American law has long accorded religion special treatment. Quaker and Mennonite objections to military service have been accommodated since colonial times. Sacramental wine was permitted during Prohibition. Today the Catholic Church is allowed to deny ordination to women despite antidiscrimination laws. Jewish and Muslim prisoners are entitled to Kosher or halal food.

This tradition has become intensely controversial of late, reflecting a growing scholarly consensus that special treatment of religion cannot be justified. While some scholars would rule out all legal accommodation, the more common view would allow it in certain cases, but under another description. It is morally arbitrary to single out “religion,” the argument goes, and so a different legal category, such as “conscience,” should be used. A second and related objection is that the bounds of “religion” are so indeterminate that the term is meaningless—a term that European colonizers, for instance, used willy-nilly to describe whatever local practices somehow reminded them of Christianity.

The singling out of religion for special legal treatment, I will argue here in response, is appropriate, and precisely because religion doesn’t correspond to any narrow category of morally salient thought or conduct; as such it is a concept flexible enough to be accommodated legally while keeping the state neutral about theological questions. Other, more specific categories are either too sectarian to be politically usable, too underinclusive, or too vague to be administrable.

First, a note on how such arguments should proceed. The philosopher John Rawls famously held that justification in political philosophy should follow the method of “reflective equilibrium,” in which we try to bring our considered moral judgments into line with our more general principles. Any general theory, he wrote in A Theory of Justice (1971), must be consistent with the specific judgments “in which we have the greatest confidence,” such as our judgments “that religious intolerance and racial discrimination are unjust.” These are “provisional fixed points,” Rawls wrote, “which we presume any conception of justice must fit.” The fixed points that theories of religious liberty must fit include such generally agreed-upon assertions as: “Quakers are entitled not to be drafted;” and “the state should not establish a religion or endorse articles of faith.” Any political philosophy that is to be viable in the United States must preserve these commitments to free exercise and disestablishment. Our question is whether this can be done without relying on the category of “religion.”

When law singles out religion for special treatment, we can reasonably ask what good or purpose it hopes to promote by doing so. The earliest and most obvious answer is that there is a good specific to religion—salvation by Christ is the classic one—and that religion should be singled out in order to promote that good. Of course, when you do that, you can end up with a pretty narrow definition of religion. Henry Fielding’s Mr. Thwackum, in the 1749 novel Tom Jones, lamented the “various sects and heresies in the world,” and declared that “[w]hen I mention religion I mean the Christian religion; and not only the Christian religion, but the Protestant religion; and not only the Protestant religion, but the Church of England.” John Locke’s classic Letter Concerning Toleration (1689) claimed that “true and saving Religion consists in the inward perswasion of the Mind,” and insisted that “frivolous things” ought not to divide “Christian Brethren, who are all agreed in the Substantial and truly Fundamental part of Religion.” But many religions have rejected such ecumenical inclusiveness and continue to proclaim themselves the only true path to heaven. Today just over a third of all Americans consider Christianity the one true religion, and half that number reject the view that all major religions contain some truth about God.

What else beyond salvation might be the good that religion supposedly delivers? Multiple candidates suggest themselves, including harmony with the transcendent origin of universal order (if it exists); courage in the face of heartbreak (if that kind of encouragement helps); a transcendent underpinning for the resolution to act morally (if that kind of underpinning helps); contact with the awesome and the indescribable (if awe is something you feel), and so on. In the cottage industry of proposals to discard the category of “religion” and substitute something else, however, these candidates haven’t gotten much attention, for the excellent reason that they are theologically loaded. It is not just that they are narrower and more specific than “religion;” it’s that their goodness is a specifically religious goodness that depends on contestable metaphysical premises. Secular liberal philosophers tend to shy away from such notions, and militant atheists are disgusted. There are, however, a lot of religious people out there. When religion is regarded as good, it is usually for religious reasons.

The fact that different religions conceive these goods differently is one reason for the state to remain vague about this question, since privileging any of them would discriminate among religions. American law’s neutrality is specifically neutrality among religions; the establishment clause of the First Amendment means, at its core, that the state may not endorse contested theological propositions. One reason for using “religion” as a proxy for specific ends such as salvation is that, if the good the state pursues is described in this vague way, the state need not assess the comparative value of those ends. There is also, of course, disagreement about which religions actually achieve these goods. If the pertinent good is salvation, for instance, then perhaps it is incumbent upon the state to figure out which religion actually delivers it. Deploying the broad category of “religion” helpfully allows the state to evade that dangerous task. It still takes the controversial position that religion is good, but embraces it abstractly enough to make room for everyone, except possibly the atheists. And (as I’ll explain shortly) it even sometimes manages to squeeze them in as well.

“RELIGION” THEN, IS a proxy for the genuine religious good (if there is one), and part of its value is that we need not agree about what exactly it is a proxy for. But what about religion as a proxy for secular goods? Secular political theorists, naturally enough, focus on secular ends. A number of these have been proposed as candidates for special accommodation in place of religion, including individual autonomy; a source of meaning inaccessible to other people; psychologically urgent needs (in effect, treating religion as analogous to a disability that needs accommodation); comprehensive views; minority culture; and conscience.

All are underinclusive. Consider conscience, the most widely advocated substitute. Conscience excludes some claims that are widely recognized as valid; and many religious claims that nearly everyone would want to accommodate are not conscientious. (A paradigm case is the ritual use of peyote by the Native American Church.) The emphasis on conscience focuses excessively on duty, while many people engage in religious practice not from a sense of duty prescribed by sacred texts, but on other grounds: adherence to custom; a need to cope with misfortune or injustice, temptation or guilt; a desire to feel connected to God. Indeed, core religious practices often have nothing to do with conscience. When a recent survey asked Catholics why they attended Mass, the largest group, 37 percent, pointed to “the feeling of meditating and communicating with God,” while only 20 percent referred to “the need to receive the Sacrament of Holy Communion,” and only 6 percent said “the church requires that I attend.” This experience-based religiosity is increasingly common in the United States. The most recent congressional pronouncement on religious liberty, the Religious Land Use and Institutionalized Persons Act of 2000, declares that “the term ‘religious exercise’ includes any exercise of religion, whether or not compelled by, or central to, a system of religious belief.”

The secular objection to the use of “religion” as a category was well stated by Christopher Eisgruber and Lawrence Sager, who in a 1994 University of Chicago Law Review article argued that “religion does not exhaust the commitments and passions that move human beings in deep and valuable ways.” This may well be true. Yet “deep” is not an administrable legal category; it is too vague for that. We are too opaque to one another, our depths too personal and idiosyncratic, for us to discern which of each other’s commitments and passions really merit respect. Reliance on imperfect proxies is an inescapable part of social life.

The difficulty of finding a suitable secular analogue for “religion” is also apparent in an issue we have not yet discussed—the disestablishment of religion. American law is clear that the state must be neutral with respect to religious questions. Government, the Supreme Court declared in 1968, “may not aid, foster, or promote one religion or religious theory against another or even against the militant opposite. The First Amendment mandates governmental neutrality between religion and religion, and between religion and nonreligion.” What could be the secular analogue to this doctrine?

Two have been proposed. One is liberal neutralitarianism, which holds that the state should be neutral toward all controversial conceptions of the good life. Instead of embracing any controversial conception of what is good, it should provide citizens with means to pursue whatever lives they happen to prefer, religious or otherwise. The difficulties of this conception are the subject of a large academic literature, but the basic problem is simple: If the state can’t promote good lives, then citizens’ lives will be impoverished. If the state can’t support the arts, for example by teaching about them in the public schools, then many citizens will never learn about Mozart and Rembrandt.

The other proposal is John Rawls’s claim that the government should be neutral toward all “comprehensive views.” A conception is comprehensive, Rawls explains in Political Liberalism, “when it includes conceptions of what is of value in human life, and ideals of personal character, as well as ideals of friendship and of familial and associational relationships, and much else that is to inform our conduct, and in the limit to our life as a whole.” And: “A conception is fully comprehensive if it covers all recognized values and virtues within one rather precisely articulated system.”

Yet many religious views are not fully comprehensive: most religious people do not rely on their religious beliefs to structure their lives in this pervasive way. Perhaps in response to this, Rawls also wants to exclude any “partially comprehensive” conception, which comprises “a number of, but by no means all, nonpolitical values and virtues and is rather loosely articulated.” This is an odd locution. It is like saying that a person with a speck of dirt on his shoulder is partially buried. Rawls is brought to this incoherent position because he is attempting to capture the moral basis of disestablishment of religion in terms that make no mention of “religion.” The fact that a philosopher as brilliant as Rawls couldn’t do it is powerful evidence that it can’t be done.

Here is the logic that makes this issue so knotty. No single-factor justification for singling out religion can succeed, since any invocation of a factor X (whether religious or secular) as a justification will logically entail substituting X for religion as a basis for special treatment, making “religion” disappear as a category of analysis. This substitution will be unsatisfactory, however, since there will be settled intuitions about establishment and accommodation that it will be unable to account for. Any X will be an imperfect substitute for religion, but a theory of religious freedom that focuses on that X will not be able to say why religion, rather than X, should be the object of solicitude.

There are two ways around this difficulty. One is to say that these are not ends the state can directly aim at, and that religion is a good proxy for them. Doing this justifies some imprecision in the law. We want to give licenses to “safe drivers,” but these are not directly detectable, so we use the potentially errant but still basically useful proxy category of “those who have passed a driving test.”

This routine feature of law is overlooked in Brian Leiter’s claim, in his book that religion could legitimately be singled out for special protection only because of “features that all and only religious beliefs have, either as a matter of (conceptual or other) necessity or as a contingent matter of fact,” or that would not merit such principled toleration if other beliefs have those same features. He is correct that no such features exist. He acknowledges the indispensability of legal proxies, but does not examine the impact of that concession on his thesis that singling out religion is unfair.

The other way is to say that religion is an adequate proxy for multiple goods, some of which cannot directly be aimed at, at least in the United States. Because “religion”—or, at least, that subset of it that is likely to come before American courts—captures multiple goods, any substitute that aims at any one of them will be underinclusive. None can fully capture our settled intuitions about accommodation. There are lots of good candidates for accommodation, in other words, and neglecting any of them is unfair. As I have argued before in these pages, religion is not just a proxy for something else; it is a proxy for many something elses. It is a bundle of proxies.

The fundamental objection to “religion” as a category, of course, is that it is itself underinclusive. That claim is hard to test, because the bounds of the category are so uncertain. The debate among legal scholars over religion is chronically confused by their failure to grasp a point familiar to scholars of religion—namely, that “religion” is a label for something that likely has no reality outside human imagination. (This is not a theological claim. The proposition that the Christian God exists outside human imagination, for instance, does not necessarily entail that “religion,” encompassing everything connoted by that word, exists similarly.) Religion—at least, as a legal category—has no essence. If it has a determinate meaning, it is simply because there is a settled and familiar practice of applying the label of “religion” in predictable ways.

Scholars of religion disagree about whether there is any identifiable essence to “religion.” Both Jonathan Z. Smith and Talal Asad claim that the term “religion” denotes an anthropological category, arising out of a particular Western practice of encountering and accounting for foreign belief systems associated with geopolitical entities the West was forced to deal with. Martin Riesebrodt, on the contrary, argues that all religions serve common functions: they promise to avert misfortune, help their followers manage crises, and bring both temporary blessings and eternal salvation. For legal purposes, it does not matter who is correct. Even if theorists could converge upon a single definition, it would not be one that American law has relied upon—nor one necessarily suited to the law’s purposes.

THE QUESTION OF religious accommodation arises in cases where a law can allow some exceptions. Many laws, such as those determining military conscription, taxes, environmental regulations, and antidiscrimination policies, will accomplish their ends even if there is some deviation from the norm they set forth, so long as that deviation does not become too great. In such cases, special treatment is sometimes appropriate. Religious exemption is the practice of singling out religion as a basis for such special treatment. Since there is no such thing as religion, if such accommodations are justified, the justification must ultimately depend on some good other than religion. Religion can only be a proxy.

The closest the Supreme Court has come to addressing the question of how to define religion for legal purposes is a pair of Vietnam-era draft-exemption cases. Both involved claimants who conscientiously objected to war but would not avow belief in God. In U.S. v. Seeger (1965), the Court responded with a functional definition of religion, holding that the crucial question is “whether a given belief that is sincere and meaningful occupies a place in the life of its possessor parallel to that filled by the orthodox belief in God of one who clearly qualifies for the exemption.” It explained that a pertinent objection “cannot be based on a ‘merely personal’ moral code,” but gave no example of the line it was drawing. In Welsh v. U.S. (1970) the court ruled further that an individual’s description of his own beliefs as nonreligious did not disqualify him from accommodation, and that such self-characterization made for “a highly unreliable guide for those charged with administering the exemption.” Conscientious-objector status, it asserted, should apply to “all those whose consciences [are] spurred by deeply held moral, ethical, or religious beliefs.” Since then the Court has offered no further clarification of what it means by “religion.”

What in fact unites such disparate worldviews as Christianity, Buddhism, and Hinduism is a well-established and well-understood semantic practice of using the term “religion” to signify them and relevantly analogous beliefs and practices. Efforts to distill this practice into a definition have been unavailing. But the common understanding of how to use the word has turned out to be all that is needed. In fact, courts rarely have any difficulty determining whether something is a religion or not. The list of reported cases that have had to determine a definition of “religion” is a remarkably short one. Browsing through the 132 volumes of Words and Phrases, a standard work of American legal research collecting brief annotations of cases from 1658 to the present, reveals that while some terms have received an enormous amount of attention from the courts—two examples drawn at random from Volume 1, “abandonment” and “abuse of discretion,” each exceed 100 pages— “religion” gets fewer than five. The question of what “religion” means may be theoretically intractable but as a practical matter it’s barely relevant. We know religion when we see it. And when we see it, we treat it as something good.

So how to assess where we are? Even if we stipulate that religion is valuable—or, more precisely, that “religion” is an irreplaceable proxy for that which is genuinely valuable—this alone does not tell us whether the American regime of special accommodation treats religion appropriately. It is possible, after all, that religion is an appropriate category of accommodation and that it is often unfairly privileged. The question in any particular case is whether the decision-maker—be it court, legislature, or administrator—is giving adequate weight to the religious interest at stake and balancing it appropriately against whatever state policy it is coming into conflict with. It’s often hard to get a clear answer to that in even a single case. Aggregating the heterogeneous universe of cases into a broad evaluation is really difficult.

We know we can rule out some answers, such as the notion that religious interests should always win, or that the state’s interest is always so weighty that accommodation is never appropriate. But that leaves all the interesting work still to be done. Injustices and errors will always exist—some state laws exempt religious day-care centers from safety inspections! But anecdotes aren’t enough to assess the general pattern. While I am not attempting to defend the current accommodation regime as a whole, I have no confidence that some widely endorsed reforms of the type discussed here would be an improvement. I suspect that from a legal and political standpoint, the best we can do is embrace the paradox of religion: protect it, because it doesn’t exist.