No, Warming Didn't Kill the Megafauna
A bunch of big animals went extinct around when humans arrived. Scholars disagree about why. But one side is much more convincing.
10,000-50,000 years ago, large animals all across Earth mysteriously went missing.
This event spanned continents. Large animals that had been around for millions of years disappeared, including woolly rhinos, giant ground sloths, saber-toothed cats, and giant wombats. All the epic, giant-ass animals that you see in museums and think “if only there were animals like this still around” were extirpated, leaving only bones.
There are two main proposals for why these creatures disappeared. First, the warming hypothesis, which says the temperature increases of the early Holocene, around 12,000 years ago, wiped these animals out. Second, overkill, which says human hunting drove these animals to extinction.
In my view, the overkill side simply has a better case. But examining this case doesn’t merely tell us about megafauna. It teaches us something important about how good reasoning works, and how you can use Bayesian reasoning to make progress on a problem where even a lot of experts get it wrong.
There are four main lines of evidence for overkill.
1 Timing
Megafauna extinctions closely tracked the arrival of humans, not climatic shifts. On continent after continent, die-offs occurred right after humans arrived. This would be a bit like noticing that people mysteriously keep being eaten right after Hannibal Lecter invites them over for dinner.
In Australia, humans arrived somewhere between 65,000 and 44,000 years ago, and between 50,000 and 40,000 years ago, 82% of megafauna were wiped out.
In North America, humans arrived roughly 12,000–20,000 years ago (possibly earlier), and extinctions occurred between 11,000 and 15,000 years ago.
Remarkably, in South America, sharp population declines came when there was a relatively stable climate, around 12,000 years ago, after human arrival.
In New Zealand, population collapse came after humans arrived around 1280, when populations had previously been increasing and there was no serious climate change. Even climate-change proponents generally admit this one was caused by human activity.
In fact, in various continents, extinctions occurred despite very different climatic conditions—climates getting wetter came with the same extinctions as climates getting drier.
If humans did not drive the megafauna extinct, it would be extremely surprising that extinction consistently happened right around the time humans arrived. The odds of this pattern are basically 100% if humans were the culprit and very near 0% if we weren’t. I think this fact is on its own enough to basically settle the debate.
Critics sometimes note that we can’t be sure all these extinctions really did follow human activity. In some cases, we don’t know exactly when they went extinct—our evidence shows extinctions and arrivals occurred in about the same time period, but there isn’t perfect overlap. I don’t find this to be a very impressive response. The fossil record is pretty patchy. So on the assumption that humans drove these animals extinct, we would expect not to find clear fossil evidence of them all going extinct right when we arrived. But it’s undeniable that extinctions consistently occurred right around the time humans appeared.
In fact, the fossil evidence we do find is about what we’d expect on the overkill hypothesis. Fossils break down over time. Once one accounts for that, thus expecting a sparser fossil record for extinctions further back in time, the fossil record closely matches what we’d expect on overkill.
2 Africa
Humans evolved in Africa alongside megafauna. For this reason, African megafauna were the least vulnerable to human hunters. On other continents, we were an invasive species. In Africa, however, large animals were prepared for us. So if humans caused megafaunal extinction, we should expect megafauna extinction to have been least severe in Africa.
This is exactly what we observe. Africa only lost about 21% of its megafauna, while North and South America lost more than half of their megafauna. The hunting hypothesis actively predicts this data, while the climate change hypothesis renders it very surprising.
Some critics have worried that there hasn’t been much analysis of precisely which traits evolved to deal with human hunting. Nikhil Brocchini (though endorsing human activity being the cause) complains that “the coevolution hypothesis has gone largely uninvestigated and rests heavily on highly contestable underlying assumptions.”
But Brocchini’s proposal for why Africa was hit least hard also favors human activity being the culprit. He argues that ecosystems with large animals are particularly vulnerable to human disruption. Africa’s ecosystems had already undergone the disruption caused by humans and restabilized before the megafauna extinction window, while other continents got hit all at once. Thus, Brocchini’s view is that African animals didn’t coevolve defenses against humans, but rather that African ecosystem dynamics had already adapted to human activity, making them less vulnerable. African megafauna only had to deal with an existing predator, while other continents had to deal with a novel invasive species that they weren’t prepared for.
Whatever precise theory one goes with, this is clear evidence for overkill. Overkill correctly predicts that animals that evolved alongside humans will be hit least hard. This holds whatever you think the mechanism is—whether coevolution or something else. While proponents of the climate hypothesis can claim that African megafauna just happened to be the least vulnerable to climate change, this would be a pretty major coincidence.
3 Similar warming periods
Megafauna survived many similar previous warming periods, having generally persisted for millions of years. So if they survived those, why not this one? As John Halstead notes:
there are many comparably dramatic climatic transitions that did not cause huge species extinctions. It is only once humans made it to different continents, in the transition to the Holocene, that the megafauna started dying off. The difference between the Pleistocene-Holocene transition and the other climate transitions just does not seem large enough to cause such a discontinuous change in species extinctions.
To illustrate the point, Halstead provides a graph of temperature over time, showing that there’s nothing particularly special about the warming period preceding megafaunal extinction.
It’s true that there are some differences between the warming period accompanying megafaunal demise and other warm periods. But any warming period will differ in some respects from others. It is very surprising, on the climate change hypothesis, that the uniquely destructive warming period came right around the time humans arrived. And remember, extinctions occurred across different continents at different times, experiencing wildly different climatic changes.
You can see the backwards Bayesian reasoning in a lot of the papers on the subject. Meltzer, for example, claims that this argument requires “two assumptions: that previous glacial−interglacial transitions were similar to the end of the Pleistocene, and that the large mammal genera survived unchanged over multiple such cycles.”
But if we’re reasoning in a Bayesian way, you don’t have to start by demonstrating that this period was different from the others. By default, it would be extremely unlikely that the climate period that hit right when humans arrived was uniquely destructive. So the starting assumption should be its non-uniqueness absent strong contrary evidence. We don’t have to prove uniqueness from the outset.
4 Selective size targeting
Humans primarily hunted large animals. Thus, if human activity drove megafauna extinct, we should expect larger animals to have been hit the hardest.
Not only were they hit the hardest, they were hit hardest to a degree that was unprecedented across 65 million years of mammalian evolution. Other mass extinctions did not hit large animals anywhere near as hard. Smaller mammals, plants, marine animals, and invertebrates were largely unaffected by the mass extinction that occurred as humans emerged. It was only large animals that were systematically wiped out.
Now, some have argued that these organisms died off mostly because of their slow reproductive rate. On this view, large animals only were killed off because they tended to reproduce more slowly, not because they were selectively targetted. But this doesn’t affect the core argument at all. Why would climate change disproportionately target slowly reproducing animals in a way that’s unprecedented across 65 million years?
5 Why do people disagree?
What are the main counterarguments?
It’s often claimed that human hunting wouldn’t have killed enough animals to cause megafaunal extinctions. However, more sophisticated population models have confirmed that extinction could occur with even relatively low levels of hunting. If organisms reproduce slowly, you don’t need super high hunting rates to drive organisms extinct.
Critics also note that the extinctions mostly occurred around the time of big climatic shifts. It is true that this is something of a coincidence, but it’s not a huge one. It’s equally a coincidence that extinctions occurred right around the time of the emergence of humans. And overkill proponents have a perfectly natural explanation of why megafauna died off at this time: that was the time humans spread throughout the globe!
People often note the rarity of kill sites. We don’t have many discovered sites where humans hunted these large animals. But given the intermittency of the fossil record, this isn’t very surprising. Even if humans did hunt these animals, it isn’t very unlikely we wouldn’t find abundant fossil evidence of it.
It’s also helpful to take a step back and just think about the evidence at a high-level. Experts on both sides disagree about whether the sparseness of kill sites is surprising on the assumption that humans drove the megafauna extinct. Now, even if you ended up thinking one side had a better case there, you shouldn’t be massively certain that they were right. So let’s be generous and assume you ended up concluding that there’s a 75% chance that the absence of kill sites is inconsistent with humans being the cause. Let’s assume you think in the remaining 25% of probability space that sparse kill sites are equally consistent with both hypotheses.
In this case, taking into account uncertainty, the rarity of kill sites will only be three times more likely given climate change-caused megafauna extinction than human-caused megafauna extinction. That’s not nothing, but it absolutely pales in comparison to the evidence for humans being the culprit. On the other side, we’re getting Bayes factors in the hundreds or thousands. The circumstantial evidence completely swamps any kind of technical nitty gritty argument that doesn’t command any kind of clear expert consensus.
Now, it might be that humans and climate change both played a role. But I think it’s clear that humans were counterfactually responsible, in the sense that if humans had never evolved, the megafauna wouldn’t have gone widely extinct. The above evidence also makes clear that humans were primarily responsible.
6 A victory for Bayes!
I’m just a philosopher. Why should you believe me when I say the case for overkill is so decisive as to be, well, overkill?
A big part of the answer is that you can simply read the arguments on both sides and see that the case for overkill just is overwhelmingly decisive. To reject overkill you have to think:
Mysteriously, across continents, megafauna repeatedly disappeared right when humans arrived, by some bizarre coincidence. Often these die-offs were tens of thousands of years apart.
This was driven by climate change, even though there had been many similar warm periods that the megafauna survived. There was just something special about the one that hit right when humans arrived. And strangely, it was driven by many different kinds of climate change across different continents at different times—some warmings, some coolings.
This also, by some strange coincidence, targeted large animals in a way that was unprecedented across 65 million years of mammalian evolution. Nothing like that had ever happened before.
Also, by chance, animals that evolved alongside humans were hit least hard.
The climate proponents can have individual piecemeal explanations of any single fact. But the global fact pattern requires a series of bizarre, unprecedented, independent coincidences. Any one of these coincidences would be close to disqualifying. By the time you’ve reached four of them, you should abandon the theory that lead you to believe them.
The other part of the reason that you should be willing to disagree with experts is that this dispute is as much about scientific methodology as about the evidence itself. Critics often worry that too much of the evidence is circumstantial. That all we have are correlations, but these don’t suffice to establish a causal pattern.
But philosophy can help illustrate why this is wrong. Some fact is evidence for a hypothesis if that fact is likelier if the hypothesis is true than false. If my DNA is at the crime scene, that’s evidence that I committed the crime. It’s not proof. It might not be reasonable overall to think I committed it. But it raises the likelihood of my being the perpetrator.
Philosophers have no special expertise in learning about fossils. But they do have expertise figuring out how evidence works. So they are often uniquely positioned to understand when a field is going wrong by thinking incorrectly about how evidence works. Circumstantial evidence can be decisive if it’s sufficiently better predicted by one theory than another. If you always develop a headache after eating a kind of mushroom, the evidence for it being causal is merely circumstantial, but still, you should think it’s very likely the cause.
John Halstead is a philosopher by training. Many of the points I make in this article are also drawn from his writing on the subject. He writes:
Scholars do not have to take a Bayesian approach to get published in academic journals, and many scholars are not Bayesian…The neglect of circumstantial evidence is especially important in this case because the circumstantial evidence in favour of the overkill hypothesis is extremely strong.
And it’s not just philosophers who think that one side has a much better case. In one of the most amusingly aggro papers I’ve ever read, titled A review of some attacks on the overkill hypothesis, with special attention to misrepresentations and doubletalk, archeologist Gary Haynes argues “The opposing Climate-Change hypothesis to explain extinctions is driven by ideology as much as by objective reasoning.” A lot of the overkill proponents really do seem to think the debate should have been over decades ago, and it’s an embarrassment to the field that it hasn’t been settled.
It’s rare you should venture outside of your field and confidently propound on other fields. But you can be reasonable justified in believing that your view is right if you’ve looked carefully into the facts and your view is:
Mainstream among experts, even if it doesn’t command a clear consensus.
Supported by many strong and specific arguments.
Rejected largely on methodological grounds, rather than because of specific counterarguments.
You have specific expertise bearing on the appropriate methodology, in a way that’s lacking inside the field.
For this reason, I think that a person who has carefully looked into the facts can be pretty sure that humans killed the megafauna—Halstead mentions his credence is 90+%, and that seems reasonable to me. The case really is lopsided, particularly when you think about evidence correctly, rather than importing confused assumptions about philosophy of science. Another victory for good old reverend Bayes.
Great piece. I looked into this about 5 years ago, more superficially, and came to the same conclusion. My guess was that opposition to overkill was politically motivated. People didn't want to "blame" indigenous populations or spoil the idea of that they lived in harmony with nature. Is there a better error theory?
I thought this was already the prevailing hypothesis. I read it in Jared Diamond book decades ago. The reason they survived in Africa is because humans co-evolved with them.
I just want to add that humans needn’t have killed large predators directly if we killed enough of the large herbivores they relied on.