How DialectGuessr works

The question bank

There is one fixed bank of 26 single-choice questions: 12 about vocabulary and 14 about pronunciation. Everyone answers the same ones, which is what makes any two people comparable and what lets a region on the map be coloured by its dominant answer.

Answers are single-select on purpose. Letting people tick several synonyms makes the regional colouring much harder to read; asking for the closest everyday term, with an escape option where one is genuinely needed, is the cleaner signal.

Every question has its own page explaining what the answers mean and who gives them — browse all 26.

How your answers are scored

Each of the 31 regional accent profiles is itself a set of expected answers to the same questions, built from primary-source regional speech recordings. Your answers are compared against every profile, and the profiles are ranked by how well they fit.

Pronunciation answers are weighted more heavily than vocabulary answers, because they are much harder to change deliberately: people pick up new words when they move, but the vowel system learned in childhood tends to persist. Features that isolate one small region — whether “fur” rhymes with “fair”, for instance — carry more weight than features shared by half the world. Where a profile genuinely varies within its region, more than one answer counts as a match, with partial credit.

The hometown you pin contributes too: profiles anchored near where you grew up get a proximity bonus, so a close answer match in the right part of the world outranks the same match on another continent.

How the map is built

The town or city you choose places an exact pin. For the coloured regional layer, responses are aggregated into H3 hexagonal cells at resolution 4 — roughly the size of a metropolitan area. A cell is only coloured once at least five people there have answered, so the map shows patterns rather than individuals.

The map is worldwide and wraps horizontally, with English and Latin-script place labels drawn from OpenStreetMap via OpenFreeMap vector tiles. Town and city data comes from the GeoNames-derived all-the-cities dataset, distributed under CC BY 4.0.

Where the method comes from

The question design follows established dialect-survey practice: ask about ordinary speech used with friends and family, use concrete everyday concepts, and keep the questions identical across locations. Starting points included the Harvard Dialect Survey, the University of Leeds Great Big Dialect Hunt, and the Cambridge English Dialects App. The bank here is an original normalised worldwide set rather than a copy of any one survey — most existing surveys cover a single country, and this one has to work from Cork to Christchurch.

Accent profiles are anchored to primary-source recordings from the International Dialects of English Archive rather than to media performances. Where a well-known speaker is shown alongside a profile, it is as an illustration of a regional pattern, not as the definition of it.

What it costs and what it collects

The quiz is free, has no account, and shows no ads. It collects your answers and the childhood town you select — not your name, email address, current location, or age. Browser session identifiers are hashed before storage, and the browser that submitted a result keeps a private token that can delete it immediately. Full detail is in the privacy notice.

Comparing with other people

After finishing you can share a link that invites someone else to take the same quiz. When they do, you get a side-by-side comparison: an overall similarity score, the questions you answered identically, and the ones where you differ most. Comparison pages are personal results, so they are deliberately kept out of search indexes.