The Indus Valley sits at the 100th percentile of a global “desert-river circumscription” index — out of every 100×100 km patch of land on Earth, none scores higher. Egypt 98.9th, Mesopotamia 99.5th. New analysis.

Robert Carneiro's 1970 theory said early civilizations formed where people couldn't disperse; fertile land hemmed in by deserts and mountains, with population pressure forcing political consolidation. The theory has problems (highland New Guinea was just as circumscribed and never produced a state) but the geography it points to is real. A new analysis tries to measure it directly.

The "Geological Circumscription Index" has three ingredients: tectonically active mountains nearby, an arid environment, and proximity to a major river. Combined multiplicatively (all three must be present), they scored every 100×100 km patch of land on Earth — 15,351 patches total, Sahara to Siberia to Antarctica.

The results:

• Indus Valley: 100th percentile. No place on Earth scores higher.
• Mesopotamia: 99.5th
• Egypt: 98.9th
• Caral-Supe (Peru): 93.6th. Partial fit, low on tectonics but high on aridity and river
• Early China, Mesoamerica, West Africa: Each missing at least one ingredient

Monte Carlo test against 100,000 random placements: p < 10⁻⁴.

The interesting part isn't that civilizations formed near rivers in deserts – that's middle-school geography. It's the shape of the result. Three sites in the top 1% of the entire planet, on a metric chosen before the data was looked at. The fourth (Caral) sits clearly above the zero-baseline but lower than the Old World three.

The paper is careful about what it doesn't claim:

• It's a measure of desert-river geometry, not "civilization" generally. China, Mesoamerica, and West Africa formed under different ecological setups. They score zero because the index isn't built to detect them.
• The Monte Carlo null includes Antarctica and the deep Sahara, which makes any habitable river valley look extreme. The authors flag this.
• The geographic pattern doesn't prove a mechanism. The paper separately proposes one — that the same geology that prevented dispersal may also have made environmental feedback unusually testable, so practical knowledge could accumulate more reliably — but treats this as a hypothesis to be tested, not a result.

Currently under peer review. Full write-up with caveats and alternative explanations here.

Preprint with code and data: https://doi.org/10.5281/zenodo.19240284.

Happy to take questions on methodology, case-set construction, or where this fits in existing circumscription literature.