concurring in part and concurring in the judgment.
I join Parts I, II, and III of the Court’s opinion and concur in the judgment. Despite the messages of bigotry and racism that may be conveyed along with religious connotations by the display of a Ku Klux Klan cross, see ante, at 771 (Thomas, J., concurring), at bottom this case must be understood as it has been presented to us — as a case about private religious expression and whether the State’s relationship to it violates the Establishment Clause. In my view, “the endorsement test asks the right question about governmental practices challenged on Establishment Clause grounds, including challenged practices involving the display of religious symbols,” County of Allegheny v. American Civil Liberties Union, Greater Pittsburgh Chapter, 492 U. S. 573, 628 (1989) (O’Connor, J., concurring in part and concurring in judgment), even where a neutral state policy toward private religious speech in a public forum is at issue. Accordingly, I see no necessity to carve out, as the plurality opinion would today, an exception to the endorsement test for the public forum context.
For the reasons given by Justice Souter, whose opinion I also join, I conclude on the facts of this case that there is “no realistic danger that the community would think that the [State] was endorsing religion or any particular creed,” Lamb’s Chapel v. Center Moriches Union Free School Dist., 508 U. S. 384, 395 (1993), by granting respondents a permit to erect their temporary cross on Capitol Square. I write separately, however, to emphasize that, because it seeks to identify those situations in which government makes “‘adherence to a religion relevant ... to a person’s standing in the political community,’ ” Allegheny, supra, at 594 (quoting *773Lynch v. Donnelly, 465 U. S. 668, 687 (1984) (O’Connor, J., concurring), the endorsement test necessarily focuses upon the perception of a reasonable, informed observer.
I
“In recent years, we have paid particularly close attention [in Establishment Clause cases] to whether the challenged governmental practice either has the purpose or effect of ‘endorsing’ religion, a concern that has long had a place in our Establishment Clause jurisprudence.” Allegheny, supra, at 592. See also Lamb’s Chapel, supra, at 395; School Dist. of Grand Rapids v. Ball, 473 U. S. 373, 390 (1985) (asking “whether the symbolic union of church and state effected by the challenged governmental action is sufficiently likely to be perceived by adherents of the controlling denominations as an endorsement, and by the nonadherents as a disapproval, of their individual religious choices”). A government statement “‘that religion or a particular religious belief is favored or preferred,’” Allegheny, supra, at 593 (quoting Wallace v. Jaffree, 472 U. S. 38, 70 (1985) (O’Connor, J., concurring in judgment), violates the prohibition against establishment of religion because such “[e]n-dorsement sends a message to nonadherents that they are outsiders, not full members of the political community, and an accompanying message to adherents that they are insiders, favored members of the political community,” Lynch, supra, at 688 (O’Connor, J., concurring). See also Allegheny, supra, at 628 (O’Connor, J., concurring in part and concurring in judgment); Wallace, supra, at 69 (O’Connor, J., concurring in judgment). Although “[experience proves that the Establishment Clause . . . cannot easily be reduced to a single test,” Board of Ed. of Kiryas Joel Village School Dist. v. Grumet, 512 U. S. 687, 720 (1994) (O’Connor, J., concurring in part and concurring in judgment), the endorsement inquiry captures the fundamental requirement of the Establishment Clause when courts are called upon to evalu*774ate the constitutionality of religious symbols on public property. See Allegheny, supra, at 593-594.
While the plurality would limit application of the endorsement test to “expression by the government itself,... or else government action alleged to discriminate in favor of private religious expression or activity,” ante, at 764, I believe that an impermissible message of endorsement can be sent in a variety of contexts, not all of which involve direct government speech or outright favoritism. See infra, at 777-778. It is true that neither Allegheny nor Lynch, our two prior religious display cases, involved the same combination of private religious speech and a public forum that we have before us today. Nonetheless, as Justice Souter aptly demonstrates, post, at 786-792, we have on several occasions employed an endorsement perspective in Establishment Clause cases where private religious conduct has intersected with a neutral governmental policy providing some benefit in a manner that parallels the instant case. Thus, while I join the discussion of Lamb’s Chapel and Widmar v. Vincent, 454 U. S. 263 (1981), in Part III of the Court’s opinion, I do so with full recognition that the factors the Court properly identifies ultimately led in each case to the conclusion that there was no endorsement of religion by the State. Lamb’s Chapel, supra, at 395; Widmar, supra, at 274. See also post, at 790-791 (Souter, J., concurring in part and concurring in judgment).
There is, as the plurality notes, ante, at 765, “a crucial difference between government speech endorsing religion, which the Establishment Clause forbids, and private speech endorsing religion, which the Free Speech and Free Exercise Clauses protect.” Board of Ed. of Westside Community Schools (Dist. 66) v. Mergens, 496 U. S. 226, 250 (1990) (plurality opinion). But the quoted statement was made while applying the endorsement test itself; indeed, the sentence upon which the plurality relies was followed immediately by the conclusion that “secondary school students are mature enough and are likely to understand that a school does not *775endorse or support student speech that it merely permits on a nondiscriminatory basis.” Ibid. Thus, as I read the decisions Justice Souter carefully surveys, our prior cases do not imply that the endorsement test has no place where private religious speech in a public forum is at issue. Moreover, numerous lower courts (including the Court of Appeals in this case) have applied the endorsement test in precisely the context before us today. See, e. g., Chabad-Lubavitch of Georgia v. Miller, 5 F. 3d 1383 (CA11 1993) (en banc); Kreisner v. San Diego, 1 F. 3d 775, 782-787 (CA9 1993), cert. denied, 510 U. S. 1044 (1994); Americans United for Separation of Church and State v. Grand Rapids, 980 F. 2d 1538 (CA6 1992) (en banc); Doe v. Small, 964 F. 2d 611 (CA7 1992) (en banc); cf. Smith v. County of Albemarle, 895 F. 2d 953 (CA4), cert. denied, 498 U. S. 823 (1990); Kaplan v. Burlington, 891 F. 2d 1024 (CA2 1989), cert. denied, 496 U. S. 926 (1990). Given this background, I see no necessity to draw new lines where “[r]eligious expression ... (1) is purely private and (2) occurs in a traditional or designated public forum,” ante, at 770.
None of this is to suggest that I would be likely to come to a different result from the plurality where truly private speech is allowed on equal terms in a vigorous public forum that the government has administered properly. That the religious display at issue here was erected by a private group in a public square available “for use by the public ... for free discussion of public questions, or for activities of a broad public purpose,” Ohio Admin. Code Ann. § 128-4-02(A) (1994), certainly informs the Establishment Clause inquiry under the endorsement test. Indeed, many of the factors the plurality identifies are some of those I would consider important in deciding cases like this one where religious speakers seek access to public spaces: “The State did not sponsor respondents’ expression, the expression was made on government property that had been opened to the public for speech, and permission was requested through the *776same application process and on the same terms required of other private groups.” Ante, at 763. And, as I read the plurality opinion, a case is not governed by its proposed per se rule where such circumstances are otherwise — that is, where preferential placement of a religious symbol in a public space or government manipulation of the forum is involved. ■ See ante, at 766.
To the plurality’s consideration of the open nature of the forum and the private ownership of the display, however, I would add the presence of a sign disclaiming government sponsorship or endorsement on the Klan cross, which would make the State’s role clear to the community. This factor is important because, as Justice Souter makes clear, post, at 785-786, certain aspects of the cross display in this case arguably intimate government approval of respondents’ private religious message — particularly that the cross is an especially potent sectarian symbol which stood unattended in close proximity to official government buildings. In context, a disclaimer helps remove doubt about state approval of respondents’ religious message. Cf. Widmar, supra, at 274, n. 14 (“In light of the large number of groups meeting on campus, however, we doubt students could draw any reasonable inference of University support from the mere fact of a campus meeting place. The University’s student handbook already notes that the University’s name will not ‘be identified in any way with the aims, policies, programs, products, or opinions of any organization or its members’ ”). On these facts, then, “the message [of inclusion] is one of neutrality rather than endorsement.” Mergens, supra, at 248 (plurality opinion).
Our agreement as to the outcome of this ease, however, cannot mask the fact that I part company with the plurality on a fundamental point: I disagree that “[i]t has radical implications for our public policy to suggest that neutral laws are invalid whenever hypothetical observers may — even reasonably — confuse an incidental benefit to religion with state *777endorsement.” Ante, at 768. On the contrary, when the reasonable observer would view a government practice as endorsing religion, I believe that it is our duty to hold the practice invalid. The plurality today takes an exceedingly narrow view of the Establishment Clause that is out of step both with the Court’s prior eases and with well-established notions of what the Constitution requires. The Clause is more than a negative prohibition against certain narrowly defined forms of government favoritism, see ante, at 766; it also imposes affirmative obligations that may require a State, in some situations, to take steps to avoid being perceived as supporting or endorsing a private religious message. That is, the Establishment Clause forbids a State to hide behind the application of formally neutral criteria and remain studiously oblivious to the effects of its actions. Governmental intent cannot control, and not all state policies are permissible under the Religion Clauses simply because they are neutral in form.
Where the government’s operation of a public forum has the effect of endorsing religion, even if the governmental actor neither intends nor actively encourages that result, see Lynch, 465 U. S., at 690 (O’Connor, J., concurring), the Establishment Clause is violated. This is so not because of “ ‘transferred endorsement,’ ” ante, at 764, or mistaken attribution of private speech to the State, but because the State’s own actions (operating the forum in a particular manner and permitting the religious expression to take place therein), and their relationship to the private speech at issue, actually convey a message of endorsement. At some point, for example, a private religious group may so dominate a public forum that a formal policy of equal access is transformed into a demonstration of approval. Cf. Mergens, 454 U. S., at 275 (concluding that there was no danger of an Establishment Clause violation in a public university’s allowing access by student religious groups to facilities available to others “[a]t least in the absence of empirical evidence that religious *778groups will dominate [the school’s] open forum”). Other circumstances may produce the same effect — whether because of the fortuity of geography, the nature of the particular public space, or the character of the religious speech at issue, among others. Our Establishment Clause jurisprudence should remain flexible enough to handle such situations when they arise.
In the end, I would recognize that the Establishment Clause inquiry cannot be distilled into a fixed, per se rule. Thus, “[e]very government practice must be judged in its unique circumstances to determine whether it constitutes an endorsement or disapproval of religion.” Lynch, 465 U. S., at 694 (O’Connor, J., concurring). And this question cannot be answered in the abstract, but instead requires courts to examine the history and administration of a particular practice to determine whether it operates as such an endorsement. I continue to believe that government practices relating to speech on religious topics “must be subjected to careful judicial scrutiny,” ibid., and that the endorsement test supplies an appropriate standard for that inquiry.
M HH
Conducting the review of government action required by the Establishment Clause is always a sensitive matter. Unfortunately, as I noted in Allegheny, “even the development of articulable standards and guidelines has not always resulted in agreement among the Members of this Court on the results in individual cases.” 492 U. S., at 623. Today, Justice Stevens reaches a different conclusion regarding whether the Board’s decision to allow respondents’ display on Capitol Square constituted an impermissible endorsement of the cross’ religious message. Yet I believe it is important to note that we have not simply arrived at divergent results after conducting the same analysis. Our fundamental point of departure, it appears, concerns the knowledge that is properly attributed to the test’s “reasonable observer [who] *779evaluates whether a challenged governmental practice conveys a message of endorsement of religion.” Id., at 630 (O’Connor, J., concurring in part and concurring in judgment). In my view, proper application of the endorsement test requires that the reasonable observer be deemed more informed than the casual passerby postulated by Justice Stevens.
Because an Establishment Clause violation must be moored in government action of some sort, and because our concern is with the political community writ large, see Allegheny, supra, at 627 (O’Connor, J., concurring in part and concurring in judgment); Lynch, supra, at 690, the endorsement inquiry is not about the perceptions of particular individuals or saving isolated nonadherents from the discomfort of viewing symbols of a faith to which they do not subscribe. Indeed, to avoid “entirely sweeping] away all government recognition and acknowledgment of the role of religion in the lives of our citizens,” Allegheny, supra, at 623 (O’Connor, J., concurring in part and concurring in judgment), our Establishment Clause jurisprudence must seek to identify the point at which the government becomes responsible, whether due to favoritism toward or disregard for the evident effect of religious speech, for the injection of religion into the political life of the citizenry.
I therefore disagree that the endorsement test should focus on the actual perception of individual observers, who naturally have differing degrees of knowledge. Under such an approach, a religious display is necessarily precluded so long as some passersby would perceive a governmental endorsement thereof. In my view, however, the endorsement test creates a more collective standard to gauge “the ‘objective’ meaning of the [government’s] statement in the community,” Lynch, supra, at 690 (O’Connor, J., concurring). In this respect, the applicable observer is similar to the “reasonable person” in tort law, who “is not to be identified with any ordinary individual, who might occasionally do unreasonable *780things,” but is “rather a personification of a community ideal of reasonable behavior, determined by the [collective] social judgment.” W. Keeton, D. Dobbs, R. Keeton, & D. Owen, Prosser and Keeton on Law of Torts 175 (5th ed. 1984). Thus, “we do not ask whether there is any person who could find an endorsement of religion, whether some people may be offended by the display, or whether some reasonable person might think [the State] endorses religion.” Americans United, 980 F. 2d, at 1544. Saying that the endorsement inquiry should be conducted from the perspective of a hypothetical observer who is presumed to possess a certain level of information that all citizens might not share neither chooses the perceptions of the majority over those of a “reasonable non-adherent,” cf. L. Tribe, American Constitutional Law 1293 (2d ed. 1988), nor invites disregard for the values the Establishment Clause was intended to protect. It simply recognizes the fundamental difficulty inherent in focusing on actual people: There is always someone who, with a particular quantum of knowledge, reasonably might perceive a particular action as an endorsement of religion. A State has not made religion relevant to standing in the political community simply because a particular viewer of a display might feel uncomfortable.
It is for this reason that the reasonable observer in the endorsement inquiry must be deemed aware of the history and context of the community and forum in which the religious display appears. As I explained in Allegheny, “the ‘history and ubiquity’ of a practice is relevant because it provides part of the context in which a reasonable observer evaluates whether a challenged governmental practice conveys a message of endorsement of religion.” 492 U. S., at 630. Nor can the knowledge attributed to the reasonable observer be limited to the information gleaned simply from viewing the challenged display. Today’s proponents of the endorsement test all agree that we should attribute to the observer knowledge that the cross is a religious symbol, that *781Capitol Square is owned by the State, and that the large building nearby is the seat of state government. See post, at 792-793 (Souter, J., concurring in part and concurring in judgment); post, at 806 (Stevens, J., dissenting). In my view, our hypothetical observer also should know the general history of the place in which the cross is displayed. Indeed, the fact that Capitol Square is a public park that has been used over time by private speakers of various types is as much a part of the display’s context as its proximity to the Ohio Statehouse. Cf. Allegheny, supra, at 600, n. 50 (noting-that “[t]he Grand Staircase does not appear-to be the kind of location in which all were free to place their displays- for weeks at a time”). This approach does not require us- to assume an “ ‘ultrareasonable observer’ who understands the vagaries of this Court’s First Amendment jurisprudence,” post, at 807 (Stevens, J., dissenting). An informed member of the community will know how the public space in question has been used in the past — and it is that fact, not that the space may meet the legal definition of a public forum, which is relevant to the endorsement inquiry.
Justice Stevens’ property-based argument fails to give sufficient weight to the fact that the cross at issue here was displayed in a forum traditionally open to the public. “The very fact that a sign is installed on public property,” his dissent suggests, “implies official recognition and reinforcement of its message.” Post, at 801. While this may be the case where a government building and its immediate curtilage are involved, it is not necessarily so with respect to those “places which by long tradition or by government fiat have been devoted to assembly and debate,. . . [particularly] streets and parks which ‘have immemorially been held in trust for the use of the public and, time out of mind, have been used for purposes of assembly, communicating thoughts between citizens, and discussing public questions.’ ” Perry Ed. Assn. v. Perry Local Educators’ Assn., 460 U. S. 37, 45 (1983) (quoting Hague v. Committee for Industrial Organization, 307 *782U. S. 496, 515 (1939)). To the extent there is a presumption that “structures on government property — and, in particular, in front of buildings plainly identified with the State— imply state approval of their message,” post, at 804 (Stevens, J., dissenting), that presumption can be rebutted where the property at issue is a forum historically available for private expression. The reasonable observer would recognize the distinction between speech the government supports and speech that it merely allows in a place that traditionally has been open to a range of private speakers accompanied, if necessary, by an appropriate disclaimer.
In this case, I believe, the reasonable observer would view the Klan’s cross display fully aware that Capitol Square is a public space in which a multiplicity of groups, both secular and religious, engage in expressive conduct. It is precisely this type of knowledge that we presumed in Lamb’s Chapel, 508 U. S., at 395, and in Mergens, 496 U. S., at 250 (plurality opinion). Moreover, this observer would certainly be able to read and understand an adequate disclaimer, which the Klan had informed the State it would include in the display at the time it applied for the permit, see App. to Pet. for Cert. A15-A16; post, at 793-794, n. 1 (Souter, J., concurring in part and concurring in judgment), and the content of which the Board could have defined as it deemed necessary as a condition of granting the Klan’s application. Cf. American Civil Liberties Union v. Wilkinson, 895 F. 2d 1098, 1104-1106 (CA6 1990). On the facts of this case, therefore, I conclude that the reasonable observer would not interpret the State’s tolerance of the Klan’s private religious display in Capitol Square as an endorsement of religion.
III
“To be sure, the endorsement test depends on a sensitivity to the unique circumstances and context of a particular challenged practice and, like any test that is sensitive to context, *783it may not always yield results with unanimous agreement at the margins.” Allegheny, 492 U. S., at 629 (O’Connor, J., concurring in part and concurring in judgment). In my view, however, this flexibility is a virtue and not a vice; “courts must keep in mind both the fundamental place held by the Establishment Clause in our constitutional scheme and the myriad, subtle ways in which Establishment Clause values can be eroded,” Lynch, 465 U. S., at 694 (O’Connor, J., concurring).
I agree that “compliance with the Establishment Clause is a state interest sufficiently compelling to justify content-based restrictions on speech.” Ante, at 761-762. The Establishment Clause “prohibits government from appearing to take a position on questions of religious belief or from ‘making adherence to a religion relevant in any way to a person’s standing in the political community.’” Allegheny, supra, at 593-594 (quoting Lynch, supra, at 687 (O’Connor, J., concurring)). Because I believe that, under the circumstances at issue here, allowing the Klan cross, along with an adequate disclaimer, to be displayed on Capitol Square presents no danger of doing so, I conclude that the State has not presented a compelling justification for denying respondents their permit.