 161 (2005) (“ ‘[sjubtle forms of bias are automatic, unconscious, and unintentional’” and “‘escape notice, even the notice of those enacting the bias’” (quoting Fiske, What’s in a Category?: Responsibility, Intent, and the Avoidability of Bias Against Outgroups, in The Social Psychology of Good and Evil 127, 127-128 (A. Miller ed. 2004))). In such circumstances, it may be impossible for trial courts to discern if a “ ‘seat-of-the-pants’ ” peremptory challenge reflects a “ ‘seat-of-the-pants’ ” racial stereotype. Batson, 476 U. S., at 106 (Marshall, J., concurring) (quoting id., at 138 (Rehnquist, J., dissenting)).

Given the inevitably clumsy fit between any objectively measurable standard and the subjective decisionmaking at issue, I am not surprised to find studies and anecdotal reports suggesting that, despite Batson, the discriminatory use of peremptory challenges remains a problem. See, e. g., Baldus, Woodworth, Zuckerman, Weiner, & Broffitt, The Use of Peremptory Challenges in Capital Murder Trials: A Legal and Empirical Analysis, 3 U. Pa. J. Const. L. 3, 52-53, 73, n. 197 (2001) (in 317 capital trials in Philadelphia between 1981 and 1997, prosecutors struck 51% of black jurors and 26% of nonblack jurors; defense counsel struck 26% of black jurors and 54% of nonblack jurors; and race-based uses of prosecutorial peremptories declined by only 2% after Bat-son)', Rose, The Peremptory Challenge Accused of Race or Gender Discrimination? Some Data from One County, 23 Law and Human Behavior 695, 698-699 (1999) (in one North Carolina county, 71% of excused black jurors were removed *269by the prosecution; 81% of excused white jurors were removed by the defense); Tucker, In Moore’s Trials, Excluded Jurors Fit Racial Pattern, Washington Post, Apr. 2, 2001, p. Al (in D. C. murder case spanning four trials, prosecutors excused 41 blacks or other minorities and 6 whites; defense counsel struck 29 whites and 13 black venire members); Mize, A Legal Discrimination; Juries Aren’t Supposed to be Picked on the Basis of Race and Sex, But It Happens All the Time, Washington Post, Oct. 8, 2000, p. B8 (authored by judge on the D. C. Superior Court); see also Melilli, Batson in Practice: What We Have Learned About Batson and Peremptory Challenges, 71 Notre Dame L. Rev. 447, 462-464 (1996) (finding Batson challenges’ success rates lower where peremptor-ies were used to strike black, rather than white, potential jurors); Brand, The Supreme Court, Equal Protection and Jury Selection: Denying That Race Still Matters, 1994 Wis. L. Rev. 511, 583-589 (examining judicial decisions and concluding that few Batson challenges succeed); Note, Batson v. Kentucky and J. E. B. v. Alabama ex rel T. B.: Is the Peremptory Challenge Still Preeminent? 36 Boston College L. Rev. 161, 189, and n. 303 (1994) (same); Montoya, The Future of the Post -Batson Peremptory Challenge: Voir Dire by Questionnaire and the “Blind” Peremptory, 29 U. Mich. J. L. Reform 981, 1006, nn. 126-127, 1035 (1996) (reporting attorneys’ views on the difficulty of proving Batson claims).

II

Practical problems of proof to the side, peremptory challenges seem increasingly anomalous in our judicial system. On the one hand, the Court has widened and deepened Bat-son’s basic constitutional rule. It has applied Batson’s anti-discrimination test to the use of peremptories by criminal defendants, Georgia v. McCollum, 505 U. S. 42 (1992), by private litigants in civil cases, Edmonson v. Leesville Concrete Co., 500 U. S. 614 (1991), and by prosecutors where the defendant and the excluded juror are of different races, Powers *270v. Ohio, 499 U. S. 400 (1991). It has recognized that the Constitution protects not just defendants, but the jurors themselves. Id., at 409. And it has held that equal protection principles prohibit excusing jurors on account of gender. See J. E. B. v. Alabama ex rel. T. B., 511 U. S. 127 (1994). Some lower courts have extended Batson’s rule to religious affiliation as well. See, e. g., United States v. Brown, 352 F. 3d 654, 668-669 (CA2 2003); State v. Hodge, 248 Conn. 207, 244-246, 726 A. 2d 531, 553 (1999); United States v. Stafford, 136 F. 3d 1109, 1114 (CA7 1998) (suggesting same); see also Davis v. Minnesota, 511 U. S. 1115, 1117 (1994) (Thomas, J., dissenting from denial of certiorari). But see Casarez v. State, 913 S. W. 2d 468, 496 (Tex. Crim. App. 1994) (en banc) (declining to extend Batson to religious affiliation); State v. Davis, 504 N. W. 2d 767, 771 (Minn. 1993) (same).

On the other hand, the use of race- and gender-based stereotypes in the jury-selection process seems better organized and more systematized than ever before. See, e. g., Post, A Loaded Box of Stereotypes: Despite ‘Batson,’ Race, Gender Play Big Roles in Jury Selection., Nat. L. J., Apr. 25, 2005, pp. 1, 18 (discussing common reliance on race and gender in jury selection). For example, one jury-selection guide counsels attorneys to perform a “demographic analysis” that assigns numerical points to characteristics such as age, occupation, and marital status — in addition to race as well as gender. See V. Starr & M. McCormick, Jury Selection 193-200 (3d ed. 2001). Thus, in a hypothetical dispute between a white landlord and an African-American tenant, the authors suggest awarding two points to an African-American venire member while subtracting one point from her white counterpart. Id., at 197-199.

For example, a bar journal article counsels lawyers to “rate” potential jurors “demographically (age, gender, marital status, etc.) and mark who would be under stereotypical circumstances [their] natural enemies and allies.” Drake, *271The Art of Litigating: Deselecting Jurors Like the Pros, 34 Md. Bar J. 18, 22 (Mar./Apr. 2001) (emphasis in original).

For example, materials from a legal convention, while noting that “nationality” is less important than “once was thought,” and emphasizing that “the answers a prospective juror gives to questions are much more valuable,” still point out that “[sjtereotypically” those of “Italian, French, and Spanish” origin “are thought to be pro-plaintiff as well as other minorities, such as Mexican and Jewish[;] [pjersons of German, Scandinavian, Swedish, Finnish, Dutch, Nordic, British, Scottish, Oriental, and Russian origin are thought to be better for the defense”; African-Americans “have always been considered good for the plaintiff,” and “[m]ore politically conservative minorities will be more likely to lean toward defendants.” Blue, Mirroring, Proxemics, Nonverbal Communication, and Other Psychological Tools, Advocacy Track — Psychology of Trial, Association of Trial Lawyers of America Annual Convention Reference Materials, 1 Ann. 2001 ATLA-CLE 153, available at WESTLAW, ATLA-CLE database (June 8, 2005).

For example, a trial consulting firm advertises a new jury-selection technology: “Whether you are trying a civil case or a criminal case, SmartJURY™ has likely determined the exact demographics (age, race, gender, education, occupation, marital status, number of children, religion, and income) of the type of jurors you should select and the type you should strike.” SmartJURY Product Information, http://www.ets-america.com/smartjury_pi.asp (as visited June 8, 2005, and available in Clerk of Court’s case file).

These examples reflect a professional effort to fulfill the lawyer’s obligation to help his or her client. Cf. J. E. B., supra, at 148-149 (O’Connor, J., concurring) (observing that jurors’ race and gender may inform their perspective). Nevertheless, the outcome in terms of jury selection is the same as it would be were the motive less benign. And as long as that is so, the law’s antidiscrimination command and *272a peremptory jury-selection system that permits or encourages the use of stereotypes work at cross-purposes.

Finally, a jury system without peremptories is no longer unthinkable. Members of the legal profession have begun serious consideration of that possibility. See, e. g., Alen v. State, 596 So. 2d 1083, 1088-1089 (Fla. App. 1992) (Hubbart, J., concurring); Broderick, Why the Peremptory Challenge Should Be Abolished, 65 Temp. L. Rev. 369 (1992) (authored by Senior Judge on the U. S. District Court for the Eastern District of Pennsylvania); Hoffman, Peremptory Challenges Should be Abolished: A Trial Judge’s Perspective, 64 U. Chi. L. Rev. 809 (1997) (authored by a Colorado state-court judge); Alschuler, The Supreme Court and the Jury: Voir Dire, Peremptory Challenges, and the Review of Jury Verdicts, 56 U. Chi. L. Rev. 153, 199-211 (1989); Amar, Reinventing Juries: Ten Suggested Reforms, 28 U. C. D. L. Rev. 1169, 1182-1183 (1995); Melilli, 71 Notre Dame L. Rev., at 502-503; Page, 85 B. U. L. Rev., at 245-246. And England, a common-law jurisdiction that has eliminated peremptory challenges, continues to administer fair trials based largely on random jury selection. See Criminal Justice Act, 1988, ch. 33, § 118(1), 22 Halsbury’s Statutes 357 (4th ed. 2003 reissue) (U. K.); see also 2 Jury Service in Victoria, Final Report, ch. 5, p. 165 (Dec. 1997) (1993 study of English barristers showed majority support for system without peremptory challenges).

Ill

I recognize that peremptory challenges have a long historical pedigree. They may help to reassure a party of the fairness of the jury. But long ago, Blackstone recognized the peremptory challenge as an “arbitrary and capricious species of [a] challenge.” 4 W. Blackstone, Commentaries on the Laws of England 346 (1769). If used to express stereotypical judgments about race, gender, religion, or national origin, peremptory challenges betray the jury’s democratic origins and undermine its representative function. See 1 A. de *273Tocqueville, Democracy in America 287 (H. Reeve transl., rev. ed. 1900) (“[T]he institution of the jury raises the people ... to the bench of judicial authority [and] invests [them] with the direction of society”); A. Amar, The Bill of Rights 94-96 (1998) (describing the Founders’ vision of juries as venues for democratic participation); see also Stevens, Foreword, Symposium: The Jury at a Crossroad: The American Experience, 78 Chi.-Kent L. Rev. 907,907-908 (2008) (citizens should not be denied the opportunity to serve as jurors unless an impartial judge states a reason for the denial, as with a strike for cause). The “scientific” use of peremptory challenges may also contribute to public cynicism about the fairness of the jury system and its role in American government. See, e. g., S. O’Connor, Juries: They May Be Broke, But We Can Fix Them, Chautauqua Institution Lecture, July 6,1995. And, of course, the right to a jury free of discriminatory taint is constitutionally protected — the right to use peremptory challenges is not. See Stilson v. United States, 250 U. S. 583, 586 (1919); see also Ross v. Oklahoma, 487 U. S. 81, 88 (1988) (defendant’s loss of a peremptory challenge does not violate his right to an impartial jury).

Justice Goldberg, dissenting in Swain v. Alabama, 380 U. S. 202 (1965), wrote, ‘Were it necessary to make an absolute choice between the right of a defendant to have a jury chosen in conformity with the requirements of the Fourteenth Amendment and the right to challenge peremptorily, the Constitution compels a choice of the former.” Id., at 244; see also Batson, 476 U. S., at 107 (Marshall, J., concurring) (same); Edmonson, 500 U. S., at 630 (opinion for the Court by Kennedy, J.) (“[I]f race stereotypes are the price for acceptance of a jury panel as fair, the price is too high to meet the standard of the Constitution”). This case suggests the need to confront that choice. In light of the considerations I have mentioned, I believe it necessary to reconsider Batson’s test and the peremptory challenge system as a whole. With that qualification, I join the Court’s opinion.

dissenting.

In the early morning hours of November 16, 1985, petitioner Thomas Joe Miller-El and an accomplice, Kennard Flowers, robbed a Holiday Inn in Dallas, Texas. Miller-El and Flowers bound and gagged hotel employees Donald Hall and Doug Walker, and then laid them face down on the floor. When Flowers refused to shoot them, Miller-El shot each twice in the back, killing Walker and rendering Hall a paraplegic. Miller-El was convicted of capital murder by a jury composed of seven white females, two white males, a black male, a Filipino male, and a Hispanic male.

For nearly 20 years now, Miller-El has contended that prosecutors peremptorily struck potential jurors on the basis of race. In that time, seven state and six federal judges have reviewed the evidence and found no error. This Court concludes otherwise, because it relies on evidence never presented to the Texas state courts. That evidence does not, much less “clear[ly] and convincingly],” show that the State racially discriminated against potential jurors. 28 U. S. C. § 2254(e)(1). However, we ought not even to consider it: In deciding whether to grant Miller-El relief, we may look only to “the evidence presented in the State court proceeding.” § 2254(d)(2). The majority ignores that restriction on our review to grant Miller-El relief. I respectfully dissent.

I

Miller-El requests federal habeas relief from a state-court judgment, and hence our review is controlled by the Antiter-rorism and Effective Death Penalty Act of 1996 (AEDPA), 110 Stat. 1214. Because Miller-El’s claim of racial discrimination in jury selection was adjudicated on the merits in Texas state court, AEDPA directs that a writ of habeas corpus “shall not be granted” unless the state court’s decision “was based on an unreasonable determination of the facts in *275light of the evidence presented in the State court proceeding” 28 U. S. C. § 2254(d)(2) (emphasis added).

To obtain habeas relief, then, Miller-El must show that, based on the evidence before the Texas state courts, the only reasonable conclusion was that prosecutors had racially discriminated against prospective jurors. He has not even come close to such a showing. The state courts held two hearings, but despite ample opportunity, Miller-El presented little evidence that discrimination occurred during jury selection. In view of the evidence actually presented to the Texas courts, their conclusion that the State did not discriminate was eminently reasonable. As a close look at the state-court proceedings reveals, the majority relies almost entirely on evidence that Miller-El has never presented to any Texas state court.

A

Jury selection in Miller-El’s trial took place over five weeks in February and March 1986. During the process, 19 of the 20 blacks on the 108-person venire panel were not seated on the jury: 3 were dismissed for cause, 6 were dismissed by the parties’ agreement, and 10 were peremptorily struck by prosecutors. Miller-El objected to 8 of these 10 strikes, asserting that the prosecutors were discriminating against black veniremen. Each time, the prosecutors proffered a race-neutral, case-related reason for exercising the challenge, and the trial court permitted the venireman to be removed. The remaining black venireman, Troy Woods, served on the jury that convicted Miller-El.

At the completion of voir dire, Miller-El moved to strike the jury under this Court’s decision in Swain v. Alabama, 380 U. S. 202 (1965), which required Miller-El to prove “systematic exclusion of black persons through the use of pe-remptories over a period of time.” Powers v. Ohio, 499 U. S. 400, 405 (1991). At the pretrial Swain hearing in March 1986, Miller-El presented three types of documentary evidence: the juror questionnaires of the 10 black veniremen *276struck by the State; excerpts from a series of newspaper articles on racial bias in jury selection; and a manual on jury selection in criminal cases authored by a former Dallas County prosecutor. The voir dire transcript was part of the official record. Miller-El, however, introduced none of the other 98 juror questionnaires, no juror cards, and no evidence related to jury shuffling. See ante, at 256-257, n. 15.

Miller-El also presented nine witnesses, five of whom had spent time as prosecutors in the Dallas County District Attorney’s (D. A.) Office and five of whom were current or former judges in Dallas County. Their testimony made three things clear. First, the D. A.’s Office had never officially sanctioned or promoted racial discrimination in jury selection, as several witnesses testified, including the county’s Chief Public Defender as well as one of the first black prosecutors to serve in the D. A.’s Office. App. 842 (Baraka); id., at 846-848 (Tait); id., at 860 (Entz); id., at 864 (Kinkeade). Second, witnesses testified that, despite the absence of any official policy, individual prosecutors had almost certainly excluded blacks in particular cases. Id., at 880, 833 (Hampton); id., at 841-842 (Baraka); id., at 846-848 (Tait); id., at 863-864 (Kinkeade). Third and most important, no witness testified that the prosecutors in Miller-El’s trial — Norman Kinne, Paul Macaluso, and Jim Nelson — had ever engaged in racially discriminatory jury selection. Id., at 843 (Baraka); id., at 859 (Entz); id., at 863 (Kinkeade). The trial court concluded that, although racial discrimination “may have been done by individual prosecutors in individual cases,” there was no evidence of “any systematic exclusion of blacks as a matter of policy by the District Attorney’s office.” Id., at 882-883.

Miller-El was then tried, convicted, and sentenced to death. While his appeal was pending, this Court decided Batson v. Kentucky, 476 U. S. 79 (1986). Batson announced a new three-step process for evaluating claims that a prosecutor used peremptory challenges to strike prospective jurors because of their race:

*277“First, a defendant must make a prima facie showing that a peremptory challenge has been exercised on the basis of race[; s]econd, if that showing has been made, the prosecution must offer a race-neutral basis for striking the juror in question!; and t]hird, in light of the parties’ submissions, the trial court must determine whether the defendant has shown purposeful discrimination.” Miller-El v. Cockrell, 537 U. S. 322, 328-329 (2003) (Miller-El I).

The Texas Court of Criminal Appeals remanded Miller-El’s case for a hearing to be held under Batson.

B

At the Batson hearing in May 1988, before the same judge who had presided over his trial, Miller-El sought to establish that prosecutors at his trial had struck potential jurors on the basis of their race. To make his prima facie case, Miller-El reintroduced some of what he had presented two years earlier at the Swain hearing: the testimony of the nine witnesses, the 10 juror questionnaires, and the excerpted newspaper articles. App. 893-895. The court instructed the State to explain its strikes. Id., at 898-899. Of the 10 peremptory strikes at issue, prosecutors had already explained 8 at trial in response to Miller-El’s objections. The State therefore called Paul Macaluso, one of the prosecutors who had conducted the voir dire, to testify regarding his reasons for striking veniremen Paul Bailey and Joe Warren.

Macaluso testified that he had struck Bailey because Bailey seemed firmly opposed to the death penalty, even though Bailey tempered his stance during voir dire. Id., at 905-906. This was accurate. Bailey expressed forceful opposition to the death penalty when questioned by Macaluso. See, e. g., 11-(A) Record of Voir Dire 4110 (hereinafter Record) (“I don’t believe in capital punishment. Like I said on [my juror questionnaire], I don’t believe anyone has the right to take another person’s life”); id., at 4112 (saying that he *278felt “[v]ery strongly” that the State should not impose the death penalty). Later, however, when questioned by defense counsel, Bailey said that he could impose the death penalty if the State proved the necessary aggravating circumstances. Id., at 4148-4150, 4152. When the trial court overruled the State’s challenge for cause, the State exercised a peremptory challenge. Id., at 4168.

Macaluso next testified that he dismissed venireman Warren because Warren gave inconsistent answers regarding his ability to apply the death penalty and because Warren’s brother had been recently convicted. App. 908-910. Maca-luso conceded that Warren was not as clearly unfavorable to the State as Bailey. Id., at 911. Nevertheless, Macaluso struck Warren because it was early in the jury selection process and the State had plenty of remaining peremptories with which it could remove marginal jurors. Macaluso candidly stated that he might not have removed Warren if fewer peremptories had been available. Id., at 910.

After the State presented nonracial, case-related reasons for all its strikes, the focus shifted to Batson’s third step: whether Miller-El had “carried his burden of proving purposeful discrimination.” Purkett v. Elem, 514 U. S. 765, 768 (1995) (per curiam); Batson, supra, at 97-98. At this point, Miller-El stood on his Swain evidence. App. 921. That evidence bore on whether some Balias County prosecutors had discriminated generally in past years; none of the evidence indicated that the prosecutors at Miller-El’s trial — Kinne, Macaluso, and Nelson — had discriminated in the selection of Miller-El’s jury. Moreover, none of this generalized evidence came close to demonstrating that the State’s explanations were pretextual in Miller-El’s particular trial.. Miller-El did not even attempt to rebut the State’s racially neutral reasons at the hearing. He presented no evidence and made no arguments. Id., at 919-922.

Nevertheless, the majority concludes that the trial judge was unreasonable in finding as a factual matter that the *279State did not discriminate against black veniremen. Ante, at 266. That is not so “in light of the evidence presented in the State court proceeding.” 28 U. S. C. § 2254(d)(2). From the scanty evidence presented to the trial court, “it is at least reasonable to conclude” that purposeful discrimination did not occur, “which means that the state court’s determination to that effect must stand.” Early v. Packer, 537 U. S. 3, 11 (2002) (per curiam).

II

Not even the majority is willing, to argue that the evidence before the state court shows that the State discriminated against black veniremen. Instead, it bases its decision on juror questionnaires and juror cards that Miller-El’s new attorneys unearthed during his federal habeas proceedings and that he never presented to the state courts.1 Ante, at 256-257, n. 15. Worse still, the majority marshals those documents in support of theories that Miller-El never argued to the state courts. AEDPA does not permit habeas petitioners to engage in this sort of sandbagging of state courts.

A

The majority discusses four types of evidence: (1) the alleged similarity between black veniremen who were struck by the prosecution and white veniremen who were not; (2) the apparent disparate questioning of black and white veniremen with respect to their views on the death penalty and their ability to impose the minimum punishment; (3) the use of the “jury shuffle” by the prosecution; and (4) evidence of historical discrimination by the D. A.’s Office in the selection of juries. Only the last was ever put before the Texas courts — and it does not prove that any constitutional viola*280tion occurred at Miller-El’s trial. The majority’s discussion of the other types of evidence relies on documents like juror questionnaires and juror cards that were added to the record before the District Court.

The majority’s willingness to reach outside the state-court record and embrace evidence never presented to the Texas state courts is hard to fathom. AEDPA mandates that the reasonableness of a state court’s factual findings be assessed “in light of the evidence presented in the State court proceeding,” 28 U. S. C. § 2254(d)(2), and also circumscribes the ability of federal habeas litigants to present evidence that they “failed to develop” before the state courts. § 2254(e)(2); Williams v. Taylor, 529 U. S. 420, 429-430 (2000). Miller-El did not argue disparate treatment or disparate questioning at the Batson hearing, so he had no reason to submit the juror questionnaires or cards to the trial court. However, Miller-El could have developed and presented all of that evidence at the Batson hearing.2 Consequently, he must satisfy § 2254(e)(2)’s requirements to adduce the evidence in federal court — something he cannot do. Williams, supra, at 437 (“Federal courts sitting in habeas are not an alternative forum for trying facts and issues which a prisoner made insufficient effort to pursue in state proceedings”). For instance, there is no doubt that Miller-El’s supplemental material could have been “previously discovered through the exercise of due diligence.” § 2254(e)(2)(A)(ii).

Just last Term, we summarily reversed the Court of Appeals for the Sixth Circuit for doing what the Court does *281here: granting habeas relief on the basis of evidence not presented to the state court. See Holland v. Jackson, 542 U. S. 649, 652 (2004) (per curiam). We reaffirmed “that whether a state court’s decision was unreasonable must be assessed in light of the record the court had before it.” Ibid.; see also Miller-El I, 537 U. S., at 348 (“[Petitioner must demonstrate that a state court’s . . . factual determination was ‘objectively unreasonable’ in light of the record before the court”). In an about-face, the majority now reverses the Court of Appeals for the Fifth Circuit for failing to grant habeas relief on the basis of evidence not before the state court. By crediting evidence that Miller-El never placed before the state courts, the majority flouts AEDPA’s plain terms and encourages habeas applicants to attack state judgments collaterally with evidence never tested by the original triers of fact.

B

The majority presents three arguments for ignoring AEDPA’s requirement that the state-court decision be unreasonable “in light of the evidence presented in the State court proceeding.” 28 U. S. C. § 2254(d)(2). None is persuasive.

1

First, without briefing or argument on the question, the majority hints that we may ignore AEDPA’s limitation on the record under § 2254(d)(2) because the parties have ignored it. Ante, at 256-257, n. 15. The majority then quickly retreats and expressly does not decide the question. Ibid. But its retreat is as inexplicable as its advance: Unless § 2254(d)(2) is waivable and the parties have waived it, the majority cannot consider evidence outside the state-court proceedings, as it concededly does.

The majority’s venture beyond the state-court record is indefensible. Even if § 2254(d) is not jurisdictional, but see Lindh v. Murphy, 521 U. S. 320, 343-344 (1997) (Rehnquist, *282C. J., dissenting), “it shares the most salient characteristic of jurisdictional statutes: Its commands are addressed to courts rather than to individuals,” id., at 344. Section 2254(d) speaks directly to federal courts when it states that a habeas application by a state prisoner “shall not be granted” except under the specified conditions. (Emphasis added); ibid. The strictures of § 2254(d) are not discretionary or waivable. Through AEDPA, Congress sought to ensure that federal courts would defer to the judgments of state courts, not the wishes of litigants.

Nevertheless, there is no need to decide whether § 2254(d)(2) may be waived, for the State has not waived it. Contrary to the majority’s assertions, ante, at 256-257, n. 15, the State has argued that § 2254(d)(2) bars our review of certain evidence not before the state trial court, Brief for Respondent 41-42, just as it did in its last appearance, see Brief for Respondent in Miller-El I, O. T. 2002, No. 01-7662, pp. 28-29, 39. The majority is correct that the State has not argued § 2254(d)(2) precludes consideration of the juror questionnaires and juror cards in particular, ante, at 256-257, n. 15, but the majority does not assert that the State may selectively invoke § 2254(d)(2) to cherry-pick only favorable evidence that lies outside the state-court record.

2

The majority next suggests that the supplemental material, particularly the juror questionnaires, might not expand on what the state trial court knew, since “the same judge presided over the voir dire, the Swain hearing, and the Bat-son hearing, and the jury questionnaires were subjects of reference at the voir dire.” Ante, at 257, n. 15. This is incorrect. At the Batson hearing, Miller-El introduced into evidence only the questionnaires of the 10 black veniremen peremptorily struck by the State. App. 893-895. The questionnaires of the other 98 veniremen — including many on which the majority relies — were never introduced into ev*283idence or otherwise placed before the trial judge. Miller-El and the State had copies; the trial judge did not.

Yet the majority insinuates that the questionnaires effectively were before the state court because they “were subjects of reference at the voir dire.” Ante, at 257, n. 15. That is extremely misleading on the facts of this case. Although counsel for Miller-El and the State questioned witnesses partially on the basis of their questionnaire responses, the lawyers’ references to questionnaires were scattered and sporadic. Even the majority does not attempt to show that the specific questionnaire responses on which it relies were called to the trial court’s attention. Clearly they were not called to the trial court’s attention at the only time that mattered: the Batson hearing.

The majority’s insinuation is doubly misleading when coupled with its insistence that “the transcript of voir dire . .. was before the state courts.” Ante, at 242, n. 2. Miller-El’s arguments gave the state court no reason to go leafing through the voir dire transcript. What is more, voir dire at Miller-El’s trial lasted five weeks, and the transcript occupies 11 volumes numbering 4,662 pages. To think that two years after the fact a trial court should dredge up on its own initiative passing references to unseen questionnaires — references buried in a more than 4,600-page transcript no less — is unrealistic. That is why § 2254(d)(2) demands that state courts be taken to task only on the basis of evidence “presented in the State court proceeding.” The 98 questionnaires before the parties, unlike the 10 questionnaires that Miller-El entered into evidence, were not “presented” to the state court.

The majority also asserts that by considering the questionnaires, it is only attempting to help the State. After all, the State claims that any disparate questioning and treatment of black and white veniremen resulted from their questionnaires, not their respective races. As the majority sees it, if the questionnaires are not properly before us, then the State cannot substantiate its defense.

*284This is a startling repudiation of both Batson and AEDPA. A strong presumption of validity attaches to a trial court’s factual finding at Batson’s third step, Hernandez v. New York, 500 U. S. 352, 364 (1991) (plurality opinion); id., at 372 (O’Connor, J., concurring in judgment); see also Batson, 476 U. S., at 98, n. 21, and that presumption is doubly strong when the Batson finding is under collateral attack in habeas, Miller-El I, 537 U. S., at 340. Thus, it is Miller-El’s burden to prove racial discrimination under Batson, and it is his burden to prove it by clear and convincing evidence under AEDPA. Without the questionnaires never submitted to the trial court, Miller-El comes nowhere near establishing that race motivated any disparate questioning or treatment, which is precisely why the majority must strain to include the questionnaires within the state-court record.

That Miller-El needs the juror questionnaires could not be clearer in light of how the Batson hearing unfolded. After offering racially neutral reasons for all of its strikes, the State could have remained silent — as Miller-El did. However, the State pointed out, among other things, that any disparate questioning of black and white veniremen was based on answers given on the juror questionnaires or during the voir dire process. App. 920-921. The State further noted that Miller-El had never alleged disparate treatment of black and white veniremen. Id., at 921. Because Miller-El did not dispute the State’s assertions, there was no need for the State to enter the juror questionnaires into the record. There was nothing to argue about. Miller-El had presented only generalized evidence of historical discrimination by the D. A.’s Office, which no one believes was sufficient in itself to prove a Batson violation. That is why Miller-El, not the State, marshaled supplemental material during his federal habeas proceedings. Without that evidence, he cannot prove now what he never attempted to prove 17 years ago: that the State’s justifications for its strikes were a pretext for discrimination.

*2853

Finally, the majority suggests that the 2-year delay between the voir dire and the post-trial Batson hearing is reason for weakened deference. See ante, at 241, n. 1. This is an argument not for setting aside § 2254(d)(2)’s limit on the record, but for relaxing the level of deference due state courts’ factual findings under §§ 2254(d)(2) and (e)(1). The presumption of correctness afforded factual findings on ha-beas review, however, does not depend on the manner in which the trial court reaches its factual findings, for reasons I have explained before. Miller-El I, supra, at 357-359 (dissenting opinion). The majority leaves those arguments unanswered.

The majority’s own argument is implausible on its face: “ ‘[T]he usual risks of imprecision and distortion from the passage of time’ ” are far greater after 17 years than after 2. Ante, at 241, n. 1 (quoting Miller-El I, supra, at 343). The majority has it just backward. The passage of time, as AEDPA requires and as this Court has held, counsels in favor of more deference, not less. At least the trial court, unlike this Court, had the benefit of gauging the witnesses’ and prosecutors’ credibility at both the Swain and Batson hearings. Miller-El 1, supra, at 339 (“Deference is necessary because a reviewing court, which analyzes only the transcripts from voir dire, is not as well positioned as the trial court is to make credibility determinations”); see also Hernandez, supra, at 364 (plurality opinion); Batson, supra, at 98, n. 21.

Ill

Even taken on its own terms, Miller-El’s cumulative evidence does not come remotely close to clearly and convincingly establishing that the state court’s factual finding was unreasonable. I discuss in turn Miller-El’s four types of evidence: (1) the alleged disparate treatment and (2) disparate questioning of black and white veniremen; (3) the prosecu*286tion’s jury shuffles; and (4) historical discrimination by the D. A.’s Office in the selection of juries. Although each type of evidence “is open to judgment calls,” ante, at 265, the majority finds that a succession of unpersuasive arguments amounts to a compelling case. In the end, the majority’s opinion is its own best refutation: It strains to demonstrate what should instead be patently obvious.

A

The majority devotes the bulk of its opinion to a side-by-side comparison of white panelists who were allowed to serve and two black panelists who were struck, Billy Jean Fields and Joe Warren. Ante, at 240-252. The majority argues that the prosecution’s reasons for striking Fields and Warren apply equally to whites who were permitted to serve, and thus those reasons must have been pretextual. The voir dire transcript reveals that the majority is mistaken.

It is worth noting at the outset, however, that Miller-El’s and the Court’s claims have always been a moving target. Of the 20 black veniremen at Miller-El’s trial, 9 were struck for cause or by the parties’ agreement, and 1 served on the jury. Miller-El claimed at the Batson hearing that all 10 remaining black veniremen were dismissed on account of race. That number dropped to 7 on appeal, and then again to 6 during his federal habeas proceedings. Of those 6 black veniremen, this Court once found debatable that the entire lot was struck based on race. Miller-El I, supra, at 343. However, 4 (Carrol Boggess, Roderick Bozeman, Wayman Kennedy, and Edwin Rand) were dismissed for reasons other than race, as the majority effectively concedes. Ante, at 252-253, n. 11; Miller-El I, supra, at 351-354 (SCALIA, J., concurring).

The majority now focuses exclusively on Fields and Warren. But Warren was obviously equivocal about the death penalty. In the end, the majority’s case reduces to a single venireman, Fields, and its reading of a 20-year-old voir dire *287transcript that is ambiguous at best. This is the antithesis of clear and convincing evidence.

1

From the outset of questioning, Warren did not specify when he would vote to impose the death penalty. When asked by prosecutor Paul Macaluso about his ability to impose the death penalty, Warren stated, “[T]here are some cases where I would agree, you know, and there are others that I don’t.” 3 Record 1526. Macaluso then explained at length the types of crimes that qualified as capital murder under Texas law, and asked whether Warren would be able to impose the death penalty for those types of heinous crimes. Id., at 1527-1530. Warren continued to hedge: “I would say it depends on the case and the circumstances involved at the time.” Id., at 1530. He offered no sense of the circumstances that would lead him to conclude that the death penalty was an appropriate punishment.

Macaluso then changed tack and asked whether Warren believed that the death penalty accomplished any social purpose. Id., at 1531-1532. Once again, Warren proved impossible to pin down: “Yes and no. Sometimes I think it does and sometimes I think it don’t. Sometimes you have mixed feelings about things like that.” Id., at 1532. Macaluso then focused on what the death penalty accomplished in those cases where Warren believed it useful. Ibid. Even then, Warren expressed no firm view:

“I don’t know. It’s really hard to say because I know sometimes you feel that it might help to deter crime and then you feel that the person is not really suffering. You’re taking the suffering away from him. So it’s like I said, sometimes you have mixed feelings about whether or not this is punishment or, you know, you’re relieving personal punishment.” Ibid.

While Warren’s ambivalence was driven by his uncertainty that the death penalty was severe enough, ante, at 250-251, *288that is beside the point. Throughout the examination, Warren gave no indication whether or when he would prefer the death penalty to other forms of punishment, specifically life imprisonment. 3 Record 1532-1533. To prosecutors seeking the death penalty, the reason for Warren’s ambivalence was irrelevant.

At voir dire, there was no dispute that the prosecution struck Warren not for his race, but for his ambivalence on the death penalty. Miller-El’s attorneys did not object to the State’s strikes of Warren or Paul Bailey, though they objected to the removal of every other black venireman. Both Bailey and Warren shared the same characteristic: It was not clear, based on their questionnaires and voir dire testimony, that they could impose the death penalty. See supra, at 277-278. In fact, Bailey was so clearly struck for nonracial reasons that Miller-El has never objected to his removal at any stage in this case. ■

There also was no question at the Batson hearing why the prosecution struck Warren. Macaluso testified:

“I thought ¡Warren’s statements on voir dire] were inconsistent responses. At one point he says, you know, on a case-by-case basis and at another point he said, well, I think — I got the impression, at least, that he suggested that the death penalty was an easy way out, that they should be made to suffer more.’’ App. 909.

In addition, Macaluso noted that Warren’s brother recently had been convicted for a crime involving food stamps. Id., at 909-910. This suggested that Warren might be more sympathetic to defendants than other jurors. Macaluso was quite candid that Warren was not as obviously disfavorable to the State as Bailey, and Macaluso stated that he might not have exercised a peremptory against Warren later in jury selection. Id., at 910-911. But Macaluso used only his 6th of 15 peremptory challenges against Warren.

*289According to the majority, Macaluso testified that he struck Warren for his statement that the death penalty was “ ‘an easy way out,’ ” ante, at 248 (quoting App. 909), and not for his ambivalence about the death penalty, ante, at 250-251. This grossly mischaracterizes the record. Macaluso specifically testified at the Batson hearing that he was troubled by the “inconsisten[cy]” of Warren’s responses. App. 909 (emphasis added). Macaluso was speaking of Warren’s ambivalence about the death penalty, a reason wholly unrelated to race. This was Macaluso’s “stated reason,” and Macaluso ought to “stand or fall on the plausibility” of this reason — not one concocted by the majority. Ante, at 252.

The majority points to four other panel members — Kevin Duke, Troy Woods, Sandra Jenkins, and Leta Girard — who supposedly expressed views much like Warren’s, but who were not struck by the State. Ante, at 248. According to the majority, this is evidence of pretext. But the majority’s premise is faulty. None of these veniremen was as difficult to pin down on the death penalty as Warren. For instance, Duke supported the death penalty. App. 373 (“I’ve always believed in having the death penalty. I think it serves a purpose”); ibid. (“I mean, it’s a sad thing to see, to have to kill someone, but they shouldn’t have done the things that they did. Sometimes they deserve to be killed”); id., at 394 (“If I feel that I can answer all three of these [special-issue] questions yes and I feel that he’s done a crime worthy of the death penalty, yes, I will give the death penalty”). By contrast, Warren never expressed a firm view one way or the other.

Troy Woods, who was black and who served on the jury, was even more supportive of the death penalty than Duke. The majority suggests that prosecutors might have allowed Woods to serve on the jury because they were running low on peremptories or they wanted to obscure a pattern of discrimination. Ante, at 249-250. That such rank conjecture *290can serve as “clear and convincing evidence” is error in its own right, but it is also belied by the record. Woods said that capital punishment was “too quick” because defendants “don’t feel the pain.” App. 409. When asked what sort of punishment defendants ought to receive, Woods said that he would “[p]our some honey on them and stake them out over an ant bed.” Ibid. He testified that he would mete out such sentences because if defendants “survive for a length of time, that would be enough punishment and . . . they wouldn’t do it again.” Id., at 410 (alteration omitted). Woods also testified that he was a lifelong believer in the death penalty, id., at 410-411; that he could impose death generally as a juror, id., at 413; and that he could impose death for murder during the course of a robbery, the specific crime of which Miller-El stood accused, ibid. It is beyond cavil why the State accepted Woods as a juror: He could impose the punishment sought by the State.

Nevertheless, even assuming that any of these veniremen expressed views similar to Warren’s, Duke, Woods, and Gi-rard were questioned much later in the jury selection process, when the State had fewer peremptories to spare. Only Sandra Jenkins was questioned early in the voir dire process, and thus only Jenkins was even arguably similarly situated to Warren. However, Jenkins and Warren were different in importánt respects. Jenkins expressed no doubt whatsoever about the death penalty. She testified that she had researched the death penalty in high school, and she said in response to questioning by both parties that she strongly believed in the death penalty’s value as a deterrent to crime. 3 Record 1074-1075, 1103-1104. This alone explains why the State accepted Jenkins as a juror, while Miller-El struck her. In addition, Jenkins did riot have a relative who had been convicted of a crime, but Warren did. At the Batson hearing, Macaluso testified that he struck Warren both for Warren’s inconsistent responses regarding the death penalty and for his brother’s conviction. Supra, at 278.

*291The majority thinks it can prove pretext by pointing to white veniremen who match only one of the State’s proffered reasons for striking Warren. Ante, at 248. This defies logic. “ ‘Similarly situated’ does not mean matching any one of several reasons the prosecution gave for striking a potential juror — it means matching all of them.” Miller-El I, 537 U. S., at 362-363 (Thomas, J., dissenting); cf. Newport News Shipbuilding & Dry Dock Co. v. EEOC, 462 U. S. 669, 683 (1983) (Title VII of the Civil Rights Act of 1964 discrimination occurs when an employee is treated ““‘in a manner which but for that person’s sex would be different” ’ ” (quoting Los Angeles Dept. of Water and Power v. Manhart, 435 U. S. 702, 711 (1978))). Given limited peremptories, prosecutors often must focus on the potential jurors most likely to disfavor their case. By ignoring the totality of reasons that a prosecutor strikes any particular venireman, it is the majority that treats potential jurors as “products of a set of cookie cutters,” ante, at 247, n. 6 — as if potential jurors who share only some among many traits must be treated the same to avoid a Batson violation. Of course jurors must not be “identical in all respects” to gauge pretext, ante, at 247, n. 6, but to isolate race as a variable, the jurors must be comparable in all respects that the prosecutor proffers as important. This does not mean “that a defendant cannot win a Batson claim unless there is an exactly identical white juror.” Ibid. It means that a defendant cannot support a Batson claim by comparing veniremen of different races unless the veniremen are truly similar.

2

The second black venireman on whom the majority relies is Billy Jean Fields. Fields expressed support for the death penalty, App. 174-175, but Fields also expressed views that called into question his ability to impose the death penalty. Fields was a deeply religious man, id., at 173-174, 192-194, and prosecutors feared that his religious convictions might make him reluctant to impose the death penalty. Those *292fears were confirmed by Fields’ view that all people could be rehabilitated if introduced to God, a fear that had special force considering the special-issue questions necessary to impose the death penalty in Texas. One of those questions asked whether there was a probability that the defendant would engage in future violence that threatened society. When they reached this question, Macaluso and Fields had the following exchange:

“[MACALUSO:] What does that word probability mean to you in that connotation?

“[FIELDS:] Well, it means is there a possibility that [a defendant] will continue to lead this type of life, will he be rehabilitated or does he intend to make this a lifelong ambition.

“[MACALUSO:] Let me ask you, Mr. Fields, do you feel as though some people simply cannot be rehabilitated? “[FIELDS:] No.

“[MACALUSO:] You think everyone can be rehabilitated?

“[FIELDS:] Yes.” Id., at 183-184.

Thus, Fields indicated that the possibility of rehabilitation was ever-present and relevant to whether a defendant might commit future acts of violence. In light of that view, it is understandable that prosecutors doubted whether he could vote to impose the death penalty.

Fields did testify that he could impose the death penalty, even on a defendant who could be rehabilitated. Id., at 185. For the majority, this shows that the State’s reason was pre-textual. Ante, at 244. But of course Fields said that he could fairly consider the death penalty — if he had answered otherwise, he would have been challengeable for cause. The point is that Fields’ earlier answers cast significant doubt on whether he could impose the death penalty. The very purpose of peremptory strikes is to allow parties to *293remove potential jurors whom they suspect, but cannot prove, may exhibit a particular bias. See Swain, 380 U. S., at 220; J. E. B. v. Alabama ex rel. T. B., 511 U. S. 127, 148 (1994) (O’Connor, J., concurring). Based on Fields’ voir dire testimony, it was perfectly reasonable for prosecutors to suspect that Fields might be swayed by a penitent defendant’s testimony.3 The prosecutors may have been worried for nothing about Fields’ religious sentiments, but that does not mean they were instead worried about Fields’ race.

As with Warren, the majority attempts to point to similarly situated nonblack veniremen who were not struck by the State, but its efforts again miss their mark for several reasons. First, the majority would do better to begin with white veniremen who were struck by the State. For instance, it skips over Penny Crowson, a white panelist who expressed a firm belief in the death penalty, but who also stated that she probably would not impose the death penalty if she believed there was a chance the defendant could be rehabilitated. Ante, at 245, n. 5; 3 Record 1211. The State struck Crowson, which demonstrates that it “was concerned *294about views on rehabilitation when the venireperson was not black.” Ante, at 245, n. 4.

Second, the nonblack veniremen to whom the majority points — Sandra Hearn, Mary Witt, and Fernando Gutierrez — were more favorable to the State than Fields for various reasons.4 For instance, Sandra Hearn was adamant about the value of the death penalty for callous crimes. App. 430, 451-452. Miller-El, of course, shot in cold blood two men who were lying before him bound and gagged. In addition, Hearn’s father was a special agent for the Federal Bureau of Investigation, and her job put her in daily contact with police officers for whom she expressed the utmost admiration. Id., at 445-446, 457-460. This is likely why the State accepted Hearn and Miller-El challenged her for cause. Id., at 447, 467.

In fact, on appeal Miller-El’s counsel had this to say about Hearn: “If ever — if ever — there was a Venireperson that, should have been excluded for cause from the Jury in this case, or any capital Murder Jury, it was Venirewoman HEARN. It is hoped that the Lord will save us from future jurors with her type of thinking and beliefs.” Id., at 1015 (emphasis added and alteration omitted); see also id., at 1010. This same juror whom Miller-El’s counsel once found so repugnant has been transformed by the majority’s revisionist history into a defense-prone juror just as objectionable to the State as Fields. Ante, at 244.

*295Mary Witt did not even have the same views on rehabilitation as Fields: She testified to the commonplace view that some, but not all, people can be rehabilitated. 6 Record 2461. Moreover, Witt expressed strong support for the death penalty. Id., at 2414-2416, 2443-2444. She testified that the death penalty was appropriate for the crime of murder in the course of a robbery, id., at 2428, or for a convict who was released from prison and committed murder (Miller-El previously had twice spent time in prison for armed robberies), id., at 2462-2463. This is likely why the State accepted Witt and Miller-El struck her. Id., at 2464-2465. Finally, Fernando Gutierrez testified that he could impose the death penalty for brutal crimes. 11-(B) Record 4391-4392. In fact, the only issue during voir dire was whether Gutierrez could apply Texas’ more lenient penalties, not its more severe ones. Id., at 4398-4399, 4413-4414, 4431. The court questioned Gutierrez at length, and ultimately he was accepted by both parties and seated on the jury. Id., at 4439-4449.

Third, Hearn, Witt, and Gutierrez were not similarly situated to Fields even apart from their views on the death penalty. Fields was dismissed not only for his prodefense views on rehabilitation, but also because his brother had several drug convictions and had served time in prison. App. 190, 199. Hearn, Witt, and Gutierrez did not have relatives with significant criminal histories. Thus, there was an additional race-neutral reason to dismiss Fields that simply was not true of the other jurors. Surely the State did not need to expend peremptories on all veniremen who expressed some faith in rehabilitation to avoid violating Batson.

The majority dismisses as “makeweight” the State’s justification as to Fields’ brother, ante, at 246, but it is the majority’s arguments that are contrived. The State questioned Fields during voir dire about his brother’s drug offenses, where the offenses occurred, whether his brother had been tried, whether his brother had been convicted, and whether *296his brother’s criminal history would affect Fields’ ability to serve on the jury. App. 190. The State did not fail to engage in a “ ‘meaningful voir dire examination/ ” as the majority contends. Ante, at 246 (quoting Ex parte Travis, 776 So. 2d 874, 881 (Ala. 2000)).

The majority also contends that the State’s justification as to Fields’ brother illustrates pretext, because the State first pointed to Fields’ views on rehabilitation as the reason for its strike. Ante, at 245-246. The timing of the State’s explanation was unexceptional. In context, the State discussed Fields’ brother at essentially the same time it discussed Fields’ religious views. The entire exchange between the State and counsel for Miller-El took place in a couple of minutes at most. App. 197-199. Thus, to call the State’s second reason an “afterthought,” ante, at 246, ignores what is obvious even from a cold record: that the State simply offered both of its reasons in quick succession.

B

Miller-El’s claims of disparate questioning also do not fit the facts. Miller-El argues, and the majority accepts, that the prosecution asked different questions at voir dire of black and nonblack veniremen on two subjects: (1) the manner of execution and (2) the minimum punishment allowed by state law. The last time this case was here, I refuted Miller-El’s claim that the prosecutors’ disparate questioning evinced racial bias, and explained why it did not even entitle him to a certificate of appealability. Miller-El I, 537 U. S., at 363-370 (dissenting opinion).

This time, the majority has shifted gears, claiming that a different set of jurors demonstrates the State’s racial bias. The majority’s new claim is just as flawed as its last. The State questioned panelists differently when their questionnaire responses indicated ambivalence about the death penalty. Any racial disparity in questioning resulted from the *297reality that more nonblack veniremen favored the death penalty and were willing to impose it.

1

While most veniremen were given a generic description of the death penalty at the outset of their voir dire examinations, some were questioned with a “graphic script” that detailed Texas’ method of execution. Ante, at 255. According to Miller-El and the majority, prosecutors used the graphic script to create cause for removing black veniremen who were ambivalent about or opposed to the death penalty. Ante, at 260. This is incorrect.

The jury questionnaires asked two questions directly relevant to the death penalty. Question 56 asked, “Do you believe in the death penalty?” It offered panelists the chance to circle “yes” or “no,” and then asked them to “[pjlease explain your answer” in the provided space. E. g., Joint Lodging 6. Question 58 asked, “Do you have any moral, religious, or personal beliefs that would prevent you from returning a verdict which would ultimately result in the execution of another human being?” and offered panelists only the chance to circle “yes” or “no.” Ibid.

According to the State, those veniremen who took a consistent stand on the death penalty — either for or against it— did not receive the graphic script. These prospective jurors either answered “no” to question 56 and “yes” to question 58 (meaning they did not believe in the death penalty and had qualms about imposing it), or answered “yes” to question 56 and “no” to question 58 (meaning they did believe in the death penalty and had no qualms about imposing it). Only those potential jurors who answered inconsistently, thereby indicating ambivalence about the death penalty, received the graphic script.

The questionnaires bear out this distinction. Fifteen blacks were questioned during voir dire. Only eight of them — or 53% — received the graphic script. All eight had *298given ambivalent questionnaire answers regarding their ability to impose the death penalty. There is no question that veniremen Baker, Bailey, Boggess, Woods, and Butler were ambivalent in their questionnaire answers. See ante, at 260, n. 27; 4 Record 1874-1875.5 The majority claims that Keaton, Kennedy, and Mackey were not ambivalent, ante, at 258-259, and nn. 17, 19, but their questionnaire answers show otherwise. For instance, Keaton circled “no” for question 56, indicating she did not believe in the death penalty, and wrote, “It’s not for me to punished [sic] anyone.” Joint Lodging 55. However, she then circled “no” for question 58, indicating that she had no qualms about imposing the death penalty. Ibid. Likewise, Mackey indicated she did not believe in the death penalty and wrote “Thou Shall Not Kill” in the explanation space. Id., at 79. Mackey then said that she had no qualms, religious or otherwise, about imposing the death penalty, even though she had just quoted one of the Ten Commandments. Ibid. Keaton’s and Mackey’s answers cannot be reconciled, and the majority makes no attempt to do so. Ante, at 258, n. 17. Kennedy wrote on his questionnaire that he would impose the death penalty “[o]nly in extreme cases, such as multiple murders.” Joint Lodging 46. This left prosecutors uncertain about whether Kennedy could impose the death penalty on Miller-El, who had murdered only one person (though he had paralyzed another).

Of the seven blacks who did not receive the graphic script, six took a stand on the death penalty — either for or against it — in their questionnaires. There was no need to use the graphic script to clarify their positions. Veniremen Boze-*299man, Fields, Rand, and Warren all answered “yes” to question 56 (indicating that they believed in the death penalty) and “no” to question 58 (indicating that they had no qualms about imposing it).6 Id., at 6 (Bozeman); id., at 14 (Fields); id., at 80 (Rand); id., at 22 (Warren). Venireman Mosley was the opposite: He said that he was opposed to the death penalty, 7 Record 2656, 2681, and that he definitely could not impose it, id., at 2669-2670. The same appears true of venireman Smith, 2 id., at 927-928, who was so adamantly opposed to the death penalty throughout her voir dire that she was struck for cause, id., at 1006. The only apparent exception is venireman Carter. She said that she believed in the death penalty, but wrote on the questionnaire, “Yes and no. It would depend on what the person had done.” 4 id., at 1993 (internal quotation marks omitted). She then answered “ ‘[y]es’ ” to question 58, indicating that she had some difficulties with imposing the death penalty. Ibid. Despite her ambivalence, Carter did not receive the full graphic script. Prosecutors told her only that Miller-El “[would] be executed by lethal injection at Huntsville.” Id., at 1952.

Thus far, the State’s explanation for its use of the graphic script fares far better than Miller-El’s or the majority’s. Questionnaire answers explain prosecutors’ use of the graphic script with 14 out of the 15 blacks, or 93%. By contrast, race explains use of the script with only 8 out of 15 veniremen, or 53%. The majority’s more nuanced explanation is likewise inferior to the State’s. It hypothesizes that the script was used to remove only those black veniremen ambivalent about or opposed to the death penalty. Ante, at 260. But that explanation accounts for only 12 out of 15 veniremen, or 80%. The majority cannot explain why prosecutors did not use the script on Mosley and Smith, who were opposed to the death penalty, or Carter, who was ambivalent. *300Because the majority does not account for veniremen like Carter, and also mischaracterizes veniremen like Keaton, Kennedy, and Mackey, it arrives at different percentages. This is not clear and convincing evidence of racial bias.

The State’s explanation also accounts for its treatment of the 12 nonblack veniremen (10 whites, 1 Hispanic, and 1 Filipino) on whom the majority relies. Granted, it is more difficult to draw conclusions about these nonblack veniremen. With the blacks, 11 of their 15 questionnaires are available; with the nonblacks, that number plummets to 3 of 12, because those veniremen were not discussed before the state court. See supra, at 279. Nevertheless, the questionnaires and voir dire permit some tentative conclusions.

First, of the five nonblacks who received the graphic script — Desinise, Evans, Gutierrez, Sztybel, and Zablan— four were ambivalent. On his questionnaire, Gutierrez answered both that he believed in the death penalty and that he had qualms about imposing it. Joint Lodging 231. Szty-bel and Zablan averred that they believed in the death penalty and could impose it, but their written answers to question 56 made it unclear under what circumstances they could vote to impose the death penalty.7 Desinise is a closer call, but he was genuinely undecided about his ability to impose the death penalty, and the parties struck him by agreement. 3 Record 1505-1506,1509,1511,1514. Of the five nonblacks who received the graphic script, Evans was the only one steadfastly opposed to the death penalty. 6 id., at 2588-2589, 2591, 2595.

Of the seven nonblacks who allegedly did not receive the graphic script, four were strongly opposed to the death penalty. See Miller-El I, 537 U. S., at 364-365 (Thomas, J., dis*301senting). Berk, Hinson, and Nelson were so opposed that they were struck for cause, and Holtz was struck by the State because he was opposed unless a policeman or fireman was murdered. Ibid. Administering the graphic script to these potential jurors would have been useless. “No trial lawyer would willingly antagonize a potential juror ardently opposed to the death penalty with an extreme portrait of its implementation.” Id., at 364.

Of the remaining three nonblacks, the majority is correct that Moses was ambivalent in her questionnaire responses, 3 Record 1140-1141, 1177, although it is not certain that Vickery was, 4 id., at 1611. Neither received the graphic script. However, the final nonblack, Girard, confirms the State’s explanation. It was not clear from Girard’s questionnaire whether she was ambivalent.8 On the stand, prosecutor Nelson started off with the abstract script. 6 id., at 2520-2521. But it quickly became apparent that Girard was “just not real sure” about her ability to impose the death penalty, and she testified that she had not decided its value as a form of punishment. Id., at 2522-2523. At that point, Nelson gave her the graphic script — for no other reason than to discern her basic reaction. Id., at 2524-2525. Not only did it succeed — Girard testified that she did not want to serve on a capital jury, id., at 2529, 2531 — but Miller-El’s attorney also used the graphic script when he questioned Girard, id., at 2553. Miller-El’s counsel was using the graphic script just as the State was: to discern a potential juror’s true feelings, not to create cause for removing a venireman. After all, Girard’s views were favorable to Miller-El.

In any event, again the State’s explanation fares well. The State’s explanation accounts for prosecutors’ choice between the abstract and graphic scripts for 9 of 12 nonblack *302veniremen, or 75%. Moses and Vickery were likely ambivalent but did not receive the graphic script, while Evans was opposed to the death penalty but did receive it. However, the majority’s theory accounts for the State’s treatment of only 6 of 12 nonblacks, or 50%. The majority can explain why jurors like Moses and Vickery did not receive the graphic script, because it believes the State was using the graphic script primarily with blacks opposed to or ambivalent about the death penalty. Ante, at 260. But the majority cannot explain the State’s use of the script with an opposed nonblack like Evans, or ambivalent nonblacks like Desinise, Girard, Gutierrez, Sztybel, and Zablan.

Finally, the majority cannot take refuge in any supposed disparity between use of the graphic script with ambivalent black and nonblack veniremen. Ante, at 257-259. The State gave the graphic script to 8 of 9 ambivalent blacks, or 88%, and 5 of 7 ambivalent nonblacks, or 71%. This is hardly much of a difference. However, when the majority lumps in veniremen opposed to the death penalty, ibid., the disparity increases. The State gave the graphic script to 8 of 11 ambivalent or opposed blacks, or 73%, and 6 of 12 ambivalent or opposed nonblacks, or 50%. But the reason for the increased disparity is not race: It is, as the State maintains, that veniremen who were opposed to the death penalty did not receive the graphic script.

In sum, the State can explain its treatment of 23 of 27 potential jurors, or 85%, while the majority can only account for the State’s treatment of 18 of 27 potential jurors, or 67%. This is a far cry from clear and convincing evidence of racial bias.

2

Miller-El also alleges that the State employed two different scripts on the basis of race when asking questions about imposition of the minimum sentence. This disparate-questioning argument is even more flawed than the last one. The evidence confirms that, as the State argues, prosecutors *303used different questioning on minimum sentences to create cause to strike veniremen who were ambivalent about or opposed to the death penalty. Brief for Respondent 33, and n. 26.

Of the 15 blacks, 7 were given the minimum punishment script (MPS). All had expressed ambivalence about the death penalty, either in their questionnaires (Baker, Boggess, and Kennedy) or during voir dire (Bozeman, Fields, Rand, and Warren).9 Woods expressed ambivalence in his questionnaire, but his voir dire testimony made clear that he was a superb juror for the State. See supra, at 289-290. Thus, Woods did not receive the MPS. There was no reason to give the MPS to Butler, Carter, Mosley, or Smith, all of whom were dismissed for cause or by agreement of the parties. That leaves Bailey, Keaton, and Mackey, all of whom were so adamantly opposed to the death penalty during voir dire that the State attempted to remove them for cause. 11-(A) Record 4112, 4120, 4142 (Bailey); id., at 4316 (Keaton); 10 id., at 3950, 3953 (Mackey). Because the State believed that it already had grounds to strike these potential jurors, it did not need the MPS to disqualify them. However, even assuming that the State should have used the MPS on these 3 veniremen, the State’s explanation still accounts for 7 of the 10 ambivalent blacks, or 70%.

The majority does not seriously contest any of this. Ante, at 261-262, and n. 34. Instead, it contends that the State used the MPS less often with nonblacks, which demonstrates that the MPS was a ruse to remove blacks. This is not true: The State used the MPS more often with ambivalent non-blacks who were not otherwise removable for cause or by agreement.

*304Of the nonblacks who reached the point in the voir dire sequence where the MPS was typically administered, the majority points to 11 whom it alleges were ambivalent and should have received the script. Ante, at 262, and n. 34. Three of these veniremen — Gibson, Gutierrez, and Holtz— were given the MPS, just like many of the blacks. Four of the remaining eight veniremen — Moses, Salsini, Vickery, and Witt — were favorable enough to the State that Miller-El peremptorily struck them.10 The State had no interest in disqualifying these jurors. Two of the remaining four veniremen — Hearn and Mazza — indicated that they could impose the death penalty, both on their questionnaires and during voir dire. The State likewise had no interest in disqualifying these jurors. Assuming that the State should have used the MPS on the two remaining veniremen, Crowson and Whaley, the State’s explanation still accounts for 9 of the 11 ambivalent nonblacks, or 81%. Miller-El’s evidence is not even minimally persuasive, much less clear and convincing.

C

Miller-El’s argument that prosecutors shuffled the jury to remove blacks is pure speculation. At the Batson hearing, Miller-El did not raise, nor was there any discussion of, the topic of jury shuffling as a racial tactic. The record shows only that the State shuffled the jury during the first three weeks of jury selection, while Miller-El shuffled the jury during each of the five weeks. This evidence no more proves that prosecutors sought to eliminate blacks from the jury, than it proves that Miller-El sought to eliminate whites even more often. Miller-El I, 537 U. S., at 360 (Thomas, J., dissenting).

*305Miller-El notes that the State twice shuffled the jury (in the second and third weeks) when a number of blacks were seated at the front of the panel. Ante, at 254. According to the majority, this gives rise to an “inference” that prosecutors were discriminating. Ante, at 255. But Miller-El should not be asking this Court to draw “inference[s]”; he should be asking it to examine clear and convincing proof. And the inference is not even a strong one. We do not know if the nonblacks near the front shared characteristics with the blacks near the front, providing race-neutral reasons for the shuffles. We also do not know the racial composition of the panel during the first week when the State shuffled, or during the fourth and fifth weeks when it did not.

More importantly, any number of characteristics other than race could have been apparent to prosecutors from a visual inspection of the jury panel. See Ladd v. State, 3 S. W. 3d 547, 563-564 (Tex. Crim. App. 1999). Granted, we do not know whether prosecutors relied on racially neutral reasons, ante, at 254-255, but that is because Miller-El never asked at the Batson hearing. It is Miller-El’s burden to prove racial discrimination, and the jury-shuffle evidence itself does not provide such proof.

D

The majority’s speculation would not be complete, however, without its discussion (block-quoted from Miller-El I) of the history of discrimination in the D. A.’s Office. This is nothing more than guilt by association that is unsupported by the record. Some of the witnesses at the Swain hearing did testify that individual prosecutors had discriminated. Ante, at 264. However, no one testified that the prosecutors in Miller-El’s trial — Norman Kinne, Paul Macaluso, and Jim Nelson — had ever been among those to engage in racially discriminatory jury selection. Supra, at 276.

The majority then tars prosecutors with a manual entitled Jury Selection in a Criminal Case (hereinafter Manual or *306Sparling Manual), authored by John Sparling, a former Dallas County prosecutor. There is no evidence, however, that Kinne, Macaluso, or Nelson had ever read the Manual— which was written in 1968, almost two decades before Miller-El’s trial.11 The reason there is no evidence on the question is that Miller-El never asked. During the entire Batson hearing, there is no mention of the Sparling Manual. Miller-El never questioned Macaluso about it, and he never questioned Kinne or Nelson at all. The majority simply assumes that all Dallas County prosecutors were racist and remained that way through the mid-1980’s.

Nor does the majority rely on the Manual for anything more than show. The Manual contains a single, admittedly stereotypical line on race: “Minority races almost always empathize with the Defendant.” App. 102. Yet the Manual also tells prosecutors not to select “anyone who had a close friend or relative that was prosecuted by the State.” Id., at 112. That was true of both Warren and Fields, and yet the majority cavalierly dismisses as “makeweight” the State’s justification that Warren and Fields were struck because they were related to individuals convicted of crimes. Ante, at 246,250, n. 8. If the Manual is to be attributed to Kinne, Macaluso, and Nelson, then it ought to be attributed in its entirety. But if the majority did that, then it could not point to any black venireman who was even arguably dismissed on account of race.

Finally, the majority notes that prosecutors “ ‘marked the race of each prospective juror on their juror cards.’ ” Ante, at 264 (quoting Miller-El I, supra, at 347). This suffers from the same problems as Miller-El’s other evidence. Prosecutors did mark the juror cards with the jurors’ race, sex, and juror number. We have no idea — and even the majority cannot bring itself to speculate — whether this was *307done merely for identification purposes or for some more nefarious reason. The reason we have no idea is that the juror cards were never introduced before the state courts, and thus prosecutors were never questioned about their use of them.

* * *

Thomas Joe Miller-El’s charges of racism have swayed the Court, and AEDPA’s restrictions will not stand in its way. But Miller-El has not established, much less established by clear and convincing evidence, that prosecutors racially discriminated in the selection of his jury — and he certainly has not done so on the basis of the evidence presented to the Texas courts. On the basis of facts and law, rather than sentiments, Miller-El does not merit the writ. I respectfully dissent.