亲属看起来最不像社会学的题目——爸妈、兄弟姐妹，那不是生物学的事吗。可这门学科在上面耗掉了一百多年，画满了三角和圆圈的图。原因很实在：在没有国家、没有公司、没有法院、没有一纸合同的地方，亲属就是那套操作系统。谁归谁养、地归谁种、打起来你站哪边、你死后谁接着替你活，全由它派。
反直觉的地方在于它派活时不看血。它先摆好一批位置，每个位置挂着一套义务，然后才决定谁站进去。这一页拆三件事：位置怎么运作，「不能娶谁」这条规矩真正在管什么，以及你嘴上那些称谓其实是一张什么图。
1930 年，一个英国人去当时英埃共管的苏丹南部（今南苏丹）研究一个叫 Nuer（努尔）的牧牛民族。他叫 Evans-Pritchard，这趟差事是那边的政府请他去的，而请他去的缘由并不体面：政府刚跟 Nuer 人打完仗。这一段他自己写进了书里——有天清晨政府部队包围了他的营地，搜捕不久前起事的两位先知，抓走人质；那阵子 Nuer 人对外人格外不友善，新败和随后的镇压让他们怨气很深。观察可以是准的，同时它被采集的位置得摆在桌面上。
他要弄明白的是：这群人靠什么组织起来。答案在牛身上。一桩婚姻要把牛交给女方亲属才算完成，而关键不是交多少头，是哪一头牛归哪一个亲属——母亲的兄弟拿哪几头、父亲那边的谁拿哪几头，都有定数。一次分牛就是把这家的亲属结构念了一遍。打起架来也一样：谁跟谁站一边不看交情，看两家在族谱上从哪一层分岔。
真正把「亲属就是血缘」顶翻的是两种婚姻。一种是鬼婚：一个男人死了、没有儿子，他的兄弟或侄子会用他的名义娶一个妻子；付牛的是活人，可孩子在法律上属于那个死人，姓他的名、继承他的牛、接他的祭祀位置。这种婚姻在 Evans-Pritchard 的记录里常见到不能当特例处理。另一种是女人娶妻：一个生不出孩子的女人可以自己付牛去娶一个妻子，于是她就是那些孩子的「父亲」；真正让妻子怀孕的那个男人在这套算法里没有父亲的位置。
这里有两个词得分开，后面一直要用：pater（社会学意义上的父亲，占着那个位置的人）和 genitor（生育者）。Evans-Pritchard 的记录很直白：只要法律上的父亲身份确立无疑，Nuer 人对生育的方式不太计较。位置在前，人在后。证据强度要说清：这是 1930 年代的单点民族志，材料来自一个人的观察，没有第三方数据可对照；Nuer 人今天在南苏丹，是活着的人，不是标本。
EN In 1930 Evans-Pritchard went to the southern Sudan to study the Nuer, cattle-keeping people of what is now South Sudan. The colonial government commissioned the work, and the reason was not creditable: it had just finished fighting them. He recorded the position himself — one morning a government force surrounded his camp hunting two prophets who had led a recent revolt and took hostages, and he noted the Nuer were unusually hostile because their defeat and the measures that followed had bred deep resentment. An observation can be accurate while the place it was gathered from stays on the table. What he wanted to know was how these people were organised, and the answer was in the cattle: a marriage is completed by handing cattle to the bride's kin, and what matters is not how many but which animal goes to which relative, so one distribution recites the whole kinship structure. Two kinds of marriage overturn any reading of kinship as biology. In ghost marriage, a man who dies without sons has a wife married in his name by a brother or nephew; the living pay the cattle, but the children are legally the dead man's, carry his name and inherit his herd. In woman-marriage, a woman who cannot bear children pays cattle to marry a wife and is thereby the father of the children, while the man who begets them holds no paternal position at all. Hence two words that matter: pater, the person occupying the position, and genitor, the begetter. Evans-Pritchard put it plainly — the Nuer paid little regard to the manner of begetting so long as legal fatherhood was well established. Position first, person second. The same logic cornered the word marriage: working on the Nayar of Kerala in 1959, Kathleen Gough found an arrangement with neither co-residence nor exclusive conjugal rights, yet perfectly clear status for children, so her definition had to be written from the child's side — a relationship that grants a woman's child the full birth-status rights of a member of that society. On evidence strength: by the time she did fieldwork in the 1940s the classical Nayar system had been dismantled by legislation, so it was reconstructed from documents and memory, not observed.
乱伦禁忌到处都有，这件事早年被当成一个谜：为什么全世界都禁？是本能吗？1949 年 Lévi-Strauss 在《亲属关系的基本结构》里换了个问法。一条禁令的背面是一条命令。禁止你娶自己的姐妹，等于规定你必须把姐妹给别人——于是你家和另一家之间有了一条必须维持下去的关系。禁忌管的不是防止什么，是强制交换。
交换有两种跑法。限定交换：两群人互换姐妹，闭合的小圈，稳，但只串得起两家。一般交换：甲给乙、乙给丙、丙给甲，开放的长圈，能把很多群体串成一片，代价是你得先把人给出去、回报要等一轮——这套东西靠赊账运转，跟「礼物与互惠」那期的延迟回报是同一副骨架。
最能说明问题的是一个细节。父亲的兄弟的女儿，和母亲的兄弟的女儿，跟你的血缘距离一模一样。可在很多父系社会里，前一个算「姐妹」，娶她就是乱伦；后一个恰恰是首选的配偶。同一个生物学距离，被切成了两个相反的类别。依据不神秘：前者跟你在同一个世系群里，后者在你要跟人家交换的那个群里。切分的尺子是交换关系，不是血缘远近。
三层批评都得摆出来。一，它把女人写成被男人交换的东西，这是最重的一刀。二，有个大范围的反例：阿拉伯—伊斯兰世界的许多地方，首选配偶恰恰是父亲的兄弟的女儿，即结构逻辑判为不可能的那一类；Korotayev 的跨文化统计把这种偏好的分布跟八世纪阿拉伯—伊斯兰化所及的范围对上了号。那套结构不是普遍法则。三，Rodney Needham 的澄清最要紧：这类规则不是「大多数人真娶了表妹」的统计偏好，而是一套把人分成可婚与不可婚两类的分类系统；真实婚配里跟族谱上那个具体表妹成婚的比例常常很低，原因扫兴——小人群里那个人可能根本不存在。证据强度：这是拿民族志报告和称谓表搭的结构模型，不是测出来的婚配行为；1973 年 Francis Korn 逐案重查，结论是分类撑不住细节。
EN The incest taboo is everywhere, and for a long time that universality was the puzzle. In The Elementary Structures of Kinship (1949) Lévi-Strauss changed the question: the reverse side of a prohibition is a command. Forbidding you to marry your sister amounts to requiring you to give her to someone else, which creates a relationship between two groups that must be maintained. The taboo does not prevent something, it forces exchange. Restricted exchange has two groups swapping sisters in a closed loop — stable, but it links only two. Generalised exchange runs A to B to C back to A, an open circle that can chain many groups together, at the price of giving first and being repaid a turn later: it runs on credit, the same skeleton as the deferred return in the topic on gifts and reciprocity. The telling detail: your father's brother's daughter and your mother's brother's daughter are exactly the same genetic distance from you, yet in many patrilineal societies the first counts as a sister and marrying her is incest, while the second is the preferred spouse. One biological distance, cut into two opposite categories — because the first is inside your own descent group and the second is in the group you exchange with. The ruler doing the cutting is the exchange relation, not degree of relatedness. Three criticisms belong on the table. The scheme writes women as objects exchanged by men, which is the heaviest cut. There is a large counterexample: across much of the Arab-Islamic world the preferred spouse is precisely the father's brother's daughter, the class the structural logic rules out, and Korotayev's cross-cultural statistics match the distribution of that preference to the reach of eighth-century Arabisation and Islamisation — so the structure is one type of system's logic, not a universal law. And Rodney Needham's clarification matters most: these rules are not statistical preferences but classifications sorting everyone into marriageable and non-marriageable, since the proportion of actual marriages to the genealogically exact cousin is often low for an unglamorous reason — in a small population that person may not exist. Evidence strength: a structural model built from ethnographic reports and terminology tables, not measured marriage behaviour; Francis Korn's case-by-case re-examination in 1973 concluded the classification does not survive the details.
回到过年那一桌。姑姑和姨、伯和叔、堂哥和表哥，你分得清清楚楚。英语里 aunt 一个词包掉四个人，cousin 一个词包掉八种人。第一个把这件事当数据处理的是 1871 年的 Lewis Henry Morgan，书叫《人类家族的血亲和姻亲制度》，由史密森学会出版。他的办法在当时很新：印好表格，托传教士、印第安事务代理人、领事寄到世界各地去填，汇了一百四十来个社会的称谓表——人类学最早的一次全球问卷。他自己的解释（称谓是过去婚姻形式的化石）被推翻了，数据活了下来。1949 年 George Murdock 把这些表整理成六型。英语属于 Eskimo 型：只分直系和旁系，父亲的兄弟和母亲的兄弟共用一个词。汉语属于 Sudanese 型，几乎每个位置一个专名，是六型里区分最细的。
机制说出来很朴素：称谓做的事是把人归进等价类——同一个词底下的人，对你的义务是同一套。哪里合并、哪里拆开都不随意。Iroquois 型里父亲和父亲的兄弟共用一个词、母亲的兄弟单独一个词，而这条线正好就是「谁家女儿能娶」那条线。所以称谓表和婚姻规则是同一件东西的两个面；也正因如此，「不能娶谁」那节的规则才跑得动：你不必记族谱，你只要知道对方该叫什么。
再往下有一记很重的自省。1984 年 David Schneider 出《对亲属研究的批评》，说这门学科一直悄悄假定亲属就是生育关系，把欧美的一套民间理论——「血」——当成了普遍骨架；他 1968 年那本《美国亲属制度》先论证了「血亲」本身就是一套文化学说，不是自然事实。这是概念批评，不是测量，学科内至今没有共识。
有意思的是：看起来最异域的那条区分，今天在法院里天天用。辅助生殖把 genitor 这个位置拆成了好几格——卵子提供者、代孕者、法律上的父母——家事法庭要判的正是 pater 与 genitor 之分，只不过写在判决书上。另一头，1991 年 Kath Weston 在旧金山湾区做田野写成《我们选择的家庭》，记录同性伴侣和朋友圈怎么用「家人」这个词办同一批事：照护、义务、谁算内圈。最日常的那层是：管不相干的人叫「叔叔」「阿姨」、师门里互称「师兄」，做的是同一件事——把人塞进一个类，类自带一套义务。
EN Back to the family table. Chinese distinguishes father's sister from mother's sister, father's elder brother from father's younger brother, paternal from maternal cousins; English packs four people into aunt and eight kinds of relative into cousin. The first person to treat this as data was Lewis Henry Morgan, whose Systems of Consanguinity and Affinity of the Human Family was published by the Smithsonian in 1871. His method was new: printed schedules mailed out through missionaries, Indian agents and consuls, yielding terminologies from some hundred and forty societies — anthropology's earliest global survey. His own interpretation, that terminologies are fossils of earlier marriage forms, was overturned; the data survived. In 1949 George Murdock sorted the tables into six types, each named for a sample society. English is the Eskimo type, distinguishing only lineal from collateral kin, with one word covering both a father's brother and a mother's brother. Chinese is the Sudanese type, with a separate term for nearly every position, the most finely divided of the six. The mechanism is plain: terminology sorts people into equivalence classes, and everyone under one term owes you, and is owed, the same set of obligations. Where terms merge and where they split is not arbitrary. In Iroquois-type systems father and father's brother share a term while mother's brother has his own, and that line is exactly the line of whose daughter you may marry. Terminology table and marriage rule are two faces of one thing, which is why the rule can actually run: you do not need to remember the genealogy, only what to call the person. Then a heavy piece of self-examination. In A Critique of the Study of Kinship (1984) David Schneider argued that the discipline had quietly assumed kinship was reproduction, taking a Euro-American folk theory — blood — as a universal skeleton; his American Kinship (1968) had done the first half of the job, arguing that "blood relation" is itself a cultural doctrine rather than a natural fact. That is a conceptual critique, not a measurement, and no consensus followed. What makes the contemporary echo interesting is that the most exotic-looking distinction is now used in court every day. Assisted reproduction splits the genitor position into several — egg donor, gestational surrogate, legal parents — and family courts rule on exactly the pater/genitor difference, only written into judgments. On the other side, Kath Weston's Families We Choose (1991), from fieldwork in the San Francisco Bay Area, recorded how lesbians and gay men used the word family to do the same jobs: care, obligation, who counts as inner circle. And the most everyday layer: calling an unrelated person uncle or auntie, or a fellow student elder brother, does the same work — putting someone into a class, and the class brings its obligations with it.