It’s not always easy to find clues to ancient campfires. Bits of charcoal, cracked bones, and discolored rocks often give a prehistoric blaze away. But not every blaze leaves such obvious traces, especially after hundreds of thousands of years.
Now, using artificial intelligence (AI) to detect the subtle ways in which extreme heat warps a material’s atomic structure, scientists have discovered the potential presence of a nearly 1-million-year-old fire featuring dozens of purportedly burnt objects buried at an archaeological site in Israel. If the technique proves reliable, the findings could shed light on when, where, and why humans first learned to harness the flame.
Richard Wrangham, an anthropologist at Harvard University, is impressed with the new method. He has long advocated that our human ancestors evolved smaller guts and larger brains once they began to cook food, perhaps about 1.8 million years ago. “We need imaginative new methods” to pinpoint ancient fires, he says. “Now, we have one.”
Most studies of fire rely on the obvious bits of charcoal and other clues. But Filipe Natalio, an archaeological biochemist at the Weizmann Institute of Science, wanted to find a way to identify the invisible evidence fire leaves behind. Previous work, led in part by forensic scientists, has shown that burning alters bone structure at the atomic level, so burnt and unburnt human bones absorb different wavelengths of the infrared spectrum. Researchers can detect a charred bone using a technique known as Fourier-transform infrared (FTIR) spectroscopy, which measures the absorption of different wavelengths of light.
Natalio and colleagues wondered whether a similar method might work for burnt stone tools, which are often more abundant than bones in very ancient sites—and are a clear sign of human presence. He and colleagues experimented by heating flint, a common toolmaking rock that can become easier to chip and shape after heating, to various temperatures in a fire, then applying spectroscopic techniques to see whether they could identify the signatures of burning. But because of natural variations in the flint, the patterns in the data were hopelessly complex.
“One peak would go up, another would go down … and the changes were so subtle that we couldn’t rely on them,” Natalio says. “That’s when we turned to artificial intelligence.”
The researchers devised a computer program to hunt for subtle patterns that would have taken ages for the scientists to find on their own, Natalio says. The AI worked. Using a technique called ultraviolet (UV) Raman spectroscopy, which measures the absorption of UV light, the AI could reliably differentiate burnt and unburnt pieces of modern flint and even reveal the temperatures at which they burned.
Next, the team applied its method to 26 flint tools, mostly small cutting edges, that had been excavated in the 1970s from Evron Quarry, a coastal site in northwestern Israel. A combination of dating methods suggested the site was between 800,000 and 1 million years old and was probably inhabited by the widespread, toolmaking human ancestor known as Homo erectus. Dozens of animal bones were found alongside the tools, but archaeologists had found no traditional evidence of fire such as charcoal or reddened sediment.
Using their new technique, Natalio and colleagues found most of the flint tools had been heated to a range of temperatures between 200°C and 600°C, they report today in the Proceedings of the National Academy of Sciences. (The average campfire burns at about 400°C.) The researchers also used FTIR spectroscopy to analyze 13 bits of tusk, from one of two elephantlike genuses known as Stegodon and Elephas, that had been found in the same sedimentary layer as the tools. The tusks, too, had been exposed to temperatures as high as 600°C.
That, Natalio says, may be evidence that the site’s inhabitants cooked their kills. If so, that would make it—along with a potential 1-million-year-old hearth in South Africa’s Wonderwerk Cave—among the oldest known cooking sites.
“It’s well done,” (the paper, not the roasted elephant) says Dennis Sandgathe, a paleoanthropologist at Simon Fraser University. “There are less than half a dozen sites in the world with [evidence for] fire that’s older than 500,000 years old. It may be because hominins were not using fire very frequently, but it may also be that we are missing some of it. So, this is really important.”
There’s still no way to definitively say whether the tools and tusks at this site burned in a natural or humanmade fire, Natalio says. Based on vegetation, fires can burn at different temperatures even within a single location. But the sheer variability of temperatures among tools so closely situated at Evron Quarry suggests to Natalio a radical notion: that the toolmakers were experimenting, heating flint cores to different temperatures to see how it affected their workability.
Sarah Hlubik, a paleoanthropologist at George Washington University who studies the origins of fire, isn’t so sure. “At the age of this site, I’d say that is unlikely but not impossible,” she says. “We don’t really see heat treatment until much later, and if the technology was being experimented with at nearly 1 million years, we would likely see it more widespread earlier than we do.”
The new technique is promising, Hlubik says. But she’d like to see the work reproduced in a wider variety of settings—and for the team to rule out other possibilities, such as naturally burnt materials from different places and times washing into the site. Until then, Hlubik says, “It’s important to take results like this with a grain of salt.”