你大概经历过这样一顿饭：有人想在开吃前说两句谢饭的话，桌上一半人低头，一半人愣一下，最后不了了之。也见过反过来的：一屋子人围着蛋糕关灯、许愿、吹蜡烛，没人觉得这是仪式，也没人问为什么非这么做。
这两幕摆在一起就是这一页的问题：宗教是在退场，还是换了个地方待着？下面讲四件事：这场争论里连最基本的数字都不稳；「宗教会消失吗」这句话里塞了三件不同的事；有一派干脆把教会当成市场上的供货方来算账；还有一派说，变的不是有多少人信，是「信」这件事的处境。
1993 年，哈达韦（C. Kirk Hadaway）、马勒（Penny Long Marler）和查维斯（Mark Chaves）做了件很笨的事：不问人，直接数。他们在俄亥俄州一个乡下县把新教教堂周日到场的人一个个点过来，又拿到十八个天主教教区的统计。民调里美国人自报的每周到场率长年在四成上下，数出来的大约只有一半。
不必说有人撒谎：「我是个去教堂的人」和「我上周去了」，在答问卷的两秒钟里很容易合成同一句。后来关于怎么数出过大量技术争论，方向只有一条：自报高于实测。这场大争论里最基本的事实，是被测量方式做出来的。
所以先别急着回答「宗教会不会消失」。伯格（Peter Berger）1968 年对《纽约时报》说：到二十一世纪，信教的人大概只剩一些小教派，抱成一团抵抗全球性的世俗文化。1999 年他编了本《世界的去世俗化》，开篇写：认为我们活在一个已经世俗化的世界里，这个假设是错的。他后来把整代人错得这么整齐的原因归到自己身上——大学里的人，身边确实都是不信的人。
把问题拆开的是卡萨诺瓦（José Casanova）。他 1994 年的《现代世界中的公共宗教》指出，「世俗化」这一个词捆着三条互不相干的命题——分化（differentiation）、衰退、私人化，各自的意思见下图。他的判断是：第一条大体站得住；第二条分地区、不统一；第三条反例太多，他考察的几个案例里宗教组织反倒重新走进公共争论，他管这叫去私人化（deprivatization）。
两边于是都觉得证据在自己手里：他们证的不是同一条，用的证据类型也不一样——制度史比较、调查数据、案例比较，本来就不该一起判决。
EN In 1993 C. Kirk Hadaway, Penny Long Marler and Mark Chaves did something laborious: instead of asking people, they counted. In one rural Ohio county they tallied the people who actually turned up at Protestant churches on a Sunday, and they obtained attendance counts from eighteen Catholic dioceses. In polls, Americans' self-reported weekly attendance had long sat at around forty per cent. What they counted came to roughly half of that. Nobody need be lying. "I am a person who goes to church" and "I went last week" collapse into one sentence rather easily in the two seconds it takes to answer a survey. A large and technically sophisticated literature has grown up since about how to count and how well; only one direction has held steady — self-reports run higher than counts. The most basic fact in this entire argument is partly a product of how it is measured. So hold off on answering whether religion is going to disappear. In 1968 Peter Berger told The New York Times that by the twenty-first century religious believers were likely to be found only in small sects, huddled together to resist a worldwide secular culture. In 1999 he edited a volume called The Desecularization of the World, which opens by stating that the assumption that we live in a secularized world is false. The person who took the question apart was José Casanova. His Public Religions in the Modern World (1994) points out that the single word "secularization" bundles three unrelated propositions — differentiation, decline, and privatization; the figure sets out what each one means. His verdict: the first holds up broadly; the second varies by region and does not generalize; the third has too many counterexamples — in the cases he examined, religious organizations moved back into public argument, which he called deprivatization. That is why both sides feel the evidence is on their side. They are not proving the same proposition, and the kinds of evidence differ — institutional history, survey data, case comparison. They were never meant to be judged together.
反对衰退论最硬的一版来自宗教市场论（religious economies）。斯塔克（Rodney Stark）与芬克（Roger Finke）1992 年的《美国的教会化》思路是：如果人对宗教的需求大体稳定，参与率的高低就不该往「人心变了」上归因，该往供给上归因——有没有人张罗、张罗得卖不卖力、有没有垄断替他挡着竞争。
他们的招牌数据是美国的教会归属率：1776 年不到 17%，到 1980 年前后约 62%。在一个据说越来越现代的社会里，参加的人是涨的。他们的解释是建国之初没有国教、谁都能开张，供货方不得不竞争，竞争把参与率抬了起来。反过来的推论是：欧洲的低参与率不是欧洲人更现代，是国教体制长期压着供给。
这一派挨的最硬一击是技术性的。2002 年沃斯（David Voas）、奥尔森与克罗克特指出：衡量「多元化」最常用的那个指数和参与率之间有一个数学上绕不开的关联——现实里就算毫无关系，算出来也会有相关。查维斯与戈尔斯基 2001 年汇总过这批研究，当时看到的就是正负都有、方向不稳。
更麻烦的是美国这块样板本身。沃斯与查维斯 2016 年的文章标题就是问句：美国是世俗化命题的反例吗？答案是不再是。看趋势不看水平：美国的宗教性几十年来一直在降，降法和西欧一样——一代比一代低，是世代更替在推。
所以这一派留下的有用部分不是「市场规律」，是那个纠正：别只从需求侧解释——管制和组织都能改动数字，这跟人心里怎么想是两件事。
EN The hardest version of the case against decline comes from what is called the religious economies approach. Rodney Stark and Roger Finke's The Churching of America (1992) runs like this: if demand for religion is roughly constant, then differences in participation should not be put down to a change of heart but to supply — whether anyone is organizing, how hard they work at it, and whether a monopoly is shielding them from competition. Their signature figure is the American rate of church adherence: under 17 per cent in 1776, about 62 per cent by around 1980. In a society supposedly growing steadily more modern, participation went up. Their explanation is that the new republic had no established church and anyone could set up shop, so suppliers had to compete, and competition pulled participation up with it. The reverse inference is that Europe's low participation is not a sign that Europeans are more modern, but that state-church arrangements sat on supply for centuries. The hardest blow this camp has taken is a technical one. In 2002 David Voas, Daniel Olson and Alasdair Crockett showed that the index most often used to measure pluralism is mathematically bound to the participation measure — even where no real relationship exists, a correlation falls out of the arithmetic. Chaves and Gorski had reviewed this body of work in 2001 and found exactly what you would expect on that account: results in both directions, no stable sign. More awkward still is the American showcase itself. Voas and Chaves titled a 2016 paper with the question: is the United States a counterexample to the secularization thesis? Their answer is that it is no longer. Look at trends rather than levels: American religiosity has been falling for decades, and falling the same way it falls in Western Europe — each cohort lower than the one before, driven by generational replacement rather than by people changing their minds mid-life. So what this camp leaves behind is not a market law but a correction: do not explain everything from the demand side. Regulation and organization can move the numbers, and that is a different matter from what people hold to be true.
泰勒（Charles Taylor）2007 年的《世俗时代》开头问了个怪问题：为什么 1500 年前后的西欧几乎不可能不信神，而 2000 年不信不但容易，对许多人来说甚至难以避免？
他的答案不是「少了什么」。他反对把现代讲成一个减法故事（subtraction story）——好像把迷信一层层剥掉，底下那个理性的人就自动露出来了。剥掉之后得有人重新搭：搭出一套不依靠超越者也能解释世界、安顿生死的做法，他管这套框架叫内在框架（immanent frame）。配套还有一对词：多孔的自我（porous self）对外开放，灵、运、诅咒都能进来；缓冲的自我（buffered self）只在自己脑子里生成意义，外面进不来。这个转变一发生，「信」就从背景变成了立场。
这个论点有个很实的现场。人类学家恩格尔克（Matthew Engelke）在英国人文主义者协会做了一年田野，自己受训考了葬礼司仪，参加过十六场人文主义葬礼和六场基督教葬礼，调查一百五十多位司仪，写成《一场诚实的死亡》。他发现最难的不是不提神，是那口棺材：它摆在那儿，逼着满屋子人处理「人死了以后剩下什么」，而司仪手上没有现成的话能用——不能许来世，不能说他在看着你们，得一场一场现造。
这正是泰勒那句话的现场版：世俗不是宗教被拿走后剩下的空白，它自己也是一套要费劲造出来的安排。阿萨德（Talal Asad）讲得更硬：世俗主义本身就是一种安排，它顺带规定了什么算「宗教」、宗教该待在哪间屋子里。
韦伯与理性化那期讲的祛魅（Entzauberung）到这里接上：祛魅说的是世界被改写成可计算的，泰勒说的是人被装修成不漏风的。
EN Charles Taylor's A Secular Age (2007) opens with an odd question: why was it virtually impossible not to believe in God in Western Europe around 1500, while in 2000 unbelief is not only easy but, for many people, hard to avoid? His answer is not that something was removed. He objects to telling the modern story as a subtraction story — as though you peel away superstition and the rational human being underneath is simply revealed. Once you peel, someone has to build: a whole way of explaining the world and settling birth and death without recourse to anything transcendent. He calls that framework the immanent frame. It comes with a pair of terms: the porous self, whose interior is open to the outside so that spirits, fortune and curses can get in; and the buffered self, which generates meaning inside its own head with the outside kept out. Once that shift happens, belief stops being the background and becomes a position. There is a very concrete site for this argument. The anthropologist Matthew Engelke spent a year doing fieldwork at the British Humanist Association, trained as a funeral celebrant himself, attended sixteen humanist funerals and six Christian ones, and surveyed more than 150 celebrants; the book is An Honest Death. What he found hardest was not avoiding mention of God but the coffin: it sits there, forcing a room full of people to deal with what is left of a person after death, and the celebrant has no ready-made line available — no promise of an afterlife, no "he is watching over you" — so the words have to be made fresh at every funeral. This is Taylor's point on the ground: the secular is not the blank left behind when religion is taken away, it is itself an arrangement somebody has to build, at some cost. Talal Asad puts it harder still: secularism is an arrangement in its own right, one that also lays down what counts as religion and which room religion belongs in. The disenchantment (Entzauberung) discussed in the piece on Weber and rationalization joins up here: disenchantment is about the world being rewritten as something calculable; Taylor is about the person being refitted so that nothing leaks in.
诺里斯（Pippa Norris）与英格尔哈特（Ronald Inglehart）2004 年的《神圣与世俗》给了个解释：宗教性大体跟着存在性安全（existential security）走——一个地方的人越不容易被饥荒、疫病、暴力、失业突然掀翻，宗教参与通常越低。但同一本书紧接着写：全世界抱传统宗教观念的人比历史上任何时候都多，因为宗教性高的社会生育率也高。
数字对得上。皮尤研究中心 2025 年 6 月比较 2010 与 2020 两个时点：基督徒从 21 亿涨到 23 亿，占全球人口的比例却从 31% 掉到 29%；没有宗教归属的从 16 亿涨到 19 亿，2020 年约占 24.2%。美国那边，2023 至 2024 年的大规模调查里 29% 的成年人没有宗教归属，基督徒比例的下滑近几年放缓；可同一批数据里，35% 的人宗教身份和小时候不一样。存量看着要稳住，流量一点没停。
最后一块拼图来自田野。马哈茂德（Saba Mahmood）1995 至 1997 年在开罗三座清真寺跟着女性的宗教学习圈做了两年田野。社会科学默认「有能动性就意味着反抗规范」，于是看见参与者主动把规范要求的举止练熟，就只剩「被支配」一种读法；她认为那份能动性长在别处。她的结论有人同意有人不同意，但田野说明的事清楚：当代的宗教复兴不是前现代的残渣，用的是现代的组织方式和媒介。所以世俗化不是一条曲线，是一组彼此独立的过程，在不同地方朝不同方向跑。
EN Pippa Norris and Ronald Inglehart's Sacred and Secular (2004) offers an explanation: religiosity broadly tracks existential security — the less likely people in a place are to be knocked over by famine, disease, violence or unemployment, the lower religious participation tends to be. But the same book adds a second sentence straight away: there are more people in the world holding traditional religious views than at any point in history, because the societies with high religiosity also have high fertility. The numbers fit. In June 2025 the Pew Research Center compared 2010 with 2020: Christians rose from 2.1 to 2.3 billion while falling as a share of world population from 31 to 29 per cent; people with no religious affiliation rose from 1.6 to 1.9 billion, about 24.2 per cent of the world in 2020. In the United States, a large survey in 2023–24 found 29 per cent of adults with no religious affiliation, and the decline in the Christian share has slowed in recent years — yet in the same data 35 per cent of adults hold a religious identity different from the one they were raised in. The stock looks as though it is settling; the flow has not stopped. The last piece comes from fieldwork. Saba Mahmood spent two years, from 1995 to 1997, with women's religious study circles at three mosques in Cairo. Social science defaults to the assumption that having agency means resisting norms, so when it sees participants deliberately practising the conduct the norms require until it becomes habit, the only available reading is domination; she argued that the agency lies elsewhere. Her conclusions have their supporters and their critics, but what the fieldwork shows is clear: contemporary religious revival is not a pre-modern leftover — it runs on modern forms of organization and modern media. So secularization is not one curve. It is a set of independent processes running at different speeds in different directions in different places.