Comparative anthropology · 1,291 societies
One partner, your own household after the wedding, parents and their children, no property crossing between the families at the wedding, cousins called something other than brothers, descent counted through the mother as well as the father. That is what the combination looks like in George Peter Murdock's Ethnographic Atlas, which coded 1,291 societies between 1962 and 1971, with reference years running from the Babylonians to 1965. That combination turns up fourteen times, among them the Spaniards, the Neapolitans, the Brazilians, the row marked "Romanians" at 1910, and four Pueblo villages in northern New Mexico. Nine hundred and nine of the 1,291 records answered all six questions, and the commonest combination among them covers only 22. There is no ordinary way to be a family.
The archive
Between 1962 and 1971 the American anthropologist George Peter Murdock published, in instalments, in the journal Ethnology, a table. Each row was a human society: a people, a village, a group described by somebody at some moment. Each column was a question: who inherits the land, what shape is the house, at what age are boys circumcised, how many levels of authority sit above the community. The entries were digits. One digit for bride-price, another for dowry, another for nothing at all.
The table reached 1,291 rows and 94 columns and is called the Ethnographic Atlas. It remains the most-used comparative database in the social sciences: for half a century almost every claim beginning "in most human societies" has been tested against it.
The one thing almost nobody does with it is the simplest thing possible: take several columns at once. Single-column statistics are everywhere: 51.7% of societies charge a price for the bride, 46.3% count descent through the father's line, and 4.9% set up house apart from both families. But a society is a combination of columns, and combinations behave very differently.
The experiment
Six of the 94 columns together describe a family: the number of spouses, where the couple moves, who makes up the household, what property crosses at marriage, what cousins are called, and which line you descend from. They are the six things you know about your own family without thinking.
Each has one answer that occurs more often than the rest, though not necessarily a majority one: the commonest descent pattern, the one that counts kin through the father, covers only 46.3% of societies. Put the six leading answers on a single index card and you get a society nobody ever imagined: occasional polygyny, a move to the husband's kin, married brothers under one roof, bride-price, cousins called brothers, the father's line. The average society is the name for this collage of six columns, a working label and not a society anyone ever saw. "Average" here means the archive's commonest answer, not an arithmetic mean.
A combination of six answers can be rare even when each answer is common, the same way flipping six heads in a row is unlikely even though each flip has even odds. The count of exact matches does not, on its own, tell you how alike societies are: for that you have to see how all the societies spread across all the possible combinations.
Search the archive for its equals, then. Five remain: the Cantonese, the Min Chinese, the Kurukh of central India, the Lepcha of Sikkim, and the islanders of Tanimbar. Not one in Africa, although Africa contributes more societies than any other region to each of these leading answers taken separately.
Now do the same with your own life. One spouse, a separate house after the wedding, parents and children, nothing to pay, cousins who are called cousins, descent from both sides. Fourteen remain. Yours is nearly three times more common than the average.
Each row adds a criterion and shows how many of the 1,291 societies survive it. The order does not change the final count. The gap comes from the second question: residence with the husband's kin is coded for 50.4% of societies, a separate household for 4.9%.
The instrument
Sixteen questions, all taken straight from Murdock's columns, all phrased the way you would answer them about your own life. Each answer shows what share of the coded societies gives it. Each choice tips out of the bowl every society that answers otherwise.
It opens on the six family questions, answered the way somebody in Europe would answer them today: fourteen societies remain. Press "Your life, 2026" to add the other ten, about inheritance, layers of authority, gods, social class, the shape of the house. One society is left out of 1,291: Brazilians, as described in 1940. Four more neither match nor drop out, because they lack a code somewhere; one of the four is Romania, with nine of the sixteen questions coded. The same sixteen questions answered with the commonest code everywhere leave nobody standing.
The orange dot is the row marked "Romanians" at 1910. In grid view the societies are arranged by Murdock's six regions, alphabetically within each. On the map each dot sits at the coordinates of the community described. The gold dots are the societies the archive will not place: they contradict none of the answers they have, but on at least one question the atlas records missing data. Their list opens under the sieve.
The main result
Of the 1,291 societies, 909 have all six columns filled in. Among them are 564 different combinations. Four hundred and thirty-eight of those are unique: a single society in the whole archive has exactly that combination of rules about marriage, household and descent.
Four hundred and thirty-eight out of 909 means nearly half the societies we can check have no twin. Unique here means no other society carries exactly the same six codes; some do share five of the six codes with others. The largest cluster has 22 members out of 1,291: sub-Saharan African societies, such as the Banyaruanda, Xhosa, Sotho and Kipsigis, plus, oddly, the Fore of the New Guinea highlands, all with general polygyny, residence with the husband's kin, independent polygynous households, bride-price, cousins called brothers and the father's line.
The next two clusters have fourteen each. One is a Muslim belt running from the western Sahara to Kashmir: Algerians, Egyptians, Syrians, Turks, Sindhi, Hazara. The other is ours.
These figures have two weaknesses known for a long time. The first is an objection from 1889: when Edward Tylor read a paper comparing hundreds of societies to the Anthropological Institute in London, Francis Galton stood up and asked whether those societies were genuinely separate cases or merely copies of the same few. The question kept his name and has never been fully answered, and it applies directly here: the four Pueblo villages of the Rio Grande valley are local variations on one cultural network, and the Cantonese and the Min Chinese share an enormous common root. The cluster of 22 in sub-Saharan Africa is, in good part, a single cultural bloc counted twenty-two times: table rows multiply faster than actual peoples do.
The second is mathematical and cuts the other way. Six variables with four to seven states each open thousands of possible combinations, and in a space that wide any particular combination becomes rare almost by definition. It is a general warning against manufacturing the average man by adding up medians taken separately. In 1952 Lieutenant Gilbert S. Daniels measured 4,063 American pilots on ten bodily dimensions and looked for how many fell in the middle range on all ten at once. He found none. Cockpits designed for the average pilot fitted nobody, and the conclusion was that the seat had to adjust. The average family is in the same position.
This is the real shape of human diversity: a field with three civilisational clusters of 22, 14 and 14, a fourth West African peak of 13, and a long queue of lonely exceptions. A single custom really can be widespread: bride-price appears in 51.7% of societies. The whole set of family rules, though, meaning who moves where, who pays, who inherits and what the relatives are called, has no dominant version anywhere: the commonest one appears in only 22 of the 909 societies we can check, or 2.4% of cases.
How the societies divide up by the number of others carrying the same family record. Almost half the archive, 438 out of 909, sits in the first column: societies with no twin at all. The last column is the largest cluster in the world, 22 societies with the same family record.
The club
The list of societies with all six traits of the modern family contains exactly what you expect and exactly what you do not. Spaniards, Neapolitans, Brazilians, New Englanders of 1920, Romanians of 1910. And beside them Taos, Picuris, San Ildefonso and Santa Clara: four Pueblo villages in northern New Mexico, coded between 1890 and 1920.
What links the row marked "Romanians" to a Tanoan pueblo on the Rio Grande is religion, not geography. Thirteen of the fourteen had, by the date of description, several centuries of Christian church behind them: the Spanish missions in New Mexico, Guam and Sonora, Orthodoxy in the Carpathians, Catholicism in Italy and Iberia, Protestantism in New England and on Tristan da Cunha. The single exception is Sumbawa, a Muslim island.
There is an explanation ready to hand. Jack Goody in 1983 and, more recently, Joseph Henrich have argued that the Church's prohibitions on marriage between kin, on adoption and on concubinage dissolved clans systematically and left behind exactly this combination: a lone couple, monogamous, bilateral.
The rest of the archive sits badly with that explanation. Of the twenty-two European societies coded in the atlas, every one with centuries of church behind it, Orthodox, Catholic or Protestant, only three carry all six traits, as the table below shows. If the Church alone had produced the combination, Catholic Hungary and Orthodox Serbia should sit beside Spain, but they do not.
The fourteen were selected precisely because they share the same family record, so the resemblance between them cannot show what produced it. It stays a striking coincidence and nothing more.
The regions are Murdock's six, which sort societies by cultural provenance rather than geography: Brazilians, New Englanders and the islanders of Tristan da Cunha are all filed under Europe and the Mediterranean.
Romania in the archive
The row marked "Romanians" in the Ethnographic Atlas carries the year 1910 and a single bibliographic reference: Beiträge zum lebenden Ehe- und Familienrecht der Rumänen, by N. Cotlarciuc, in the Zeitschrift für österreichische Volkskunde, volume 19, 1913, pages 16–32 and 93–104.
Twenty-nine pages, in German, in a Viennese folklore journal. The author is Nicolae Cotlarciuc, a Bukovinian theologian teaching at Chernivtsi, who would become Metropolitan of Bukovina in 1924. From his article, Murdock assembled Romania's record. The atlas nevertheless places the society at 45°N, 26°E, in Wallachia, some five hundred kilometres from Chernivtsi.
Of the 94 columns, the Romanian row has 31 filled in. The median for the archive is 74. On coverage, Romania ranks 1,236th out of 1,291. From this archive we do not know what shape the roof of a Romanian house had, who inherited the land, what was expected of girls before marriage, or how many levels of authority stood above the village. These gaps still have a row each in the table, marked missing data instead of a number: somebody looked, and recorded that it was not known. The thirty-one filled-in columns cover mostly kinship and marriage, and all of them come from one man.
There is a second row: "Moldovans", year 1900, coded in 2005 by a Russian team working from ten sources. On four of the six family columns the two rows agree; on two they do not. The Moldovans are coded virilocal, meaning the bride moves to the husband's kin, and with dowry rather than "nothing substantial". The two rows describe different populations and different decades: one rests on a single 1913 study, the other on ten Russian sources about Bessarabia at 1900, on the far side of an imperial border. But they are neighbours speaking the same language, and the archive gives them opposite rules about where a young couple makes its home and who hands over property at the wedding.
Take all nine questions to which the Romanian row has an answer and run them through the sieve: five societies remain, Brazilians, the Camba of Bolivia, the Chamorro of Guam, Santa Clara Pueblo, and Romania. That is the whole of our resemblance in the archive of humanity, after nine questions asked by one man a hundred and thirteen years ago.
Was Romania modern earlier than its neighbours? The archive cannot answer that: the question "where did newlyweds live in the Romanian village of 1900" has two contradictory answers here, each from a team that read something different.
Romanian ethnography, from Henri H. Stahl to the monographs of the Bucharest Sociological School, describes both nuclear households and periods of living with the husband's parents, with wide regional variation. Murdock's coding picks one for the entire country. That is how the whole archive works.
The neighbourhood
If the Romanian row is right, then in the archive Romania sits closer to New England than to Serbia. The Serbs are coded with small extended families, meaning the zadruga: married brothers and their wives under a single roof, moved into the husband's parents' house. So are the Bulgarians. The Magyars have large extended families. Czechs and Ukrainians have stem families: parents with one married child. The Gheg Albanians are the only Europeans coded with ordinary polygyny.
Green marks the answers identical to a European's today. Three societies have all six: Romania, the Neapolitans and the Spaniards. Russians, Hutsuls and Sami have five. Serbs and Gheg Albanians have one each. Ossetians, none.
The year in the second column is the year the description refers to, not the year of publication. The decades separating the rows are real and make the comparison more fragile than it looks.
The atlas
Search by name or filter by region. The "coded" column shows how many of the 94 columns a society has filled in; the "sources" column shows how many works the entire row rests on. Three hundred and six societies in the archive have a single source. The median is two.
| Society | Year | Region | Coded | Sources | Household | Residence | Spouses | Marriage payment | Descent |
|---|
What the archive cannot see
The Ethnographic Atlas is an archive of descriptions: every row is somebody who wrote something, at some point, about somebody else. The reference years run from 2000 BC (the Babylonians) to 1965 (the Yanomami), nearly four millennia, laid out in the same table as though they were one present. Almost every describer is a man, European or American; many are missionaries or colonial administrators; and the questions they were answering were Murdock's questions, not the questions of the people described.
The coding inevitably simplifies: it picks one answer for an entire people and freezes it at one year. The "ethnographic present", the tense most of these descriptions are written in, implies a society standing still, and standing still for ever. None of them stood still.
And the coverage is thin everywhere, not only in Romania. The median is two sources per society. Three hundred and six societies, almost a quarter of the archive, enter comparative anthropology through a single work each. Any percentage of the form "in such-and-such per cent of human societies" rests on an archive whose median society has two works behind it, some of them written a hundred and twenty years ago.
With those limits, the archive stays what it is: the best inventory we have of the ways people have organised family life, with a precision that never exceeds the precision of whoever wrote the book. The main result does not depend on any single row: remove Romania, or any other society, and it does not move, because 438 of the 909 complete records are not repeated anywhere else. Every family is, statistically speaking, an exception to a norm that does not exist.
If this archive had a row for the Europe of 2026, the subsistence column would stay empty. Murdock provided digits for gathering, hunting, fishing, animal husbandry and agriculture. For somebody who works at a computer there is no code.