with whom Justice Stevens and Justice Ginsburg join, dissenting.
The First Amendment’s Establishment Clause prohibits Congress (and, by incorporation, the States) from making any law respecting an establishment of religion. It has been *868held to prohibit not only the institution of an official church, but any government act favoring religion, a particular religion, or for that matter irreligión. Thus, it bars the use of public funds for religious aid.
The establishment prohibition of government religious funding serves more than one end. It is meant to guarantee the right of individual conscience against compulsion, to protect the integrity of religion against the corrosion of secular support, and to preserve the unity of political society against the implied exclusion of the less favored and the antagonism of controversy over public support for religious causes.
These objectives are always in some jeopardy since the substantive principle of no aid to religion is not the only limitation on government action toward religion. Because the First Amendment also bars any prohibition of individual free exercise of religion, and because religious organizations cannot be isolated from the basic government functions that create the civil environment, it is as much necessary as it is difficult to draw lines between forbidden aid and lawful benefit. For more than 50 years, this Court has been attempting to draw these lines. Owing to the variety of factual circumstances in whieh the lines must be drawn, not all of the points creating the boundary have enjoyed self-evidence.
So far as the line drawn has addressed government aid to education, a few fundamental generalizations are nonetheless possible. There may be no aid supporting a sectarian school’s religious exercise or the discharge of its religious mission, while aid of a secular character with no discernible benefit to such a sectarian objective is allowable. Because the religious and secular spheres largely overlap in the life of many such schools, the Court has tried to identify some facts likely to reveal the relative religious or secular intent or effect of the government benefits in particular circumstances. We have asked whether the government is acting neutrally in distributing its money, and about the form of the aid itself, its path from government to religious institution, *869its divertibility to religious nurture, its potential for reducing traditional expenditures of religious institutions, and its relative importance to the recipient, among other things.
In all the years of its effort, the Court has isolated no single test of constitutional sufficiency, and the question in every ease addresses the substantive principle of no aid: what reasons are there to characterize this benefit as aid to the sectarian school in discharging its religious mission? Particular factual circumstances control, and the answer is a matter of judgment.
In what follows I will flesh out this summary, for this case comes at a time when our judgment requires perspective on how the Establishment Clause has come to be understood and applied. It is not just that a majority today mistakes the significance of facts that have led to conclusions of unconstitutionality in earlier cases, though I believe the Court commits error in failing to recognize the divertibility of funds to the service of religious objectives. What is more important is the view revealed in the plurality opinion, which espouses a new conception of neutrality as a practically sufficient test of constitutionality that would, if adopted by the Court, eliminate enquiry into a law’s effects. The plurality position breaks fundamentally with Establishment Clause principle, and with the methodology painstakingly worked out in support of it. I mean to revisit that principle and describe the methodology at some length, lest there be any question about the rupture that the plurality view would cause. From that new view of the law, and from a majority’s mistaken application of the old, I respectfully dissent.
I
The prohibition that "Congress shall make no law respecting an establishment of religion,” U. S. Const., Arndt. 1, eludes elegant conceptualization simply because the prohibition applies to such distinct phenomena as state churches and aid to religious schools, and as applied to school aid has *870prompted challenges to programs ranging from construction subsidies to hearing aids to textbook loans. Any criteria, moreover, must not only define the margins of the establishment prohibition, but must respect the succeeding Clause of the First Amendment guaranteeing religion’s free exercise. Ibid. It is no wonder that the complementary constitutional provisions and the inexhaustably various circumstances of their applicability have defied any simple test and have instead produced a combination of general rules often in tension at their edges. If coherence is to be had, the Court has to keep in mind the principal objectives served by the Establishment Clause, and its application to school aid, and their recollection may help to explain the misunderstandings that underlie the majority’s result in this case.
A
At least three concerns have been expressed since the founding and run throughout our First Amendment jurisprudence. First, compelling an individual to support religion violates the fundamental principle of freedom of conscience. Madison’s and Jefferson’s now familiar words establish clearly that liberty of personal conviction requires freedom from coercion to support religion,1 and this means that the government can compel no aid to fund it. Madison put it simply: “[T]he same authority which can force a citizen to *871contribute three pence only of his property for the support of any one establishment, may force him to conform to any other establishment.” Memorial and Remonstrance ¶ 3, reprinted in Everson v. Board of Ed. of Ewing, 330 U. S. 1,64, 65-66 (1947). Any tax to establish religion is antithetical to the command “that the minds of men always be wholly free.” Id., at 12 (discussing Madison’s Memorial and Remonstrance); id., at 13 (noting Jefferson’s belief that “compel[ling] a man to furnish contributions of money for the propagation of opinions which he disbelieves, is sinful and tyrannical;. . . even the forcing him to support this or that teacher of his own religious persuasion, is depriving him of the comfortable liberty of giving his contributions to the particular pastor, whose morals he would make his pattern” (internal quotation marks omitted)); see also Rosenberger v. Rector and Visitors of Univ. ofVa., 515 U. S. 819, 868-874 (1995) (Souter, J., dissenting).
Second, government aid corrupts religion. See Engel v. Vitale, 370 U. S. 421, 431 (1962) (“[The Establishment Clause’s] first and most immediate purpose rested on the belief that a union of government and religion tends to destroy government and to degrade religion”); Everson, supra, at 53 (Rutledge, J., dissenting). Madison argued that establishment of religion weakened the beliefs of adherents so favored, strengthened their opponents, and generated “pride and indolence in the Clergy; ignorance and servility in the laity; [and] in both, superstition, bigotry and persecution.” Memorial and Remonstrance ¶7, quoted in Everson, 330 U. S., at 67. “[Experience witnesseth that ecclesiastical establishments, instead of maintaining the purity and efficacy of Religion, have had a contrary operation.” Ibid. In a variant of Madison’s concern, we have repeatedly noted that a government’s favor to a particular religion or sect threatens to taint it with “corrosive secularism.” Lee v. Weisman, 505 U. S. 577, 608 (1992) (internal quotation marks and citations omitted); see also Illinois ex rel. McCollum v. Board *872of Ed. of School Dish No. 71, Champaign Cty., 333 U. S. 203, 228 (1948).
“[Government and religion have discrete interests which are mutually best served when each avoids too close a proximity to the other. It is not only the nonbeliever who fears the injection of sectarian doctrines and controversies into the civil polity, but in as high degree it is the devout believer who fears the secularization of a creed which becomes too deeply involved with and dependent upon the government.” School Dist. of Abington Township v. Schempp, 374 U. S. 203, 259 (1963) (Brennan, J., concurring).
See also Rosenberger, supra, at 890-891 (Souter, J., dissenting).
Third, government establishment of religion is inextricably linked with conflict. Everson, supra, at 8-11 (relating colonists’ understanding of recent history of religious persecution in countries with established religion); Engel, supra, at 429 (discussing struggle among religions for government approval); Lemon v. Kurtzman, 403 U. S. 602, 623 (1971). In our own history, the turmoil thus produced has led to a rejection of the idea that government should subsidize religious education, id., at 645-649 (opinion of Brennan, J.) (discussing history of rejection of support for religious schools); McCol-lum, supra, at 214-217 (opinion of Frankfurter, J.), a position that illustrates the Court’s understanding that any implicit endorsement of religion is unconstitutional, see County of Allegheny v. American Civil Liberties Union, Greater Pittsburgh Chapter, 492 U. S. 573, 592-594 (1989).2
*873B
These concerns are reflected in the Court’s classic summation delivered in Everson v. Board of Education, swpra, its first opinion directly addressing standards governing aid to religious schools:3
“The ‘establishment of religion’ clause of the First Amendment means at least this: Neither a state nor the Federal Government can set up a church. Neither can pass laws which aid one religion, aid all religions, or prefer one religion over another. Neither can force nor influence a person to go to or to remain away from church against his will or force him to profess a belief or disbelief in any religion. No person can be punished for entertaining or professing religious beliefs or disbeliefs, for church attendance or non-attendance. No tax in any amount, large or small, can be levied to support any religious activities or institutions, whatever they may be called, or whatever form they may adopt to teach or practice religion. Neither a state nor the Federal Government can, openly or secretly, participate in the affairs of any religious organizations or groups and vice versa. In the words of Jefferson, the clause against establishment of religion by law was intended to erect ‘a wall of separation between church and State.’” 330 U.S., at 15-16 (quoting Reynolds v. United States, 98 U. S. 145, 164 (1879)).
The most directly pertinent doctrinal statements here are these: no government “can pass laws which aid one religion *874[or] all religions .... No tax in any amount . . . can be levied to support any religious activities or institutions . . . whatever form they may adopt to teach . .. religion.” 330 U. S., at 16. Thus, the principle of “no aid,” with which no one in Everson disagreed.4
Immediately, however, there was the difficulty over what might amount to “aid” or “support.” The problem for the Everson Court was not merely the imprecision of the words, but the “other language of the [First Amendment that] commands that [government] eannot hamper its citizens in the free exercise of their own religion,” ibid., with the consequence that government must “be a neutral in its relations with groups of religious believers and non-believers,” id., at 18. Since withholding some public benefits from religious groups could be said to “hamper” religious exercise indirectly, and extending other benefits said to aid it, an argument-proof formulation of the no-aid principle was impossible, and the Court wisely chose not to attempt any such thing. Instead it gave definitive examples of public benefits provided pervasively throughout society that would be of some value to organized religion but not in a way or to a degree that could sensibly be described as giving it aid or violating the neutrality requirement: there was no Establishment Clause concern with “such general government services as ordinary police and fire protection, connections for sewage disposal, public highways and sidewalks.” Id., at 17-18. These “benefits of public welfare legislation,” id., at 16, extended in modern times to virtually every member of the population and valuable to every person and association, were the paradigms of advantages that religious organiza*875tions could enjoy consistently with the prohibition against aid, and that governments could extend without deserting their required position of neutrality.
But paradigms are not perfect fits very often, and government spending resists easy classification as between universal general service or subsidy of favoritism. The 5-to-4 division of the Everson Court turned on the inevitable question whether reimbursing all parents for the cost of transporting their children to school was close enough to police protection to tolerate its indirect benefit in some degree to religious schools, with the majority in Everson thinking the reimbursement statute fell on the lawful side of the line. Although the state scheme reimbursed parents for transporting children to sectarian schools, among others, it gave “no money to the schools. It [did] not support them. Its legislation [did] no more than provide a general program to help parents get their children, regardless of their religion, safely and expeditiously to and from accredited schools.” Id., at 18. The dissenters countered with factual analyses showing the limitation of the law’s benefits in fact to private school pupils who were Roman Catholics, id., at 20 (Jackson, J., dissenting), and indicating the inseparability of transporting pupils to school from support for the religious instruction that was the school’s raison d’etre, id., at 45-46 (Rutledge, J., dissenting).
Everson is usefully understood in the light of a successor case two decades later, Board of Ed. of Central School Dist. No. 1 v. Allen, 392 U. S. 236 (1968), in which the challenged government practice was lending textbooks to pupils of schools both public and private, including religious ones (as to which there was no evidence that they had previously supplied books to their classes and some evidence that they had not, id., at 244, n. 6). By the time of Allen, the problem of classifying the state benefit, as between aid to religion and general public service consistent with government neutral*876ity, had led to the formulation of a “test” that required secular, primary intent and effect as necessary conditions of any permissible scheme. Id., at 243. Again the Court split, upholding the state law in issue, but with Everson’s majority author, Justice Black, now in dissent. What is remarkable about Allen today, however, is not so much its division as its methodology, for the consistency in the way the Justices went about deciding the ease transcended their different conclusions. Neither side rested on any facile application of the “test” or any simplistic reliance on the generality or even-handedness of the state law. Disagreement concentrated on the true intent inferrable behind the law, the feasibility of distinguishing in fact between religious and secular teaching in church schools, and the reality or sham of lending books to pupils instead of supplying books to schools. The majority, to be sure, cited the provision for books to all schoolchildren, regardless of religion, 392 U. S., at 243, just as the Everson majority had spoken of the transportation reimbursement as going to all, 330 U. S., at 16, in each case for the sake of analogy to the provision of police and fire services.5 But the stress was on the practical significance of the actual benefits received by the schools. As Everson had rested on the understanding that no money and no support went to the school, id., at 18, Allen emphasized that the savings to parents were devoid of any measurable effect in teaching religion, 392 U. S., at 243-244. Justice Harlan, concurring, summed up the approach with his observations that the required government “[neutrality is ... a coat of many colors,” and quoted Justice Goldberg’s conclusion, that there was “ 'no simple and clear measure’... by which this or any [religious school aid] case may readily be decided,” id., at 249 (quoting Schempp, 374 U. S., at 306).
*877After Everson and Allen, the state of the law applying the Establishment Clause to public expenditures producing some benefit to religious schools was this:
1. Government aid to religion is forbidden, and tax revenue may not be used to support a religious school or religious teaching.
2. Government provision of such paradigms of universally general welfare benefits as police and fire protection does not count as aid to religion.
3. Whether a law’s benefit is sufficiently close to universally general welfare paradigms to be classified with them, as distinct from religious aid, is a function of the purpose and effect of the challenged law in all its particularity. The judgment is not reducible to the application of any formula. Evenhandedness of distribution as between religious and secular beneficiaries is a relevant factor, but not a sufficiency test of constitutionality. There is no rule of religious equal protection to the effect that any expenditure for the benefit of religious school students is necessarily constitutional so long as public school pupils are favored on ostensibly identical terms.
4. Government must maintain neutrality as to religion, “neutrality” being a conclusory label for the required position of government as neither aiding religion nor impeding religious exercise by believers. “Neutrality” was not the name of any test to identify permissible action, and in particular, was not synonymous with even-handedness in conferring benefit on the secular as well as the religious.
Today, the substantive principle of no aid to religious mission remains the governing understanding of the Establishment Clause as applied to public benefits inuring to religious schools. The governing opinions on the subject in the 85 years since Allen have never challenged this principle. The *878eases have, however, recognized that in actual Establishment Clause litigation over school aid legislation, there is no pure aid to religion and no purely secular welfare benefit; the effects of the laws fall somewhere in between, with the judicial task being to make a realistic allocation between the two possibilities. The Court’s decisions demonstrate its repeated attempts to isolate considerations relevant in classifying particular benefits as between those that do not discernibly support of threaten support of a school’s religious mission, and those that cross or threaten to cross the line into support for religion.
II
A
The most deceptively familiar of those considerations is “neutrality,” the presence or absence of which, in some sense, we have addressed from the moment of Everson itself. I say “some sense,” for we have used the term in at least three ways in our cases, and an understanding of the term’s evolution will help to explain the concept as it is understood today, as well as the limits of its significance in . Establishment Clause analysis. “Neutrality’ has been employed as a term to describe the requisite state of government equipoise between the forbidden encouragement and discouragement of religion; to characterize a benefit or aid as secular; and to indicate evenhandedness in distributing it.
As already mentioned, the Court first referred to neutrality in Everson, simply stating that government is required “to be a neutral” among religions and between religion and nonreligion. 330 U. S., at 18.. Although “neutral” may have carried a hint of inaction when we indicated that the First Amendment “does not require the state to be [the] adversary” of religious believers, ibid., or to cut off general government services from religious organizations, Everson provided no explicit definition of the term or further indication of what the government was required to do or not do to be *879“neutral” toward religion. In practical terms, “neutral” in Everson was simply a term for government in its required median position between aiding and handicapping religion. The second major ease on aid to religious schools, Allen, used “neutrality” to describe an adequate state of balance between government as ally and as adversary to religion, see 392 U. S., at 242 (discussing line between “state neutrality to religion and state support of religion”). The term was not further defined, and a few subsequent school cases used “neutrality” simply to designate the required relationship to religion, without explaining how to attain it. See, e. g., Tilton v. Richardson, 403 U. S. 672, 677 (1971) (describing eases that “see[k] to define the boundaries of the neutral area between [the Religion Clauses] within which the legislature may legitimately act”); Roemer v. Board of Public Works of Md., 426 U. S. 736,747 (1976) (plurality opinion of Blackmun, J.) (“Neutrality is what is required. The State must confine itself to secular objectives, and neither advance nor impede religious activity. Of course, that principle is more easily stated than applied”); see also Committee for Public Ed. & Religious Liberty v. Nyquist, 413 U. S. 756, 782 (1973) (describing “neutral posture” toward religion); Roemer, swpra, at 745-746 (opinion of Blackmun, J.) (“The Court has enforced a scrupulous neutrality by the State, as among religions, and also as between religious and other activities”); cf. Wolman v. Walter, 433 U. S. 229, 254 (1977) (quoting Lemon and noting difficulty of religious teachers’ remaining “‘religiously neutral’ ”).
The Court began to employ “neutrality” in a sense different from equipoise, however, as it explicated the distinction between “religious” and “secular” benefits to religious schools, the latter being in some circumstances permissible. See infra, at 884-899 (discussing considerations). Even though both Everson and Allen had anticipated some such distinction, neither ease had used the term “neutral” in this way. In Everson, Justice Black indicated that providing *880police, fire, and similar government services to religious institutions was permissible, in part because they were “so separate and so indisputably marked off from the religious function.” 330 U. S., at 18. Allen similarly focused on the fact that the textbooks lent out were “secular” and approved by secular authorities, 392 U. S., at 245, and assumed that the secular textbooks and the secular elements of education they supported were not so intertwined with religious instruction as “in fact [to be] instrumental in the teaching of religion,” id., at 248. Such was the Court’s premise in Lemon for shifting the use of the word “neutral” from labeling the required position of the government to describing a benefit that was nonreligious. We spoke of “[o]ur decisions from Everson to Allen [as] permitting] the States to provide church-related schools with secular, neutral, or nonideological services, facilities, or materials,” 403 U. S., at 616, and thereafter, we regularly used “neutral” in this second sense of “secular” or “nonreligious.” See, e. g., Tilton, supra, at 687-688 (characterizing subsidized teachers in Lemon as “not necessarily religiously neutral,” but buildings as “religiously neutral”); Meek v. Pittenger, 421 U. S. 349, 365-366 (1975) (describing instructional materials as “ ‘secular, nonideologi-eal and neutral’ ” and “wholly neutral”); id., at 372 (describing auxiliary services as “religiously neutral”); Roemer, supra, at 751 (opinion of Blaekmun, J.) (describing Tilton’s approved buildings as “neutral or nonideologieal in nature”); 426 U. S., at 754 (describing Meek’s speech and hearing services as “neutral and nonideologieal”); Zobrest v. Catalina Foothills School Dish, 509 U. S. 1, 10 (1993) (discussing translator as “neutral service”); Agostini v. Felton, 521 U. S. 203, 232 (1997) (discussing need to assess whether nature of aid was “neutral and nonideologieal”); cf. Levitt v. Committee for Public Ed. & Religious Liberty, 413 U. S. 472, 478 (1973) (noting that District Court approved testing cost reimbursement as payment for services that were “ ‘secular, neutral, or nonideologieal’ ” in character, citing Lemon, 403 U. S., at *881616); Wolman, supra, at 242 (quoting Lemon, supra, at 616 (describing permitted services aid as “secular, neutral, or nonideologieal”)).
The shift from equipoise to secular was not, however, our last redefinition, for the Court again transformed the sense of “neutrality” in the 1980’s. Reexamining and reinterpreting Everson and Allen, we began to use the word “neutral” to mean “evenhanded,” in the sense of allocating aid on some common basis to religious and secular recipients. Again, neither Everson nor Allen explicitly used “neutral” in this manner, but just as the label for equipoise had lent itself to referring to the secular characteristic of what a government might provide, it was readily adaptable to referring to the generality of government services, as in Everson’s paradigms, to which permissible benefits were compared.
The increased attention to a notion of evenhanded distribution was evident in Nyquist, where the Court distinguished the program under consideration from the government services approved in Allen and Everson, in part because “the class of beneficiaries [in Everson and Allen] included all schoolchildren, those in public as well as those in private schools.” 413 U. S., at 782, n. 38. Nyquist then reserved the question whether “some form of public assistance . . . made available generally without regard to the sectarian-nonsectarian, or public-nonpublie nature of the institution benefitted” would be permissible. Id., at 783, n. 38 (citations omitted). Subsequent eases continued the focus on the “generality^ of the approved government services as an important characteristic. Meek, for example, characterized Everson and Allen as approving “a general program” to pay bus fares and to lend school books, respectively, 421 U. S., at 360; id., at 360, n. 8 (approving two similar “general program[s]” in New York and Pennsylvania), and Wolman upheld diagnostic services described as “ ‘general welfare services for children,’” 433 U.S., at 243 (quoting Meek, supra, at 371, n. 21).
*882Justice Blaekmun, writing in Roemer, first called such a “general” or evenhanded program “neutral,” in speaking of “facial neutrality’ as a relevant consideration in determining whether there was an Establishment Clause violation. “[RJeligious institutions need not be quarantined from public benefits that are neutrally available to all.” 426 U. S., at 746-747; see also id., at 746 (discussing buses in Everson and school books in Allen as examples of “neutrally available” aid). In Mueller v. Allen, 463 U. S. 388 (1983), the Court adopted the redefinition of neutrality as evenhandedness, citing Nyquist, 413 U. S., at 782, n. 38, and alluding to our discussion of equal access in Widmar v. Vincent, 454 U. S. 263 (1981). The Court upheld a system of tax deductions for sectarian educational expenses, in part because such a “facially neutral law,” 463 U. S., at 401, made the deduction available for “all parents, including those whose children attend public schools and those whose children attend nonsectarian private schools or sectarian private schools,” id., at 397. Subsequent cases carried the point forward. See, e. g., Witters v. Washington Dept. of Servs. for Blind, 474 U. S. 481, 487 (1986) (quoting Nyquist and characterizing program as making aid “available generally”); Zobrest, swpra, at 8-9 (discussing “government programs that neutrally provide benefits to a broad class of citizens defined without reference to religion” and citing Mueller and Witters); Agos-tini, supra, at 231 (discussing aid allocated on the basis of “neutral, secular criteria that neither favor nor disfavor religion, . . . made available to both religious and secular beneficiaries on a nondiscriminatory basis”); see also Rosen-berger, 515 U. S., at 839 (“[T]he guarantee of neutrality is respected, not offended, when the government, following neutral criteria and evenhanded policies, extends benefits to recipients whose ideologies and viewpoints, including religious ones, are broad and diverse”).
In sum, “neutrality” originally entered this field of jurisprudence as a eonelusory term, a label for the required rela*883tionship between the government and religion as a state of equipoise between government as ally and government as adversary. Reexamining Everson’s paradigm cases to derive a prescriptive guideline, we first determined that "neutral” aid was secular, nonideologieal, or unrelated to religious education. Our subsequent reexamination of Everson and Allen, beginning in Nyquist and culminating in Mueller and most recently in Agostini, recast neutrality as a concept of "evenhandedness.”
There is, of course, good reason for considering the generality of aid and the evenhandedness of its distribution in making close calls between benefits that in purpose or effect support a school’s religious mission and those that do not. This is just what Everson did. Even when the disputed practice falls short of Everson’s paradigms, the breadth of evenhanded distribution is one pointer toward the law’s purpose, since on the face of it aid distributed generally and without a religious criterion is less likely to be meant to aid religion than a benefit going only to religious institutions or people. And, depending on the breadth of distribution, looking to evenhandedness is a way of asking whether a benefit can reasonably be seen to aid religion in fact; we do not regard the postal system as aiding religion, even though parochial schools get mail. Given the legitimacy of considering evenhandedness, then, there is no reason to avoid the term “neutrality” to refer to it. But one crucial point must be borne in mind.
In the days when "neutral” was used in Everson’s sense of equipoise, neutrality was tantamount to constitutionality; the term was eonclusory, but when it applied it meant that the government’s position was constitutional under the Establishment Clause. This is not so at all, however, under the most recent use of "neutrality” to refer to generality or evenhandedness of distribution. This kind of neutrality is relevant in judging whether a benefit scheme so characterized should be seen as aiding a sectarian school’s religious *884mission, but this neutrality is not alone sufficient to qualify the aid as constitutional. It is to be considered only along with other characteristics of aid, its administration, its recipients, or its potential that have been emphasized over the years as indicators of just how religious the intent and effect of a given aid scheme really is. See, e. g., Tilton, 403 U. S., at 677-678 (opinion of Burger, C. J.) (acknowledging “no single constitutional caliper”); Meek, 421 U. S., at 358-359 (noting considerations as guidelines only and discussing them as a matter of degree); School Dist. of Grand Rapids v. Ball, 473 U. S. 373, 383 (1985) (quoting Meek), overruled in part by Agostini, 521 U. S., at 203; Board of Ed. of Kiryas Joel Village School Dist. v. Grumet, 512 U. S. 687, 720 (1994) (opinion of O’Connor, J.) (“Experience proves that the Establishment Clause, like the Free Speech Clause, cannot easily be reduced to a single test”); Rosenberger, 515 U. S., at 847-849 (O’Connor, J., concurring) (discussing need for line-drawing); id., at 852 (noting lack of a single “Grand Unified Theory” for Establishment Clause and citing Kiryas Joel)', ef. Agostini, supra, at 232-233 (examining a variety of factors). Thus, the basic principle of establishment scrutiny of aid remains the principle as stated in Everson, that there may be no public aid to religion or support for the religious mission of any institution.
B
The insufficiency of evenhandedness neutrality as a standalone criterion of constitutional intent or effect has been clear from the beginning of our interpretative efforts, for an obvious reason. Evenhandedness in distributing a benefit approaches the equivalence of constitutionality in this area only when the term refers to such universality of distribution that it makes no sense to think of the benefit as going to any discrete group. Conversely, when evenhandedness refers to distribution to limited groups within society, like groups of schools or schoolchildren, it does make sense to regard the benefit as aid to the recipients. See, e. g., Everson, 330 U. S., *885at 16 (discussing aid that approaches the “verge” of forbidden territory); Lemon, 403 U. S., at 612 (“[W]e can only dimly perceive the lines of demarcation in this extraordinarily sensitive area of constitutional law”); Nyquist, 413 U. S., at 760-761 (noting the “most perplexing questions” presented in this area and acknowledging “‘entangling] precedents’”); Mueller, 463 U. S., at 393 (quoting Lemon); Witters, 474 U. S., at 485 (quoting Lemon).
Hence, if we looked no further than evenhandedness, and failed to ask what activities the aid might support, or in fact did support, religious schools could be blessed with government funding as massive as expenditures made for the benefit of their public school counterparts, and religious missions would thrive on public money. This is why the consideration of less than universal neutrality has never been recognized as dispositive and has always been teamed with attention to other facts bearing on the substantive prohibition of support for a school’s religious objective.
At least three main lines of enquiry addressed particularly to school aid have emerged to complement evenhandedness neutrality. First, we have noted that two types of aid recipients heighten Establishment Clause concern: pervasively religious schools and primary and secondary religious schools. Second, we have identified two important characteristics of the method of distributing aid: directness or indirectness of distribution and distribution by genuinely independent choice. Third, we have found relevance in at least five characteristics of the aid itself: its religious content; its cash form; its divertibility or actually diversion to religious support; its supplantation of traditional items of religious school expense; and its substantiality.
1
Two types of school aid recipients have raised special concern. First, we have recognized the fact that the overriding religious mission of certain schools, those sometimes called *886“pervasively sectarian,” is not confined to a discrete element of the curriculum, Everson, 330 U. S., at 22-24 (Jackson, J., dissenting); id., at 45-47 (Rutledge, J., dissenting), but permeates their teaching.6 Walz v. Tax Comm’n of City of New York, 397 U. S. 664, 671 (1970); Lemon, supra, at 636-637 (“A school which operates to commingle religion with other instruction plainly cannot completely secularize its instruction. Parochial schools, in large measure, do not accept the assumption that secular subjects should be unrelated to religious teaching”); see also Bowen v. Kendrick, 487 U. S. 589, 621-622 (1988) (discussing pervasively sectarian private schools). Based on record evidence and long experience, we have concluded that religious teaching in such schools is at the core of the instructors’ individual and personal obligations, cf. Canon 803, §2, Text & Commentary 568 (“It is necessary that the formation and education given in a Catholic school be based upon the principles of Catholic doctrine; teachers are to be outstanding for their correct doctrine and integrity of life”), and that individual religious teachers will teach religiously.7 Lemon, 403 U. S., at 615-620; id., at 635-*887641 (Douglas, J., concurring); Levitt, 413 U. S., at 480; Meek, 421 U. S., at 369-371; Wolman, 433 U. S., at 249-250 (discussing nonseverability of religious and secular education); Ball, 473 U. S., at 399-400 (O’Connor, J., concurring in judgment in part and dissenting in part), overruled in part by Agostini, 521 U. S., at 236. As religious teaching cannot be separated from secular education in such schools or by such teachers, we have concluded that direct government subsidies to such schools are prohibited because they will inevitably and impermissibly support religious indoctrination. Zobrest, 509 U. S., at 12 (discussing Meek and Ball).
Second, we have expressed special concern about aid to primary and secondary religious schools. Tilton, 403 U. S., at 685-686. On the one hand, we have understood how the youth of the students in such schools makes them highly susceptible to religious indoctrination. Lemon, supra, at 616 (“This process of inculcating religious doctrine is, of course, enhanced by the impressionable age of the pupils, in primary schools particularly”). On the other, we have recognized that the religious element in the education offered in most sectarian primary and secondary schools is far more intertwined with the secular than in university teaching, where the natural and academic skepticism of most older students may separate the two, see Tilton, supra, at 686-689; Roermer, 426 U. S., at 750. Thus, government benefits accruing to these pervasively religious primary and secondary schools raise special dangers of diversion into support for the religious indoctrination of children and the involvement of government in religious training and practice.
*8882
We have also evaluated the portent of support to an organization’s religious mission that may be inherent in the method by which aid is granted, finding pertinence in at least two characteristics of distribution. First, we have asked whether aid is direct or indirect, observing distinctions between government schemes with individual beneficiaries and those whose beneficiaries in the first instance might be religious schools. Everson, supra, at 18 (bus fare supports parents and not schools); Allen, 392 U. S., 243-244, and n. 6 (textbooks go to benefit children and parents, not schools); Lemon, supra, at 621 (invalidating direct aid to schools); Levitt, supra, at 480, 482 (invalidating direct testing aid to schools); Witters, 474 U. S., at 487-488 (evaluating whether aid was a direct subsidy to schools). Direct aid obviously raises greater risks, although recent eases have discounted this risk factor, looking to other features of the distribution mechanism. Agostini, supra, at 225-226.8
*889Second, we have distinguished between indirect aid that reaches religious schools only incidentally as a result of numerous individual choices and aid that is in reality directed to religious schools by the government or in practical terms selected by religious schools themselves. Mueller, 463 U. S., at 399; Witters, supra, at 488; Zobrest, supra, at 10. In these cases, we have declared the constitutionality of programs providing aid directly to parents or students as tax deductions or scholarship money, where such aid may pay for education at some sectarian institutions, Mueller, supra, at 399; Witters, 474 U. S., at 488, but only as the result of “genuinely independent and private choices of aid recipients,” id., at 487. We distinguished this path of aid from the route in Ball and Wolman, where the opinions indicated that “[wjhere ... no meaningful distinction can be made between aid to the student and aid to the school, the concept of a loan to individuals is a transparent fiction.” 474 U. S., at 487, n. 4 (citations and internal quotation marks omitted).9
3
In addition to the character of the school to which the benefit accrues, and its path from government to school, a number of features of the aid itself have figured in the elassifica-*890tions we have made. First, we have barred aid with actual religious content, which would obviously run afoul of the ban on the government’s participation in religion, Everson, 330 U. S., at 16; Walz, 397 U. S., at 668; cf. Lemon, 403 U. S., at 617 (discussing variable ideological and religious character of religious teachers compared to fixed content of books). In eases where we have permitted aid, we have regularly characterized it as “neutral” in the sense (noted supra, at 879-881) of being without religious content. See, e. g., Tilton, 403 U. S., at 688 (characterizing buildings as “religiously neutral”); Zobrest, 509 U. S., at 10 (describing translator as “neutral service”); Agostini, 521 U. S., at 232 (discussing need to assess whether nature of aid was “neutral and nonideological”). See also ante, at 820 (plurality opinion) (barring aid with religious content).10
Second, we have long held government aid invalid when circumstances would allow its diversion to religious education. The risk of diversion is obviously high when aid in the form of government funds makes its way into the coffers of religious organizations, and so from the start we have understood the Constitution to bar outright money grants of aid to religion.11 See Everson, 830 U. S., at 16 (“[The State] *891cannot consistently with the ‘establishment of religion’ clause of the First Amendment contribute tax-raised funds to the support of an institution which teaches the tenets and faith of any church”); id., at 18 (“The State contributes no money to the schools. It does not support them”); Allen, 892 U. S., at 243-244 (“[N]o funds or books are furnished to parochial schools, and the financial benefit is to parents and children, not schools”); Walz, supra, at 675 (“Obviously a direct money subsidy would be a relationship pregnant with involvement and, as with most governmental grant programs, could encompass sustained and detailed administrative relationships for enforcement of statutory or administrative standards”); Lemon, supra, at 612 (identifying “three main evils” against which Establishment Clause was to protect as “sponsorship, financial support, and active involvement of the sovereign in religious activity,” citing Walz); 403 U. S., at 621 (distinguishing direct financial aid program from Everson and Allen and noting problems with required future surveillance); Nyquist, 413 U. S., at 762, 774 (striking down “direct money grants” for maintaining buildings because there was no attempt to restrict payments to those expenditures related exclusively to secular purposes); Levitt, 413 U. S., at 480, 482 (striking down “direct money grant” for testing expenses);12 Hunt v. *892McNair, 413 U. S. 734, 745, n. 7 (1973) (noting approved aid is “no expenditure of public funds, either by grant or loan”); Wolman, 433 U. S., at 239, and n. 7 (noting that “statute does not authorize any payment to nonpublic school personnel for the costs of administering the tests”); Agostini, 521 U. S., at 228-229 (emphasizing that approved services are not “distributed ‘directly to the religious schools.’ ... No Title I funds ever reach the coffers of religious schools, and Title I services may not be provided to religious schools on a school-wide basis” (citations omitted)); Bowen, 487 U. S., at 614-615; Rosenberger, 515 U. S., at 842 (noting that “we have recognized special Establishment Clause dangers where the government makes direct money payments to sectarian institutions”); ef. Lemon, 403 U. S., at 619-620 (noting that safeguards and accounting inspections required to prevent government funds from supporting religious education will cause impermissible entanglement); Roemer, 426 U. S., at 753-757 (approving segregated funds after finding recipients not pervasively religious); Ball, 473 U. S., at 392-393 (noting that “[w]ith but one exception, our subsequent eases have struck down attempts by States to make payments out of *893public tax dollars directly to primary or secondary religious educational institutions”)? overruled in part by Agostini, supra, at 236; Witters, 474 U. S., at 487 (“It is equally well-settled . .. that the State may not grant aid to a religious school, whether cash or in kind, where the effect of the aid is that of a direct subsidy to the religious school” (internal quotation marks and citations omitted)); Rosenberger, supra, at 851-852 (O’CONNOR, J., concurring) (noting that student fee was not a tax).
Divertibility is not, of course, a characteristic of cash alone, and when examining provisions for ostensibly secular supplies we have considered their susceptibility to the service of religious ends.13 In upholding a scheme to provide students with secular textbooks, we emphasized that “each book loaned must be approved by the public school authorities; only secular books may receive approval.” Allen, 392 U. S., at 244-245; see also Meek, 421 U. S., at 361-362 (opinion of Stewart, J.); Wolman, supra, at 237-238. By the same token, we could not sustain provisions for instructional materials adaptable to teaching a variety of subjects.14 Meek, supra, at 363; Wolman, supra, at 249-250. While the textbooks had a known and fixed secular content not readily di~ *894vertible to religious teaching purposes, the adaptable materials did not.15 So, too, we explained the permissibility of busing on public routes to schools but not busing for field trips designed by religious authorities specifically because the latter trips were components of teaching in a pervasively religious school. Compare Everson, 330 U. S., at 17 (noting wholly separate and secular nature of public bus fare to schools), with Wolmcm, 433 U. S., at 254 (“The field trips are an integral part of the educational experience, and where the teacher works within and for a sectarian institution, an unacceptable risk of fostering of religion is an inevitable byproduct” (citation omitted)). We likewise were able to uphold underwriting the expenses of standard state testing in religious schools while being forced to strike down aid for testing designed by the school officials, because the latter tests could be used to reinforce religious teaching. Compare id., at 240 (“[T]he State provides both the schools and the school district with the means of ensuring that the minimum standards are met. The nonpublic school does not control the content of the test or its result. This serves to prevent the use of the test as , part of religious teaching, and thus avoids that kind of direct aid to religion found present in Levitt”); Committee for Public Ed. and Religious Liberty v. Regan, 444 U. S. 646, 661-662 (1980) (same), with Levitt, 413 U. S., at 480 (“We cannot ignore the substantial risk that these examinations, prepared by teachers under the authority of réligious institutions, will be drafted with an eye, unconsciously or otherwise, to inculcate students in the religious precepts of the sponsoring church”).
*895With the same point in mind, we held that buildings constructed with government grants to universities with religious affiliation must be barred from religious use indefinitely to prevent the diversion of government funds to religious objectives. Tilton, 403 U. S., at 683 (plurality opinion) (“If, at the end of 20 years, the building is, for example, converted into a chapel or otherwise used to promote religious interests, the original federal grant will in part have the effect of advancing religion. To this extent the Act therefore trespasses on the Religion Clauses”); see also Hunt, 413 U. S., at 743-744. We were accordingly constrained to strike down aid for repairing buildings of noiipub-lic schools because they could be used for religious education. Nyquist, 413 U. S., at 776-777.
Divertibility was, again, the issue in an order remanding an as-applied challenge to a grant supporting counseling on teenage sexuality for findings that the aid had not been used to support religious education. Bowen, 487 U. S., at 621; see also id., at 623 (O’Connor, J., concurring). And the most recent example of attention to the significance of divertibility occurred in our explanation that public school teachers could be assigned to provide limited instruction in religious schools in Agostini, 521 U. S., at 223-227, a majority of the Court rejecting the factual assumption that public school teachers could be readily lured into providing religious instruction.16
*896Third, our cases have recognized the distinction, adopted by statute in the Chapter 2 legislation, between aid that merely supplements and aid that supplants expenditures for offerings at religious schools, the latter being barred. Although we have never adopted the position that any benefit that flows to a religious school is impermissible because it frees up resources for the school to engage in religious indoctrination, Hunt, supra, at 743, from our first decision holding it permissible to provide textbooks for religious schools we have repeatedly explained the unconstitutionality of aid that supplants an item of the school’s traditional expense. See, e. g., Cochran v. Louisiana Bd. of Ed., 281 U. S. 370, 375 (1930) (noting that religious schools “are not the beneficiaries of these appropriations. They obtain nothing from them, nor are they relieved of a single obligation because of them” (internal quotation marks omitted)); Everson, 330 U. S., at 18 (specifically noting that bus fare program did not support or fund religious schools); Allen, 392 U. S., at 244 (stating that “the financial benefit [of providing the textbooks] is to parents and children, not to schools” (footnote omitted)); id., *897at 244, n. 6 (explicitly recognizing that “the record contains no evidence that any of the private schools in appellants’ districts previously provided textbooks for their students”); Lemon, 403 U. S., at 656 (opinion of Brennan, J.) (noting no aid to schools was involved in Allen). We ignored this prohibition only once, in Regan, 444 U. S., at 646; see also ante, at 16, n. 7, where reimbursement for budgeted expenses of required testing was not struck down, but we then quickly returned to the rule as a guideline for permissible aid.17 In Zobrest, 509 U. S., at 12, the Court specifically distinguished Meek and Ball by explaining that the invalid programs in those eases “relieved sectarian schools of costs they otherwise would have borne in educating their students.” In Agostini, the Court made a point of noting that the objects of the aid were “by law supplemental to the regular curricula” and, citing Zobrest, explained that the remedial education services did not relieve the religious schools of costs they would otherwise have borne. 521 U. S., at 228 (citing Zobrest, supra, at 12). The Court explicitly stated that the *898services in question did not “supplant the remedial instruction and guidance counseling already provided in New York City’s sectarian schools.” 521 U. S., at 229.
Finally, we have recognized what is obvious (however imprecise), in holding “substantial” amounts of aid to be unconstitutional whether or not a plaintiff can show that it supplants a specific item of expense a religious school would have borne.18 In Meek, 421 U. S., at 366, we invalidated the loan of instructional materials to religious schools because “faced with the substantial amounts of direct support authorized by [the program], it would simply ignore reality to attempt to separate secular educational functions from the predominantly religious role performed by many of Pennsylvania’s church-related elementary and secondary schools and then characterize [the program] as channeling aid to the secular without providing direct aid to the sectarian.” Id., at 365. See id., at 366 (“Substantial aid to the educational function of such schools ... necessarily results in aid to the sectarian school enterprise as a whole”); see also Nyquist, 413 U. S., at 783; Wolman, 433 U. S., at 250-251. In Witters, 474 U. S., at 488, the Court asked whether the aid in question was a direct subsidy to religious schools and addressed the substantiality of the aid obliquely in noting that “nothing in the record indicates that. . . any significant portion of the *899aid expended under the Washington program as a whole will end up flowing to religious education.” In Zobrest, supra, at 12, the Court spoke of the substantiality test in Meek, noting that “[disabled children, not sectarian schools, are the primary beneficiaries of the [Individuals with Disabilities Act (IDEA)]; to the extent sectarian schools benefit at all from the IDEA, they are only incidental beneficiaries.”
C
This stretch of doctrinal history leaves one point clear beyond peradventure: together with James Madison we have consistently understood the Establishment Clause to impose a substantive prohibition against public aid to religion and, hence, to the religious mission of sectarian schools. Even-handedness neutrality is one, nondispositive pointer toward an intent and (to a lesser degree) probable effect on the permissible side of the line between forbidden aid and general public welfare benefit. Other pointers are facts about the religious mission and education level of benefited schools and their pupils, the pathway by which a benefit travels from public treasury to educational effect, the form and content of the aid, its adaptability to religious ends, and its effects on school budgets. The object of all enquiries into such matters is the same whatever the particular circumstances: is the benefit intended to aid in providing the religious element of the education and is it likely to do so?
The substance of the law has thus not changed since Ever-son. Emphasis on one sort of fact or another has varied depending on the perceived utility of the enquiry, but all that has been added is repeated explanation of relevant considerations, confirming that our predecessors were right in their prophecies that no simple test would emerge to allow easy application of the establishment principle.
The plurality, however, would reject that lesson. The majority misapplies it.
*900III
A
The nub of the plurality’s new position is this:
“[I]f the government, seeking to further some legitimate secular purpose, offers aid on the same terms, without regard to religion, to all who adequately further that purpose, then it is fair to say that any aid going to a religious recipient only has the effect of furthering that secular purpose. The government, in crafting such an aid program, has had to conclude that a given level of aid is necessary to further that purpose among secular recipients and has provided no more than that same level to religious recipients.” Ante, at 810 (citation omitted).
As a break with consistent doctrine the plurality’s new criterion is unequaled in the history of Establishment Clause interpretation. Simple on its face, it appears to take even-handedness neutrality and in practical terms promote it to a single and sufficient test for the establishment constitutionality of school aid. Even on its own terms, its errors are manifold, and attention to at least three of its mistaken assumptions will show the degree to which the plurality’s proposal would replace the principle of no aid with a formula for generous religious support.
First, the plurality treats an external observer’s attribution of religious support to the government as the sole impermissible effect of a government aid scheme. See, e. g., ante, at 809 (“[N]o one would conclude that any indoctrination that any particular recipient conducts has been done at the behest of the government”). While perceived state endorsement of religion is undoubtedly a relevant concern under the Establishment Clause, see, e. g., Allegheny County, 492 U. S., at 592-594; see also Capitol Square Review and Advisory Bd. v. Pinette, 515 U. S. 753, 772-774 (1995) (O’Connor, J., concurring in part and concurring in *901judgment); id., at 786-787 (Souter, J., concurring in part and concurring in judgment), it is certainly not the only one. Everson made this clear from the start: secret aid to religion by the government is also barred. 330 U. S., at 16. State aid not attributed to the government would still violate a taxpayer’s liberty of conscience, threaten to corrupt religion, and generate disputes over aid. In any event, since the same-terms feature of the scheme would, on the plurality’s view, rule out the attribution or perception of endorsement, adopting the plurality’s rule of facial evenhandedness would convert neutrality into a dispositive criterion of establishment constitutionality and eliminate the effects enquiry directed by Allen, Lemon, and other eases. Under the plurality’s rule of neutrality, if a program met the first part of the Lemon enquiry, by declining to define a program’s recipients by religion, it would automatically satisfy the second, in supposedly having no impermissible effect of aiding religion.19
Second, the plurality apparently assumes as a fact that equal amounts of aid to religious and nonreligious schools will have exclusively secular and equal effects, on both external perception and on incentives to attend different schools. See ante, at 809-810, 813-814. But there is no reason to believe that this will be the case; the effects of same-terms aid may not be confined to the secular sphere at all. This is the reason that we have long recognized that unrestricted aid to religious schools will support religious teaching in ad*902dition to secular education, a fact that would be true no matter what the supposedly secular purpose of the law might be.
Third, the plurality assumes that per capita distribution rules safeguard the same principles as independent, private choices. But that is clearly not so. We approved university scholarships in Witters because we found them close to giving a government employee a paycheck and allowing him to spend it as he chose, but a per capita aid program is a far cry from awarding scholarships to individuals, one of whom makes an independent private choice. Not the least of the significant differences between per capita aid and aid individually determined and directed is the right and genuine opportunity of the recipient to choose not to give the aid.20 To hold otherwise would be to license the government to donate funds to churches based on the number of their members, on the patent fiction of independent private choice.
The plurality’s mistaken assumptions explain and underscore its sharp break with the Framers’ understanding of establishment and this Court’s consistent interpretative course. Under the plurality’s regime, little would be left of the right of conscience against compelled support for religion; the more massive the aid the more potent would be the influence of the government on the teaching mission; the more generous the support, the more divisive would be the resentments of those resisting religious support, and those religions without school systems ready to claim their fair share.
B
The plurality’s conception of evenhandedness does not, however, control the case, whose disposition turns on the misapplication of accepted categories of school aid analysis. The facts most obviously relevant to the Chapter 2 scheme *903in Jefferson Parish are those showing divertibility and actual diversion in the circumstance of pervasively sectarian religious schools. The type of aid, the structure of the program, and the lack of effective safeguards clearly demonstrate the divertibility of the aid. While little is known about its use, owing to the anemic enforcement system in the parish, even the thin record before us reveals that actual diversion occurred.
The aid that the government provided was highly susceptible to unconstitutional use. Much of the equipment provided under Chapter 2 was not of the type provided for individual students, App. to Pet. for Cert. 140a; App. 262a-278a, but included “slide projectors, movie projectors, overhead projectors, television sets, tape recorders, projection screens, maps, globes, filmstrips, cassettes, computers,” and computer software and peripherals, Helms v. Cody, No. 85-5533, 1990 WL 36124 (ED La., Mar. 27, 1990); App. to Pet. for Cert. 140a; App. 90a, 262a-278a, as well as library books and materials, id., at 56a, 126a, 280a-284a. The videocassette players, overhead projectors, and other instructional aids were of the sort that we have found can easily be used by religious teachers for religious purposes. Meek, 421 U. S., at 363; Wolman, 433 U. S., at 249-250. The same was true of the computers, which were as readily employable for religious teaching as the other equipment, and presumably as immune to any countervailing safeguard, App. 90a, 118a, 164a-165a. Although library books, like textbooks, have fixed content, religious teachers can assign seeular library books for religious critique, and books for libraries may be religious, as any divinity school library would demonstrate. The sheer number and variety of boohs that could be and were ordered gave ample opportunity for such diversion.
The divertibility thus inherent in the forms of Chapter 2 aid was enhanced by the structure of the program in Jefferson Parish. Requests for specific items under Chapter 2 came not from seeular officials, ef. Allen, 392 U. S., at 244-*904245, but from officials of the religious schools (and even parents of religious school pupils), see ante, at 803 (noting that private religious schools submitted their orders to the government for specific requested items); App. 156a-158a. The sectarian schools decided what they wanted and often ordered the supplies, id., at 156a-159a, 171a-172a, to be forwarded directly to themselves, id., at 156a-159a. It was easy to select whatever instructional materials and library books the schools wanted, just as it was easy to employ computers for the support of the religious content of the curriculum infused with religious instruction.
The concern with divertibility thus predicated is underscored by the fact that the religious schools in question here covered the primary and secondary grades, the grades in which the sectarian nature of instruction is characteristically the most pervasive, see Lemon, 403 U. S., at 616; cf. Tilton, 403 U. S., at 686-689, and in which pupils are the least critical of the schools’ religious objectives, see Lemon, swpra, at 616. No one, indeed, disputes the trial judge’s findings, based on a detailed record, that the Roman Catholic schools,21 which made up the majority of the private schools participating,22 were pervasively sectarian,23 that *905their common objective and mission was to engage in religious education,24 and that their teachers taught religiously,25 making them precisely the kind of primary and *906secondary religious schools that Taise the most serious Establishment Clause concerns. See Watz, 397 U. S., at 671; Hunt, 413 U. S., at 743; Lemon, supra, at 636-637. The threat to Establishment Clause values wáfe accordingly at its highest in the circumstances of this case. Such precautionary features as there were in the Jefferson Parish scheme were grossly inadequate to counter the threat. To be sure, the disbursement of the aid was subject to statutory admonitions against diversion, see, e. g., 20 U. S. C. §§7332, 8897, and was supposedly subject to a variety of safeguards, see ante, at 802-803, 832-834. But the provisions for onsite monitoring visits, labeling of government property, and government oversight cannot be accepted as sufficient in the face of record evidence that the safeguard provisions proved to be empty phrases in Jefferson Parish. Cf. Agos-tini, 521 U. S., at 228-229; Zobrest, 509 U. S., at 13 (accepting precautionary provisions in absence of evidence of them uselessness).
The plurality has already noted at length the ineffectiveness of the government’s monitoring program. Ante, at 832-834; see also App. Illa (“A system to monitor nonpublie schools was often not in operation and therefore the [local educational agency] did not always know: (a) what was purchased or (b) how it was utilized”). Monitors visited a non-publie school only sporadically, discussed the program with a single contact person, observed nothing more than attempts at recordkeeping, and failed to inform the teachers of the restrictions involved. Id., at 154a-155a. Although Chapter 2 required labeling of government property, it occurred haphazardly at best, id., at 113a, and the government’s sole monitoring system for computer use amounted to nothing more *907than questioning school officials and examining the location of computers at the schools, id., at 118a. No records of software and computer use were kept, and no such recordkeep-ing was even planned. Id., at 118a, 164a-166a. State and local officials in Jefferson Parish admitted that nothing prevented the Chapter 2 computers from being used for religious instruction, id., at 102a, 118a, 164a-166a, and although. they knew of methods of monitoring computer usage, such as locking the computer functions, id., at 165a-166a, they implemented no particular policies, instituted no systems, and employed no technologies to minimize the likelihood of diversion to religious uses,26 id., at 118a, 165a-166a. The watchdogs did require the religious schools to give not so much as an assurance that they would use Chapter 2 computers solely for secular purposes, Helms v. Picard, 151 F. 3d 347, 368 (1998), amended, 165 F. 3d 311 (CA5 1999); App. 94a-95a. Government officials themselves admitted that there was no way to tell whether instructional materials had been diverted, id., at 118a, 139a, 144a-145a, and, as the plurality notes, the only screening mechanism in the library book scheme was a review of titles by a single government official, ante, at 832-833, n. 15; see App. 137a. The government did not even have a policy on the consequences of noncompliance. Id., at 145a.
The risk of immediate diversion of Chapter 2 benefits had its complement in the risk of future diversion, against which the Jefferson Parish program had absolutely no protection. By statute all purchases with Chapter 2 aid were to remain the property of the United States, 20 U.S.C. §7372(e)(1), merely being “lent” to the recipient nonpublic schools. In actuality, however, the record indicates that nothing in the *908Jefferson Parish program stood in the way of giving the Chapter 2 property outright to the religious schools when it became older. Although old equipment remained the property of the local education agency, a local government administrative body, one agency employee testified that there was no set policy for dealing with old computers, which were probably given outright to the religious schools. App. 161a-162a. The witness said that government-funded instructional materials, too, were probably left with the religious schools when they were old, and that it was unclear whether library books were ever to be returned to the government. Ibid.
Providing such governmental aid without effective safeguards against future diversion itself offends the Establishment Clause, Tilton, 403 U.S., at 682-684; Nyquist, 413 U. S., at 776-777, and even without evidence of actual diversion, our cases have repeatedly held that a “substantial risk” of it suffices to invalidate a government aid program on establishment grounds. See, e. g., Wolman, 433 U. S., at 254 (invalidating aid for transportation on teacher-accompanied field trips because an “unacceptable risk of fostering of religion” was “an inevitable byproduct”); Meek, 421 U. S., at 372 (striking down program because of a “potential for impermissible fostering of religion”); Levitt, 413 U. S., at 480 (invalidating aid for tests designed by religious teachers because of “the substantial risk that. .. examinations, prepared by teachers under the authority of religious institutions, will be drafted with an eye, unconsciously or otherwise, to inculcate students in the religious preeepts of the sponsoring church”); Lemon, 403 U. S., at 619 (finding invalid aid with a “potential for impermissible fostering of religion”); ef. Bowen, 487 U. S., at 621 (noting that where diversion risk is less clearly made out, a case may be remanded for findings on actual diversion of aid to religious indoctrination); Began, 444 U. S., at 656 (characterizing as “minimal” the chance that state-drafted tests with “complete” safeguards would be adopted to reli*909gious testing). A substantial risk of diversion in this ease was more than clear, as the plurality has conceded. The First Amendment was violated.
But the record here goes beyond risk, to instances of actual diversion. What one would expect from such paltry efforts at monitoring and enforcement naturally resulted, and the record strongly suggests that other, undocumented diversions probably occurred as well. First, the record shows actual diversion in the library book, program. App. 132a-133a. Although only limited evidence exists, it contrasts starkly with the records of the numerous textbook programs that we have repeatedly upheld, where there was no evidence of any actual diversion. See Allen, 392 U. S., at 244-245; Meek, supra, at 361-362; Wolman, supra, at 237-238. Here, discovery revealed that under Chapter 2, nonpublic schools requested and the government purchased at least 191 religious books with taxpayer funds by December 1985.27 App. 133a. Books such as A Child’s Book of Prayers, id., at 84a, and The Illustrated Life of Jesus, id., at 132a, *910were discovered among others that had been ordered under the program. See also id., at 59a-62a.
The evidence persuasively suggests that other aid was actually diverted as well. The principal of one religious school testified, for example, that computers lent with Chapter 2 funds were joined in a network with other non-Chapter 2 computers in some schools, and that religious officials and teachers were allowed to develop their own unregulated software for use on this network. Id., at 77a. She admitted that the Chapter 2 computer took over the support of the computing system whenever there was a breakdown of the master computer purchased with the religious school’s own funds. Ibid. Moreover, as the plurality observes, ante, at 833-834, n. 17, comparing the records of considerable federal funding of audiovisual equipment in religious schools with records of the schools’ use of unidentified audiovisual equipment in religion classes strongly suggests that film projectors and videotape machines purchased with public funds were used in religious indoctrination over a period of at least seven years. App. 205a, 210a, 206a-207a; see also id., at 108a (statement of second-grade teacher indicating that she used audiovisual materials in all classes).
Indeed, the plurality readily recognizes that the aid in question here was divertible and that substantial evidence of actual diversion exists. Ante, at 832-834, and nn. 14-17. Although Justice O’Connor attributes limited significance to the evidence of divertibility and actual diversion, she also recognizes that it exists. Ante, at 864-865 (opinion concurring in judgment). The Court has no choice but to hold that the program as applied violated the Establishment Clause.28
*911IV
The plurality would break with the law. The majority misapplies it. That misapplication is, however, the only consolation in the case, which reaches an erroneous result but does not stage a doctrinal coup. But there is no mistaking the abandonment of doctrine that would occur if the plurality were to become a majority. It is beyond question that the plurality’s notion of evenhandedness neutrality as a practical guarantee of the validity of aid to sectarian schools would be the end of the principle of no aid to the schools’ religious mission. And if that were not so obvious it would become so after reflecting on the plurality’s thoughts about diversion *912and about giving attention to the pervasiveness of a school’s sectarian teaching.
The plurality is candid in pointing out the extent of actual diversion of Chapter 2 aid to religious use in the case before us, ante, at 882-834, and n. 17, and equally candid in saying it does not matter, ante, at 820-825, 833-834. To the plurality there is nothing wrong with aiding a school’s religious mission; the only question is whether religious teaching obtains its tax support under a formally evenhanded criterion of distribution. The principle of no aid to religious teaching has no independent significance.
And if this were not enough to prove that no aid in religious school aid is dead under the plurality’s First Amendment, the point is nailed down in the plurality’s attack on the legitimacy of considering a school’s pervasively sectarian character when judging whether aid to the school is likely to aid its religious mission. Ante, at 826-829. The relevance of this consideration is simply a matter of common sense: where religious indoctrination pervades school activities of children and adolescents, it takes great care to be able to aid the school without supporting the doctrinal effort. This is obvious. The plurality nonetheless condemns any enquiry into the pervasiveness of doctrinal content as a remnant of anti-Catholie bigotry (as if evangelical Protestant schools and Orthodox Jewish yeshivas were never pervasively sectarian29), and it equates a refusal to aid religious schools with hostility to religion (as if aid to religious teaching were not *913opposed in this very case by at least one religious respondent30 and numerous religious amici curiae31 in a tradition claiming descent from Roger Williams). My concern with these arguments goes not so much to their details32 as it does to the fact that the plurality’s choice to employ imputations of bigotry and irreligión as terms in the Court’s debate makes one point clear: that in rejecting the principle of no aid to a school’s religious mission the plurality is attacking the most fundamental assumption underlying the Establishment Clause, that government can in fact operate with neutrality in its relation to religion. I believe that it can, and so respectfully dissent.
Jefferson’s Virginia Bill for Establishing Religious Freedom provided “[t]hat no man shall be compelled to frequent or support any religious worship, place, or ministry whatsoever-” Jefferson, A Bill for Establishing Religious Freedom, in 5 The Founder’s Constitution 84 (R Kurland & R. Lerner eds. 1987); see also Rosenberger v. Rector and Visitors of Univ. of Va., 515 U. S. 819, 870-872 (1995) (Souter, J., dissenting). We have “previously recognized that the provisions of the First Amendment, in the drafting and adoption of which Madison and Jefferson played such leading roles, had the same objective and were intended to provide the same protection against governmental intrusion on religious liberty as the Virginia statute.” Everson v. Board of Ed. of Ewing, 330 U. S. 1, 13 (1947).
The plurality mistakes my recognition of this fundamental concern. Ante, at 825-826. The Court may well have moved away from considering the political divisiveness threatened by particular instances of aid as a practical criterion for applying the Establishment Clause case by case, but we have never questioned its importance as a motivating concern behind the Establishment Clause, nor could we change history to find that sectarian conflict did not influence the Framers who wrote it.
The Court upheld payments by Indian tribes to apparently Koman Catholic schools in Quick Bear v. Leupp, 210 U. S. 50 (1908), suggesting in dicta that there was no Establishment Clause problem, but it did not squarely face the question. Nor did the Court address a First Amendment challenge to a state program providing textbooks to children in Cochran v. Louisiana Bd. of Ed., 281 U. S. 370 (1930); it simply concluded that the program had an adequate public purpose. The Court first squarely faced the issue in Everson.
While Everson’s dissenters parted company with the majority over the specific question of school buses, the Court stood as one behind the principle of no aid for religious teaching. 330 U. S., at 15-16; id., at 25-26 (Jackson, J., dissenting); id., at 28-29, 31-32 (Rutledge, J., dissenting).
In fact, religious education in Roman Catholic schools is defined as part of required religious practice; aiding it is thus akin to aiding a church service. See 1988 Code of Canon Law, Canon 798, reprinted in The Code of Canon Law: A Text and Commentary 566 (1985) (hereinafter Text & Commentary) (directing parents to entrust children to Roman Catholic schools or otherwise provide for Roman Catholic education); Canon 800, §2, Text & Commentary 567 (requiring the faithful to support establishment and maintenance of Roman Catholic schools); Canons 802, 804, Text & Commentary 567, 568 (requiring diocesan bishop to establish and regulate schools “imparting an education imbued with the Christian spirit”).
Although the Court no longer assumes that public school teachers assigned to religious schools for limited purposes will teach religiously, see Agostini v. Felton, 521 U. S. 203, 228-228 (1997), we have never abandoned the presumption that religious teachers will teach just that way. Lemon v. Kurtzman, 403 U. S. 602, 615-620 (1971); id., at 635-641 (Douglas, J., concurring); Levitt v. Committee for Public Ed. & Religious Liberty, 413 U. S. 472, 480 (1973); Meek v. Pittenger, 421 U. S. 349, 369-371 (1975); Wol-*887man v. Walter, 433 U. S. 229, 249-250 (1977); School Dist. of Grand Rapids v. Ball, 473 U. S. 373, 399-400 (1985) (O’Connor, J., concurring in judgment in part and dissenting in part), overruled in part by Agostini, supra, at 236. Cf. NLRB v. Catholic Bishop of Chicago, 440 U. S. 490, 504 (1979) (“The church-teacher relationship in a church-operated school differs from the employment relationship in a public or other nonreligious school”).
In Agostini, the Court indicated that “we have departed from the rule relied on in Ball that all government aid that directly assists the educational function of religious schools is invalid,” 521 U. S., at 225, and cited Witters v. Washington Dept. of Servs. for Blind, 474 U. S. 481 (1986), and Zobrest v. Catalina Foothills School Dist., 509 U. S. 1 (1998). However, Agostini did not rely on this dictum, instead clearly stating that “[w]hile it is true that individual students may not directly apply for Title I services, it does not follow from this premise that those services are distributed ‘directly to the religious schools.’ In fact, they are not. No Title I funds ever reach the coffers of religious schools, and Title I services may not be provided to religious schools on a sehoolwide basis.” 521 U. S., at 228-229 (citations omitted). Until today, this Court has never permitted aid to go directly to schools on a schoolwide basis.
The plurality misreads our precedent in suggesting that we have abandoned directness of distribution as a relevant consideration. See ante, at 815-818. In Wolman, we stated that nominally describing aid as to students would not bar a court from finding that it actually provided a subsidy to a school, 433 U. S., at 250, but we did not establish that a program giving “direct” aid to schools was therefore permissible. In Witters, we made the focus of Wolman dear, continuing to examine aid to determine *889if it was a “direct subsidy” to a school, 474 U. S., at 487, and distinguishing the aid at issue from impermissible aid in Ball and Wolman precisely because the designation of the student as recipient in those eases was only nominal. 474 U. S., at 487, n. 4. Our subsequent cases have continued to ask whether government aid programs constituted impermissible “direct subsidies” to religious schools even where they are directed by individual choice. Zobrest, supra, at 11-13; Mueller v. Allen, 463 U. S. 388, 399 (1983); Agostini, supra, at 226.
We have also permitted the government to supply students with public-employee translators, Zobrest, supra, at 10, and public-employee special education teachers, Agostini, supra, at 226,228, who directly provided .them with government services in whatever schools those specific students attended, public or nonpublic. I have already noted Agostini’s limitations. See n. 8, supra.
1 agree with the plurality that the Establishment Clause absolutely prohibits the government from providing aid with dear religious content to religious, or for that matter nonreligious, schools. Ante, at 822-825. The plurality, however, misreads our precedent as focusing only on affirmatively religious content. At the very least, a building, for example, has no such content, but we have squarely required the government to ensure that no publidy financed building be diverted to religious use. Tilton v. Richardson, 403 U. S. 672, 681-684 (1971). See also Bowen v. Kendrick, 487 U. S. 589, 623 (1988) (O’CONNOR, J., concurring) AJny use of public funds to promote religious doctrines violates the Establishment Clause”).
We have similarly noted that paying salaries of parochial school teachers creates too much of a risk that such support will aid the teaching of religion, striking down such programs because of the need for pervasive monitoring that would be required. See Lemon, 403 U. S., at 619 (“We do not assume, however, that parochial school teachers will be unsuccessful *891in their attempts to segregate their religious beliefs from their secular educational responsibilities. But the potential for impermissible fostering of religion is present. The [state legislature] has not, and could not, provide state aid on the basis of a mere assumption that secular teachers under religious discipline can avoid conflicts. The State must be certain, given the Religion Clauses, that subsidized teachers do not inculcate religion- A comprehensive, discriminating, and continuing state surveillance will inevitably be required to ensure that these restrictions are obeyed and the First Amendment otherwise respected”).
It is true that we called the importance of the cash payment consideration into question in Committee for Public Ed. and Religious Liberty v. Regan, 444 U. S. 646, 657-659 (1980) (approving program providing religious school with “direct cash reimbursement” for expenses of standardized testing). In that case, we found the other safeguards against the diversion of such funds to religious uses sufficient to allow such aid: “A *892contrary view would insist on drawing a constitutional distinction between paying the nonpublic school to do the grading and paying state employees or some independent service to perform that task, even though the grading function is the same regardless of who performs it and would not have the primary effect of aiding religion whether or not performed by nonpublic school personnel.” Id., at 658. Aside from this isolated circumstance, where we found ironclad guarantees of nondiversion, we have never relaxed our prohibition on direct cash aid to pervasively religious schools, and have in fact continued to acknowledge the concern. See Agostini, 521U. S., at 228-229; cf. Rosenberger, 515 U. S., at 842.
The plurality concedes this basic point. See ante, at 818-819. Given this, I find any suggestion that this prohibition has been undermined by Mueller or Witters without foundation. See ante, at 819-820, n. 8. Those cases involved entirely different types of aid, namely, tax deductions and individual scholarship aid for university education, see also n. 16, infra, and were followed by Rosenberger and Agostini, which continued to support this absolute restriction.
I reject the plurality’s argument that divertibility is a boundless principle. Ante, at 824-825. Our long experience of evaluating this consideration demonstrates its practical limits. See infra this page and 894-895. Moreover, the Establishment Clause charges us with making such enqui-ries, regardless of their difficulty. See supra, at 875-878, 884-885. Finally, the First Amendment’s rule permitting only aid with fixed secular content seems no more difficult to apply than the plurality’s rule prohibiting only aid with fixed religious content.
Contrary to the plurality’s apparent belief, Lamb’s Chapel v. Center Moriches Union Free School Dist., 508 U. S. 384 (1993), sheds no light on the question of divertibility and school aid. Ante, at 822, n. 9. The Court in that case clearly distinguished the question of afterschool access to public facilities from anything resembling the school aid cases: “The showing of this film series would not have been during school hours, would not have been sponsored by the school, and would have been open to the public, not just to chinch members.” 508 U. S., at 395.
In Lemon, we also specifically examined the risk that a government program that paid religious teachers would support religious education; the teachers posed the risk of being unable to separate secular from religious education. Although we invalidated the program on entanglement grounds, we suggested that the monitoring the State had established in that ease was actually required to eliminate the risk of diversion. See 403 U. S., at 619; see also n. 11, supra.
The plurality is mistaken in its reading of Zobrest. See ante, at 820-821. Zobrest does not reject the principle of divertibility. There the government provided only a translator who was not considered divertible because he did not add to or subtract from the religious message. The Court approved the translator as it would approve a hearing aid, health services, diagnostics, and tests. See Zobrest, 509 U. S., at 13, and n. 10. Cf Bradfield v. Roberts, 175 U. S. 291, 299-300 (1899); Wolman, 433 U. S., at 244. Zobrest thus can be thought of as akin to our approval of diagnostic services in Wolman, supra, at 244, which we considered to have “little or no educational content[,] not [to be] closely associated with the educational mission of the nonpublie school,” and not to pose “an impermissible *896risk of the fostering of ideological views.” The fact that the dissent saw things otherwise (as the plurality points out, ante, at 821) is beside the point here.
Similarly, the plurality is mistaken in reading our holdings in Mueller and Witters, see ante, at 821, to undermine divertibility as a relevant principle. First, these cases approved quite factually distinct types of aid; Mueller involving tax deductions, which have a quite separate history of approval, see 468 U. S., at 396, and nn. 5, 6 (citing Walz v. Tax Comm’n of City of New York, 397 U. S. 664 (1970», and Witters involving scholarship money distributed to a university, not a primary or secondary school, see Tilton, 403 U. S., at 685-686, that was not significant enough as a whole to support that institution, Witters, 474 U. S., at 488. Second, in neither case did the program at issue provide direct aid on a schoolwide basis (as Chapter 2 does here); in both we found a distinction based on the genuinely independent, private choices which allocated such very different types of aid (tax deductions and university scholarship money that did not amount to substantial support of the university). See Mueller, supra, at 399; Witters, supra, at 488.
Our departure from this principle in Regan is not easily explained, but it is an isolated holding surrounded by otherwise unbroken adherence to the no-supplanting principle. Long after Regan we have continued to find the supplement/supplant distinction, like the bar to substantial aid, to be an important consideration. See Zobrest, supra, at 12; Agostini, 521 U. S., at 228; c£ Witters, supra, at 487-488 (discussing rule against "direct subsidy”). The weight that the plurality places on Regan is thus too much for it to bear. See ante, at 815, n. 7. Moreover, the apparent object of the Regan Court's concern was vindicating the principle that aid with fixed secular content was permissible, distinguishing it from the divertible testing aid in Levitt. Regan, 444 U. S., at 661-662 (citing Wolman, supra, at 263); c£ Levitt, 413 U. S., at 480. The plurality provides no explanation for our continued reference to the principle of no-supplanting aid in subsequent cases, such as Zobrest and Agostini, which it finds trustworthy guides elsewhere in its discussion of the First Amendment. See ante, at 822-823, 825, 827, 829-832. Nor does the plurality explain why it places so much weight on Regan’s apparent departure from the no-supplanting rule while it ignores Regan’s core reasoning that the testing aid there was permissible because, in direct contrast to Levitt, the aid was not divertible.
1 do not read the plurality to question the prohibition on substantial aid. The plurality challenges any rule based on the proportion of aid that a program provides to religious recipients, citing Witters and Agostini. See ante, at 812, n. 6. I reject the plurality’s reasoning. The plurality misreads Witters; Justice Marshall, writing for the Court in Witters, emphasized that only a small amount of aid was provided to religious institutions, 474 U. S., at 488, and no controlling majority rejected the importance of this fact. The plurality also overreads Agostini, supra, at 229, which simply declined to adopt a rule based on proportionality. Moreover, regardless of whether the proportion of aid actually provided to religious schools is relevant, we have never questioned our holding in Meek that substantial aid to religious schools is prohibited.
Adopting the plurality’s rule would permit practically any government aid to religion so long as it could be supplied on terms ostensibly comparable to the terms under which aid was provided to nonreligious recipients. As a principle of constitutional sufficiency, the manipulability of this rule is breathtaking. A legislature would merely need to state a secular objective in order to legalize massive aid to all religions, one religion, or even one sect, to which its largess could be directed through the easy exercise of crafting facially neutral terms under which to offer aid favoring that religious group. Short of formally replacing the Establishment Clause, a more dependable key to the public fisc or a cleaner break with prior law would be difficult to imagine.
Indeed, the opportunity for an individual to choose not to have her religious school receive government aid is just what at least one of the respondents seeks here. See Brief for Respondents 1, and n. 1.
Litigation, discovery, and the opinions below focused almost exclusively on the aid to the 34 Roman Catholic schools. Consequently, I will confine my discussion to that information. Of course, the same concerns would be raised by government aid to religious schools of other faiths that a court found had similar missions of religious education and religious teachers teaching religiously.
The Jefferson Parish Chapter 2 program included 46 nonpublic schools, of which 41 were religiously affiliated. Thirty-four of these were Roman Catholic, seven others were religiously affiliated, and five were not religiously affiliated. App. to Pet. for Cert. 143a-144a.
The trial judge found that the Roman Catholic schools in question operate under the general supervision and authority of the Archbishop of New Orleans and their parish pastors, and are located next to parish churches and sometimes a rectory or convent. Id., at 144a. The schools include religious symbols in their classrooms, App. 75a, require attendance at daily religion classes, id., at 76a, conduct sacramental preparation *905classes during the schoolday, require attendance at mass, and provide extracurricular religious activities. At least some exercise a religious preference in accepting students and in charging tuition. App. to Pet. for Cert. 145a.
The District Court found that the mission of the Roman Catholic schools is religious education based on the Archdiocese’s and the individual schools’ published statements of philosophy. For example, the St. Anthony School Handbook, cited by the District Court, reads:
“Catholic education is intended to make men’s faith become living, conscious and active through the light of instruction. The Catholic school is the unique setting within which this ideal can be realized in the lives of the Catholic children and young people.
“Only in such a school can they experience learning and living fully integrated in the light of faith.... Here, too, instruction in religious truth and values is an integral part of the school program. It is not one more subject along side the rest, but instead it is perceived and ftmctions as the underlying reality in which the student’s experiences of learning and living achieve their coherence and their deepest meaning.” Ibid.
The Handbook of Policies and Regulations for Elementary Schools of the Archdiocese of New Orleans indicates that the operation of the Roman Catholic schools is governed by canon law. It also lists the major objectives of those schools as follows:
“To work closely with the home in educating children towards the fullness of Christian life.
“To specifically teach Catholic principles and Christian values.” Id., at 146a.
The mission statements and objectives outlined by the other Roman Catholic schools also support the conclusion that these institutions’ primary objective is religious instruction. See also App. 65a, 71a.
The Archdiocese’s official policy calls for religious preferences in hiring and the contracts of principals and teachers in its schools contain a provision allowing for termination for lifestyle contrary to the teachings of the Roman Catholic church. App. to Pet. for Cert. 145a. One of the objectives of the handbook is “[t]o encourage teachers to become committed Christians and to develop professional competence.” Id., at 146a. Other record evidence supports the conclusion that these religious schoolteachers teach religiously. See, e. g., App. 125a (deposition of president of sectarian high school) (“Our teachers, whether they are religion teachers *906or not, are certainly instructed that when issues come up in the classroom that have a religious, moral, or value concept, that their answers be consistent with the teachings of the Catholic Church and that they respond in that way to the students, so that there can be opportunities in other classes other than religion where discussion of religio[n] could take place, yes, sir”); id., at 73a, 74a.
The Government's reliance on U.S. Department of Education Guidance for Title VI of the Elementary and Secondary Education Act (Feb. 1999) is misplaced. See App. to Brief for Secretary of Education la. It was not in place when discovery dosed in this matter, and merely highlights the reasons for a lack of evidence on diversion or compliance.
The plurality applies inconsistent standards to the evidence. Although the plurality finds more limited evidence of actual diversion sufficient to support a general finding of diversion in the computer and instructional materials context, even in the face of Justice O’Connor’s objections, it fails to find a violation of the prohibition against providing aid with religious content based on the more stark, undisputed evidence of religious books. Compare ante, at 832-834,'and nn. 14-17, with ante, at 834-835. As a matter of precedent, the correct evidentiary standard is clearly the former: “[A]ny use of public funds to promote religious doctrines violates the Establishment Clause.” Bowen, 487 U. S., at 623 (O’Connor, J., concurring). We have never before found any actual diversion or allowed a risk of it; we have struck down policies that might permit it, e. g., Tilton, 403 U. S., at 682-684, or have remanded for specific factual findings about whether diversion occurred, Bowen, rnpra, at 621. See supra, at 890-895. As a matter of principle, this low threshold is required to safeguard the values of the First Amendment. Madison’s words make cleai* that even a small infringement of the prohibition on compelled aid to religion is odious to the freedom of conscience. No less does it open the door to the threat of corruption or to a return to religious conflict.
Since the divertibility and diversion require a finding of unconstitutionality, I will not explore other grounds, beyond noting the likelihood that unconstitutional supplantation occurred as well. The record demonstrates that Chapter 2 aid impermissibly relieved religious schools of some costs that they otherwise would have borne, and so unconstitutionally sup*911planted support in some budgetary categories. The record of affidavits and evaluation forms by religious schoolteachers and officials indicates that Chapter 2 aid was significant in the development of teaching cur-rículums, the introduction of new programs, and the support of old ones. App. 105a-108a, 184a-185a. The evidence shows that the concept of supplementing instead of supplanting was poorly understood by the sole government official administering the program, who apparently believed that the bar on supplanting was nothing more than a prohibition on paying for replacements of equipment that religious schools had previously purchased. Id., at 167a. Government officials admitted that there was no way to determine whether payments for materials, equipment, books, or other assistance provided under the program reduced the amount of money budgeted for library and educational equipment, id., at 145a-146a, and the 1985 Monitoring Report shows that the officials of at least one religious school admitted that the government aid was used to create the library, with the school’s regular funds, when occasionally available, used merely to supplement the government money, Fine Deposition, id., at 63a. The use records for audiovisual materials at one religious high school revealed that Chapter 2 funds were essential to the school’s educational process, id., at 187a, and a different school, as already noted, used a Chapter 2 computer to support its computer network when its own computers failed, id., at 77a. The record is sparse, but these incidents suggest that the constitutional and statutory prohibition on supplanting expenses may have been largely aspirational. It seems that the program in Jefferson Parish violated the statute and ran afoul of the Constitution. Cf. Nyquist, 413 U. S., at 783; Zobrest, 509 U. S., at 12.
Indeed, one group of amici curiae, which consists of “religious and educational leaders from a broad range of both Eastern and Western religious traditions, and Methodist, Jewish and Seventh-day Adventist individuals” including “church administrators, administrators of religious elementary and secondary school systems; elementary and secondary school teachers at religious schools; and pastors and laity who serve on church school boards,” identifies its members as having “broad experience teaching in and administering pervasively sectarian schools.” Brief for Interfaith Keligious Liberty Foundation et al. as Amici Curiae 1.
One of the respondents describes herself as a “life-long, committed member of the Roman Catholic Church” who “objects to the government providing benefits to her parish school” because “[s]he has seen the chilling effect such entangling government aid has on the religious mission of schools run by her church.” Brief for Respondents 1. She has been a member of the church for about 36 years, and six of her children attended different Jefferson Parish Catholic run schools. Id., at 1, n. 1.
E. g., Brief for Baptist Joint Committee on Public Affairs as Amicus Curiae; Brief for Interfaith Religious Liberty Foundation et al. as Amici Curiae; Brief for National Committee for Public Education and Religious Liberty et al. as Amici Curiae.
1 do not think it worthwhile to comment at length, for example, on the plurality’s dear misunderstanding of our access-to-public-forum cases, such as Lamb’s Chapel and Widmar v. Vincent, 454 U. S. 263 (1981), as “decisions that have prohibited governments from discriminating in the distribution of public benefits based on religious status or sincerity,” ante, at 828, when they were decided on completely different and narrowly limited free-speech grounds. Nor would it be worthwhile here to engage in extended discussion of why the goal of preventing courts from having to “trol[l] through a person’s or institution’s religious beliefs,” ibid., calls for less aid and commingling of government with religion, not for tolerance of their effects.