你请邻居代收了个快递。想塞给他二十块，他多半不高兴——不是嫌少，是钱一付事就结了，而本来他手里还捏着一个你欠他的小人情。
经济学教科书的开头是另一个故事：很久以前人们以物易物，太麻烦，于是发明了钱。人类学家找了一百多年，没找到那个社会。这一页拆三件事：那个故事错在哪、「欠着」为什么比「结清」更能维持关系、最早的钱本来是拿来干什么的。
教科书的故事大致是这样：面包师想要肉，屠夫想要鞋，两边凑不上，于是大家约定用某种人人都要的东西当中介。亚当·斯密 1776 年写过类似的段落，之后两百多年被反复转述——可这个社会一直没被找到。剑桥的人类学家 Caroline Humphrey 1985 年写过一句很硬的话：从来没有哪个纯粹的以物易物经济被描述过，更谈不上从中长出货币。她那篇文章叫《以物易物与经济解体》——观察正相反：大规模的以物易物通常出现在货币体系崩掉之后，是解体的产物，不是起点。
机制上这不难理解。以物易物要凑一个「需求的双重巧合」：我手上有你要的，你手上恰好也有我要的，还得是同一时刻。但熟人之间根本不需要凑——肉你先拿去，这事记着就行。赊账不需要双重巧合，只需要你们还会再见面。
考古那边的次序也对得上。美索不达米亚的神庙和宫殿从约公元前三千年起就在泥板上记账：用银的重量单位舍客勒（shekel）当记账尺度，谷物和银之间有约定折算，借贷和利息都是常事。最早的铸币要到公元前七世纪末的吕底亚（今天土耳其西部）才出现，用金银合金打制。写下来的债比铸出来的钱早了两千多年，而且泥板上很多笔根本没有金属经手，只是记着谁欠谁多少。
证据强度要说清楚：Humphrey 那句是否定性论断，「没找到」不等于「不存在」。David Graeber 2011 年的《债：第一个五千年》把这条推得很广，书里不少具体历史断言被专家挑过错，读时要分开看；但「以物易物神话不成立、信用早于铸币」这条主线争议不大。
EN The textbook story runs: barter was clumsy, so money was invented to fix it. Adam Smith told a version of it in 1776 and it has been retold ever since. The society in question has never been found. Caroline Humphrey wrote in 1985 that no example of a barter economy, pure and simple, has ever been described, let alone the emergence of money from it — and her paper, titled Barter and Economic Disintegration, argues that large-scale barter appears after monetary systems collapse, as a product of breakdown rather than a starting point. The mechanism is simple: barter needs a double coincidence of wants, mine matching yours at the same moment, whereas people who will meet again simply keep a running tally. The archaeology agrees. Mesopotamian temples and palaces kept accounts on clay from roughly 3000 BCE, using the shekel — a weight of silver — as a unit of reckoning, with conventional grain-to-silver rates, loans and interest all routine. The earliest coins appear only in late seventh-century BCE Lydia, in an alloy of gold and silver. Recorded debt precedes struck coin by more than two thousand years, and many of those tablets involved no metal at all. On evidence strength: Humphrey's claim is a negative one, and while Graeber's Debt (2011) pushed the argument broadly and has been faulted by specialists on particular historical claims, this main line is not much contested in economic anthropology.
回到那个快递。你要给邻居二十块，他不高兴。本来他手里捏着一个你欠他的小人情，那意味着你们还会有下一次往来；钱一付，这条线就断了。这是 Graeber 那本书里最站得住的一条：债是一段尚未完成的交换，而正是「尚未完成」在维持关系。礼物与互惠那期讲过为什么白给比交易更有约束力，这里是同一条机制的另一头——能一次结清的东西，结清之后就不欠了，也就不必再见面。
田野上有个很有分量的例子。Evans-Pritchard 1940 年对南苏丹努尔人（Nuer）的民族志里写到，杀了人要赔牛，数目相当大，通常在四十头上下，由一位不属于任何一方的「豹皮首领」居中调解。关键不在于一条命折算成多少头牛。那批牛交出去之后常被当作新娘财用掉，两个原本要互相复仇的家族因此结成姻亲。赔偿要办的事不是了结，是把一段会互相杀下去的关系，换成一段会长期往来的关系。
那债在什么时候变成另一种东西？Graeber 的关键一步在这里：一份义务要变成债，得凑齐三样——一个可计量的数额、一个确定的债权人、一个可以清偿的时刻。三样齐了，它就可以被转让、被买卖、被交给第三方强制执行。你欠邻居的人情不能转让给别人去收，它跟「是他」绑在一起；写成数字的欠款可以，谁拿着条子谁就是债权人。量化不只是把义务记清楚，是把它从这段具体关系里拆下来。
这条最该挨的批评是容易滑成「量化即堕落」。反方同样有力：模糊的人情账恰恰可以被无限索取，因为没有任何数额能说「到此为止」；写下来的数目对欠债那一方反而是保护，它给出了终点。历史上的债务赦免——美索不达米亚的清账诏令、犹太传统里的禧年——也说明可计量才让「一笔勾销」成为可能。两种读法各对一半，这一页不选边。证据强度：民族志个案加历史文本解释，没有可做的实验。
EN Offer your neighbour cash for taking in a parcel and the offer tends to offend. Not because the sum is small: paying closes the matter, and what he was holding — a small unfinished obligation — was the thing that implied a next time. This is the most durable line in Graeber's book: debt is an exchange not yet completed, and the incompleteness is what keeps the relationship running. The same mechanism appeared from the other side in the piece on gifts and reciprocity. A weighty field case: in Evans-Pritchard's 1940 ethnography of the Nuer of South Sudan, a killing is settled by cattle — typically around forty head — mediated by a leopard-skin chief belonging to neither side. The point is not what a life is worth in cattle. Those cattle were commonly used as bridewealth, so two families bound to feud became affines instead. Compensation does not end the relation; it converts a killing relation into a long-running one. So when does debt become something else? Graeber's decisive step: an obligation becomes a debt once three things are in place — a quantified amount, a determinate creditor, and a moment at which it can be discharged. With those, it can be transferred, sold, and handed to a third party for enforcement. A favour owed to a neighbour cannot be assigned to a collector; a written sum can. Quantification does not merely record the obligation, it detaches it from the particular relationship. The strongest objection is that this slides into treating measurement itself as corruption. The other side is equally strong: a vague debt of gratitude has no figure that can say enough, whereas a written amount gives the debtor an endpoint — and historical debt cancellations, from Mesopotamian clean-slate edicts to the Jubilee, work only because amounts are countable. No side is taken here. Evidence strength: ethnographic cases plus textual history, with no available experiment.
一根手指多少钱？这问题听着冒犯，可几乎每套早期法典都认真回答过。约公元六百年肯特王国的《埃塞尔伯特法典》——现存最早的英语成文法——一条条列着：打掉一颗牙赔多少、砍掉一根手指赔多少、把人耳朵打聋赔多少。日耳曼法里还有个专门的词 wergild（人命金），字面意思就是「人的价钱」。
钱币史学者 Philip Grierson 1977 年提出：这类赔偿价目表很可能是「价值尺度」最早的来源之一。它要办的事不是买卖，是止住血仇——两家要谈得下去，得先有一张双方都认的表。人类学里把这类支付手段叫社会性货币：用于婚姻、赔偿、罚金，但不用来买今晚的饭。
更完整的一个田野在尼日利亚中部。Paul Bohannan 1950 年代在提夫人（Tiv）中间描述了一套分层的交换圈：最底下是日常生计，薯蓣、谷物、小牲口、农具；中间是威望物，铜条、牛、白布（历史上还包括被当作财产的人）；最上面是与婚姻相关的权利。圈内互换是家常便饭，跨圈往上换是成就，往下换是丢人。圈与圈之间没有汇率——正因为没有，一个人不可能靠卖光粮食换娶亲的资格。
殖民时期通用货币进来之后，墙塌了。Bohannan 1959 年那篇讲的就是后果：所有东西落到同一把尺子上，粮食换成钱、钱换成铜条、铜条走向婚姻那一层。年轻人多了一条路，长辈原本靠掌握上层交换物维持的位置被抽空了。通用货币最要紧的属性是「通用」，不是「便于交易」——它把原本不可通约的几类东西放进同一个数轴，而这一步是社会性的，不是技术性的。
批评也要给：这个三层模型后来被指太整齐，后续研究认为提夫人实际的跨层换算比模型灵活，殖民之前也已有货币化因素。当图式用很有力，当历史实况要打折扣。照今天也一样：你不会给帮你搬家的朋友转账，也不会给孩子的成绩标价——那是还立着的墙；把注意力、排队的时间、一次推荐换算成积分，是刚被拆掉的那几堵。哪些墙该立着这一页不回答，只说一件可观察的事：墙一拆，原本被挡住的流动会真的发生，靠那堵墙站着的人会真的失位。
EN What is a finger worth? The question sounds offensive, yet nearly every early law code answered it in detail. The Law of Æthelberht of Kent, around 600 CE and the earliest written law in English, tariffs injuries item by item — a knocked-out tooth, a severed finger, an ear struck deaf. Germanic law had a word for the whole idea: wergild, man-price. The numismatist Philip Grierson argued in 1977 that such compensation tariffs are a likely early source of the very idea of a scale of value: their business is not trade but stopping a feud, and two families cannot negotiate without a schedule both accept. Anthropologists call these social currencies — used for marriage, compensation and fines, never for buying dinner. A fuller case comes from central Nigeria, where Paul Bohannan described Tiv exchange in the 1950s as ranked spheres: subsistence at the bottom (yams, grain, small livestock, tools), prestige goods in the middle (brass rods, cattle, white cloth, and historically people held as property), and rights in marriage at the top. Exchange within a sphere was routine, moving up was an achievement and moving down a disgrace — and there was no rate between spheres, which is exactly why no one could sell enough grain to buy a wife. When general-purpose money arrived under colonial rule the walls fell, and Bohannan's 1959 paper is about the consequences: everything landed on one scale, the young gained a route upward, and elders whose standing rested on controlling the upper spheres were hollowed out. The decisive property of general-purpose money is that it is general, not that it is convenient: it places incommensurable classes of things on a single axis, and that step is social rather than technical. Criticism and evidence strength both belong here: Bohannan's three-sphere model has been faulted as too tidy, with later work showing Tiv conversions were more flexible and monetisation partly predated colonial rule. It works well as a schema and should be discounted as history. The same applies to reading it forward: you do not transfer money to a friend who helped you move, nor price a child's grades — walls still standing — while attention, queue time and a recommendation are being converted into points and payments right now. Which walls ought to stand is not answered here. The observable claim is narrower: remove a wall and flows that were blocked really do happen, and whoever was standing on that wall really does lose their footing.