Our species walked out of Africa and spread across Eurasia on feet, stopped at the edge of water, and eventually crossed it. The Wallace Line, the invisible biogeographic boundary that runs between Bali and Lombok, marks where the animals of Asia end and the animals of Australia begin, and it is a line nobody walked. To cross it you had to float. Modern humans crossed the Wallace Line at least 50,000 years ago, reaching the ancient supercontinent of Sahul, which connected modern Australia and New Guinea when sea levels were lower. From Sahul's northeastern coast, they kept going. Some of them reached an enlarged island we now call Greater Bougainville, which connected Buka, Bougainville, Shortland, Choiseul, Santa Isabel, and Nggela into one landmass during the glacial maximum around 20,000 years ago. Kilu Cave sits near the northern tip of what was then that larger island, and inside it, the oldest radiocarbon dates come from a marine snail shell, a Nerita, carbon-dated to 28,740 years Before Present, calibrated to sometime between 29,850 and 31,560 BC.

The layers at Kilu Cave read like a menu for people who knew their coast. Shark bones make up the largest share of the fish remains, with a striking 20 percent of the Pleistocene fish bones coming from pelagic species, the tuna family, dolphinfish, and jacks that swim in deep water far from shore. These are fish you catch from a boat. Reef fish dominate the rest. Thousands of shells of Nerita undata and Nerita plicata, small tidal snails, pile up in the cave's layers; lizards, especially monitors and skinks; turtles and frogs; bats and rats. Mammals are nearly absent from what they hunted, because there were almost no mammals to hunt. Buka had no wallabies, no tree kangaroos, nothing like the faunal bounty that humans had encountered in Australia or mainland New Guinea. The people at Kilu worked with what the sea and the limestone karst gave them, and it was enough.

Here is what makes Kilu unusual among paleolithic sites in Melanesia, perhaps anywhere. On seventeen of the stone tools from the Pleistocene layer, archaeologists found starch grains from taro, both Colocasia and Alocasia varieties. They also found evidence for galip nut, Canarium indicum and its relatives, and coconut, Cocos nucifera. This is no minor detail. Taro is not a wild root that people happened to dig up. It is a cultivated plant, or at least an intensively managed one, and its presence on 29,000-year-old tools pushes back the story of plant food in the tropical Pacific by tens of thousands of years before anyone imagined. Kilu Cave is currently the only site in Melanesia with evidence for plant use by the region's initial inhabitants. The discovery rewrote what we thought paleolithic people did in forests. They were not just hunters and shellfish gatherers. They knew the plants of this place in detail, and they brought pieces of it home.

Then, around 20,000 years ago, people stopped visiting the cave. The occupation goes quiet for a long time, not because the people disappeared, but because the shoreline they had relied on moved. As sea levels rose, Kilu found itself stranded further from the coast, and the marine resources became inconvenient. The cave was reoccupied more intensively during the Holocene, roughly 9,000 to 5,000 years ago. Post-Lapita pottery in the upper stratum, dating after 2,500 BP, shows it was still in use when the Lapita culture, ancestors of today's Polynesians and Micronesians, swept through with their distinctively decorated ceramics around 3,000 years ago. The Lapita arrival also coincides with the extinction of a range of endemic birds and mammals on Buka. Of eighteen landbird species identified from the cave's bones, seven are now extinct and eleven are gone from Buka.

Kilu Cave is simultaneously the oldest known human site in the Solomon Islands archipelago, the oldest proof of paleolithic open-ocean navigation, and the longest documented paleolithic sea crossing. It is the only site in Melanesia with direct evidence for plant use by the initial inhabitants. It holds the type specimens of two rat species new to science, Solomys spriggsarum and Melomys spechti, described from fossil bones. The cave itself is unimpressive from outside, a notch at the base of a cliff on a small island most maps render as an afterthought. But when archaeologists first sifted its floor in 1987, they found that Buka had been lived in for 29,000 years. Everyone who arrived after that was, in a sense, late.