JUSTICE O'CONNOR, with whom JUSTICE SOUTER, JUSTICE GINSBURG, and JUSTICE BREYER join, dissenting.
In a long line of cases spanning nearly three decades, this Court has applied heightened scrutiny to legislative classifications based on sex. The Court today confronts another statute that classifies individuals on the basis of their sex. While the Court invokes heightened scrutiny, the manner in which it explains and applies this standard is a stranger to our precedents. Because the Immigration and N aturalization Service (INS) has not shown an exceedingly persuasive justification for the sex-based classification embodied in 8 U. S. C. § 1409(a)(4)-i. e., because it has failed to establish at least that the classification substantially relates to the achievement of important governmental objectives-I would reverse the judgment of the Court of Appeals.
I
Sex-based statutes, even when accurately reflecting the way most men or women behave, deny individuals opportunity. Such generalizations must be viewed not in isolation, but in the context of our Nation's "'long and unfortunate history of sex discrimination.'" J. E. B. v. Alabama ex rel. T. B., 511 U. S. 127, 136 (1994) (quoting Frontiero v. Richardson, 411 U. S. 677, 684 (1973) (plurality opinion)). Sex-based generalizations both reflect and reinforce "fixed notions concerning the roles and abilities of males and females." Mississippi Univ. for Women v. Hogan, 458 U. S. 718, 725 (1982).
For these reasons, a party who seeks to defend a statute that classifies individuals on the basis of sex "must carry the burden of showing an 'exceedingly persuasive justification' for the classification." Id., at 724 (quoting Kirchberg v. Feenstra, 450 U. S. 455, 461 (1981)); see also United States v. Virginia, 518 U. S. 515, 531 (1996). The defender of the classification meets this burden "only by showing at least that the classification serves 'important governmental objectives and that the discriminatory means employed' are 'substantially related to the achievement of those objectives.''' Mississippi Univ. for Women, supra, at 724 (quoting Wengler v. Druggists Mut. Ins. Co., 446 U. S. 142, 150 (1980)); see also Virginia, 518 U. S., at 533.
Our cases provide significant guidance concerning the meaning of this standard and how a reviewing court is to apply it. This Court's instruction concerning the application of heightened scrutiny to sex-based classifications stands in stark contrast to our elucidation of the rudiments of rational basis review. To begin with, under heightened scrutiny, "[t]he burden of justification is demanding and it rests entirely on [the party defending the classification]." Ibid. Under rational basis scrutiny, by contrast, the defender of the classification "has no obligation to produce evidence to sustain the rationality of a statutory classification." Heller v. Doe, 509 U. S. 312, 320 (1993). Instead, "[t]he burden is on the one attacking the legislative arrangement to negative every conceivable basis which might support it, whether or not the basis has a foundation in the record." Id., at 320321 (internal quotation marks and citation omitted).
Further, a justification that sustains a sex-based classification "must be genuine, not hypothesized or invented post hoc in response to litigation." Virginia, supra, at 533. "[T]he mere recitation of a benign, compensatory purpose is not an automatic shield which protects against any inquiry into the actual purposes underlying a statutory scheme." Weinberger v. Wiesenfeld, 420 U. S. 636, 648 (1975). Under rational basis review, by contrast, it is "'constitutionally irrelevant [what] reasoning in fact underlay the legislative decision.''' Railroad Retirement Bd. v. Fritz, 449 U. S. 166, 179 (1980) (quoting Flemming v. Nestor, 363 U. S. 603, 612 (1960)).
Heightened scrutiny does not countenance justifications that "rely on overbroad generalizations about the different talents, capacities, or preferences of males and females." Virginia, supra, at 533. Rational basis review, by contrast, is much more tolerant of the use of broad generalizations about different classes of individuals, so long as the classification is not arbitrary or irrational. See, e. g., Kimel v. Florida Bd. of Regents, 528 U. S. 62, 84 (2000); Fritz, supra, at 177.
Moreover, overbroad sex-based generalizations are impermissible even when they enjoy empirical support. See, e. g., J. E. B., supra, at 139, n. 11; Craig v. Boren, 429 U. S. 190, 199 (1976); Wiesenfeld, supra, at 645. Under rational basis scrutiny, however, empirical support is not even necessary to sustain a classification. See, e. g., FCC v. Beach Communications, Inc., 508 U. S. 307, 315 (1993) ("[A] legislative choice is not subject to courtroom factfinding and may be based on rational speculation unsupported by evidence or empirical data").
The different burdens imposed by these equal protection standards correspond to the different duties of a reviewing court in applying each standard. The court's task in applying heightened scrutiny to a sex-based classification is clear:
"Focusing on the differential treatment or denial of opportunity for which relief is sought, the reviewing court must determine whether the proffered justification is 'exceedingly persuasive.''' Virginia, 518 U. S., at 532-533. In making this determination, the court must inquire into the actual purposes of the discrimination, for "a tenable justification must describe actual state purposes, not rationalizations for actions in fact differently grounded." Id., at 535-536; see also id., at 533; Wiesenfeld, supra, at 648; Califano v. Goldfarb, 430 U. S. 199, 212-217 (1977) (plurality opinion); id., at 219-221 (STEVENS, J., concurring in judgment). The rational basis standard, on the other hand, instructs that "a classification 'must be upheld against equal protection challenge if there is any reasonably conceivable state of facts that could provide a rational basis for the classification.'" Heller, supra, at 320 (quoting Beach Communications, supra, at 313). This standard permits a court to hypothesize interests that might support legislative distinctions, whereas heightened scrutiny limits the realm of justification to demonstrable reality.
These different standards of equal protection review also set different bars for the magnitude of the governmental interest that justifies the statutory classification. Heightened scrutiny demands that the governmental interest served by the classification be "important," see, e. g., Virginia, supra, at 533, whereas rational basis scrutiny requires only that the end be "legitimate," see, e. g., Nordlinger v. Hahn, 505 U. S. 1, 10 (1992).
The most important difference between heightened scrutiny and rational basis review, of course, is the required fit between the means employed and the ends served. Under heightened scrutiny, the discriminatory means must be "substantially related" to an actual and important governmental interest. See, e. g., Virginia, supra, at 533. Under rational basis scrutiny, the means need only be "rationally related" to a conceivable and legitimate state end. See, e. g., Cleburne v. Cleburne Living Center, Inc., 473 U. S. 432, 440 (1985).
The fact that other means are better suited to the achievement of governmental ends therefore is of no moment under rational basis review. See, e. g., Vance v. Bradley, 440 U. S. 93, 103, n. 20 (1979) ("Even were it not irrelevant to [rational basis review] that other alternatives might achieve approximately the same results… "); Massachusetts Bd. of Retirement v. Murgia, 427 U. S. 307, 316 (1976) (per curiam) ("[T]he State perhaps has not chosen the best means to accomplish this purpose. But where rationality is the test, a State 'does not violate the Equal Protection Clause merely because the classifications made by its laws are imperfect'" (quoting Dandridge v. Williams, 397 U. S. 471, 485 (1970))). But because we require a much tighter fit between means and ends under heightened scrutiny, the availability of sex-neutral alternatives to a sex-based classification is often highly probative of the validity of the classification. See, e. g., Wengler, 446 U. S., at 151 (invalidating a sex-based classification where a sex-neutral approach would completely serve the needs of both classes); Orr v. Orr, 440 U. S. 268,281 (1979) (finding "no reason, therefore, to use sex as a proxy for need" where the alimony statute already provided for individualized hearings that took financial circumstances into account); Wiesenfeld, 420 U. S., at 653 (finding a gender-based distinction to be "gratuitous" where "without it, the statutory scheme would only provide benefits to those men who are in fact similarly situated to the women the statute aids").
II
The Court recites the governing substantive standard for heightened scrutiny of sex-based classifications, see ante, at 60-61, 70, but departs from the guidance of our precedents concerning such classifications in several ways. In the first sentence of its equal protection analysis, the majority glosses over the crucial matter of the burden of justification. Ante, at 60 ("For a gender-based classification to withstand equal protection scrutiny, it must be established… "); see also ante, at 70. In other circumstances, the Court's use of an impersonal construction might represent a mere elision of what we have stated expressly in our prior cases. Here, however, the elision presages some of the larger failings of the opinion.
For example, the majority hypothesizes about the interests served by the statute and fails adequately to inquire into the actual purposes of § 1409(a)(4). The Court also does not always explain adequately the importance of the interests that it claims to be served by the provision. The majority also fails carefully to consider whether the sex-based classification is being used impermissibly "as a 'proxy for other, more germane bases of classification,'" Mississippi Univ. for Women, 458 U. S., at 726 (quoting Craig, 429 U. S., at 198), and instead casually dismisses the relevance of available sex-neutral alternatives. And, contrary to the majority's conclusion, the fit between the means and ends of § 1409(a)(4) is far too attenuated for the provision to survive heightened scrutiny. In all, the majority opinion represents far less than the rigorous application of heightened scrutiny that our precedents require.
A
According to the Court, "[t]he first governmental interest to be served is the importance of assuring that a biological parent-child relationship exists." Ante, at 62. The majority does not elaborate on the importance of this interest, which presumably lies in preventing fraudulent conveyances of citizenship. Nor does the majority demonstrate that this is one of the actual purposes of § 1409(a)(4). Assuming that Congress actually had this purpose in mind in enacting parts of § 1409(a)(4), cf. Miller v. Albright, 523 U. S. 420, 435-436 (1998) (opinion of STEVENS, J.), the INS does not appear to rely on this interest in its effort to sustain § 1409(a)(4)'s sexbased classification. Cf. Brief for Respondent 11 (claiming that § 1409 serves "at least two important interests: first, ensuring that children who are born abroad out of wedlock have, during their minority, attained a sufficiently recognized or formal relationship to their United States citizen parent-and thus to the United States-to justify the conferral of citizenship upon them; and second, preventing such children from being stateless"). In light of the reviewing court's duty to "determine whether the proffered justification is 'exceedingly persuasive,'" Virginia, 518 U. S., at 533, this disparity between the majority's defense of the statute and the INS' proffered justifications is striking, to say the least.
The gravest defect in the Court's reliance on this interest, however, is the insufficiency of the fit between § 1409(a)(4)'s discriminatory means and the asserted end. Section 1409(c) imposes no particular burden of proof on mothers wishing to convey citizenship to their children. By contrast, § 1409(a)(1), which petitioners do not challenge before this Court, requires that "a blood relationship between the person and the father [be] established by clear and convincing evidence." Atop § 1409(a)(1), § 1409(a)(4) requires legitimation, an acknowledgment of paternity in writing under oath, or an adjudication of paternity before the child reaches the age of 18. It is difficult to see what § 1409(a)(4) accomplishes in furtherance of "assuring that a biological parent-child relationship exists," ante, at 62, that § 1409(a)(1) does not achieve on its own. The virtual certainty of a biological link that modern DNA testing affords reinforces the sufficiency of § 1409(a)(1). See Miller, supra, at 484-485 (BREYER, J., dissenting).
It is also difficult to see how § 1409(a)(4)'s limitation of the time allowed for obtaining proof of paternity substantially furthers the assurance of a blood relationship. Modern DNA testing, in addition to providing accuracy unmatched by other methods of establishing a biological link, essentially negates the evidentiary significance of the passage of time. Moreover, the application of § 1409(a)(1)'s "clear and convincing evidence" requirement can account for any effect that the passage of time has on the quality of the evidence.
The Court criticizes petitioners' reliance on the availability and sophistication of modern DNA tests, ante, at 63, but appears to misconceive the relevance of such tests. No one argues that § 1409(a)(1) mandates a DNA test. Legitimation or an adjudication of paternity, see §§ 1409(a)(4)(A), (C), may well satisfy the "clear and convincing" standard of § 1409(a)(1). (Satisfaction of § 1409(a)(4) by a written acknowledgment of paternity under oath, see § 1409(a)(4)(B), would seem to do little, if anything, to advance the assurance of a blood relationship, further stretching the means-end fit in this context.) Likewise, petitioners' argument does not depend on the idea that one particular method of establishing paternity is constitutionally required. Petitioners' argument rests instead on the fact that, if the goal is to obtain proof of paternity, the existence of a statutory provision governing such proof, coupled with the efficacy and availability of modern technology, is highly relevant to the sufficiency of the tailoring between § 1409(a)(4)'s sex-based classification and the asserted end. Because § 1409(a)(4) adds little to the work that § 1409(a)(1) does on its own, it is difficult to say that § 1409(a)(4) "substantially furthers" an important governmental interest. Kirchberg, 450 U. S., at 461.
The majority concedes that Congress could achieve the goal of assuring a biological parent-child relationship in a sex-neutral fashion, but then, in a surprising turn, dismisses the availability of sex-neutral alternatives as irrelevant. As the Court suggests, "Congress could have required both mothers and fathers to prove parenthood within 30 days or, for that matter, 18 years, of the child's birth." Ante, at 64 (citing Miller, supra, at 436 (opinion of STEVENS, J.)). Indeed, whether one conceives the majority's asserted interest as assuring the existence of a biological parent-child relationship, ante, at 62, or as ensuring acceptable documentation of that relationship, ante, at 63, a number of sex-neutral arrangements-including the one that the majority offerswould better serve that end. As the majority seems implicitly to acknowledge at one point, ante, at 62, a mother will not always have formal legal documentation of birth because a birth certificate may not issue or may subsequently be lost. Conversely, a father's name may well appear on a birth certificate. While it is doubtless true that a mother's blood re lation to a child is uniquely "verifiable from the birth itself" to those present at birth, ibid., the majority has not shown that a mother's birth relation is uniquely verifiable by the INS, much less that any greater verifiability warrants a sexbased, rather than a sex-neutral, statute.
In our prior cases, the existence of comparable or superior sex-neutral alternatives has been a powerful reason to reject a sex-based classification. See supra, at 78. The majority, however, turns this principle on its head by denigrating as "hollow" the very neutrality that the law requires. Ante, at 64. While the majority trumpets the availability of superior sex-neutral alternatives as confirmation of § 1409(a)(4)'s validity, our precedents demonstrate that this fact is a decided strike against the law. Far from being "hollow," the avoidance of gratuitous sex-based distinctions is the hallmark of equal protection. Cf. J. E. B., 511 U. S., at 152-153 (KENNEDY, J., concurring in judgment) (" 'At the heart of the Constitution's guarantee of equal protection lies the simple command that the Government must treat citizens as individuals, not as simply components of a racial [or] sexual… class'" (quoting Metro Broadcasting, Inc. v. FCC, 497 U. S. 547, 602 (1990) (O'CONNOR, J., dissenting))).
The majority's acknowledgment of the availability of sexneutral alternatives scarcely confirms the point that "[t]he differential treatment is inherent in a sensible statutory scheme." Ante, at 64. The discussion instead demonstrates that, at most, differential impact will result from the fact that "[f]athers and mothers are not similarly situated with regard to the proof of biological parenthood." Ante, at 63. In other words, it will likely be easier for mothers to satisfy a sex-neutral proof of parentage requirement. But facially neutral laws that have a disparate impact are a different animal for purposes of constitutional analysis than laws that specifically provide for disparate treatment. We have long held that the differential impact of a facially neutral law does not trigger heightened scrutiny, see, e. g., Washington v. Davis, 426 U. S. 229 (1976), whereas we apply heightened scrutiny to laws that facially classify individuals on the basis of their sex. See, e. g., United States v. Virginia, 518 U. S. 515 (1996); see also J. E. B., supra, at 152 (KENNEDY, J., concurring in judgment) ("[O]ur case law does reveal a strong presumption that gender classifications are invalid"); Parham v. Hughes, 441 U. S. 347, 351 (1979) (plurality opinion) ("Not all legislation, however, is entitled to the same presumption of validity…. [T]he presumption of statutory validity may also be undermined when a State has enacted legislation creating classes based upon certain other immutable human attributes" (citing, inter alia, Reed v. Reed, 404 U. S. 71 (1971))).
If rational basis scrutiny were appropriate in this case, then the claim that "[t]he Constitution… does not require that Congress elect one particular mechanism from among many possible methods of establishing paternity," ante, at 63, would have much greater force. So too would the claim that "[t]he requirement of § 1409(a)(4) represents a reasonable conclusion…. " Ibid. But fidelity to the Constitution's pledge of equal protection demands more when a facially sex-based classification is at issue. This is not because we sit in judgment of the wisdom of laws in one instance but not the other, cf. Beach Communications, 508 U. S., at 313, but rather because of the potential for "injury… to personal dignity," J. E. B., supra, at 153 (KENNEDY, J., concurring in judgment), that inheres in or accompanies so many sex-based classifications.
B
The Court states that "[t]he second important governmental interest furthered in a substantial manner by § 1409(a)(4) is the determination to ensure that the child and the citizen parent have some demonstrated opportunity or potential to develop not just a relationship that is recognized, as a formal matter, by the law, but one that consists of the real, everyday ties that provide a connection between child and citizen parent and, in turn, the United States." Ante, at 64-65. The Court again fails to demonstrate that this was Congress' actual purpose in enacting § 1409(a)(4). The majority's focus on "some demonstrated opportunity or potential to develop… real, everyday ties" in fact appears to be the type of hypothesized rationale that is insufficient under heightened scrutiny. See supra, at 75-77.
The INS asserts the governmental interest of "ensuring that children who are born abroad out of wedlock have, during their minority, attained a sufficiently recognized or formal relationship to their United States citizen parentand thus to the United States-to justify the conferral of citizenship upon them." Brief for Respondent 11. The majority's asserted end, at best, is a simultaneously watereddown and beefed-up version of this interest asserted by the INS. The majority's rendition is weaker than the INS' in that it emphasizes the "opportunity or potential to develop" a relationship rather than the actual relationship about which the INS claims Congress was concerned. The majority's version is also stronger in that it goes past the formal relationship apparently desired by the INS to "real, everyday ties."
Assuming, as the majority does, that Congress was actually concerned about ensuring a "demonstrated opportunity" for a relationship, it is questionable whether such an opportunity qualifies as an "important" governmental interest apart from the existence of an actual relationship. By focusing on "opportunity" rather than reality, the majority presumably improves the chances of a sufficient means-end fit. But in doing so, it dilutes significantly the weight of the interest. It is difficult to see how, in this citizenship-conferral context, anyone profits from a "demonstrated opportunity" for a relationship in the absence of the fruition of an actual tie. Children who have an "opportunity" for such a tie with a parent, of course, may never develop an actual rela tionship with that parent. See Miller, 523 U. S., at 440 (opinion of STEVENS, J.). If a child grows up in a foreign country without any postbirth contact with the citizen parent, then the child's never-realized "opportunity" for a relationship with the citizen seems singularly irrelevant to the appropriateness of granting citizenship to that child. Likewise, where there is an actual relationship, it is the actual relationship that does all the work in rendering appropriate a grant of citizenship, regardless of when and how the opportunity for that relationship arose.
Accepting for the moment the majority's focus on "opportunity," the attempt to justify § 1409(a)(4) in these terms is still deficient. Even if it is important "to require that an opportunity for a parent-child relationship occur during the formative years of the child's minority," ante, at 68, it is difficult to see how the requirement that proof of such opportunity be obtained before the child turns 18 substantially furthers the asserted interest. As the facts of this case demonstrate, ante, at 57, it is entirely possible that a father and child will have the opportunity to develop a relationship and in fact will develop a relationship without obtaining the proof of the opportunity during the child's minority. After his parents' relationship had ended, petitioner Nguyen lived with the family of his father's new girlfriend. In 1975, before his sixth birthday, Nguyen came to the United States, where he was reared by his father, petitioner Boulais. In 1997, a DNA test showed a 99.98% probability of paternity, and, in 1998, Boulais obtained an order of parentage from a Texas court.
Further underscoring the gap between the discriminatory means and the asserted end is the possibility that "a child might obtain an adjudication of paternity 'absent any affirmative act by the father, and perhaps even over his express objection.''' Miller, 523 U. S., at 486 (BREYER, J., dissenting) (quoting id., at 434 (opinion of STEVENS, J.)). The fact that the means-end fit can break down so readily in theory, and not just in practice, is hardly characteristic of a "substantial" means-end relationship.
Moreover, available sex-neutral alternatives would at least replicate, and could easily exceed, whatever fit there is between § 1409(a)(4)'s discriminatory means and the majority's asserted end. According to the Court, § 1409(a)(4) is designed to ensure that fathers and children have the same "opportunity which the event of birth itself provides for the mother and child." Ante, at 67. Even assuming that this is so, Congress could simply substitute for § 1409(a)(4) a requirement that the parent be present at birth or have knowledge of birth. Cf. Miller, supra, at 487 (BREYER, J., dissenting). Congress could at least allow proof of such presence or knowledge to be one way of demonstrating an opportunity for a relationship. Under the present law, the statute on its face accords different treatment to a mother who is by nature present at birth and a father who is by choice present at birth even though those two individuals are similarly situated with respect to the "opportunity" for a relationship. The mother can transmit her citizenship at birth, but the father cannot do so in the absence of at least one other affirmative act. The different statutory treatment is solely on account of the sex of the similarly situated individuals. This type of treatment is patently inconsistent with the promise of equal protection of the laws. See, e. g., Reed, 404 U. S., at 77 ("By providing dissimilar treatment for men and women who are thus similarly situated, the challenged section violates the Equal Protection Clause").
Indeed, the idea that a mother's presence at birth supplies adequate assurance of an opportunity to develop a relationship while a father's presence at birth does not would appear to rest only on an overbroad sex-based generalization. A mother may not have an opportunity for a relationship if the child is removed from his or her mother on account of alleged abuse or neglect, or if the child and mother are separated by tragedy, such as disaster or war, of the sort apparently present in this case. There is no reason, other than stereotype, to say that fathers who are present at birth lack an opportunity for a relationship on similar terms. The "[p]hysical differences between men and women," Virginia, 518 U. S., at 533, therefore do not justify § 1409(a)(4)'s discrimination.
The majority later ratchets up the interest, for the sake of argument, to "the establishment of a real, practical relationship of considerable substance between parent and child in every case, as opposed simply to ensuring the potential for the relationship to begin." Ante, at 70. But the majority then dismisses the distinction between opportunity and reality as immaterial to the inquiry in this case. Ibid. The majority rests its analysis of the means-end fit largely on the following proposition: "It is almost axiomatic that a policy which seeks to foster the opportunity for meaningful parent-child bonds to develop has a close and substantial bearing on the governmental interest in the actual formation of that bond." Ibid. A bare assertion of what is allegedly "almost axiomatic," however, is no substitute for the "demanding" burden of justification borne by the defender of the classification. Virginia, supra, at 533.
Moreover, the Court's reasoning hardly conforms to the tailoring requirement of heightened scrutiny. The fact that a discriminatory policy embodies the good intention of "seek[ing] to foster" the opportunity for something beneficial to happen is of little relevance in itself to whether the policy substantially furthers the desired occurrence. Whether the classification indeed "has a close and substantial bearing" on the actual occurrence of the preferred result depends on facts and circumstances and must be proved by the classification's defender. Far from being a virtual axiom, the relationship between the intent to foster an opportunity and the fruition of the desired effect is merely a contingent proposition. The majority's sweeping claim is no surrogate for the careful application of heightened scrutiny to a particular classification.
The question that then remains is the sufficiency of the fit between § 1409(a)(4)'s discriminatory means and the goal of "establish[ing]… a real, practical relationship of considerable substance." Ante, at 70. If Congress wishes to advance this end, it could easily do so by employing a sex-neutral classification that is a far "more germane bas[i]s of classification" than sex, Craig, 429 U. S., at 198. For example, Congress could require some degree of regular contact between the child and the citizen parent over a period of time. See Miller, 523 U. S., at 470 (GINSBURG, J., dissenting).
The majority again raises this possibility of the use of sex-neutral means only to dismiss it as irrelevant. The Court admits that "Congress could excuse compliance with the formal requirements when an actual father-child relationship is proved," but speculates that Congress did not do so "perhaps because of the subjectivity, intrusiveness, and difficulties of proof that might attend an inquiry into any particular bond or tie." Ante, at 69. We have repeatedly rejected efforts to justify sex-based classifications on the ground of administrative convenience. See, e. g., Wengler, 446 U. S., at 152; Frontiero, 411 U. S., at 690-691. There is no reason to think that this is a case where administrative convenience concerns are so powerful that they would justify the sex-based discrimination, cf. Wengler, supra, at 152, especially where the use of sex as a proxy is so ill fit to the purported ends as it is here. And to the extent Congress might seek simply to ensure an "opportunity" for a relationship, little administrative inconvenience would seem to accompany a sex-neutral requirement of presence at birth, knowledge of birth, or contact between parent and child prior to a certain age.
The claim that § 1409(a)(4) substantially relates to the achievement of the goal of a "real, practical relationship" thus finds support not in biological differences but instead in a stereotype-i. e., "the generalization that mothers are significantly more likely than fathers… to develop caring relationships with their children." Miller, supra, at 482483 (BREYER, J., dissenting). Such a claim relies on "the very stereotype the law condemns," J. E. B., 511 U. S., at 138 (internal quotation marks omitted), "lends credibility" to the generalization, Mississippi Univ. fo