It’s known as the mammoth cemetery for good reason. Along the banks of a Siberian river not far from the Arctic Ocean lie thousands of bones, most of them belonging to the giant, shaggy relatives of today’s elephants. A new study argues that such mysterious graveyards were not the results of a natural catastrophe, but rather the work of early human hunters—who may have had help from some of the world’s first dogs.

“This is the first time that someone’s gone out on a limb and suggested something different than what we thought before,” says Angela Perri, a postdoctoral fellow at the Max Planck Institute for Evolutionary Anthropology in Leipzig, Germany, and an expert on dog domestication. “But it’s still very speculative at this stage.”

Study author Pat Shipman first became interested in what she calls “mammoth megasites” in 2009. About 30 such spots have been unearthed in central Europe and North Asia, some with tens of thousands of bones packed tightly on top of each other across areas as small as 60 square meters. The massive tusks and femurs of mammoths jut out among the remains of wild horses, deer, foxes, and other animals. “They’re crazy sites,” says Shipman, an anthropologist at Pennsylvania State University, University Park. “The sheer number of dead mammoths is astounding.” More than 160 of the tusked goliaths lie in the mammoth cemetery—a site known as Berelekh—alone.

How did they get there? Some scientists think it was an act of nature—perhaps a flood that swept dozens of animals to a particular spot, or an unlucky herd that fell through thin ice. But recent evidence has suggested that people may to be blame. Shipman says the mammoth megasites begin to appear about 44,000 years ago, just about the time that modern humans entered this part of the world. What’s more, archaeologists have found evidence of huts made of mammoth bones at some of these locations, as well as cuts and burn marks on the bones that could only have been made by people.

To get a clearer picture, Shipman combed through the literature on more than a dozen mammoth megasites, paying particular attention to the age and sex of the mammoths unearthed there. She then compared these demographics with those seen with the deaths of large numbers of elephants, the mammoth’s closest living relative. Natural disasters such as droughts kill the youngest and oldest elephants, but other sudden die-offs—such as a herd falling through ice or a cull of elephants to control their population—kill indiscriminately, leaving behind the carcasses of young and old, male and female. Elephant hunters, meanwhile, tend to kill each animal in a different place. “To my surprise, hardly anything matched these patterns,” Shipman says of the mammoth bones. What’s more, the dating on the bones indicated that they had been laid down over hundreds of years. That suggests that the animals were killed over and over in the same spot over many generations, she reports in Quaternary International. “There’s something that’s drawing them to that location.”

Shipman says the data point to a scenario in which humans killed the mammoths, but not in the way people do today. Instead of culling them or hunting them across vast plains, ancient peoples may have ambushed the creatures. The reason so many bones are found in the same location may be that these spots were ideal for such ambushes. Perhaps they were surrounded by thick brush, in which spear-hurling humans could hide, or maybe they lay along a commonly traveled migration route. Shipman also thinks the hunters may have had some help from dogs.

It’s still unclear exactly when or where dogs became domesticated, but some recent archaeological evidence suggests it may have happened around the same time and place as the mammoth megasites. A skull recovered from a cave in southern Belgium, for example, has both wolf- and doglike features, and it dates to about 32,000 years ago. Though genetic evidence indicates that this animal may not have been an ancestor of today’s dogs, the find suggests that the process of canine domestication could have begun tens of thousands of years ago. Significantly, Shipman says, similar skulls have been found among the mammoth bones at several megasites. Many of the skulls bear healed fractures, a possible indication that these animals were cared for by humans.

Shipman speculates that the mammoth megasites may be the first significant evidence of a cooperative relationship between man and dog. The canines could have corralled the mammoths at the ambush sites and held the prey in place while human hunters moved in for the kill, Shipman says. Once the mammoths were dead, the dogs could have protected the sites from scavengers. “All of that mammoth meat would have brought predators from miles around,” she says. In return, the humans may have provided these canines with food and protection. And slowly, a closer relationship may have begun to form.

Finding more large and strong doglike animals at these sites would support her hypothesis, Shipman says. Such finds will be necessary to convince archaeologists like Nicholas Conard that the new work is more than just a leap of faith. “I like it as an idea, but there’s no smoking gun,” says Conard, who works at the University of Tübingen in Germany and who has personally excavated mammoth megasites. Perri agrees. “We don’t know enough about what early dogs—or even the wolves of the time—looked like,” she says. “This is extrapolating from too few examples.” Still, Conard says, “there are so few ideas about how these sites formed, and what Shipman is arguing is possible and testable. It’s a move in the right direction.”