You arrive somewhere new and someone asks where you are from. The answer shifts with whoever is asking: to someone from the same city you name a district, to someone from another province you name the province, to a foreigner you name the country. None of them is false and none is the whole thing. And you barely think about it — your mouth picks the level on its own.
This page takes apart three things: what holds an ethnic boundary in place (a hint: not how much the two sides differ culturally); when the name of an ethnic group actually becomes a set of people who can act together; and what happens when the same population is cut by two different classification systems. The page only lays out the strongest version of each account and labels how hard the evidence is; the judgement is yours.
For a long time the field assumed an ethnic group was a bag: language, dress, custom and marriage rules inside it, with the edge of the bag wherever those things start to differ. In 1969 the Norwegian anthropologist Fredrik Barth edited a volume that reversed the order.
Gunnar Haaland's chapter in it reports fieldwork in Darfur, Sudan, in the 1960s. Two populations lived side by side: the Fur, settled millet farmers, and the Baggara, who moved with cattle. A Fur household that saved up a few head of cattle had nowhere to put them inside the village — no more land, no room for a large herd — so the paying move was to leave and follow the herding routes. A few years on, that household spoke Arabic day to day, its children grew up among the cattle, the neighbours counted them as Baggara, and so did they. People crossed; the boundary did not move an inch.
Barth's own chapter is about why people cross. Being a Pathan has to be performed continuously: you hold land, you host in your own men's house (hujra), you can speak in the council of elders (jirga). A man who loses his land is a failure by that standard and cannot recover — the organisation is egalitarian and has no place for a landless member. On the Baluch side to the south there is hierarchy, and outsiders can be absorbed through contracts of clientage: a landless man there is a chief's dependant rather than a failed Pathan. So movement runs one way only. What decides the direction is not cultural similarity but which side has a position you can stand in.
Barth's conclusion, quoted for half a century since, runs roughly: what to study is the boundary that defines the group, not the cultural stuff it encloses. A boundary holds only because two acts keep being performed — I assign you to a category, and you assign yourself to one (the terms are ascription and self-ascription). While those acts continue, the two sides can intermarry, trade and share a language and the boundary still stands. Cultural difference is less the cause of the boundary than its product.
The heaviest criticism is that this reads too much like accounting: if switching can be weighed up, why do people die for ethnicity? The replies do not deny that boundaries are maintained; they argue over how costly the maintenance is — in some places it is a door, in others a wall. Evidence: ethnographic cases from the 1960s, not statistically representative; and boundaries in that same region hardened greatly over the following decades, which itself suggests hardness is not intrinsic to a boundary.
When a report says relations between group A and group B are tense, the subject of the sentence is two groups, as if each were something that can deliberate, decide and act together. Rogers Brubaker's 2002 article "Ethnicity without Groups" aimed at exactly that step: it is given away for free and has to be earned.
Two things have to be pulled apart. An ethnic category is a name, a column, a box on a form. An ethnic group is a set of people who can genuinely act in concert. Categories are everywhere; groups are rare and take a great deal of work to produce. So groupness is a variable: it happens, in bursts, rather than sitting there waiting to be activated.
He and several colleagues took the claim into the field: around 2000 they spent several years interviewing and observing in a Romanian- and Hungarian-speaking town in Romania (Cluj/Kolozsvár). Public political language there was intensely ethnicised at the time, down to what colour the lamp-posts were painted. But in everyday life — shopping, working, complaining about prices — most of the time ethnicity simply went unmentioned, and people used other categories (colleague, neighbour, who has money). Which language got spoken mostly depended on what the other person understood, not on any stance.
What pushes ethnicity to the front is usually an organised party: political parties, associations, newspapers, how a school sorts its classes. Hence a warning that carries a lot of weight: that an organisation speaks for an ethnic group is not evidence that the group is a group — the organisation is real, and its representativeness has to be shown separately. This runs with the page on deviance and labelling: who picks a category up matters more for the consequences than whether it is accurate.
The criticism is strong too. First, this is easily misread as "ethnicity does not matter", which Brubaker repeatedly denies: a category need not be a group to determine how a person is treated, and the next section is entirely about that. Second, where violence recurs, calling groupness a mere variable sounds weightless to the people living there. What holds is narrower: high groupness is the thing that needs explaining and must not be treated as the default state. Evidence: years of fieldwork in one town, not generalisable; and the claim works less as a falsifiable proposition than as a discipline of analysis — it forces "the group acted" to be unpacked into who organised and who responded.
In 1976 Brazil's national statistics agency ran its household survey with the colour question left open — no options, describe yourself — and got back 136 distinct terms. The official form has five boxes: white, black, brown (parda), yellow, indigenous.
The United States side followed descent rather than appearance. Late-nineteenth-century census forms still had boxes for "quadroon" and "octoroon"; by around 1930 this had collapsed into a single rule, under which any non-white ancestry placed a person on the non-white side (commonly called the one-drop rule). Only in 2000 was a person first allowed to mark more than one box.
The boxes also move on their own: in Brazil's 2022 national census the share self-reporting as parda (45.3%) exceeded the share reporting branca (about 43.5%) for the first time, with no demographic change on anything like that scale in between. What changed is how people fill the form, not the people. The sociologist Edward Telles adds another finding: self-classification and the interviewer's classification often disagree, and the direction of the disagreement correlates with the person's education and income.
From here it is tempting to conclude that the categories are fake and should not be taken seriously. The evidence does not support that step. Quillian and colleagues' 2017 meta-analysis pooled United States résumé field experiments from 1989 to 2015: identical résumés carrying names that cue different ethnicities, and white names drew on average about 36% more callbacks than African American names, essentially unchanged across twenty-five years. The same team's 2019 cross-national study covers nine countries in Europe and North America, 97 experiments and over 200,000 applications: in the highest-discrimination countries white natives received nearly twice the callbacks of minorities, in the lowest only about 25% more. The categories are cut and the consequences are real; and a gap that varies several-fold across institutional settings is being produced by something rather than fixed.
Produced by what? Two non-exclusive accounts are still contested: that the person screening résumés holds a direct negative evaluation of certain names, or that they are using the category as an informational shortcut (this is called statistical discrimination). Callback data alone cannot separate them, which is where this page most needs to say there is no verdict yet; and what it measures is callbacks, not final hiring.
Classification systems are themselves policy instruments, so both strongest cases go on the table. France essentially does not collect ethnic statistics (a 1978 data law forbids collecting such data, and in 2007 the Constitutional Council struck down a provision that would have opened a path for measuring discrimination), on grounds that combine the republican principle of counting citizens without regard to origin with the memory of registration being used for persecution during the Second World War: once a state starts registering people by ethnicity it is producing the very classification it wants to dissolve. Britain, the United States and others run the opposite way on an equally hard argument: what is not collected cannot be measured, and gaps that stay invisible are nobody's responsibility. There is no free option here, and this page does not adjudicate.
These three sections yield no "therefore we should" of any kind — that is a value judgement, outside this discipline's measuring range. What can be said is that any proposal has to name the mechanism it targets: the two acts that maintain a boundary, the organisations that turn a category into a group, or differential treatment at the point of assignment. Aim wrong and the indicator moves while the mechanism does not.