The question of how to define ‘religion’ used to be a common topic in the philosophy of religion. There may be many reasons for the question’s falling into desuetudeFootnote 1; among them is not the question’s inherent interest. We confidently, with an enormous amount of agreement, classify a set of human practices as religions; it seems reasonable to ask what it is about those practices that makes us so categorize them (Wittgensteinian concerns about this way of posing the question are addressed below). We may put the question this way: Is there any essence to religions; that is, is there anything they all have in common, the having of which makes a practice a religion, and the lacking of which keeps a practice from being a religion? For the purpose of this discussion, let us call any proposal about what that essence is a theory of religion.

Theories can be given for a variety of purposes, and so what counts as an adequate theory will depend on who is asking the question. An adequate sociological theory of religion will be one that is useful for further sociological work. The question I am interested in is the general philosophical, or Socratic, question: What is the essence, if any, of religion? The quest for an adequate philosophical theory of religion is not simply the quest for an adequate definition of the English word ‘religion,’ nor is it necessarily the quest for the origins of religious practices, whether those origins are historical, psychological, or sociological. These related areas of inquiry may well give useful insights to the philosopher, but they cannot substitute for the philosophical question.

Likewise, the question cannot be answered wholly by an analysis of religious feelings or sentiments. Though it is possible that an understanding of religious feelings will give us insight into what religions are, we are interested in the religions themselves, the social institutions that have arisen in all cultures. The philosophical theory of religion is the theory of Christianity, Buddhism, and the like; these are socially instantiated practices, not personal feelings or aggregations of personal feelings.

In this essay I will argue that the correct theory of religion is an eliminativist one. There is no essence to religion, and we therefore should accept that there is no such thing. In part one, I will outline some desiderata for a philosophical theory of religion and argue for the claim that some representative popular proposals fail to satisfy these desiderata. In the process, I will offer a general reason why any such proposal will fail. In the second section, I will address competitors to eliminativism as an account of that failure. The conclusion is that, because it is unlikely for any proposal to be able to meet all the desiderata, eliminativism is the best theoretical choice.

Desiderata for Theories of Religion

One obvious desideratum is extensional accuracy. What we take to be the essence of religion should be something that all actual religions have, and no non-religion has. While it is difficult, if not impossible, to give necessary and sufficient conditions, and so yield a judgment for any possible practice whether it is a religion, it ought to be easier to give conditions that sort actual practices correctly. Various obvious first tries at a theory that meets these fail to be extensionally accurate; that is, there are things that we unproblematically call religions that don’t meet the proposed theory. We can’t say a religion is a doctrine about God or the gods, because some Buddhist sects deny that there are gods. We can’t say a priesthood or hierarchy is necessary, or scriptures, or doctrine of life after death. It’s child’s play to find counterexamples to that kind of theory.

To give a concrete example, Paul J. Griffiths defines religion as ‘a form of life that seems to those who belong to it to be comprehensive, incapable of abandonment, and of central importance.’Footnote 2 This is too narrow, in that religions frequently seem capable of abandonment, even to those most centrally involved with them, at the height of their involvement; that is, apostasy is a real concern. Griffiths’s gloss on this is that a practice does seem to be incapable of abandonment if it seems to those who belong to it to be so important that ‘leaving it is impossible without also leaving themselves.’Footnote 3 If this is any more than a metaphor (and involves an actual break in personal identity), then it seems clear that almost no practice seems that way to those who belong to it. If I consider the possibility of leaving my current religion, I conceive of that event as my leaving the religion, not my ceasing to exist and having a successor of some sort in another religion, or no religion. It is also too broad, in that a host of non-religious practices can seem to those involved with them to be both comprehensive and of central importance, for example, the study of ethics, or the physical sciences. To be fair, Griffiths does admit that his definition allows for religions of uxoriousness (for the devoted husband) or sports (for the devoted fan). That being said, his theory will not serve our purposes, since our aim is to find a theory of the social practices normally called ‘religions.’

The way to check for extensional accuracy is to start with a list of paradigm cases and judge the theory according to whether it classifies all those paradigms as religions, and no paradigmatically non-religious institutions as religions. The list of paradigms should include Christianity, Judaism, Islam, Hinduism, Buddhism, Taoism and Confucianism (or perhaps Chinese religion, instead of Taoism and Confucianism), at a minimum. It is easy to expand the list of paradigms, but the list above is probably adequate. It is important to note that we may end up changing our minds along the way about whether a given item really is a paradigm after consideration of several proposed theories. In other words, we should treat our list of paradigms as defeasible.Footnote 4 Long reflection may show that something we took to be a paradigm is really so different from the other items on the list that it doesn’t really count. Many people have this reaction to Confucianism, for example. Nevertheless, that a practice is a paradigm is a matter to be given a lot of weight.

Another desideratum is neutrality. The theory should not entail that any religion is the true religion, or that any religion’s central claims are false. This is a desideratum of our theorizing, because our classifying some human practice as a religion should not beg any important questions about its value. In other words, in order to evaluate a religion as true or false, we must first decide what a religion is; categorizing should come first, and evaluating after. In order to see that this is really desirable, one need only imagine a practice arising that has all the features of a religion except that its claims (including claims about its origin) happen to have the opposite truth value of the one required by the theory. If it is otherwise indistinguishable from a practice that is clearly a religion, it would be perverse to deny that the new practice is a religion. This desideratum rules out many attempted theories. Some theories of religion are privileging theories, according to which one can understand correctly what a religion is only by accepting one of them as the truth. Many early Christians adopted the view that the pagan religions extant in their day were the result of demons deliberately deceiving human beings.Footnote 5 These writers were not seeking a theory of religion as I have defined it, but were speculating on the origin of pagan religions. However, if this theory were adopted as a theory of religion, it would fail to meet the neutrality requirement.

Similarly, some theories are debunking theories, according to which one must see that all religions are false in order to see what they all have in common. The theories of FreudFootnote 6 and MarxFootnote 7 are frequently offered as theories of religion (though I do not wish to claim that Freud and Marx themselves so offered them; they were only trying to say something true about religion, and not necessarily to describe its essence). If Freud is right and religion consists in acting on a vague memory of infantile ‘oceanic’ feeling, then religions are all illusions, and so either false or only accidentally true, because they have no contact with any external reality. Similarly, if Marx is right that religion is the spontaneous unconscious creation of the oppressed working class, it is either false or only accidentally true. There may be some truth to the claims of privileging theories or debunking theories, but neither of these kinds of theory is satisfactory. Categorizing comes before evaluating. For this reason, our theory of religion ought to be neutral among religions.

A third desideratum concerns what sort of properties the theory should focus on. A satisfactory theory of religion must be phenomenological. We can recognize whether a practice in a community is a religious practice without making inquiries into historical, sociological, or psychological origins. To put it another way, there cannot be two institutions that behave exactly alike, but one is a religion and the other isn’t because of facts that are not readily available to the spectator. Of course, that is not to say that we should be able to tell whether a practice is religious after a cursory and superficial examination; however, the features of the practice that make it a religion should be available to some kind of empirical investigation of its current features. This is so because our classification of human practices as religions doesn’t depend on discovering any of these deep facts. Whatever the concept is that we are wielding, it doesn’t seem to have that kind of content.

The final desideratum is non-circularity. A satisfactory theory of religion must be illuminating, and so it should be possible to state it without using irreducibly religious language. It is easy to find adequate definitions of the word ‘religion’—even by the standards described above—but these definitions frequently contain terms that are every bit as much in need of explication as the term ‘religion’ itself. Definitions in terms of the sacred, or the supernatural, or the spiritual then may be adequate as lexical definitions, but they are unsatisfactory as theories, because they don’t really succeed in theorizing. They don’t tell us what it is about a practice of institution that makes it a religious one. For example, the sociologist Ronald L. Johnstone defines ‘religion’ thus: ‘Sociologically viewed, then, ... religion can be defined as a system of beliefs and practices by which a group of people interprets and responds to what they feel is sacred, and, usually, supernatural as well.’Footnote 8 Johnstone is not offering a philosophical theory of religion, as the beginning of the quoted sentence shows. This theory may well be adequate for sociological inquiry; but since it uses terms from the religious ambit, it does not serve our purpose of illuminating what makes a practice religious.

Most currently popular theories of religion fail this non-circularity test. Here follow two examples from twentieth century literature. William Tremmel, in Religion: What is it? Footnote 9, offers a theory of religion that includes the following clauses:

  1. 1.

    Religion is a complex form of human behavior whereby a person (or community of persons) is prepared intellectually and emotionally to deal with those aspects of human experience that are overwhelming, critically threatening, and beyond control.

  2. 2.

    Doing so from a conviction that there is at the center of human experience, and even of all reality, a being, or beings, or process (a divine reality) in which and through which a person (or community of persons) can transcend the life-negating traumas of human existence and overcome the sense of finitude.

This fragment of Tremmel’s theory of religion shows that he fails to meet the non-circularity requirement. While he avoids parochialism by making the ontological commitment of religion sufficiently vague, he can avoid being too broad only by including terms which are themselves just as in need of analysis as the term ‘religion’ itself. What does it mean for a being, beings, or process to be ‘at the center of reality?’ This is a metaphor at best, as are all references to transcendence, or things above or beyond ordinary experience. It’s difficult to see how to restate such claims in non-metaphorical terms. While there is nothing wrong with metaphors per se, their centrality to a theory of religion only serves to show that we do not yet understand what it is literally, and frequently when the metaphor is cashed out in literal language, or the vagueness is eliminated, we are left with something either circular or extensionally inadequate. In the above attempt, Tremmel gives the game away by inserting the parenthetical phrase ‘a divine reality.’ What is a divine reality, if not something that is by definition the central concern of a religion?

One might be tempted to think that the idea of the supernatural doesn’t have these disabilities, since it can be understood by contrast with the natural. Then if we understand religions as concerned with supernatural agents, there is no problem with vagueness or metaphor, and the proposal seems to include no essentially religious vocabulary. Unfortunately, this proposal won’t work, either. For one thing, understanding ‘supernatural’ merely by contrasting it with ‘natural’ loses the ‘super’ part of the idea.Footnote 10 All we are left with is non-natural agents, and that’s no longer clearly a religious idea. But even if ‘non-natural’ were a sufficient gloss for ‘supernatural,’ it’s still not clear that the idea would be extensionally adequate. In recent discussions of naturalism, it has been noted that it is extremely difficult to cash out the idea of naturalism in such a way that it both rules out what is usually thought of as supernatural and rules in all the things countenanced by physics.Footnote 11 If the natural is what is countenanced by current physics, then some future discoveries would count as non-natural or supernatural. If the natural is what is countenanced by any future science, then it is not at all clear that angels, souls, gods, or demons won’t be part of our best scientific theory at some point.Footnote 12

Paul Tillich, in his famous analysis of the concept of religion, falls prey to the same error. He famously says, ‘Religion is the state of being grasped by an ultimate concern, a concern which qualifies all other concerns as preliminary and which itself contains the answer to the question of the meaning of life.’Footnote 13 This statement was challenged for its ambiguity almost as soon as it was published.Footnote 14 Is an ‘ultimate concern’ what concerns a person above all other concerns? Then everybody has a religion—something, or some set of things, has to come in first—so it fails the extensionality test. Is it a concern with what is really ultimate? Then, presuming a concern with the laws of physics and such can’t be really ultimate, it fails the non-circularity test, since it is hard to spell out what ‘ultimate’ means without dragging in terms like divine, spiritual, or supernatural to do the heavy lifting. The same sort of trouble is going to bedevil James’s ‘belief in an unseen order’ (do atoms count?), Otto’s ‘numinous,’Footnote 15 and Schleiermacher’s ‘feeling of absolute dependence.’Footnote 16

Alternative Conclusions

The difficulty seems to be that the heterogeneous collection of practices we call religions are so different from one another that it is impossible to give an account that is specific enough to avoid the circularity problem and also general enough to pass the extensionality test. The impossibility of fulfilling the desideratum of non-circularity drives some to give up that requirement. Perhaps the essence of religion is not something that can be understood in non-religious terms. Some essentially circular theories are undoubtedly unproblematic, but for them we have some other grip (perceptual, for example) on the notions required. For example, we probably can’t give a theory of colors (as perceptually experienced, not physical reflectances) without using some color terms, but we have general agreement about the reference of color terms to anchor our understanding. It’s not clear (mystics not withstanding, since they don’t agree among themselves about the reference of key terms) that we have any other road into understanding religious notions. We could defer to the mystics and say that religion is a set of social practices centered on what they are experiencing, but that would make the essence of religion turn on the tendentious question as to whether all mystics are experiencing the same thing. There is certainly reason to think that Nirguna Brahman cannot be the same thing as the Trinity, and neither can be the same as sunyata.Footnote 17

Another popular reaction is to give up on the idea that there is any essence to religion for us to uncover, but that does not keep us from having a useful philosophical theory of religion. On this view, there is nothing they all have in common, so there is no true theory in the sense I am employing. The term ‘religion’ is a family-resemblance term. This amounts to saying that the first of my desiderata, extensional accuracy, is no desideratum at all. Wittgenstein, in the Philosophical Investigations,Footnote 18 famously offers such a view of games, on his way to make the same point about language and meaning. He says:

Consider for example the proceedings that we call ‘games.’ I mean board-games, card-games, ball-games, Olympic games, and so on. What is common to all of them?—Don’t say: ‘There must be something common, or they would not be called “games”’—but look and see whether there is anything common to all.—for if you look at them you will not see something that is common to all, but similarities, relationships, and a whole series of them at that. ...I can think of no better expression to characterize these similarities than ‘family resemblances’...

This may be true, of games and of religions, but it amounts to admitting that religions don’t form a kind, whether natural, social, or functional, which is just my point. Further, to be interesting as a philosophical theory of religion, a family-resemblance theory should specify what the resemblances are in a way that respects the other desiderata. Suppose the list of resemblances includes predicates like ‘postulates the existence of gods,’ ‘has authoritative priests,’ and the like. There is no way to spell out what these terms mean non-circularly. How are we to distinguish gods from very powerful aliens without appealing to terms like ‘spirit,’ ‘supernatural,’ and the like? How are we to distinguish priests from people with other kinds of authority without similar problems? I conclude that not only do religions not form a kind, they do not even have any interesting family resemblances; all the family resemblances, once circularity is eliminated, are superficial. Consequently, our best theory is the null theory. There is no such thing as religion.

This need not be a bleak prospect. It may yet be possible to theorize usefully about all those things we call religions, just not all in one group. For example, there may be smaller sub-groups that do have some kind of essence. The ‘Abrahamic’ religions—Judaism, Christianity, and Islam—have a lot in common, and it is reasonable to treat them (along with others appropriately related to themFootnote 19) as forming a kind about which interesting questions can be asked. Likewise, Hinduism, Jainism and Buddhism have a great deal in common, and they may form a kind. The only thing we are barred from doing is trying to say anything about all of them together. And that is OK, because, if I am right, there is nothing useful to be said about them all, anyway.