What Is an Ethnic Group? The Six-Part Test That Actually Works

“Ethnic group” gets used loosely enough that it has come to mean roughly “any group of people who are somehow different.” That’s not useful, and it isn’t how the term works in either sociology or law.

An ethnic group is a named human population that shares a myth of common descent, historical memories, elements of culture, a link to a homeland, and a sense of solidarity. The value of that definition is that it functions as a test. Some communities pass it, some fail, and the ones that fail in interesting ways teach you more about the concept than the ones that pass.

Smith’s six attributes

The most widely used framework comes from Anthony D. Smith, who identified six attributes of what he called an ethnie:

• A collective proper name — the group calls itself something, and others use that name.
• A myth of common ancestry — a shared story of origin. “Myth” is not an insult here; it means a founding narrative, whether or not it is historically verifiable.
• Shared historical memories — remembered events, heroes, golden ages, catastrophes.
• Elements of a common culture — typically language, religion, custom, dress, or law.
• A link with a homeland — a territory the group considers its own, even if it no longer lives there.
• A sense of solidarity — a felt sense of mutual obligation among at least significant sections of the group.

Note what is missing: genetics, physical appearance, and political independence. None of the six requires a shared biology, and none requires a state.

The legal version: Mandla v. Dowell Lee

Courts have needed a workable definition too, and the sharpest one comes from the UK.

In 1978, a Sikh boy was refused a place at a Birmingham school because his father would not cut his son’s hair or remove his turban. The case turned on a narrow question: does the Race Relations Act, which protects “racial groups” including ethnic ones, cover Sikhs at all? The Court of Appeal said no — Sikhs, it reasoned, were a religious community, not an ethnic one.

The House of Lords reversed it in Mandla v. Dowell Lee (1983). Lord Fraser set out a test that has governed UK discrimination law ever since. Two conditions are essential:

• a long shared history, of which the group is conscious as distinguishing it from other groups
• a cultural tradition of its own, including family and social customs and manners

And five further characteristics are relevant but not required:

• common geographical origin, or descent from a small number of common ancestors
• a common language, not necessarily peculiar to the group
• a common literature peculiar to the group
• a common religion different from that of neighbouring groups
• being a minority, or an oppressed or dominant group, within a larger community

Sikhs satisfied both essentials and most of the rest. They were an ethnic group in law.

Two things about this test are worth noticing. It requires consciousness of a shared history — the group has to know it is a group. And it treats religion as one possible ingredient rather than a disqualifier, which is precisely where the Court of Appeal had gone wrong.

Running the test on hard cases

The framework earns its keep on borderline communities.

Roma pass on all six. A collective name, an origin narrative tracing to northern India, a vivid history of persecution, distinct language (Romani) and custom, a remembered homeland, and strong internal solidarity. No state, no territory — still unambiguously an ethnic group.

Cajuns pass. Descent from Acadian exiles expelled from Nova Scotia in 1755, a remembered catastrophe (Le Grand Dérangement), a distinct French dialect, a homeland in south Louisiana, and a name they claim themselves.

Appalachians are genuinely contested. There is a region, a shared history of extraction and poverty, and recognisable cultural forms. What’s thin is the descent myth — Appalachian identity is fundamentally regional rather than genealogical. Most scholars treat it as a regional culture, and the disagreement is itself informative.

Deaf communities are the sharpest test. They have a name, a rich culture, a language in ASL or BSL, deep solidarity, a literature, and a history of oppression — that’s five of six, and both of Lord Fraser’s essentials arguably met. What they lack is descent: over 90% of deaf children are born to hearing parents, so the community reconstitutes each generation rather than inheriting. This is exactly why “Deaf culture” is the settled phrase and “Deaf ethnicity” is not, and it isolates common descent as the one criterion doing the heaviest lifting.

Ethnic groups are nested, not parallel

A second common error is imagining ethnic groups as a flat list of mutually exclusive boxes, the way a census form implies.

They are hierarchical. Igbo sits inside a broader West African grouping and contains dozens of sub-identities below it. A Bavarian is German is European. A Cantonese speaker in Guangzhou is Han is Chinese. Which level a person foregrounds depends entirely on who is asking and why — the same person answers “Bavarian” at home and “German” abroad.

Neither answer is more correct. Ethnic identity is scalar, and the level that becomes salient is the level at which a boundary is currently being drawn.

Why the count keeps changing

If you have ever wondered why sources disagree wildly on how many ethnic groups exist — several hundred? several thousand? — this is why. The answer depends entirely on where you cut the hierarchy.

China recognises 56 groups officially. Its 1953 census collected more than 400 self-reported names, which state ethnologists then consolidated. Nigeria is usually described as having around 250. Neither number was discovered; both were decided, and both could have been drawn differently.

Every official list of ethnic groups is a policy artifact, produced by a state for administrative purposes. That is worth remembering whenever you see a fixed set of categories presented as natural fact — including the category list returned by any classification tool.

What this means for identifying your own

Run the six criteria on yourself and the results are often clarifying. Most people find they belong to several ethnic groups at different scales, that some criteria are strong for them and others nearly absent, and that the boundaries are fuzzier than a checkbox implies.

That is the normal condition, not a failure to have a real identity. If you want a starting point rather than an answer, an ethnicity guesser will tell you which populations a model associates with your facial features. Treat that as one data point about phenotype — the six criteria above are what actually determine group membership, and only you and your community can assess those.