你第一次走进一个陌生场合——很贵的餐厅、别人公司的会议室、一场谁都比你熟的家宴。没人发说明书，但你几秒内就把自己调过了：走路的速度、说话的音量、手放哪儿、什么时候笑。事后你说不出依据的是什么规则。
这一页拆三件事：那套说不出来的东西怎么装进身体里；它在学校和公司这类机构里怎么被读成「个人素质」；以及为什么谈品味时，「什么算过头」比「什么算好」更能定位一个人。
Pierre Bourdieu 的起点不是理论，是两片田野。1950 年代末他在阿尔及利亚的卡比利亚（Kabylia）山区做人类学调查；回到法国后，1960 年代初他回自己长大的贝亚恩（Béarn）——法国西南的山村——弄清一件当地人自己也困惑的事：长子娶不到妻子。
按当地继承规矩，农庄归长子，照规矩推他该是最有条件的对象。可 1950 年代城市化加速，姑娘们更愿意嫁进城里，于是按规矩最占优的那批人反而大量单身。Bourdieu 注意到的是：如果「规则」能解释行为，就不该长出这个结果。
真正在运作的是另一层东西——什么样的亲事「说得过去」，乡下舞会上谁敢去邀谁跳舞，一个农民走路和站立的样子在城里长大的姑娘眼里意味着什么。这些没一条写在纸上，却决定了谁结成婚。他 1962 年写贝亚恩单身现象那篇文章里第一次用了 habitus（惯习）这个词，并明确接上人类学家 Mauss 讲过的「身体技术」：连怎么走路都是学来的。
惯习不是一份行为清单，是一台生成器：长期处境沉淀下来的一套身体化倾向——姿态、口音、什么值得追求的直觉、什么让你觉得别扭。关键性质是生成性：你没遇过的场合它也能替你即时生成反应。所以它比规则难改——规则可以背，倾向只能练。这也解释了那种熟悉的窘迫：你明明知道该怎么做，身体还是慢半拍。
这套说法最常挨的批评很硬：它有循环的危险。任何行为都能事后说成「这是惯习」，那它就没禁止任何事，也谈不上被证伪；而且它极难测量。证据强度要说清：惯习是从民族志材料里长出来的概念，不是一条可检验的预测。它的价值在于让人看见——一批原本被当成「个人风格」的差异其实有系统来源。
EN Bourdieu's starting point was fieldwork, not theory: Kabylia in Algeria in the late 1950s, then his own village region of Béarn in south-west France in the early 1960s, where he investigated something locals themselves found puzzling — eldest sons, who inherited the farm and by the rules should have been the best matches, were staying single in large numbers as 1950s urbanisation drew young women toward town husbands. If rules explained conduct, that outcome should not exist. What actually operated was a feel for which match was thinkable, who dared ask whom to dance, what a peasant's way of walking meant to a girl raised in town. His 1962 article on Béarn celibacy was the first place he used the word habitus, explicitly borrowing Mauss on techniques of the body. Habitus is not a list of behaviours but a generator: a system of embodied dispositions laid down by a durable position — posture, accent, intuitions about what is worth wanting, what feels awkward — able to improvise a response in situations never encountered. That is why it resists change: rules can be memorised, dispositions only trained. The standard objection is serious — the concept risks circularity, forbids nothing, and is very hard to measure. Evidence: a concept grown from ethnographic material, not a testable prediction.
家长会上老师说「这孩子表达能力挺好」。表达能力是哪门课教的？成绩单上没有这一栏，可它进了评价。
Bourdieu 和 Jean-Claude Passeron 1964 年的《继承人》、1970 年的《再生产》盯的就是这个缺口：法国高等教育里不同出身的学生比例悬殊，而学校真诚地相信自己在公平选拔。他们的论点不是学校作弊，是学校奖励的能力有一部分不在它的课程里——那部分在家里顺带学会，不交学费但很花钱（花的是时间、藏书、大人的谈话方式）。社会学管这类东西叫文化资本（cultural capital）。「社会分层」那期讲精英招聘时那套「文化匹配」，是同一台机器在职场端的样子。
Bourdieu 把它分成三种状态，这个区分比概念本身更有用：身体化的（口音、姿态、什么话题算得体——只能长年浸出来）；客体化的（家里的书、乐器、画——可以买，但买了不等于用得上）；制度化的（文凭）。第三步是关键的一跃：文凭把「我家里有这些」换成「我有这个资格」，来源在兑换的瞬间被切断了。
能不能量出来？Paul DiMaggio 1982 年在《美国社会学评论》上试过：用 1960 年 Project Talent 调查里十一年级学生的数据，把「想听古典音乐会、读严肃文学、自认有教养」这类回答做因子分析当作度量。控制了能力测验分数和家庭地位之后，它仍与在校成绩相关，且效应在男女生之间不一样。证据强度：单次横截面相关，1960 年美国数据；后来在别国复制，方向常在、大小差得多——不是一条已被牢固复制的定律。
更有画面的是 Annette Lareau 的《Unequal Childhoods》（2003）。她和助手访谈了 88 个孩子及其家长，其中 12 家深度跟踪。最关键的场合之一是儿科诊室：中产家庭的孩子进去前被父母「教练」过，会主动向医生提问、甚至纠正医生；工人阶级和贫困家庭的孩子被教导在权威面前保持安静和尊重。两种教养她分别叫协作培养（concerted cultivation）和自然成长的达成（accomplishment of natural growth）。
她提醒的一点常被引述者扔掉：后一种在很多方面对孩子更好——自由玩耍多得多、和亲属更亲。它的问题不在孩子身上，在于它在学校和医院这类机构里兑不出好处。证据强度：小样本民族志，机制看得清，比例别外推。
EN A teacher says a child "expresses herself well" — which class taught that? Bourdieu and Passeron (Les Héritiers, 1964; La Reproduction, 1970) argued that part of what schools reward is not in their curriculum: it is picked up at home, tuition-free but not cost-free. Cultural capital comes in three states, and the distinction matters more than the term. Embodied: accent, bearing, a feel for what topic is appropriate — only absorbed over years. Objectified: the books, instruments and pictures in a house — purchasable, though owning is not using. Institutionalised: the credential, and this is the decisive step, because a diploma converts "my family had these things" into "I hold this qualification", severing the origin at the moment of exchange. DiMaggio (American Sociological Review, 1982) tried to measure it, factor-analysing eleventh-graders' reported interest in classical concerts, serious literature and art in the 1960 Project Talent data; net of ability scores and family status it still predicted grades, with different effects for boys and girls. Evidence: one cross-sectional correlation on 1960 US data, replicated abroad with consistent direction but widely varying magnitude. Lareau's Unequal Childhoods (2003) — 88 children interviewed, 12 families followed closely — caught the mechanism in a paediatrician's office: middle-class children coached beforehand to question and even correct the doctor, working-class and poor children taught to be quiet and deferential before authority. She calls these concerted cultivation and the accomplishment of natural growth, and insists on a point often dropped: the second gives children far more free play and closer kin ties. Its disadvantage is not in the child but in the fact that institutions do not cash it. Evidence: small-sample ethnography.
有人说某个东西「有点土」「太用力了」。你大概能感觉到他指什么，但谁也说不出标准在哪。Bourdieu 1979 年那本《区隔》（La Distinction，英译 1984）整本都在拆这个感觉。
数据是实打实的问卷：1963 年和 1967–68 年两次调查合计 1217 人，问音乐、绘画、摄影、家具、饮食、穿着。结果是偏好与教育资本、与父辈职业系统对应。手头一直紧的处境里长出来的，他叫「必需品的品味」——实在、够量、功能优先，看照片要看得出拍的是什么；教育资本高的一端更看形式而非内容，能对着一张「没什么可看的」照片谈很久。
他最反直觉的一句是关于机制的：品味主要靠否定来运作。审美上的对立首先是社会性的对立——说「太甜」「太花哨」时，划的是自己和谁不同。所以问一个人喜欢什么，答案常常模糊；问他受不了什么，位置立刻清楚。而且哪种品味算「正当」不是自然长出来的，由学校、博物馆、评论界背书。「戈夫曼的拟剧论」那期讲的前台表演是有意识地管理印象，这里不同——你觉得什么过头，不是演出来的，是你没法不这么觉得。
这套说法今天还成立多少？最重要的修正来自 Richard Peterson 和 Roger Kern 1996 年那篇文章：比对美国 1982 年和 1992 年两次「公众艺术参与调查」（Survey of Public Participation in the Arts），他们发现高地位者不再只消费高雅文化，消费面明显更宽，而且这变化既来自世代替换也来自同一批人自己变了。他们把这种人叫文化杂食者（cultural omnivore）。这不推翻区隔，改的是区隔的形式：区分点从「听什么」移到「听多少种、以什么姿态听」——什么都听且每种都说得头头是道，本身成了新的稀缺。
最强的反方值得原样呈现。John Goldthorpe 2007 年的批评有两条：文化资本这概念形式化不清、因果链条弱；解释教育不平等根本不需要它。他和 Richard Breen 1997 年提出的相对风险规避模型只假设一件事——各阶层家庭在升学的每个分岔口上首要目标都是避免向下流动，而「向下」对不同起点意味着不同距离，光这个不对称就能推出差异化选择。两者被直接比过：2007 年一篇用阿姆斯特丹中学生数据的研究结论是两套机制管的环节不一样——文化资本更多与在校成绩相关，相对风险规避更多解释成绩相同时选哪条路。证据强度：单一城市样本，各国结论不一致。这场争论没有收敛，两边都别当定论。
EN La Distinction (1979) rests on surveys of 1,217 people in 1963 and 1967–68 covering music, painting, photography, furniture, food and dress; preferences map systematically onto educational capital and parental occupation. Constrained circumstances breed what Bourdieu calls the taste of necessity — substantial, functional, a photograph should show what it is of — while the high-education pole attends to form over content. His most counter-intuitive claim is about mechanism: taste works chiefly by refusal. Aesthetic opposition is social opposition first, so asking what someone likes yields vagueness, while asking what they cannot stand locates them at once — and which taste counts as legitimate is underwritten by schools, museums and critics. Unlike the managed front-stage performance in the dramaturgy topic, this is not acted: you cannot help finding certain things overdone. The main revision: Peterson and Kern (1996), comparing the 1982 and 1992 Survey of Public Participation in the Arts, found high-status Americans consuming a far wider range — cultural omnivores — driven both by cohort replacement and by change within cohorts. That relocates distinction rather than dissolving it: the marker moves from what you listen to toward how many kinds and in what stance. The strongest counter-position is Goldthorpe's (2007): the concept is loosely formalised, its causal basis weak, and educational inequality can be explained without it. The relative risk aversion model of Breen and Goldthorpe (1997) assumes only that families at each educational branching point aim above all to avoid downward mobility, an asymmetry that generates class-differentiated choices with no cultural mechanism. Studies comparing the two, including a 2007 analysis of Amsterdam secondary students, suggest they govern different links in the chain — cultural capital more closely tied to school performance, relative risk aversion to which track is chosen at equal performance. Evidence: single-city samples, inconsistent across countries; the dispute has not converged.