Is mushroom rock art found where the mushrooms grow? A test

The question

Several published claims say ancient art shows psychoactive mushrooms: a Spanish rock shelter (Selva Pascuala), the Sahara plateau of Tassili, Arctic Siberia, Guatemala and Mexico among them. If those artists painted a mushroom they knew, that mushroom might be more common around the site than around other places.

That is the one claim tested here. The test was run by an AI model (Claude) on public data, at the direction of the author Paul L. G. Harris. It is a history and botany test only, and it does not ask whether the shapes are mushrooms at all.

The test

The rules were fixed before the counting, apart from one disclosed test query to check the database worked.

The sites. Eleven places where a published source proposes that an artwork or object shows a psychoactive mushroom, and where a coordinate could be found: Selva Pascuala (Spain), Tassili n'Ajjer (Algeria), the Pegtymel River (Russia), Mont Bégo (France), Stonehenge (UK), Åby (Sweden), Vinča (Serbia), Kaminaljuyu (Guatemala), Tzintzuntzan (Mexico), the Kerala megaliths (India) and the Plaincourault chapel (France). Each was tested against the mushroom its own claim names: Psilocybe and Panaeolus, fly agaric, or either where no species is named.

The data. GBIF, the world's largest open database of species records. For each place the study counted records in a box about 220 km from north to south. A raw count mostly maps where mushroom-hunters live (the Stonehenge box holds 542,259 records of gilled mushrooms), so the measure was the claimed mushroom's share of all gilled-mushroom records in the box. A box with fewer than 20 such records counted as "insufficient".

The comparison. Each claimed site was set against places nobody has claimed anything about: 18 ordinary rock-art and prehistoric sites in the same regions (Newgrange and Tikal among them) and 33 random land points, 150 to 600 km away. That is 62 places in all. The statistic was each claimed site's percentile rank within its own group (0.5 is the middle, higher favours the claim), averaged, then tested against 100,000 shuffles of the "claimed" label.

As a quality check, six stored counts were fetched again by a different route. All six matched.

What came out

Two of the most famous sites could not be tested. The Tassili box holds 3 gilled-mushroom records and the Pegtymel box holds none. A blank means nobody has collected, not that nothing grows. The Tassili paintings are also dated (by Giorgio Samorini) to a green Sahara 9,000 to 7,000 years ago, which no modern record could represent.

That left nine sites:

The average was 0.43, slightly below the middle, with p = 0.79: in about 79 of every 100 shuffles the claimed sites did at least as well. Only three of the nine rank above the middle. The pre-registered reading is "no detectable difference". The variants agree. Ignoring recording effort gives 0.55 (p = 0.31), rock-art controls alone 0.47 (p = 0.66), and random points alone 0.42 (p = 0.85).

Simple presence tells us nothing. The claimed mushroom group appears in the records at all 49 usable places (9 claimed, 15 rock-art, 25 random), so "it grows near the site" is true almost everywhere anyone has looked.

The test was blunt, too. To reach p below 0.05 the claimed sites needed an average of at least 0.656, roughly "usually in the top third". A weak tendency would have gone unseen.

One observation was not in the plan. The 2011 paper on Selva Pascuala named Psilocybe hispanica, but GBIF holds only three records of that species worldwide, and the nearest with a location is 284 km from the site. GBIF can neither confirm nor refute that it grows there.

What this does not show

• It does not show whether any painted shape is a mushroom. Identifications are contested (Letcher 2011 on Selva Pascuala, Panofsky on Plaincourault). Even a positive result would only have shown the mushroom was available, not that it was painted.
• It does not show that ancient artists never painted mushrooms. It shows that today's records give these eleven claims no support, and for two of them say nothing.
• Modern records, mostly from recent decades, stand in for ancient landscapes. The boxes are coarse, three sites use stand-in coordinates, and the list mixes rock art, portable objects and a medieval fresco.
• Nine usable sites is a small sample.
• Several claims reached the study through one survey by Samorini. Four (Kaplan 1975, Dikov 1971, Gimbutas 1974, Guzmán 1997) were not checked against the originals.
• Database counts were read through a summarising tool, and only six were re-checked.

Where it goes in the books

This is the "where the mushrooms grow" test in the Interlude of The Oldest Medicine, Book 4 of Hidden in Plain Sight. Chapter 11, "Mushrooms on the Wall", sets out the claims and the arguments about them.

You are invited to check it. Every count came from the public GBIF occurrence API on 1 October 2026, and all 124 query URLs are stored with the study, so anyone can run them again (the database grows, so expect small differences). If you find a slip, say so.