What killed the large animals?

Most of Australia’s losses came before the last glacial maximum. A 2016 analysis of dated fossils puts their peak about 42,000 years ago, some thirteen thousand years after its estimate for when people arrived. North America lost its own between about 13,800 and 11,400 years ago. South America’s losses are poorly dated and ran on into a relatively stable Holocene climate. Madagascar and New Zealand lost theirs within the last two thousand years.

No single global climate event lines up with all of that. The spread of people does, more loosely than the word arrival suggests. In Australia people and the large animals overlapped for thousands of years; in Madagascar most populations crashed about a thousand years ago, and how long people had been there by then is argued; and in North America the losses fall in a time of severe climate change as well.

A 2014 analysis modelled both factors together across 229 regions of the world. Which hominins had been present, and so for how long, accounted for 64 per cent of the variation in extinction; the best climate-only model accounted for 20 per cent, and its signal appeared in Eurasia and nowhere else.

A 2004 review set out the state of the debate: a climate camp, a human camp, and the difficulty that in the northern hemisphere human impacts and a pronounced change of climate overlap.

It concluded that human hunting was not solely responsible everywhere, and that firmer chronologies, better ecological models and regional palaeoecology were still needed before the details of the worldwide pattern could be understood.

The 2014 analysis is not simply a continuation of it. Barnosky read the northern hemisphere as climate and people acting together, and that is the reading Sandom argues against, on the grounds that the climate signal is absent in exactly the places modern humans reached first.

One objection to the human explanation is scale: small groups of foragers with stone weapons, spread across a continent of large animals.

The answer offered is demographic rather than dramatic. A simulation published in 2001 reproduced the North American losses, calling 32 of 41 prey species right, from slow human population growth, hunting at random and a low ceiling on hunting effort, without invoking climate change at all.

That is a model rather than an observation. Whether the archaeological record has too few kill sites to carry it is itself argued: the scarcity is the standard objection, and others have found kill sites no rarer than expected.

Eurasia is the one region where the 2014 analysis finds a climate signal, and northwest Eurasia is where it sits. The 2004 review read the northern hemisphere the same way, as human pressure meeting a severe climate change.

Even there the reading is contested. Sandom takes the Eurasian signal as partly about people too: on its reading, cold held archaic humans out of the far north, which left those regions as refuges until modern humans arrived.

North America is where a rival reading is strongest. A 2021 analysis of megafauna and human radiocarbon dates found megafauna declines tracking falling temperatures, and no persistent relationship with human numbers. Its authors expect the question to stay contentious.

So the honest version is not that climate played no part. It is that the pattern across the whole planet points to people, and climate alone does not reproduce it.