The idea of separating church and state – now entrenched in many modern liberal democracies – is in some ways strange. After all, governments must keep themselves informed and inform their citizens. That’s why they turn to the science of meteorology to predict weather patterns, to economics to understand patterns of growth and decline, and to the medical sciences to improve public health. So why wouldn’t governments also keep close alliances with religious institutions to know what’s going on with God?
Separating church and state appears strange from other perspectives as well. In the historical sweep of large-scale human societies, it’s a relative newcomer, often associated with the United States’ founding father Thomas Jefferson. In 1802, Jefferson asserted the principle of separation in a letter to the Danbury Baptist Association in Connecticut:
Believing with you that religion is a matter which lies solely between man and his God … that the legitimate powers of government reach actions only, and not opinions, I contemplate with sovereign reverence that act of the whole American people which declared that their legislature would ‘make no law respecting an establishment of religion, or prohibiting the free exercise thereof,’ thus building a wall of separation between Church and State.
We’re fans of Jefferson’s principle, one of us (Neil Van Leeuwen) being a philosophy professor who grew up in the Christian Reformed Church, and the other (Jacob Eisler) being a law professor who regularly attends synagogue. But Jefferson’s letter makes putting the principle into practice sound simpler than it is – as though governments could be entirely hands-off when it comes to people’s beliefs, or ‘opinions’. That, however, can’t be right: engaging with and shaping political, social and other kinds of beliefs is actually a core function of government. Public education guides what people believe about everything from mathematics to history to astronomy. Governments also shape beliefs for policy reasons. If fire department officials learn that people believe a building is safe when it isn’t, they will try to change what people believe about the building’s safety. Belief is also a category of legal interest. If a psychotherapist, for example, believes a patient is a danger to himself or others, the therapist is obligated by law to report that belief.
All this means that Jefferson’s ‘wall of separation’ can’t sensibly prohibit governments from intervening on beliefs of any type whatsoever. Rather, Jefferson – along with advocates of the principle who came after him – must have presupposed that there is something distinctive about religious beliefs that makes it reasonable to give them a different legal status.
But what is this distinctive thing? Is it that religious beliefs concern God? No, because Buddhism is a religion with its own beliefs, yet Buddhism has no supreme deity. Is it that religious beliefs concern the supernatural generally? No, because many superstitions that aren’t part of a religion concern the supernatural too. Is it that religious beliefs concern morality, as the biologist Stephen Jay Gould once suggested? No, because many religious beliefs aren’t about morality, and many moral beliefs aren’t religious. In any case, any system of government must presuppose some moral beliefs – eg, that murder is wrong – to be a government at all. These considerations might make one despair of articulating the psychological distinction of interest – what’s different about religious ‘belief’? – that’s needed to make sense of separating church and state in the first place.
Such attempts at characterising religious belief fail, we think, because they focus on belief contents, or what the believing is about. But since any given idea can be believed in a religious way or not, that won’t work. Our suggestion, then, is that religiously believing is a different attitude from ordinary matter-of-fact believing – a different way of relating to ideas – and that this is what’s distinctive about religious belief. In other words, sacralising an idea is different from just taking it to describe how things are. Furthermore, empirical research in cognitive science and anthropology of religion, along with accompanying philosophical analysis, can help reveal characteristic features of the difference in attitude we’re after. To have specific terms, we call that the difference between religious credence and factual belief. Understanding this difference, we’ll see, helps make sense of Jefferson’s wall of separation, and supports the rationale for separating church and state in liberal democracies more generally.
Whatever they are, ‘beliefs’ come in different varieties. The word ‘belief’ is polysemous – meaning it can refer to a range of different-but-related things – and it subsumes a motley crew: from humdrum factual beliefs (say, about where the ice-cream store is) to superstitions (like those found in astrology), and from moral beliefs about murder to philosophical beliefs about how ‘all is one’. As mentioned earlier, beliefs can also be categorised by their contents (what is the belief about? Is it about politics? Or maybe about a probable scientific theory?) Or by the attitude they embody (how does a person relate to given ideas or contents?) Here, however, we focus on the differences between matter-of-fact believing and religious credence as distinct ways in which people process ideas. There are, of course, more varieties of belief-like attitudes. But we home in on factual belief and religious credence since the differences in those attitudes matter most to the psychology of separating church and state.
Ethnographies of religion, like the anthropologist Tanya Luhrmann’s When God Talks Back (2012), feature the devout talking of their ‘beliefs’ as a matter of choice. One of the congregants from the Vineyard Church whom Luhrmann interviewed told her: ‘There is always a choice to believe what it is.’ And a common theme in the Vineyard is choosing Jesus, which means choosing to believe in Jesus. Nor is choosing ‘belief’ unique to Christianity. People around the world switch religions for reasons of romance or for other social incentives, which means voluntarily choosing what religious credence to have.
But suppose someone offered you $1 million to believe, in a matter-of-fact way, that lions are vegetarians. Could you do it? We doubt it, even if you wanted the money. You could voluntarily imagine that lions are vegetarians, or you might pretend to believe it. But if you know that lions eat meat, you can’t just choose to believe the contrary. This is one illustration of how factual belief is not under voluntary control. Or, as one of us (Neil) has put it in Religion as Make-Believe (2023): if you factually believe it, you can’t help believing it. Try for a moment to factually believe that New York City is in Quebec, that Iceland is in the Southern Hemisphere, or that Justin Bieber speaks Swahili as his mother tongue.
This gives us our first processing difference between religious credence and matter-of-fact belief:
Difference number 1: religious credence is voluntary, while factual belief is involuntary.
But if matter-of-fact belief isn’t voluntary, or responsive to the will, what does it respond to? Numerous psychological pressures come into play, including distortions like biases and misremembering. But that shouldn’t obscure the main driver of factual belief: evidence. Examples are so simple they’re easy to overlook. If you think, in a matter-of-fact way, that you started the washing machine and then, an hour later, see that the clothes you loaded are dry and still dirty, the evidence of your eyes extinguishes your prior factual belief. If you think the supermarket is open and then arrive to find its lights are all out, that evidence changes your factual belief.
Noah was an exemplar of faith for building an enormous boat when there was no evidence of an impending flood
Empirical research reveals such evidence-responsiveness already in children and even infants. According to The Scientist in the Crib (1999) by the developmental psychologist Alison Gopnik and colleagues, young children and even babies learn about the world by forming mini-theories in different domains, and testing them – much like scientists do. According to the research of the developmentalists Paul Harris and Melissa Koenig, children are also evidentially discriminating in how they learn from other people. Among other things, children are more likely to believe sources that have been accurate in the past than sources that have been inaccurate.
Is religious credence similarly responsive to evidence?
Consider faith. In Hebrews 11:1, we read: ‘Now faith is the assurance of things hoped for, the conviction of things not seen.’ If that’s right, then religious credence doesn’t depend on evidence in the way factual belief does. In the Gospel of John, for example, we read about Doubting Thomas, who demands evidence that the man before him is Jesus resurrected; he consequently is an exemplar of what faith is not. In the Old Testament, Noah was an exemplar of faith for building an enormous boat when there was no visible evidence of an impending flood. Examples can be multiplied. If faith is to stand strong, religious credence must endure the slings and arrows of absent, or even contrary, evidence.
Contemporary cognitive science of religion bears out this point. In a series of cross-cultural studies, researchers have found that, from a young age, people, even in religious societies like Iran and the US, report lower levels of confidence in the invisible religious entities they profess, like God and angels, than in invisible scientific entities, like germs and oxygen. That difference in confidence is presumably due to the felt poverty of evidence for the invisible religious entities. Yet people in some sense nonetheless ‘believe’ (read: religious credence) in the invisible religious entities. The psychologists Emily Liquin, S Emlen Metz and Tania Lombrozo conducted studies finding that participants indicated a lower need for explanation when it came to religious beliefs as opposed to scientific beliefs, while at the same time having a higher mystery acceptance for religious beliefs. So people are often fine treating their religious credences as inexplicable and mysterious.
That gives us our second difference between religious credence and matter-of-fact belief:
Difference number 2: factual belief is highly vulnerable to evidence, while religious credence is not vulnerable to evidence.
There are, of course, complications here. Religious apologists, for example, claim evidence for the stories and doctrines they profess. But claiming evidence is not the same as having one’s credences be vulnerable to it, and apologists are outliers anyway against the broad background of faithful who don’t usually pretend to treat their religious credence in an evidence-based way.
Those differences, voluntary choice versus evidence-responsiveness, are differences in how religious credence versus factual belief arise in the first place. But what about downstream differences, like how they bring about action and behaviour?
This brings us to the third difference. Have you ever heard the expression ‘once-a-week Christian’? The phrase describes people who act as though their Christian ‘beliefs’ are true one day a week – and then largely leave them dormant otherwise. Having grown up and participated in different religious traditions ourselves, we can say from personal experience that more people fit this compartmentalised model than would like to admit. Even those who practise devotion daily may act as if God is real specifically in sacred or ritual times and places, but then not really act this way otherwise. If one asks religious people whether God is ever-present in their lives, they might say yes. But such probes might be like opening the refrigerator to see whether the light is always on. In Breaking the Spell: Religion as a Natural Phenomenon (2006), the philosopher Daniel Dennett highlights this behavioural pattern in even blunter terms. He points out that people do things with God watching that they wouldn’t do with their own mothers watching, and then asks pointedly: ‘Do you really believe God is watching you? Perhaps not.’ We would say that people who profess to religious credence do typically have religious credence in a watchful deity – but, to Dennett’s point, not factual belief.
Empirical studies and ethnographic observations further exemplify the compartmentalisation of religious credence. According to one study, people in the Bible Belt start subscriptions to pornography sites at the same rate as in less conservative regions of the US – though not on Sundays, when the rate is significantly lower. According to another study, charitable giving among the religious is not significantly different from among the secular – except for on Sundays. Turning to a different religious context, the experimental economist Erik Duhaime found that, in an economic game, 100 per cent of Moroccan shopkeepers in the Medina of Marrakesh chose the most charitable option while the call to prayer was audible, dropping to 50 per cent in the 20 minutes following it.
Even if one pretends that a slab of concrete is a soft bed, one still stretches out on it slowly so as not to injure oneself
Another way of putting the point about compartmentalisation would be to say that religious credences tend to guide people’s symbolic actions – prayer, liturgies, rituals or even charity – but not their everyday instrumental actions, like commuting to work or feeding the dog. In those cases, the world is just a world, and supernatural beliefs are largely dormant. And even if a religious tradition says its religious ‘beliefs’ should govern actions all the time, the fact that such norms exist and are hard to follow is actually evidence that such religious credences usually don’t guide action outside of sacred settings. (If they did, why have such a norm?)
Matter-of-fact believing, however, is not compartmentalised in this way. Factual belief, in the sense at work here, is a person’s default cognitive map of the world, and it is always on, whenever we act. Examples of ideas most people factually believe help to illustrate this point. Dried concrete is always hard. Electricity just exists. Wood comes from trees. And so on – at least for people who believe those things in a matter-of-fact way. This lack of compartmentalisation can also be illustrated through a contrasting reference to pretend play: even if one pretends that a slab of concrete is a soft bed, one still stretches out on it slowly so as not to injure oneself, rather than flopping down on it. That’s because the factual belief that concrete is hard still guides action even when one is pretending otherwise. Whenever you perform actions, you act on a mental model of how things just are, and a person’s default mental model of what the world is like is what we are calling factual belief.
This gives us the following:
Difference number 3: religious credence is compartmentalised, while factual belief is uncompartmentalised.
This is all just a snapshot of some of the psychological processing differences between religious credence and factual belief. There are, of course, complications and details to sort out – including hard-to-categorise examples, like conspiracy ‘beliefs’, along with quasi-religious ideological credences – but those shouldn’t obscure the overall picture. Religious people, by and large, have a two-map cognitive structure with factual belief and religious credence playing distinct roles. Factual belief is the evidence-responsive default cognitive map that guides action across settings, and religious credence is a chosen, compartmentalised cognitive map that becomes active in sacralised ritual settings. Those settings may occur as often as several times a day, just one day a week, or even more rarely than that, but compartmentalisation is there in any case.
Before returning to the issue of separating church and state, let’s just address one more question about this overall psychological picture: are lay people (not just people who read and write essays like this one) aware of the psychological difference between factual belief and religious credence?
The answer, the evidence suggests, is yes – at least in a vague way.
To start, people often express their religious credences by saying things like ‘I personally believe…’ But one would not express a matter-of-fact belief, say, that their bicycle is in the garage, by saying: ‘I personally believe my bicycle is in the garage.’ They would rather say: ‘I think my bicycle is in the garage’ or just ‘It’s in the garage.’ The inclination to use ‘personally’ suggests people have an intuitive awareness that the kind of ‘belief’ being expressed is a different attitude, a different way of relating to ideas from matter-of-fact belief. And there is more systematic empirical evidence for this point, as well. One of us (Neil) together with colleagues has conducted experimental studies showing that English speakers tend to use the word ‘believes’ when it comes to describing religious attitudes but just ‘thinks’ when it comes to describing matter-of-fact beliefs.
Furthermore, such linguistic differentiation has parallels in other languages, including Fante (a dialect of the Akan people of Ghana), Thai, Mandarin and Bislama (a creole spoken in Vanuatu). Further studies show that people’s use of the word ‘believes’ correlates with seeing an attitude as being held voluntarily and not for evidential reasons – for example, out of loyalty – and as giving rise to symbolic action, like public pronouncements. People’s use of the word ‘thinks’, on the other hand, correlates with seeing an attitude as being held for evidential reasons and as giving rise to truth-dependent instrumental action (like buying property or making investments). All of that suggests that most people are at some level tracing the differences that the foregoing theory makes explicit – and they’re often using the words ‘believe’ versus ‘think’ to report religious credence versus factual belief.
Let’s return to Jefferson’s idea and to the political realm more generally. How does the psychological difference between factual belief and religious credence support the political principle of separating church and state? We think that understanding the difference – psychological dynamics and all – helps us see two justifications that we couldn’t have articulated clearly otherwise.
First, the fact that religious credences are not vulnerable to evidence helps justify the idea that liberal democracies should not establish state religions, in the sense of compelling people to adhere to them. (This is the part of the ‘wall of separation’ captured in the so-called ‘establishment clause’ of the First Amendment to the US Constitution.) The reasoning here is simple. It starts by recognising that any actions a government takes will, in point of fact, be based on, and justified by, ‘beliefs’ of some sort. For example, quarantining people who return from the Democratic Republic of Congo at the time of an Ebola outbreak is based on officials’ factual belief – justified in this case – that those people risk spreading the disease.
Importantly, if the beliefs in question are vulnerable to evidence, as we argued that factual beliefs are, then they will be susceptible to change through the kind of dialogue-based argument and reasoning that is the deliberative heart of democracy. Basically, factual belief-based reasoning and policy-making allows everyone to participate, since factual beliefs respond to evidence, and evidential reasons are the property of everyone.
But, conversely, if a ‘belief’ is not responsive to evidence, as we argued religious credences are not, then it won’t be movable by falsification-oriented argumentative reasoning. Only people who already have the ‘belief’ will be convinced by decisions based on it. Thus, a policy explicitly religious in nature – imposing religious practice, for example – would lack tractability to the defining practices of democratic decision-making. In short, liberal democracies should exclude religion-establishing policies in order to safeguard the rational communal deliberation that is at their centre.
Private motivation and public justification are different things
This perspective illuminates why policy is justified the way it is in liberal democracies. Even when they happen to be motivated by spiritual convictions, citizens tend to, and legislators must (if they are to be constitutionally respectable), frame the descriptive claims in their arguments in terms of matter-of-fact, universally tractable evidence. Such framing yields public explanations that are better suited to liberal democratic coexistence than the mere direct clash of disparate religious ‘beliefs’. This is apparent in one of the most controversial areas of contemporary policymaking: abortion law. Right now in the US, abortion policy runs from fairly permissive in some states to materially restrictive in others. Yet, while it’s true that many political actors are motivated to take positions on this issue that align with their spiritual beliefs and religious affiliations, they still attempt to justify those positions by appeal to beliefs that can be contested by reason and evidence. As a result, debate over how to respect ideals like sanctity of life can, at least in principle, be a matter of public dialogue that includes everyone.
We’ve focused on the US case here, because it is prominent in the history of separating church and state. But commitment to the kind of separation spelled out in the US with the First Amendment is mirrored across the globe, from Japan to Mexico and from Portugal to Namibia. Even the United Kingdom, which has state religions in the form of the Church of England and the Church of Scotland, places strict limits on what those state religions can and can’t do in the public sphere. Importantly, again, none of this prevents citizens and leaders from being motivated to prefer one policy over another due to their religious commitments. But private motivation and public justification are different things. Our present point is that the psychological difference between factual belief and religious credence helps explain why a principle that teases apart those two things makes sense.
Second, the fact that religious credences are psychologically compartmentalised helps justify the idea that, in a liberal society, ‘free exercise’ of religion is a reasonable ideal. (This is the other side of the ‘wall of separation’, also captured in the First Amendment: government shouldn’t stop people from choosing which religion to follow, if any.)
Here’s how this argument works. The compartmentalisation of religious credence, along with the fact that it operates as a ‘map’ distinct from factual belief, means that people who have religious credences are at least capable of setting them aside when it comes to interacting with people outside their own religion. (For example, people who have religious credence in the existence of demons don’t typically perform exorcisms on strangers in the supermarket.) So, since people are capable of setting their religious credences aside in the public sphere, they should be free to keep and venerate them in private, in light of general considerations of privacy and autonomy. Public deliberation can then proceed on the basis of evidence-responsive factual beliefs, as well as moral beliefs that respond to reasons. In short, the default compartmentalisation to which religious credence is prone makes it possible for people with a plurality of religious ‘beliefs’ to co-exist in a free society without a need for the state to challenge any of those ‘beliefs’ per se.
To put this all together, our argument is this: the deal that separation of church and state offers to the religious – namely, that everyone gets religious freedom in exchange for abiding by a constitution that prevents government from establishing their specific religion – can work only because religious credences are psychologically compartmentalised and operate at a different level from factual belief. But the very fact of that default compartmentalisation is precisely what makes such a deal worth striking. Jefferson, we suspect, intuitively sensed this, when he composed his letter to the Danbury Baptists, quoted at the outset, assuring them that they wouldn’t be oppressed for their religious commitments. And perhaps we hear a foreshadow of the very idea of compartmentalisation in his line: ‘religion is a matter which lies solely between man and his God.’
The possessive pronoun in that phrase (‘man and his God’) bespeaks the personal nature of religious credence. Religious credence, after all, is voluntary. That personal nature makes religious credence inapt for entry into public deliberation, but it also makes it ripe for receiving the freedom from government interference into private lives that characterises liberal societies generally.