We tend to think of religion as an age-old feature of human existence. So it can be startling to learn that the very concept dates to the early modern era. Yes, you find gods, temples, sacrifices and rituals in the ancient Mediterranean, classical China, pre-Columbian Mesoamerica. What you don’t find is a term that quite maps onto ‘religion’.
What about the Romans, to whom we owe the word? Their notion of religio once meant something like scruples or exactingness, and then came to refer, among other things, to a scrupulous observance of rules or prohibitions, extending to worship practices. It was about doing the right thing in the right way. The Romans had other terms as well for customs, rites, obligations, reverence and social protocols, including cultus, ritus and superstitio. Yet they weren’t cordoned off into a realm that was separate from the workaday activities of public life, civic duty and family proprieties. What the Romans encountered abroad were, in their eyes, more or less eccentric versions of cultic life, rather than alien ‘religions’, in our sense. It was assumed that other localities would have other divinities; in times of war, you might even summon them, via evocatio, to try to get them to switch sides. But the local gods and rites of foreigners could be assessed without categorising them as instances of a single universal genus.
Even after the empire became officially Christian, you still don’t get our sense of ‘religions’. The Romans don’t start sorting the world into bounded systems analogous to ‘Christianity’, ‘Judaism’, ‘Manichaeism’, ‘Islam’ and so on. They have other, older sorting mechanisms, as Brent Nongbri elaborates in his terrific study Before Religion (2013). When Lactantius, in the 4th century, contrasts vera religio with falsae religiones, he means to distinguish right worship from wrong worship; he isn’t identifying other self-contained systems that might be lined up on a chart for comparison. The Christians of late antiquity didn’t view themselves as possessing one religion among many; they viewed themselves as possessing the truth.
The 1709 edition of De veritate religionis Christianae (1627) by Hugo Grotius. Courtesy the Internet Archive
To arrive at the modern category of religion, scholars now tend to think, you needed a complementary ‘secular’ sphere: a sphere that wasn’t, well, religious. That’s why the word’s modern, comparative sense wasn’t firmly established until the 17th century – Hugo Grotius’s De veritate religionis Christianae (1627) is one touchstone – at a time when European Christendom was both splintering and confronting unfamiliar worlds through exploration and conquest. Even as religion could be conceived as a special domain that might be isolated from law and politics, the traffic with ancient and non-European cultures forced reflection on what counted as ‘true religion’. It’s just that, when Europeans looked at India, Africa, China or the ancient Mediterranean, they sifted for Christian-like (and often Protestant-like) elements: a sacred text to anchor authority, a prophetic founder to narrate origins, a set of theological doctrines to sort out orthodoxy and heresy, and perhaps duties that offered a path to salvation. If a tradition didn’t provide these, scholars might helpfully supply them. In time, ‘world religions’ could be conjured up as bounded systems with creeds and essences, even when the local practices they subsumed were profoundly heterogeneous. Traditions with no founders were given founders; traditions with no single scripture were assigned canonical texts; diverse local rites were bundled into overarching systems.
As world religions took hold as a subject of academic study in the later 19th century, European scholars did their systematic best to treat disparate systems of practice and thought as members of a class. Buddhism became one test case. To call it a single ‘religion’, scholars first had to unify various practices of South, Central and East Asia, and then to decide whether a sometimes godless tradition could qualify. Such struggles over classification exposed a deeper uncertainty: how was ‘religion’ to be defined?
The great minds of the era had ideas. John Stuart Mill held that a religion must unite creed, sentiment and moral authority. Herbert Spencer thought that what religions shared was ‘the tacit conviction that the existence of the world with all it contains and all which surrounds it, is a mystery ever pressing for interpretation.’ The anthropologist Edward B Tylor proposed, as a minimum definition, ‘belief in spiritual beings’. The philologist Max Müller called religion a ‘mental faculty’, separate from ‘sense and reason’, by which humans apprehend the Infinite. For the Old Testament scholar and Orientalist William Robertson Smith, the true foundation of religious life was ritual – the binding force of collective acts. The sociologist Émile Durkheim’s own definition, in his classic The Elementary Forms of Religious Life (1912), joined belief to behaviour and belonging: religion, he wrote, was ‘a unified system of beliefs and practices relative to sacred things’ that united its adherents ‘into one moral community, called the Church’.
These definitions came up short because they excluded too much or included too much. Either they failed to net the fish you were after or they netted too much bycatch. Mill wanted creed, emotion and moral suasion in one package, but many traditions that Europeans encountered in the 19th century didn’t distribute those elements in anything like that pattern. Did a religion involve a metaphysical stance on the cosmos and our place within it – was it driven by the ever-pressing ontological mysteries that Spencer considered central? What we’d call ancient Judaism had very little of that; the biblical writers do not stand before the universe feeling compelled to develop a worldview; they stand within a covenantal drama, entwining law, story and communal identity. And then Müller’s definition could apply to a Romantic poet. (Wilhelm Müller, Max’s father, was a great one.) Dubious of belief-based accounts like Tylor’s, Robertson Smith had concluded that ‘the antique religions had for the most part no creed; they consisted entirely of institutions and practices,’ and ‘while the practice was rigorously fixed, the meaning attached to it was extremely vague.’ Robertson Smith’s own corrective faltered in the face of practices that were communal but not in any obvious way ‘sacred’, or traditions in which doctrine mattered intensely. Durkheim’s formula fatefully relied on a sharp division between sacred and profane that countless ethnographies would undermine.
Georg Simmel, writing around the turn of the 20th century, had already dismissed the ‘Open Sesame’ dream that a single word could unlock the mystery: ‘No light will ever be cast in the sibyllic twilight that, for us, surrounds the origin and nature of religion as long as we insist on approaching it as a single problem requiring only a single word for its solution.’ A few years later, William James complained about ‘verbal’ disputation, but then fell back on a recognisably Protestant formula, defining religion as ‘the feelings, acts, and experiences of individual men in their solitude, so far as they apprehend themselves to stand in relation to whatever they may consider the divine.’ The linguist Jane Ellen Harrison, in her study Themis (1912), refused to define religion at all: a definition, she said, ‘desiccates its object’.
In ‘traditional religions’, there’s a continuity between what we’d distinguish as the natural and the supernatural realm
In the decades that followed, followers of Durkheim foregrounded function, treating religion as a mechanism that bound together societies, comforted individuals, marked transitions, legitimised power. But saying what religion does wouldn’t necessarily tell you what religion was, and, anyway, these functions weren’t peculiar to religion. Clifford Geertz’s elegant formula from the 1960s cast religion as a ‘system of symbols’, one that establishes ‘powerful, pervasive, and long-lasting moods and motivations’. Yet this formula likewise went too big, opening the door to all sorts of political ideologies.
Evolutionary and cognitive theorists since have offered definitions of their own. The evolutionary psychologist Robin Dunbar, for instance, suggested that religion may amount to ‘belief in some kind of transcendental world … inhabited by spirit beings or forces (that may or may not take an interest in and influence the physical world …).’ Inevitably, these belief-oriented accounts run into the same complaints that earlier doxastic definitions had: they seem awfully Protestant, privileging inner conviction over outward form. Even if you bought into the ‘belief’ part, though, you could baulk at the ‘transcendental’ part. In many ‘traditional religions’, there’s a deep continuity between what we’d distinguish as the natural and the supernatural realm. In the Akan region of Ghana where I spent much of my childhood, people would appease or reproach their ancestors in the same spirit that they might wheedle or berate someone at a municipal office. As the anthropologist Robin Horton observed, so-called traditional religions are less like the Western notion of religion than they are like science: they aim at explanation, prediction and control. True, where science posited impersonal forces, traditional thought posited personal ones. But the underlying move from observed regularities to theoretical constructs was similar; what Europeans wanted to call religion was a pragmatic explanatory framework, reasonable given the available evidence, and part of the same conceptual space as folk biology, folk psychology and everyday causal reasoning.
By the late 20th century, hopes for a definition had faded. Some theorists turned to Ludwig Wittgenstein’s notion of ‘family resemblance’. The thought is that traditions can belong to the same conceptual family because they overlap in crisscrossing ways – like cousins who share a nose here, a chin there, without any feature that they all have in common. It’s a permissive approach: you map the ripple of resemblances and give up on strict boundaries. Unfortunately, those resemblances always depend on what you pick as your prototype. If you start with Protestant Christianity, you’ll find resemblances that matter to Protestants; begin instead with Yoruba orisha devotion, and you’ll trace a very different set of likenesses.
The anthropologist Talal Asad influentially and illuminatingly traced both ‘religion’ and ‘the secular’ to the political and intellectual habits of Western modernity. Yet in his account, religion sometimes seems more an effect of those forces than a cause, more a product of power rather than a power in itself. And even if you think that the phenomena we cluster under the term have been sorted and named by Western modernity, you could wonder how we could be sure that they’re examples of the same thing.
Was the category beyond redemption? The scholar and minister Wilfred Cantwell Smith, whose book The Meaning and End of Religion (1962) had meticulously detailed the belated emergence of the ‘religion’ concept in Europe, long maintained that talk of ‘religion’ conflated too many things not to cause mischief, and urged that we give up such talk altogether; we should, instead, speak of faith and ‘cumulative tradition’. The anthropologist and historian Daniel Dubuisson, who anathematised ‘religion’ as a 19th-century Western imposition on non-Western worlds, urged that it be replaced with ‘cosmographic formation’. These evasive manoeuvres, in turn, have met with scepticism. As the social theorist Martin Riesebrodt drily observed, neologisms like Dubuisson’s could doubtless be shown to ‘have also been “constructed” through historically specific discourses’ and revealed as ‘instruments in the linguistic battle between classes or cultures.’ Besides, he pointed out, those who would eliminate the term ‘religion’ seldom manage long without it.
So how has ‘religion’, as a concept and category, endured in the absence of a stable definition? To answer that question, it may help to think about how referring expressions do their referring. Some terms keep their grip on the world even as our understanding of what they denote changes radically; others, once central to serious thought, fall away when their supposed referents are deemed illusions. What distinguishes the survivors from the casualties?
Think about our names for ‘natural kinds’. These are meant to pick out groupings that are found not just in our heads but in nature: bosons, barium, bonobos, beech trees. The things these names designate are thought to have causal powers, explanatory roles or underlying properties that justify treating them as more than convenient fictions. When we name a natural kind, what we’re naming is really out there in the world. Anyway, that’s the aim. How do we decide when we’ve got it right?
Start with chemistry, and the question of what counts as an acid. When the term was first used, it referred simply to substances that tasted sour, or acidus. Later they were marked out by what they did: etching metal, losing their bite in contact with alkalis. In 1777, the French chemist Antoine Lavoisier was convinced that acidity came from a common ingredient he called oxygen – oxygène, the ‘acid-producer’. He was wrong. Yet we’d say that when Lavoisier spoke of acids, he was referring to the same class of things we mean by the word.
A century on, chemists refined the concept. Svante Arrhenius defined acids by their propensity to dissociate in water and release hydrogen ions; in 1923, Johannes Nicolaus Brønsted and Thomas Martin Lowry each reconceived them as proton donors; Gilbert Lewis broadened the net again by calling acids electron-pair acceptors. Each shift expanded the boundaries, but none made the term obsolete. The word survived because its targets – the substances doing the dissolving and reacting – were real enough to anchor it even as its theoretical profile changed.
It’s the difference between a bad map of a real country and a map of Atlantis. Only the first can be fixed
Not every scientific term has been so lucky. In 1774, Joseph Priestley isolated a gas he took to be ‘dephlogisticated air’. Phlogiston was supposed to be a substance released during combustion, the invisible essence of burning. What he had actually found, we’d say, was what we know as oxygen, the name derived from that discarded theory of Lavoisier’s. Unlike oxygen, nothing in the world behaved as phlogiston was said to behave. Indeed, it was Lavoisier who brought the curtain down on phlogiston; closed-system experiments, which he conducted with his wife and lab assistant Marie-Anne Paulze Lavoisier, showed that combustion involved the gain of a component of air (namely, oxygen) rather than the loss of an invisible essence. The phlogiston concept evaporated because chemists came to see that it referred to nothing at all. Priestley’s ‘dephlogisticated air’, by contrast, referred successfully despite being misdescribed: his experiments had latched on to a real thing, even if his theory of it was wrong.
This difference between a term that refers despite error and one that refers to nothing is the difference between a bad map of a real country and a map of Atlantis. Only the first can be fixed. Philosophers have used such cases to argue that successful reference doesn’t depend on getting the description right. What matters is the causal connection between our words and the things they’re meant to denote. The strategy is straightforward enough: if you want to know what object a word refers to, find the thing that gives the best causal explanation of the central features of uses of that word. The features that drove Lavoisier’s acid-talk were produced by substances we still recognise as acids, which is why we don’t treat him as having been talking about some other thing, or about nothing at all. Causal theories of reference explain why our words can target the same class of object even when our conception of it shifts, and when the boundaries of the class shift, too. Pluto can stop being a planet without shaking the foundations of ‘planet’ talk. In such theories of reference, a word continues to refer, so long as it stands in the right causal relation to the entity that gives rise to its use. Misdescribed objects can survive conceptual upheavals; nonexistent ones can’t.
Even in the natural sciences, though, classes of things can fall between those stools. ‘Luminiferous ether’ is a case in point: an invisible medium once thought to carry light waves, it was indispensable to 19th-century physics yet eventually dissolved into what came to be called electromagnetic fields. Was ‘ether’ simply a phantasm? Some philosophers think we could well have retained the term, redefining it to mean the very fields that replaced it. Albert Einstein himself, who once helped kill the ether idea, later repurposed the term as the relativistic ether of spacetime, a field with its own geometry. Other theorists suspect that our ‘electromagnetic fields’ may eventually go the way of ether.
If there can be uncertainty about objects within the natural sciences, the wicket gets stickier when we move into the historical and social realm. Here the things we name – revolutions, nations, money, marriage, religion – are doubly human products, being products first of our collective activity, then of our collective description. These entities are what the philosopher Sally Haslanger would call ‘socially founded’ (a term she uses to sidestep the confusions associated with ‘socially constructed’). Many philosophers of language now call such entities social kinds.
To approach religion as a social kind isn’t to say that it’s as referentially sound as other familiar examples of this sort. Religion may, in fact, be in worse shape than most. It belongs to that subcategory of social kinds that living people apply to themselves. Some social kinds, like ‘recession’, can be defined externally, without the participation of those they describe. Economists can declare one to have happened in the 1870s, even if no one at the time felt it by that name. Others, like ‘wedding’, depend on shared recognition: you cannot hold one without a community that believes in weddings. ‘Religion’, like many social kinds, functions in both ways. Anthropologists can use the term to describe practices that their participants would never call religions, yet, once the label circulates, it acquires a reflexive power: believers come to organise their self-understanding around it. In this respect, religion is a product of classification that helps to shape the reality it describes.
The philosopher Ian Hacking captured this feedback loop with his idea of dynamic nominalism – the process by which classifications and people classified reshape one another. Categories create kinds. The heavy drinker is seen, and sees himself, as an alcoholic. The word doesn’t merely label the phenomenon – it helps to constitute it. Hacking later preferred to call this ‘dialectical realism’, on the grounds that what emerges from the loop (labels affecting those labelled, which then affects the label) is, by any reasonable measure, real enough. When you’ve been told that what you have is a religion, what’s affected isn’t just how you relate to it but what you think you are.
If ‘religion’ endures, it’s because the word still does work, practical and theoretical
Where does this leave someone trying to understand human life through such refractory terms? We might concede that ‘religion’ resists a unitary meaning and proceed case by case, choosing the angle that best reveals what we need to make visible. When speaking of the Abrahamic faiths, a practice-centred approach may capture the lived textures of ritual and observance. The propositions of the Nicene or the Athanasian Creed are, after all, obscure and arguably incoherent, but the act of avowing them carries weighty significance. When we’re turning to the ‘traditional’ thought of the Azande, the Nuer or the Asante, by contrast, a belief-centred, even neo-Tylorian, lens may illuminate elements that the modern Christian model hides from view. Each emphasis is bound to clarify something that the other leaves obscure.
The larger truth is that we’ve always navigated the world with models that merely approximate it, with varying degrees of adequacy. As Hans Vaihinger argued in The Philosophy of ‘As If’ (1911), we often reason through fictions we judge ‘true enough’, because making use of them helps us act, anticipate and understand. The map may not be the territory, but we’d be lost without it. And the sciences, social and natural alike, advance through such tolerable falsehoods. Their worth lies in the utility of their results.
If ‘religion’ endures, it’s because the word still does work, practical and theoretical. It orders law and policy, directs research, and shapes the inner lives of those who use it. Sociologists can enquire into its relation to charity or suicide; psychologists can study its connection to prejudice or wellbeing. In the United States, legislators and judges must have a sufficient grasp of the category that they can balance the Constitutional dos and don’ts of ‘accommodation’ and ‘non-establishment’. For the religionist, meanwhile, it continues to name a space where meaning is made, defended or denied. Whatever else it may be, ‘religion’ remains a category with too many stakeholders to be fired by fiat. When it comes to what the word means, no one gets to say, and everyone gets a say.
Of course, scholarship itself requires observance – with respect to its own standards of evidence, and on the discipline of paying attention. To be observant, in this sense, is to watch the world closely without pretending to stand outside it. And so we try to use our terms with care, aware of what they can hide from sight and of how much they still let us see. We begin where we are, with the tools our history leaves us, and we make do, even if we suspect that our models may someday be replaced. For now, religion endures as a shared act of attention: one of those serviceable maps by which we try to find our bearings, and to keep faith with the world.
