OPINION

Petitioner-Appellant Joaquin Leso Fernandez appeals the district court’s order denying his 28 U.S.C. § 2254 petition for a writ of habeas corpus challenging his 1995 state jury trial conviction for second degree murder. Petitioner contends that under Batson v. Kentucky, 476 U.S. 79 , 106 S.Ct. 1712 , 90 L.Ed.2d 69 (1986), the prosecutor impermissibly exercised peremptory challenges to exclude four Hispanic and two African-American jurors in violation of the Equal Protection Clause. After de novo review pursuant to Wade v. Terhune, 202 F.3d 1190 (9th Cir.2000), we hold that the district court erred in finding that petitioner had not made a prima facie showing of discrimination. Accordingly, we reverse and remand.

I

Petitioner was a member of a Latino gang in Corona, California. In 1995, he was tried in California Superior Court in Riverside County for his alleged role in the killing of a member of another Latino gang. He was convicted of second degree murder and conspiracy to commit murder, and was sentenced to a term of fifteen years to life in prison.

Approximately sixty potential jurors were called for voir dire for petitioner’s trial. Of these, seven were Hispanic, and two were African-American. The prosecutor exercised peremptory challenges to exclude four of the seven Hispanics. Of the remaining three, the trial judge excused one for cause; the defense exercised a peremptory challenge against one; and one was seated on the jury. The prosecutor exercised peremptory challenges against both of the African-Americans.

In the first round of challenges, without any direct questioning on voir dire by either the court or the prosecutor, the prosecutor struck Mr. Sanchez (2nd pros-ecutorial peremptory, 1st prosecutorial peremptory against a Hispanic). In the next round, the prosecutor struck Mr. Morales (9th peremptory, 2nd Hispanic) and Mr. Merendon (11th peremptory, 3rd Hispanic); both were struck without any direct questioning almost immediately after being seated in the jury box. In the third round, again without any questioning, the prosecutor struck Mr. Salcido (14th peremptory, 4th Hispanic).

After the prosecutor’s challenge against Mr. Salcido, the defense objected under People v. Wheeler, 22 Cal.3d 258 , 148 Cal.Rptr. 890 , 583 P.2d 748 (1978), to all the prosecutor’s peremptory challenges against the prospective Hispanic jurors. “In California, a Wheeler motion is the procedural equivalent of a federal Batson challenge.” Tolbert v. Gomez, 190 F.3d 985, 987 (9th Cir.1999) (citing People v. Jackson, 10 Cal.App.4th 13 , 21 n. 5, 12 Cal.Rptr.2d 541 (1992)). The trial court did not request an explanation from the prosecutor of his challenges, but the prosecutor volunteered that the challenges were based on scores he had assigned to the juror questionnaires before voir dire had commenced. The record is unclear whether the identification of the prospective jurors’ ethnicity had been requested on the questionnaires, but it is clear that their surnames did appear. The trial court did not review any of the juror questionnaires.

The trial court denied the Wheeler motion, stating that it did not appear that the prosecutor was systematically seeking to exclude Hispanics from the jury. The court relied on the fact that one Hispanic woman, Ms. Loya, had not yet been challenged and remained in the' jury box, though subject to future challenge. The trial judge admonished, howeyer, that “if there are any further Hispanics excused from the jury, that would cause me to conclude that a prima facie showing has been made.” After this exchange, the prosecutor did not object to Ms. Loya, and she was seated on the jury.

When voir dire continued, the prosecutor struck without any direct questioning Ms. Pleasant, the only prospective black juror in the jury box at the time (17th peremptory, 1st African-American). Then, during selection of the alternate jurors, the prosecution struck without any direct questioning Ms. Carter (20th overall, 2nd African-American). Ms. Carter was the only other African-American prospective juror in the venire.

After the peremptory challenge against Ms. Carter, the defense brought a second Wheeler motion, based on the State’s exclusion of “people of color.” The prosecutor stated that he had assigned Ms. Carter a low score and that she appeared disinterested in the process. The trial court concluded that no prima facie case of discrimination against African-Americans had been shown and denied the motion.

After trial, the jury returned a verdict of guilty for second degree murder and conspiracy to commit murder. The California Court of Appeal reversed the conspiracy conviction but otherwise affirmed. Applying the Wheeler standard, it held that petitioner had not shown a “strong likelihood” that the prosecutor had challenged the prospective jurors on account of their race or ethnicity, and therefore petitioner had not established a prima facie case of discrimination. The Court of Appeal did not apply the standard of Batson v. Kentucky, 476 U.S. 79, 96 , 106 S.Ct. 1712 , 90 L.Ed.2d 69 (1986), which requires only that a criminal defendant show facts and circumstances that “raise an inference” of exclusion based on race or ethnicity in order to establish a prima facie case.

Petitioner filed two unsuccessful petitions for habeas corpus in the California state courts. In 1997, petitioner timely filed the present petition for habeas corpus in federal district court, alleging, inter alia, that the state trial court erred in denying petitioner’s motions objecting to the prosecutor’s peremptory challenges. The district court adopted the Magistrate’s Report and Recommendation, holding that the trial court did not err in finding that petitioner failed to establish a prima facie case of discrimination under Batson , and denying the petition with prejudice. Petitioner timely appealed.

II

We review de novo the decision of the district court to grant or deny a petition for writ of habeas corpus. Wade, 202 F.3d at 1194 . Under the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), we may disturb a state court’s determinations of law only if they are “contrary to” or “involved an unreasonable application of’ clearly established federal law as determined by the United States Supreme Court. 28 U.S.C. § 2254 (d)(1). On a petition for habeas corpus, we normally review the state trial court’s fact-specific determination of whether a defendant has made a prima facie case of a Batson violation deferentially, applying AEDPA’s “statutory presumption of correctness.” Wade, 202 F.3d at 1195 (quoting Tolbert v. Page, 182 F.3d 677, 685 (9th Cir.1999) (en banc)). However, where the trial court has applied the wrong legal standard, AEDPA’s rule of deference does not apply. Under these circumstances, we review de novo the question whether a defendant made a prima facie showing of a Batson violation. Id. at 1199; Cooperwood v. Cambra, 245 F.3d 1042, 1046-47 (9th Cir.2001).

As we recently held in Wade , California state courts following Wheeler have erroneously required a defendant to show a “strong likelihood” of discrimination in order to establish a prima facie case rather than just an “inference” of discrimination as required by Batson. Wade, 202 F.3d at 1197 . In this case, the state trial court did not expressly articulate the “strong likelihood” standard in finding that petitioner had not established a prima facie case, but it treated both motions as Wheeler , rather than Batson , motions. When the California Court of Appeal affirmed the trial court’s denials of petitioner’s claims, it expressly employed the “strong likelihood” language of Wheeler. See People v. Fernandez, No. E016744 (Cal.Ct.App. Feb. 14, 1997) (unpublished disposition) (“To make a prima facie case, the party ... ‘must show a strong likelihood that such persons are being challenged because of their group association rather than because of any specific bias.’ ”) (citation omitted and emphasis added). The Court of Appeal did not mention Bat-son or its more lenient “inference” test. Because the California state courts applied an incorrect legal standard, contrary to federal law as pronounced in Batson , we review petitioner’s Batson claims de novo. Cooperwood, 245 F.3d at 1047 ; Wade, 202 F.3d at 1197 .

Ill

Under Batson , a prosecutor’s racially discriminatory use of peremptory challenges violates the Fourteenth Amendment’s Equal Protection Clause. To bring a successful Batson challenge, a defendant must first establish a prima facie case by showing that (1) the defendant is a member of a cognizable group; (2) the prosecution has removed members of such a group; and (3) circumstances raise an “inference” that the challenges were motivated by race. 476 U.S. at 96 , 106 S.Ct. 1712 . The burden then shifts to the prosecutor to articulate a race-neutral basis for the peremptory challenges. Id. at 97 , 106 S.Ct. 1712 . Finally, the trial court must determine, in light of the prima facie case and the prosecutor’s explanation, whether the defendant has proven purposeful discrimination. Id. at 98 , 106 S.Ct. 1712 .

Petitioner argues that he established a prima facie case of a Batson violation. Both Hispanics and African-Americans constitute “cognizable groups” for Batson purposes. In Powers v. Ohio, 499 U.S. 400, 409-16 , 111 S.Ct. 1364 , 113 L.Ed.2d 411 (1991), the Supreme Court held that a criminal defendant may object to race-based exclusion of jurors whether or not the defendant and the excluded jurors are of the same race. Petitioner, a Hispanic, thus permissibly can raise a Bat-son challenge based on the exclusion not only of Hispanics, but also of African-Americans, from the jury.

[4] The specific question before us is whether the circumstances of the prosecutor’s challenges “raise an inference” of exclusion based on race, such that inquiry into the prosecution’s motives is required under Batson . A pattern of exclusionary strikes is not necessary for finding an inference of discrimination. See United States v. Vasquez-Lopez, 22 F.3d 900, 902 (9th Cir.1994) (“[T]he Constitution forbids striking even a single prospective juror for a discriminatory purpose.”). But “[a] pattern of exclusion of minority veni-repersons provides support for an inference of discrimination.” Turner v. Marshall, 63 F.3d 807, 812 (9th Cir.1995), overruled on other grounds by Tolbert v. Page, 182 F.3d 677 (9th Cir.1999) (en banc). In Turner , we found a prima facie Batson violation where the prosecution exercised peremptory challenges to exclude five out of a possible nine (56%) African-American jurors. Id. In a number of other cases, with less striking disparities, we have assumed the existence of a prima facie case. See, e.g., United States v. Lorenzo, 995 F.2d 1448, 1453-54 (9th Cir.1993) (three of nine Hawaiian jurors stricken); United States v. Bishop, 959 F.2d 820, 822 (9th Cir.1992) (two of four African-American jurors stricken). Other circuits have found an inference of discrimination under similar circumstances. See United States v. Alvarado, 923 F.2d 253, 255 (2d Cir.1991) (finding prima facie case of discrimination when prosecutor struck four of seven minority venirepersons).

The Turner court relied not only on the high proportion of African-Americans stricken, but also on the dispropoiiionate rate of strikes against African-Americans. Turner, 63 F.3d at 813 . Although only about 30% of those called for voir dire had been African-American, the prosecution had used 56% of its peremptory challenges against African-Americans. In Alvarado , the prosecution used 50% of its challenges against minority venirepersons, who represented only 29% of the pool. Alvarado, 923 F.2d at 255-56 . In Turner , we held that “[s]uch a disparity also supports an inference of discrimination.” 63 F.3d at 813 . Indeed, we relied only on the statistical disparities described above in finding a prima facie Batson violation. Id.

In applying Batson to this case, we focus first on the prospective Hispanic jurors. The statistical evidence in this case is comparable to Turner as to both the proportion of available minorities stricken and the relative rate of such strikes. The prosecutor struck four out of seven (57%) Hispanics, slightly greater than the percentage in Turner , thus supporting an inference of discrimination. While Hispanics constituted only about 12% of the venire, 21% (four out of nineteen) of the prospective juror challenges were made against Hispanics. At the time of the first Wheeler motion, after which the judge in effect warned the prosecutor not to strike any more Hispanics, the prosecutor had exercised -29% (four out of fourteen) of his challenges against Hispanics. Therefore, the prosecutor disproportionately struck Hispanics from the jury box, resulting in a statistical disparity similar to that in Turner . Those challenges, standing alone, are enough to raise an inference of racial discrimination.

Focusing next on the African-American venirepersons, we note that the prosecutor struck the only two prospective African-American jurors. Two challenges out of two venirepersons are not always enough to establish a prima facie case. Because the numbers are so small (and, hence, potentially unreliable), two such challenges, standing alone, may not be sufficient to support an inference of discrimination. See United States v. Vaccaro, 816 F.2d 443, 457 (9th Cir.1987), overruled on other grounds by Huddleston v. United States, 485 U.S. 681 , 108 S.Ct. 1496 , 99 L.Ed.2d 771 (1988). In this case, however, the two challenges against African-Americans do not stand alone. Under Batson , we must consider “all relevant circumstances” surrounding the challenges. Batson, 476 U.S. at 96-97 , 106 S.Ct. 1712 . Even before the prosecutor struck the first African-American venireperson (Ms. Pleasant), the prosecutor’s prior use of peremptory challenges against the prospective Hispanic jurors had given rise to a 'prima facie case of racial discrimination. Given the circumstances — that the prosecutor’s behavior had already supported an inference of discrimination and that the trial court had expressly warned him against striking any more Hispanics — the prosecutor’s subsequent strikes against the only two African-American venireper-sons also support an inference of racial discrimination.

The State defends the finding that Fernandez failed to establish a prima facie case by relying on two points: the fact that one Hispanic juror was ultimately seated, and the fact that the prosecutor said he relied on scores assigned to the juror questionnaires. That one Hispanic juror remained on the jury, though helpful to the State, is not dispositive. See Turner, 63 F.3d at 814 (“In denying a Batson motion, ... a trial court may not rely solely on the fact that some African-Americans remain on the jury.”); Montiel v. City of Los Angeles, 2 F.3d 335, 340 (9th Cir.1993) (stating that trial court “clearly erred” by refusing to conduct Batson inquiry where five of seven peremptory challenges were exercised against minorities even though one African-American woman remained on the jury, finding her presence to “only nominally” weigh against a prima facie showing). Indeed, in Turner , we found a prima facie case of discrimination despite the presence of four African-Americans on the seated jury. See Turner, 63 F.3d at 811-13 . Further, the lone Hispanic juror’s presence on the jury here is less helpful than the presence of African-Americans on the jury in Turner in light of the trial judge’s explicit warning to the prosecutor that any additional challenges against Hispanics would trigger a prima facie finding of discrimination.

In relying on the prosecutor’s explanation of his system for peremptory challenges based on the jury questionnaires, the State compares this case to United States v. Ponce, 51 F.3d 820 (9th Cir.1995), in which the prosecutor offered its juror evaluations based on jury questionnaires. However, in Ponce , we were reviewing for “clear error,” unlike the de novo review applied here. Id. at 830 . Also, the trial court here never actually looked at the questionnaires to verify the prosecutor’s claims that the low-scoring questionnaires indicated a bias against the judicial system or a desire to free the defendant. Moreover, the prosecutor’s claim that he did not know the ethnicity of the prospective jurors when evaluating their responses may be questioned, given that their surnames appeared on the questionnaires. Finally, we note that the prosecutor failed to engage in meaningful questioning of any of the minority jurors.

Case law suggests that we should not even consider the prosecutor’s unsubstantiated explanations at the stage of determining whether a prima facie case exists. See Alvarado, 923 F.2d at 255 (“[T]he initial question is whether appellants presented a prima facie case sufficient to require explanations; that determination must be made before the explanations are considered.”) (emphasis added). Upon our de novo review, we therefore are left with a bare record of statistical disparities of peremptory strikes against Hispanic and African-American venirepersons. Because a pattern of strikes against prospective jurors of a particular race can give rise to an inference of discrimination, see Batson, 476 U.S. at 97 , 106 S.Ct. 1712 , and because the statistical disparities here indicate such a pattern, see Turner, 63 F.3d at 813 , we find that petitioner has established a 'pri-ma facie showing of discrimination under Batson .

Conclusion

For the foregoing reasons, we VACATE the district court’s denial of Petitioner’s habeas petition as to the Batson claim, and REMAND to the district court to conduct an evidentiary hearing, including review of the jury questionnaires, to determine whether there was a Batson violation in the trial court.

VACATED AND REMANDED.

. The trial court did, however, enter the questionnaires into the record for purposes of appeal, and they were filed in the Superior Court under seal. So far as we can determine, neither the California Court of Appeal nor the United States District Court has reviewed the questionnaires.