Appellant, Terry Darnell Williams, alleges that during his state trial for second-degree robbery, he made a prima facie showing of discrimination under Batson v. Kentucky, 476 U.S. 79 , 106 S.Ct. 1712 , 90 L.Ed.2d 69 (1986), when he objected to the prosecutor’s use of three of four peremptory challenges to excuse African-Americans. We hold that the district court, not having the guidance of the Supreme Court’s opinions in Johnson v. California, — U.S.-, 125 S.Ct. 2410 , 162 L.Ed.2d 129 (2005), and Miller-El v. Dretke, — U.S.-, 125 S.Ct. 2317 , 162 L.Ed.2d 196 (2005), failed to appreciate the import of Williams’ showing of statistical disparity. Accordingly, we reverse and remand.

Williams, an African-American, was charged in Los Angeles Superior Court with second-degree robbery. The State alleged that Williams and his accomplice, James E. Roberts, robbed a mini-mart at gunpoint. Williams’ defense was that he had been with his fiancee at the time of the robbery and that another person had accompanied Roberts.

Williams was tried with Roberts in March 1998. When the court indicated that counsel could make peremptory challenges, the prosecutor stated that he would accept the jury. Defense counsel proceeded to excuse potential jurors and the prosecutor then exercised four peremptory challenges. After the fourth challenge, defense counsel objected, stating:

The dialogue continued as follows:

After this exchange, peremptory challenges continued. The defense excused three jurors and the prosecutor passed twice. The court then called in another twenty-two prospective jurors. On the next day, defense counsel excused another seven potential jurors. The prosecutor excused one juror and passed seven times. The defense did not renew its Wheeler challenge. The case proceeded to trial and the jury returned a verdict of guilty. Williams was sentenced to a term of thirty-four years to life.

Williams’ conviction was affirmed by the California Court of Appeal in an unpublished opinion. One of the three issues he raised on appeal was that the trial court had erred in denying his Wheeler motion. In rejecting Williams’ Wheeler/Batson claim, the appellate court held that, “from all the circumstances of the case,” Williams had not shown “a strong likelihood that such persons are being challenged because of their group association rather than because of any specific bias.” In addition, the appellate court indicated that because a Wheeler motion calls upon the trial judge’s personal observations, it reviewed the trial judge’s decision with considerable deference, and because the record suggested grounds upon which the prosecutor might reasonably have challenged the jurors, it affirmed. The California Supreme Court denied Williams’ petition for review.

Williams filed a petition for a writ of habeas corpus in the United States District Court for the Central District of California. He alleged, inter alia, that the prosecutor had improperly used his peremptory challenges to eliminate African-Americans from the jury. The district judge denied relief concluding that Williams had “not come forward with sufficient evidence to show a reasonable inference of purposeful discrimination arose solely on the basis of statistics that would have required the trial judge to perform the entire Batson analysis at the time the objection was made.”

Williams filed a timely notice of appeal and the district court granted a certificate of appealability on Williams’ Batson claim.

We review de novo a district court’s denial of a habeas petition. Medina v. Hornung, 386 F.3d 872, 876 (9th Cir.2004). Moreover, although Williams’ petition is subject to the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”), Pub.L. No. 104-132, 104 , 110 Stat. 1214 , 1218-19 (1996), we have held that where the state court used the “strong likelihood” standard for reviewing a Batson claim, the state court’s findings are not entitled to deference and our review is de novo. Paulino v. Castro, 371 F.3d 1083, 1090 (9th Cir.2004).

Ill

In resolving Williams’ appeal we must determine whether he made a prima facie showing of purposeful discrimination when he raised his Wheeler/Batson objection. Our inquiry into this matter requires that we first ascertain the proper.legal standard for reviewing a claim of purposeful discrimination based on statistical disparity, and then apply that standard to Williams’ case. In considering both of these matters, we, unlike the district court, have the guidance of the Supreme Court’s recent opinions in Johnson and Miller-El.

A The Legal Standard

125 S.Ct. at 2416 (footnote omitted).

Addressing the first step of the Batson test, the Court noted that “an inference of discriminatory purpose” is a lesser standard for a prima facie showing than the “more likely than not” standard. Id. It explained that it “did not intend the first step to be so onerous that a defendant would have to persuade the judge — on the basis of all the facts, some of which are impossible for the defendant to know with certainty — 'that the challenge was more likely than not the product of purposeful discrimination.” Id. at 2417. Rather, a defendant satisfies the requirements of Batson’s first step “by producing evidence sufficient to permit the trial judge to draw an inference that discrimination has occurred.” Id.

The Court explained that, although the burden of persuasion remains with the defendant, it is not until the third step of the Batson procedure that the persuasiveness of the prosecutor’s justification becomes relevant. Id. at 2418. The framework was “designed to produce actual answers to suspicions and inferences that discrimination may have infected the jury selection process.” Id.

Here, Williams established that he is African-American and that the prosecutor used three of his first four peremptory challenges to remove African-Americans from the jury. In addition, it appears that only four of the first forty-nine potential jurors were African-American.

These bare facts present a statistical disparity. We have held that a defendant can make a prima facie showing based on a statistical disparity alone. In Paulino, we concluded there was an inference of bias where the prosecutor had used five out of six peremptory challenges to strike African-Americans. 371 F.3d at 1091 . In Fernandez v. Roe, 286 F.3d 1073, 1077-80 (9th Cir.2002), we found an inference of bias where four of seven Hispanics and two African-Americans were excused by the prosecutor. In Turner v. Marshall, 63 F.3d 807, 812 (9th Cir.1995), overruled on other grounds by Tolbert v. Page, 182 F.3d 677, 681 (9th Cir.1999) (en banc), we determined there was a prima facie showing of discrimination where the prosecutor exercised peremptory challenges to exclude five out of a possible nine African-Americans. In addition, we have noted that the “Constitution forbids striking even a single prospective juror for a discriminatory purpose.” United States v. Vasquez-Lopez, 22 F.3d 900, 902 (9th Cir.1994).

Prior to the issuance of Johnson , we had held that, although a statistical disparity could be sufficient to make a prima facie inference of bias, such a presumption could be dispelled by other relevant circumstances. See Paulino, 371 F.3d at 1091 (‘We sometimes consider whether the context in which a defendant made a Batson objection changes the significance of a statistical pattern in the exercise of peremptory challenges.”); see also Fernandez, 286 F.3d at 1079 (“Under Batson, we must consider ‘all relevant circumstances’ surrounding the challenges.”).

Batson held that the use of race-based peremptory challenges to excuse prospec tive jurors violates the Equal Protection Clause of the Fourteenth Amendment. Batson, 476 U.S. at 89 , 106 S.Ct. 1712 . In Johnson , the Supreme Court reiterated the now-familiar three-step procedure as follows:

Johnson implicitly reaffirms this perspective. In Batson , the Court noted that, in “deciding whether the defendant has made the requisite showing, the trial court should consider all relevant circumstances,” and noted that a “prosecutor’s questions and statements during voir dire examination and in exercising his challenges may support or refute an inference of discriminatory purpose.” 476 U.S. at 96-97 , 106 S.Ct. 1712 (emphasis added). In Johnson , the Court reiterates that a defendant may rely on “any other relevant circumstances” to raise an inference of discriminatory purpose. 125 S.Ct. at 2417. It follows that, when reviewing a Batson claim, a court should continue to consider “any other relevant circumstances” brought to their attention that may support or refute an inference of discriminatory purpose.

Johnson , while not constricting what circumstances a court may consider in reviewing a Batson claim, did clarify the equation into which the circumstances are factored. The Court emphasized that a defendant need only show “an inference of discriminatory purpose” and could not be required to show that a “challenge was more likely than not the product of purposeful discrimination.” Id. at 2416-17. The Johnson Court further noted, citing our decision in Paulino, 371 F.3d at 1090 , that it “does not matter that the prosecutor might have had good reasons[; W]hat matters is the real reason [potential jurors] were stricken.” 125 S.Ct. at 2418. This concern was reiterated by the Court in Miller-El :

125 S.Ct. at 2332. Accordingly, to rebut an inference of discriminatory purpose based on statistical disparity, the “other relevant circumstances” must do more than indicate that the record would support race-neutral reasons for the questioned challenges.

B. Williams’ Showing of an Inference of Bias

The underlying facts in this case bear a resemblance to those in Johnson . There, as here, the prosecutor used peremptory challenges to remove prospective African-American jurors. The state trial judge did not ask the prosecutor to explain the rationale for his strikes, but summarily determined that the prosecutor’s strikes could be justified by race-neutral reasons. 125 S.Ct. at 2414. Similarly, in this case, the trial judge did not require that defense counsel elaborate on his Wheeler challenge, did not allow the prosecutor to proffer an explanation for his peremptory challenges, and summarily found that Williams had failed to make a prima facie showing of bias.

In each case, the trial judge’s actions limit the scope of appellate review. We cannot determine the reasonableness of the prosecutor’s challenges, but can only review the record to determine whether “other relevant circumstances” eroded the premises of Williams’ allegations of discrimination based on statistical disparity.

The district court, however, as well as the California Court of Appeal, addressed a different issue: whether the record could support race-neutral grounds for the prosecutor’s peremptory challenges. Although their conclusion that the record supported such grounds for the peremptory challenges may have been reasonable, the Supreme Court’s clarification of Batson in Johnson , and its review of the record in Miller-El , lead to the conclusion that this approach did not adequately protect Williams’ rights under the Equal Protection Clause of the Fourteenth Amendment or “ ‘public confidence in the fairness of our system of justice.’ ” Johnson, 125 S.Ct. at 2418 (quoting Batson, 476 U.S. at 87 , 106 S.Ct. 1712 ); see also Miller-El, 125 S.Ct. at 2323-24 .

Our review of the existing record, informed by Johnson and Miller-El , fails to disclose a refutation of the inference of bias raised by the statistical disparity. Accordingly, we vacate the district court’s denial of Williams’ habeas petition and remand. It is true that the prosecutor accepted the jury, including African-American members, several times before he exercised his first peremptory challenge. This, however, does not refute the inference that when the prosecutor did make peremptory challenges, he did so in a purposefully discriminatory manner as evidenced by his use of three of his first four peremptory challenges to dismiss African-American jurors. Similarly, the fact that the prosecutor did not use a peremptory challenge against an African-American juror after Williams’ Wheeler objection is not enough in this case to refute the inference.

The district court and the California Court of Appeal also reviewed all the evidence in the record concerning the challenged jurors and determined that the record contained evidence for each juror that would support peremptory challenges on non-objectionable grounds. This, however, does not measure up to the Supreme Court’s pronouncement that the question is not whether the prosecutor might have had good reasons, but what were the prosecutor’s real reasons for the challenges. Johnson, 125 S.Ct. at 2418 ; see also Miller-El, 125 S.Ct. at 2332 (“A Batson challenge does not call for a mere exercise in thinking up any rational basis.”). Furthermore, although in some instances the evidence in support of race-neutral reasons for the peremptory challenges may dispel any inference of bias, we cannot — on this record — so conclude.

Finally, we note that the district court, having determined that Williams had failed “to show a reasonable inference of purposeful discrimination that would have required the trial judge to perform the entire Batson analysis at the time the objection was made,” opined that it could not order an evidentiary hearing under the AEDPA, 28 U.S.C. § 2254 (e). It stated that this result flowed from Williams’ “ne gleet in presenting information regarding the composition of the venire to the state courts during his direct appeal and state habeas petition.” We disagree. First, as previously noted, because the state appellate court used the improper “strong likelihood” standard for evaluating Williams’ Wheeler/Batson claim, the district court was required to review the matter de novo. Paulino, 371 F.3d at 1090 . Second, as made clear by Johnson , Williams, having presented a statistical disparity based on the information then known to him, cannot be charged, prior to the prosecutor’s explanation of his challenges, with developing a record that might refute the prosecutor’s possible explanations. Instead, it appears that if there are other relevant circumstances that might dispel the inference, it was the state’s responsibility to create a record that dispels the inference.

We conclude that the state appellate court and the district court, not having the benefit of the Supreme Court’s recent opinions in Johnson and Miller-El , failed to appreciate that (1) Williams’ showing of statistical disparity was only required to raise an inference of purposeful discrimination, and (2) refutation of the inference requires more than a determination that the record could have supported race-neutral reasons for the prosecutor’s use of his peremptory challenges on prospective African-American jurors. Accordingly, the district court’s denial of Williams’ habeas petition is VACATED and the case is REMANDED to the district court.

. People v. Wheeler, 22 Cal.3d 258, 148 Cal.Rptr. 890 , 583 P.2d 748 (1978), is California's counterpart to Batson v. Kentucky, 476 U.S. 79 , 106 S.Ct. 1712 , 90 L.Ed.2d 69 (1986). A Wheeler motion is considered the procedural equivalent to a challenge made under Batson . Paulino v. Castro, 371 F.3d 1083 , 1088 n. 4 (9th Cir.2004). . The other major issues raised on appeal were (1) whether the trial court erred in denying Williams' motion for mistrial when jurors witnessed an altercation, involving a key defense witness outside the courtroom, and (2) whether the trial court erred in instructing the jury that it could disregard the testimony of the defense alibi witness because Williams had not disclosed the witness to the prosecution thirty days before trial. .Our explanation in Paulino is also applicable to this case: We would normally be required to defer to the court of appeal’s factual finding that there was no prima facie showing of bias. See Tolbert v. Page, 182 F.3d 677, 682 (9th Cir. 1999) (en banc). In making this finding, however, the court of appeal explicitly required Paulino to “show a strong likelihood” ... of .bias under Wheeler, and did not mention Batson or its inference standard. We have held that the Wheeler standard "is impermissibly stringent in comparison to the more generous Batson 'inference' test.” Wade v. Terhune, 202 F.3d 1190, 1197 (9th Cir.2000). Thus, "California courts in following the 'strong likelihood’, language of Wheeler are not applying the correct legal standard for a prima facie case under Batson." Id. Because the court of appeal employed the incorrect legal standard in reviewing Paulino's claim, we examine his Batson claim de novo. Id.; see also Fernandez v. Roe, 286 F.3d 1073, 1077 (9th Cir.2002); Cooperwood v. Cambra, 245 F.3d 1042, 1046 (9th Cir.2001). 371 F.3d at 1090 (footnote omitted). . Our determination that we must apply a de novo standard of review distinguishes this case from Kesser v. Cambra, 392 F.3d 327 (9th Cir.), rehearing en banc granted, 425 F.3d 1230 (9th Cir.2005). . These cases apply to Williams’ habeas appeal as the Supreme Court clearly indicates in Johnson that it is clarifying Batson , not making new law. Johnson, 125 S.Ct. at 2416-17 . . In Johnson , the Court reiterated that, to establish an inference, the defendant first must show that he is a member of a cognizable racial group, and that the prosecutor has exercised peremptory challenges to remove from the venire members of the defendant's race. Second, the defendant is entitled to rely on the fact, as to which there can be no dispute, that peremptory challenges constitute a jury selection practice that permits “those to discriminate who are of a mind to discriminate.'' Finally, the defendant must show that these facts and any other relevant circumstances raise an inference that the prosecutor used that practice to exclude the veniremen from the petit jury on account of their race. [Batson, 476 U.S.] at 96, 106 S.Ct. 1712 , 90 L.Ed.2d 69 (citations omitted) (quoting Avery v. Georgia, 345 U.S. 559, 562 , 73 S.Ct. 891 , 97 L.Ed. 1244 (1953)). 125 S.Ct. at 2416-17. . The Court in Johnson further observed: The inherent uncertainty present in inquiries of discriminatory purpose counsels against engaging in needless and imperfect speculation when a direct answer can be obtained by asking a simple question. See Paulino v. Castro, 371 F.3d 1083, 1090 (C.A.9 2004) (“[I]t does not matter that the prosecutor might have had good reasons ... [w]hat matters is the real reason they were stricken” (emphasis deleted)); Holloway v. Horn, 355 F.3d 707, 725 (C.A.3 2004) (speculation “does not aid our inquiry into the reasons the prosecutor actually harbored” for a peremptory strike). The three-step process thus simultaneously serves the public purposes Batson is designed to vindicate and encourages “prompt rulings on objections to peremptory challenges without substantial disruption of the jury selection process.” Hernandez v. New York, 500 U.S. 352, 358-359 , 111 S.Ct 1859 , 114 L.Ed.2d 395 (1991) (opinion of KENNEDY, J.). 125 S.Ct. at 2418-19. . After the trial court rejected Williams' Wheeler objection, the trial court called in another twenty-two potential jurors. The racial composition of these twenty-two jurors is not in the record. . One such circumstance may be the timing of the Batson objection. We noted in Paulino, 371 F.3d at 1091 , and in Wade, 202 F.3d at 1198, that if an African-American juror is the first person called and struck, at that time all of the prosecutor's peremptory challenges will have been exercised against African-Americans, whereas if the same juror is called and struck at the end of voir dire, the percentage would be far lower. . The cases have several differences relevant to the determination of whether a defendant has made a prima facie showing of purposeful discrimination. Johnson concerned a prosecutor's use of three of twelve peremptory challenges which removed all African-Americans from the jury. 125 S.Ct. at 2414. In the instant case, the prosecutor used three of his four peremptory challenges on African-Americans, but the record is not clear whether another African-American remained on the jury. In Johnson , the trial judge proffered a reason for his rejection of the Wheeler challenge: the jurors in question "had offered equivocal or confused answers in their written questionnaires.” Id. Here, the trial judge stated only that he was not going to find "a prima facie Wheeler violation.” . The record here indicates that, following the denial of Williams' Wheeler objection, the prosecutor only used one more peremptory challenge. The record is not clear as to the excused juror’s race or the race of the remaining individuals in the jury pool. . For example, the California Court of Appeal opined with regard to Juror 18, one of the excused African-American jurors, that the "fact that he may be a loner’ and had no 'track record' as a juror could have caused the prosecutor to question his ability to effectively function in a group decision-making process, an essential part of jury service.” It further honestly observed, "[o]f course, it is entirely possible the prosecutor's decision to challenge this juror was motivated by something as simple as the prospective juror’s demeanor in the courtroom, a circumstance which frequently would not be apparent to someone reading the cold record, but one which would fully support the exercise of a peremptory challenge.” This speculation is not consistent with the Supreme Court's admonition in Johnson that the first step in the Batson test does not require that “a defendant would have to persuade the judge — on the basis of all the facts, some of which are impossible for the defendant to know with certainty — that the challenge was more likely than not the product of purposeful discrimination.” 125 S.Ct. at 2417. Moreover, even accepting that being a "loner” or not having previously served on a jury can be a race-neutral basis for exercising a peremptory challenge, it is not the type of reason that weighs against an inference of bias. Indeed, were the state court’s rationale on Juror 18 accepted, it is difficult to imagine how any defendant could prevail on a Batson claim following a trial court’s summary rejection of the Batson challenge at the first step of the Batson test. The Supreme Court’s exhaustive review of the record in Miller-El forecloses such an approach. . Thus, the proscription set forth in 28 U.S.C. § 2254 (e)(1) and (2) is not applicable for two reasons. First, because the state court used the wrong legal standard its factual determination is not "presumed to be correct," or alternatively, Williams has rebutted "the presumption of correctness.” 28 U.S.C. § 2254 (e)(1). Second, as Williams' legal claim was not dependent on his further development of the record, he cannot be said to have "failed to develop the factual basis of a claim in State court proceedings." 28 U.S.C. § 2254 (e)(2). . As the district court did not have the benefit of the Supreme Court’s opinions in Johnson and Miller-El , we vacate and remand, even though the state represented to the district court that the prosecutor no longer remembers why he utilized his peremptory challenges and could not locate the jury selection notes.