Method

What counts as a claim

Every fact, date, range, route and number on the site is recorded as a claim in a plain file. A claim has a sentence, a confidence level, and one or more sources. For each source we record where in it the support is found and what the source says, in our own words.

The site is built from those files, and the build is the gatekeeper. It stops if a claim has no source, if a paragraph of a story carries no citation, if a map site or route has no citation, if an image has no provenance record, or if a claim marked contested cites only one side.

The site currently holds 663 claims resting on 439 sources.

Confidence

Established

Broad agreement among researchers. 165 claims.

Supported

Good evidence, with some debate, or resting on a single strong study. 425 claims.

Contested

Researchers actively disagree. Both sides are cited and the disagreement is described. 73 claims.

The color of each citation marker in the text shows the level, and opening the marker gives the reasoning.

Sources

We prefer peer-reviewed research and datasets, then books and review articles. Popular articles are not cited for facts. Where a source reports someone else's finding, the claim says whose it is.

No specialist has yet reviewed the content. If you find an error, the sources are there so that it can be shown and fixed.

Dates

Content files record dates the way the source reports them: radiocarbon ages, calibrated years before present (counted back from 1950), or calendar years. The site converts them to plain forms such as "about 6,250 years ago". Most early dates are ranges, and they are shown as ranges.

The timeline compresses deep time so that the last five centuries get room. On it, a soft bar is an uncertain date or a span, and a sharp diamond is a date known to the year.

The map

Soft washes

Wild ranges are painted from GBIF occurrence records: stronger where there are more records, fading where there are few. Areas of domestication are outlined loosely because the studies name a region and do not draw a boundary.

Sites

Each marker is a dated find with its own citation. Where a report gives no coordinates, the marker sits at the nearest town and says so.

Routes

Solid lines are well supported, dashed lines are proposed, dotted lines with a question mark are contested. All routes are schematic: they join dated finds and do not trace a known path.

Production

Hatching and column height show metric tons grown per country, from FAOSTAT. Column height is proportional to metric tons.

The base map is Natural Earth, in the public domain. The engraved styling is drawn by the site and is not a reproduction of any historical globe.

Images

We use only images whose license allows free public use: public domain, CC0, CC BY, and CC BY-SA case by case. The license is checked on the holding institution's own page, which is linked from every credit. Each image carries a badge saying whether it is historical, adapted or a schematic model. See Image credits.

The 3D models

The four ears in the maize story are drawn in code at one scale from cited measurements. Their shapes and colors are schematic, and anything reconstructed, such as kernels on a cob that survives only bare, is marked.

The eight ears in the wheat story are built the same way, from cited ear lengths, spikelet counts and awn lengths. Ear widths are mostly our estimates, and one wild grass is modeled from a table that gave no units. Each model says which of its numbers are measured.

The nine panicles in the rice story are built from cited panicle lengths, counts of branches and spikelets, grain sizes and awn lengths. Most of the counts are averages that we worked out from published data files, and they come from plants grown in trial plots, not in farmers' fields or in the wild. How wide each panicle spreads and how far it nods are drawn from descriptions. Each model says which of its numbers are measured.

Where a model has not been made yet, a plain gray block stands in for it. Every form of Brassica oleracea is a stand-in at present. A stand-in is labeled as one wherever it appears and is not to scale.

Family trees

Some crops were bred in several directions from one plant. Their forms are drawn as a tree, and each line of descent is a claim like any other: it has a citation, and it is drawn solid, dashed or dotted by how well it is supported. The date beside a form is its first secure written record, which can be centuries later than the form itself. Where studies disagree about a line, the page says so and cites each side.

Indigenous knowledge and names

These crops were made by Indigenous farmers, and the site says so in the story itself. Where we give a name in an Indigenous language we cite a dictionary or linguistic source for it, and say when that source is a teaching resource and not a scholarly one.

Known gaps

• The wild-range layers were retrieved from the GBIF search service. Before publication they should be re-issued as a registered GBIF download so the data has its own DOI.
• Several site coordinates are approximate, as marked on each site.
• No free image of George Beadle was found, so the chapter on the origin debate has no portrait.
• The Chinese record of 1555 is cited through a secondary report that gives the province inconsistently.
• Wheat's measurements lean on one monograph of 1921.
• Several early wheat sites have no published coordinates. Their markers sit on a nearby village or basin and say so. Three sites named in the text (Tasbas, Zhaojiazhuang and Houshiliang) have no marker at all.
• We found no free image of a wild emmer plant, a Neolithic sickle, a Roman harvest or the dwarf wheats of the Green Revolution, so those chapters have no plate or use a later one.
• The licenses of the Metropolitan Museum images were confirmed through the museum's open-access data, not by opening each object page.
• Rice has no agreed family tree. The lines we draw between indica, japonica, aus and aromatic rice follow recent genome studies that disagree with one another, and each line says how firmly it is held.
• The pin on the Ganges plain marks one model of how indica rice arose, not an agreed place of domestication. The pin on the Niger marks a proposal of 1962 that some genome studies support and others do not.
• Many early rice sites are placed from a gazetteer, on the nearest town or on a museum, because the papers cited give no coordinates. Mahagara in India, named in the text, has no marker at all.
• The wild rice records on the map cannot tell a wild plant from one that has crossed with the crop, which is common in wild rice. GBIF also files the annual wild rice, Oryza nivara, under Oryza rufipogon, so the two are drawn as one layer.
• The early Portuguese accounts of West African rice are cited as quoted in one modern paper, and the story of Champa rice in 1012 through one modern summary of an older study.
• Words for rice in Greek, Dyula and several other languages are cited from Wiktionary, which is not a scholarly dictionary. Each such claim says so.
• We found no free image of rice in early modern Europe, of IR8 or the first semi-dwarf rices, or of the crop today, so those chapters have no plate or use an older one. The picture beside "Rice today" is a woodcut of 1696.
• The licenses of the three Naturalis herbarium sheets were confirmed through the museum's open-data records.
• For cabbage and its kin, the date given for the arrival of Chinese kale in China rests on a Chinese study that is cited through two papers that summarize it.
• Several first records of cabbage forms are taken from one compiler of 1919 whose own sources are not cited directly. Each such claim says so.
• FAO's figure for cabbages very probably includes Chinese cabbage, a different species.
• No free image was found of kale as a whole plant, of Chinese kale, or of any ancient or medieval picture of the crop whose license we could confirm with its holder.
• The wild cabbage records on the map cannot tell a truly wild plant from an escaped crop.

Licenses

The text, maps and code of Wild Type are © 2026 Jack Carroll, all rights reserved. Images carry their own licenses, shown beside each one on the image credits page. The base map is Natural Earth, in the public domain.