About this site

How we research

The rules every volume is held to, so you can judge a claim here by the standard we use.

Last changed 5 September 2026

Local sources first

The rule that decides whether any of this work is worth anything: sources from inside the tradition come before sources about it. In order of preference, that means writing published in the country concerned, its national and regional press, its food writers and cookbook authors, its university and agricultural research, its archives and its nutrition bodies; then sources in the local language, searched in that language and quoted with the original terms; then cooks and writers from inside the tradition wherever they live, including the diaspora; then academic ethnography and food history by scholars from the region. Writing about a cuisine from outside it is used to fill gaps, never to settle them.

What we will not build on

Western generalist food blogs, recipe aggregators, content farms written for search engines, listicles, travel site round ups and tourist board copy are excluded as sources for any factual claim. English language writing about most of these cuisines is unusually polluted: a large share of what ranks first is generated or copied, often with a recent date to look current. If a widely repeated claim exists only in Western secondary media, we do not make it. We say so instead.

Searching in the language of the place

Research is run in the language of the place, not in English with a local word dropped in. For the African volumes that means French for West and Central Africa, Portuguese for Cape Verde, Guinea-Bissau, Sao Tome, Angola and Mozambique, Amharic and Tigrinya for the Horn, Arabic and Somali for the Swahili coast's northern neighbours, Arabic and French for North Africa, Swahili for East Africa, French and Malagasy for Madagascar, and Afrikaans and isiZulu for Southern African dish names. The rule for every region since is the same: Chinese for Sichuan, Hindi and Urdu for Lucknow, Spanish for the Andes. The single biggest difference between a researched volume and a plausible one is whether anyone read something written by someone who lives there.

Contested claims

Where a region covers several traditions with competing claims to a dish, each claim is attributed to who makes it and from where they speak, and we do not adjudicate what is genuinely unresolved. One community's writers are not allowed to speak for another's food. Where a claim is disputed, the sentence that makes it says so, not a footnote.

Claims we checked and would not make

Every volume ends with a section under that heading. It lists the claims that are widely repeated about the cuisine and did not survive checking: origin stories with no source older than a travel article, ages of dishes that turn out to be twentieth century inventions, national dishes that belong to a neighbour, and anything for which the only sources are Western. It is the most important section in the volume, because it is the record of what we chose not to tell you and why. If a fact you expected is missing, look there first.

Recipes

The rule for recipes is real quantities and real timings from credible sources inside the tradition. Many dishes were first written as the tradition's method, with the ingredients named but not weighed, and each dish page says which it carries: a recipe to scale, or a method to read. We are adding quantities volume by volume from in-tradition sources, lead dishes first. A time shown on a dish comes from a source that states one, from a tested recipe or a cook on camera, or, where the page says so, from comparable dishes in this atlas. Each dish is given first as the tradition builds it, then with the changes a modern chef would make, and each change names the kitchen or the reasoning it comes from rather than importing a technique because it is fashionable. The reader this is written for owns a high power blender, a sous vide bath, a dehydrator and a spice grinder and wants restaurant level flavour at home; the recipes assume that and say when a piece of kit is essential rather than convenient.

Today 1,541 of 1,783 dishes carry full quantities; the rest are outlines, or cooked by cue, and say so.

Provenance

Every research brief behind a volume carries a source provenance section: roughly how many of its sources are local or in tradition and how many are Western and secondary, which claims rest on weak sourcing, and which local sources could not be reached. Where the local record was thin, the chapter is plainer and more hedged, and says which country or community it is thin for. Honest gaps are preferred to false completeness. Making that provenance visible on each volume's page is planned; until it is, the not asserted section is the reader's best guide to how firm a chapter's ground is.

Corrections

If you know this food or this field better than we have written it, the correction link on every volume and dish opens a short form that arrives with the page already named. Corrections from inside a tradition are the most valuable thing this site receives. Each one is checked against sources in the same order of preference described above, and where it stands the chapter is changed and the change is recorded. Paid review of each volume by readers from inside its tradition is planned, and a volume that has had it says so on its page.