The eruption has a date and an address. It happened about 74,000 years ago on the island of Sumatra, in what is now Indonesia, at roughly 2.7 degrees north and 98.9 degrees east. The looser figure in our own research file is 74,000 years give or take 5,000. The tightest published dating, an argon-argon measurement of the erupted tuff by Mark and colleagues in 2014, pins it more precisely at 74,000 years give or take 2,000. Either way, this is one of the better-fixed moments in the last hundred thousand years of Earth history.

It was enormous, the largest eruption anywhere on Earth in the last two million years. Toba rates a full VEI-8, the maximum value on the Volcanic Explosivity Index, and it ejected on the order of 2,800 cubic kilometers of material measured as dense rock. Numbers that large stop meaning anything without a yardstick, so here is the one our file uses. Mount St. Helens in 1980 put out about 1 cubic kilometer. Tambora in 1815, the largest eruption in recorded human history, put out about 160. Toba was therefore roughly 2,800 times the volume of Mount St. Helens and about 17.5 times the volume of Tambora.

The hole it left is now a lake. The Toba caldera, the basin that dropped when the emptied magma chamber collapsed, is about 100 kilometers long and 30 kilometers wide, and it holds Lake Toba, the largest volcanic lake on Earth. Samosir Island, the large island near its center, is not a separate feature at all but the caldera floor itself, slowly pushed back up after the eruption. The mountain did not simply explode. It swallowed the ground it had been standing on.

The ash reached almost incomprehensibly far. The Youngest Toba Tuff, the technical name for the eruption's fallout, blankets South and Southeast Asia, the floor of the Indian Ocean, the Arabian Sea, and East Africa. Deposits more than 15 centimeters thick have been mapped across roughly 4 million square kilometers. In central India, some 2,500 kilometers from the volcano, the ash still reaches up to 6 meters thick, and it turns up at archaeological sites, including one called Jwalapuram that will matter a great deal later in this story.

What it did to the sky is harder to measure, and the file is honest about the uncertainty. The eruption is estimated to have injected somewhere between 1 and 10 gigatonnes of sulfur dioxide into the stratosphere, which could have cut incoming sunlight by anywhere from 25 to 90 percent. That is an enormous range, and the width of it is the point. Reconstructing the atmosphere of a single year 74,000 years gone is genuinely hard, and no one should pretend the figure is tighter than it is.

Toba belongs to a small and dreaded class. A VEI-8 eruption, sometimes called mega-colossal, ejects more than 1,000 cubic kilometers of material and recurs, very roughly, on the order of once every hundred thousand years somewhere on Earth. Only a handful of such eruptions are known across the whole geological record, and the still-active example everyone worries about is the Yellowstone hotspot, which shares Toba's VEI-8 rating exactly. When people reach for a comparison to Toba, Yellowstone is the one they reach for, because there is almost nothing else in the same category.

Start with a real and slightly unsettling fact about our species, because Ambrose's argument is built on it. For a mammal spread across the entire planet, human beings carry startlingly little genetic diversity. Our effective population size, the geneticist's measure of how much variation we actually harbor, sits at only about 10,000, far below what our headcount would predict. Two chimpanzee troops living on opposite banks of a single African river can hold more genetic diversity between them than the whole human species does. In population genetics, a sharp drop in numbers that strips out this kind of variation is called a bottleneck, and its mathematics was worked out by Sewall Wright and others decades before anyone thought to point it at a volcano.

Ambrose's move was to connect that genetic thinness to the eruption. If humanity had passed through a severe bottleneck sometime between roughly 50,000 and 100,000 years ago, it would explain why we are all so genetically alike today. Genetic calculations put the effective breeding population at the low point at something like 3,000 to 10,000 individuals, and the timing of that squeeze overlaps with Toba at about 74,000 years ago. Ambrose read the overlap as cause and effect. The volcanic winter, he argued, drove a near-extinction, and every living human descends from the few thousand who made it through. It was an elegant hypothesis, and it stitched geology, climate science, and genetics together in a way that drove decades of research.

The mechanism he proposed was ecological collapse from the bottom up. A years-long volcanic winter would have wrecked vegetation, thinned out animal populations, and dried up water sources, starving the human food web until it broke. He also noted that different human populations carry slightly different bottleneck signatures, which he read as evidence of survival in scattered refugia, small pockets of humanity riding out the cold in sheltered places rather than one clean global event.

The three positions line up like this. The Catastrophist reading, associated above all with Stanley Ambrose, holds that Toba caused a six to ten year volcanic winter, a global ecological collapse, and a human crash to a few thousand breeding individuals. The Continuist reading, argued most forcefully by the archaeologist Michael Petraglia, holds that humans rode out the eruption without catastrophic disruption and that the genetic bottleneck, real as it is, has some other cause. The Moderate reading, associated with Clive Oppenheimer and Martin Williams, sits between them: the eruption was regionally devastating across South and Southeast Asia but not a global near-extinction, and some populations carried on with little disruption. The single most important thing to notice is the direction of travel. Over the last two decades the field has moved away from the Catastrophist end and toward the other two.

Several specific findings pull against the strong version. In Africa, the likely homeland of modern humans at 74,000 years ago, the archaeological record shows continuity, not collapse. If our species was doing fine in the place most of us came from, a truly global near-extinction becomes hard to sustain. Nor is there any clear mass die-off of other species at 74,000 years ago. A catastrophe severe enough to nearly finish humanity should have left similar bottleneck scars across many mammals, and few of them show any.

There is also a subtler objection that cuts to the heart of the claim. The human bottleneck might be entirely real and Toba might still have nothing to do with it. The genetic squeeze may predate the eruption, or it may be the signature of the Out of Africa migration itself, a founder effect produced when a small band left the continent and seeded the rest of the world. On that reading the bottleneck is genuine, but pinning it on the volcano is a separate and unproven step.

Jwalapuram, in the Kurnool District of Andhra Pradesh, sits about 2,500 kilometers from Toba, squarely inside the ash-fall zone, with a Youngest Toba Tuff layer about 2 meters thick lying visibly in its stratigraphy. That ash is a clock frozen in the ground. Below it, archaeologists find Middle Paleolithic stone tools made with the Levallois technique and clear signs of human occupation. Above it, they find the same tools, made the same way, with no obvious break in occupation and no change in where the toolmakers sourced their raw stone. To Petraglia the reading is straightforward: the people at Jwalapuram lived through the ashfall and simply carried on, with no sign of collapse, flight, or cultural rupture. Nor is Jwalapuram alone. At a second Indian site, Dhaba, stone-tool assemblages span the Toba ash layer without significant change as well, published by Clarkson and colleagues in 2020, giving the subcontinent a second continuous record straight across the eruption.

Ambrose has a rebuttal, and it is not a weak one. A few surviving sites, he argues, do not disprove a wider crash. Jwalapuram might record rapid recolonization rather than unbroken occupation, because the sediments may lack the time resolution to catch a brief abandonment. A layer that looks continuous to us could hide a gap of decades. This is exactly why the site settles nothing on its own: the same stratigraphy reads as survival to one camp and as a resolution problem to the other, and no one has yet found a way to force the ground to choose.

The original figures were genuinely frightening. In 1992 Michael Rampino and Stephen Self modeled the eruption's aerosol cloud and proposed 3 to 5 degrees Celsius of global cooling, a volcanic winter lasting six to ten years, and the possibility of a deeper volcanic freeze. Those numbers are real, in the sense that they were honestly published and widely cited for years. They are also, on the current evidence, too high. What follows is not a scandal. It is how the process is supposed to work.

Follow the shrinkage. Robock and colleagues, rerunning the eruption through an updated general circulation model in 2009, found cooling nearer 1.5 to 3 degrees, not enough to trigger global glaciation and looking more regional than planetary. Timmreck and colleagues went further in 2010: once you model the way sulfur aerosols clump together and fall out faster than the older models assumed, the cooling drops to something like 1 to 2 degrees for about five years, and the old nuclear-winter language starts to look overblown.

Then the modeling met the mud, and the mud did not cooperate. Lane and colleagues pulled sediment cores from Lake Malawi in East Africa in 2013 and found no volcanic-winter signal in them at all, no ashy fingerprint of a frozen world at the very latitudes where our ancestors were living. And in 2018 Smith and colleagues, working the archaeology at Pinnacle Point on the South African coast, found people not merely surviving but thriving straight through the eruption, with no evidence of ecological disruption at all. The test was run, in exactly the place a global winter should have shown up. It came back empty.

Those results do not stand alone. Curtis Marean's long excavation of the South African coast at Pinnacle Point points to continuous human occupation straight across the 74,000-year window, and Chad Yost's analysis of phytoliths and charcoal from Lake Malawi found no sign of catastrophic environmental disruption in East Africa either. Taken together, the honest reading of all this is not that Ambrose was foolish. It is that the bottleneck, if it happened, was likely either geographically selective, hammering South and Southeast Asia while African populations carried on, or milder than the extreme figure of a few thousand breeding pairs. The catastrophe kept surviving contact with the data by getting smaller.

The cheetah is genetically almost a copy of itself. You can graft skin from one unrelated cheetah onto another and the recipient will not reject it, because their immune systems can barely tell each other apart. Cheetah sperm is poor across the whole species. Genome sequencing has confirmed why: cheetahs passed through a severe bottleneck around the end of the last ice age, one that flattened their genetic diversity down to levels you would otherwise only find in inbred laboratory mice. This is what Ambrose's mechanism looks like when it genuinely happens. The cheetah case is disputed by no one. It is the clean control that shows the human question is at least a reasonable one to ask, even though the answer for us keeps coming back uncertain.

Mitochondrial Eve is the most recent woman from whom every living person inherits their mitochondrial DNA, usually placed around 150,000 to 200,000 years ago. Y-Chromosomal Adam is the equivalent most recent common ancestor down the male line, commonly cited between about 120,000 and 156,000 years ago, though the estimate shifts with the method. Here is what they are not. They were not the first woman and man. They were not a couple. They did not even live at the same time. And, crucially for this story, neither one is a population count. They are statistical facts about which genetic lineages happened to survive, not evidence that humanity was ever reduced to two people, or to any particular number at all. Mitochondrial Eve had a great many contemporaries. She is simply the one whose unbroken maternal line reaches all the way to the present.

The genetics do point to a real founder event, but perhaps not Toba's. When modern humans dispersed out of Africa, sometime around 70,000 to 50,000 years ago, they did so as a founding group estimated at only about 1,000 to 5,000 effective individuals. That migration left a clear fingerprint: human genetic diversity decreases steadily the farther a population lives from Africa, exactly the pattern a serial founder effect produces as each new group is seeded from the last. This Out of Africa bottleneck may be large enough to account for the low diversity of our species all by itself. If it is, then the volcano is not needed to explain the genetics at all, and the very number Ambrose leaned on has an entirely non-volcanic home.

One last caution belongs here, about the machinery behind every one of these numbers. Ancient genetic material is short, chemically damaged, and desperately easy to contaminate, so any claimed population size from tens of thousands of years ago rests on careful authentication: the characteristic decay patterns in the DNA, the tiny fragment lengths, clean-room extraction, contamination estimates. Modern paleogenomics has largely replaced the old picture of human evolution as a single marching line with something messier and truer, full of interbreeding, population structure, and repeated demographic turnover. The Toba bottleneck debate lives inside that messiness, not above it.

The open one is a genuinely interesting idea about what a near-extinction might have done to the survivors. Ambrose and others have suggested that the bottleneck, if it happened, could have accelerated the arrival of behavioral modernity, forcing small, isolated bands into tighter cooperation, richer symbolic communication, and faster technological invention as the very price of survival. The burst of symbolic artifacts in the African record after about 70,000 years ago, the beads and pigments and engravings at sites like Blombos Cave, has been hung on this idea. It is a good story. It is also explicitly speculative, a causal chain no one has come close to demonstrating, and it rests on a bottleneck that may not have happened in the first place.

The closed one is simpler. Toba did not cause the ice ages. The Quaternary glaciations began about 2.6 million years ago, driven by the slow arithmetic of Earth's orbit, the Milankovitch cycles, and by tectonic change, long before Toba existed. The eruption may have deepened the cooling of one glacial phase, Marine Isotope Stage 4, but it did not start the ice ages, and the claim that it did is simply wrong.