你填过的每一张表，开头都是那几个格子：姓名、住址、证件号。看着天经地义。可固定的、父子相传的姓氏在世界很多地方只有一两百年历史，而且它出现的理由写得明明白白——为了收税和征兵。
这一页拆三件事：一个政府为了把治下看清楚都做过什么；那些怎么也写不进表格的知识长什么样；以及，有没有人干脆把日子过成了不值得被登记的样子。
十八世纪后期的普鲁士和萨克森，林务官手上有了个新工具，叫标准木（Normalbaum）：先量出一棵「标准」树能出多少材积，再据此推算整片林子值多少钱。这是个漂亮的发明——森林第一次变成了可以写进预算的数字。
接下来那一步才要命。台账要准，最费劲的办法是改进测量；最省事的办法是改造森林。于是灌丛清掉、杂树伐掉、落叶耙走，换上同一树种、同一年龄、成行成列的人工林，主要是云杉和赤松。走进去像走在货架中间。第一茬长得极好，产量可预测，官员很满意。
第二茬出事了。大约一百年后，同样的地块产量掉了两到三成，德国林学界给最坏的情形造了个词：Waldsterben（森林死亡）。后来重建的解释是：第一茬吃的是老林子几百年攒下的土壤本钱——腐殖质、菌根、昆虫、鸟、倒木。这些东西一样都没进台账，因为它们不出木料。台账不是漏记了它们，是根本没有装它们的格子。
证据强度得说清楚：这是历史个案加事后重建，不是对照实验；而且 Waldsterben 这个词在 1980 年代主要被用来指酸沉降造成的森林损害，跟单一栽培那件事不完全是一回事，两者别混着讲。
斯科特（James C. Scott）1998 年那本书把这件事提炼成一条机制，叫清晰化（legibility）。要点不在「国家想多知道点」——谁都想多知道点。要点在于：国家能用的知识必须可数、可比、能在两百公里外的办公室核对。一个只对本地人成立的事实，对那间办公室等于不存在。
姓氏就是这么来的。1849 年西班牙殖民当局在菲律宾颁了克拉韦里亚法令，随令发下一本六万多个姓氏的册子，各省按镇、按字母分配，一家一个。法令把理由写得很直白：便于登记人口、征税、征兵。
话要说全：同一套工程也是公共卫生、普及教育、疫苗记录和灾害救援的前提——找不到你，就没法给你打疫苗。清晰化不是坏事，是有代价的事。代价在下一节。
EN In late-eighteenth-century Prussia and Saxony, foresters acquired a new instrument: the Normalbaum, a standard tree whose timber volume let them extrapolate what a whole forest was worth. For the first time a forest could be written into a budget. The decisive move came next. To make the ledger accurate, improving measurement is laborious; remaking the forest is cheap. Undergrowth was cleared, litter raked, mixed stands replaced by single-species, same-age rows, mostly spruce and pine. The first rotation grew superbly. About a century later the second rotation on the same ground yielded twenty to thirty percent less, and German forestry coined a word for the worst cases: Waldsterben, forest death. The reconstruction is that the first crop had been living on soil capital the old forest had banked over centuries — humus, mycorrhizal fungi, insects, birds, fallen wood — none of which appeared in the ledger, because none of it yields timber. The ledger had not omitted them; it had no column for them. Evidence strength: a historical case with retrospective reconstruction, not a controlled experiment, and the word Waldsterben in the 1980s referred mainly to acid-deposition damage, a different matter that should not be conflated with this one. James C. Scott's 1998 book distils this into one mechanism he calls legibility. The point is not that states want to know more — everyone does. It is that usable state knowledge must be countable, comparable and verifiable from an office two hundred kilometres away; a fact that holds only locally does not exist for that office. Surnames arrived the same way: in 1849 the Spanish colonial administration in the Philippines issued the Clavería decree with a catalogue of more than sixty thousand surnames, distributed town by town and alphabetically, one per household, for the stated purposes of registration, taxation and conscription. Say the whole thing, though: the same apparatus underwrites public health, mass schooling, vaccination records and disaster relief. Legibility is not a bad thing. It is a costly thing.
有一种罢工不停工。工人照常上班，只是把所有规章一条不落地执行——法语叫 grève du zèle，热心罢工。结果是生产垮掉。这件事比任何论证都直接：写下来的规程从来没有自足过，它一直跑在一大堆没写下来的判断上面。
斯科特给这堆判断起了个名字：米提斯（mētis），借自古希腊语，取自德蒂安与韦尔南 1974 年讲希腊「狡黠智慧」的那本书。特征很好认：在地、随情境、只能靠做出来学会、一写成手册就掉一半。最古典的例子是引航——大船进港要请本地人上船，不是因为外来船长技术差，是因为那条水道的沙洲每年都在动，而「今年沙洲在哪」不在任何一张海图上。
农业上有个扎实的田野。Paul Richards 1985 年的《本土的农业革命》写西非（主要在塞拉利昂）农民的混作：一块地上同时种好几样作物，成熟期错开、根系深浅不一、高的给矮的遮阴。殖民时期和之后的农艺官长期把这看成落后，推行单一作物成行播种。理由不全是偏见——单作能测产、能横向比较、能派人下去督导，混作这三样全做不到。Richards 指出，在当地的土壤与降雨条件下，混作往往更稳：抗旱、抗病、劳力摊得开。证据强度：田野研究加农学比较，不是普遍结论——换个条件，单作的产量确实更高。
规模最大的一次是坦桑尼亚 1973 到 1976 年的村庄化。分散的农户被迁进规划好的新村，人数口径差别很大，从五百多万到上千万的估计都有。目的不坏，也部分做到了：住得集中才好建学校、通水、派农技员。但新村的位置是按图纸和路线定的，不是按各家几十年攒下的「哪块地哪年种什么、哪个坡先旱、哪片洼地涝年反而保收」。产量下去了。
这本书最可迁移的是四个条件同时出现才出灾难：行政上的简化、一种「我们已经看懂了」的自信、一个足以强推的权力、一个无力抵抗的社会。少一样，通常只是个难看的项目，不是灾难。
批评也要给。有人说他对国家能力过于悲观，也有人说他把米提斯浪漫化了——本地知识同样会错、同样会固守有害的旧法，而且「说不清楚」不等于「是对的」。这一页不选边：两种读法各自有站得住的案例。
EN One kind of strike involves no stoppage at all. Workers turn up and follow every rule to the letter — the French call it grève du zèle, a zeal strike — and production collapses. Nothing argues the point more directly: written procedure has never been self-sufficient, because it runs on a mass of unwritten judgement. Scott names that judgement mētis, borrowing the Greek term from Detienne and Vernant's 1974 study of cunning intelligence. Its marks are easy to spot: local, situational, learned only by doing, and half-destroyed by being written into a manual. The classic case is river pilotage — a large ship takes a local pilot aboard not because the visiting captain lacks skill, but because the sandbars in that channel shift every year and where they are this year is on no chart. A solid field case comes from agriculture. Paul Richards's Indigenous Agricultural Revolution (1985) describes intercropping among West African, largely Sierra Leonean, farmers: several crops in one plot, maturing at different times, rooting at different depths, the tall shading the short. Colonial and later agronomists long read this as backward and pushed single crops in rows. The reason was not purely prejudice: monocrops can be yield-measured, compared across sites and supervised, and mixed plots allow none of the three. Richards argues that under local soils and rainfall, intercropping is often the more robust system — against drought, against disease, and in spreading labour across the season. Evidence strength: field research plus agronomic comparison, not a universal finding; under other conditions monocropping does yield more. The largest case is Tanzania's villagisation of 1973–76, when dispersed households were moved into planned villages; the counts vary widely, with estimates from somewhat over five million to above ten million people. The aims were not bad ones and were partly met, since concentration is what makes schools, water and extension services feasible. But the sites were chosen from plans and road lines rather than from what each household had learned over decades about which plot to plant in which year and which slope dries first. Yields fell. The book's most transferable passage is its four conditions, which produce catastrophe only in combination: administrative simplification, a confidence that the thing is now understood, a power able to impose it, and a society unable to resist. Remove one and you usually get a bad project rather than a disaster. Criticism belongs here too: Scott is read by some as too pessimistic about state capacity, and by others as romanticising mētis — local knowledge also errs, also clings to harmful practice, and being hard to articulate is not the same as being right. No side is taken here.
先看两种作物的差别。稻谷麦子长在地上，同时成熟，耐存耐运——收税的人骑马过来，一眼估得出这片地能征多少，而且非得在收割那几天来。薯类木薯这类块茎长在地下，什么时候要什么时候刨，刨出来放不久——要征收就得一块地一块地挖。同样是种地，一种几乎是为征税准备的，另一种几乎是征不动的。
Zomia 这个名字是历史学家范申德尔 2002 年提出的，取自当地藏缅语族里一个表示「山地人」的说法，指东南亚大陆海拔三百米以上那一大片山地，约两百五十万平方公里，跨着好几个国家，住着约一亿人。斯科特 2009 年那本《逃避统治的艺术》给了个挑衅的读法：这些人群身上被谷地视为「落后」的特征——游耕而不定居、口传而不用文字、谱系说改就改、社会结构分散没有中心——可能不是「还没发展到那一步」，而是次生的、策略性的，每一样都在降低被征税、被征兵、被抓去服徭役的概率。
批评挨得很重，要原样交代。历史学家指出，山地与谷地长期互市、进贡、通婚，不少山地群体自己也建过政体，「逃避」远不是唯一解释；把几百个来历各异的群体统一读成「逃避者」，本身就是一次简化——正是「为了把树数清楚，他们重新种了一整片森林」那节批评的那种简化。斯科特自己说这个论点对 1945 年之后基本不适用：现代国家的道路、通讯和测绘把摩擦地形抹平了。证据强度：历史人类学的解释性论证，不是可检验的因果推断。
可以带走的不是「山民真高明」——那只是把「落后」换成「高明」，人照样没有自己的历史。可以带走的是一条更冷的规律：任何东西一旦被用来征收或考核，它就会开始朝难以被计量的方向变形。组织与制度那期讲过组织为什么长得越来越像；这里是反过来的一半——被计量的那一方也在动，而且往往动得更快。
EN Start with two kinds of crop. Grain grows above ground, ripens at once and stores and travels well: a tax collector riding out can estimate the yield at a glance, and must come during the few days of harvest. Tubers such as yams and cassava grow underground, are dug when wanted and keep badly, so collecting them means digging plot by plot. The same farming labour is, in one case, almost purpose-built for taxation and, in the other, almost impossible to tax. The name Zomia was proposed in 2002 by the historian Willem van Schendel, from a Tibeto-Burman term for highlander, for the belt of mainland Southeast Asian uplands above three hundred metres — some 2.5 million square kilometres across several countries, home to roughly a hundred million people. Scott's The Art of Not Being Governed (2009) offers a provocative reading: the traits the lowlands read as backward — shifting cultivation rather than settlement, oral transmission rather than writing, genealogies revised at will, dispersed social structures without a centre — may not be stages not yet reached but secondary and strategic, each lowering the odds of being taxed, conscripted or seized for corvée labour. State power, on this account, does not fall off evenly with distance but with the cost of extraction, which Scott calls the friction of terrain; premodern borders look more like contour lines than like the straight lines on maps. The criticism is heavy and belongs here in full. Historians point out that uplands and lowlands traded, paid tribute and intermarried over long periods, that a number of upland groups built polities of their own, and that evasion is far from the only explanation. Reading hundreds of differently descended groups as a single category of evaders is itself a simplification — exactly the kind the forestry section criticises. Scott himself holds that the argument largely stops applying after 1945, when roads, communications and survey flattened the friction of terrain. Evidence strength: an interpretive argument in historical anthropology, not a testable causal claim. What travels is not admiration for highlanders, which merely swaps backward for clever and still leaves them without a history of their own. What travels is colder: once something is used for extraction or assessment, it starts deforming towards whatever is hard to measure. The piece on organisations and institutions covered why organisations come to look alike; this is the other half, in which the measured party moves too, and often moves faster.