What if the assumption that monument-building only began after the glaciers melted has it exactly backwards? What if the builders came first, and the rising ocean erased them? That is the inverted possibility lurking 25 meters beneath the waves off Yonaguni, the westernmost speck of Japan, closer to Taiwan than to Tokyo. A sport diver named Kihachiro Aratake, looking for hammerhead sharks in 1986, instead found a vast terraced mass of stone with flat platforms, sheer right-angled walls, and what looked unsettlingly like steps.

Here is what is not in dispute. The structure is real, enormous, and submerged. It sits in the Yaeyama Group bedrock, sandstones and mudstones laid down in the Early Miocene roughly 20 million years ago. The relevant point is sea level: at the height of the last glaciation, around 10,000 to 12,000 years ago, the ocean stood far lower, and whatever Yonaguni is, it stood above water then. So if human hands touched it, they touched it deep in the Ice Age, before Gobekli Tepe, before the pyramids, before the conventional clock of civilization is supposed to have started ticking. That is the entire stakes of the argument.

The case for human involvement is carried mostly by one man: marine geologist Masaaki Kimura of the University of the Ryukyus, who has logged years of dives and built scale models of the site. Kimura points to features he reads as deliberate: paired holes he interprets as post sockets, a sharp trench, what he calls quarry marks, and stones he believes were dressed and moved. In his reading this is not a single monument but the wreckage of a settlement, with roads and an arena, drowned and forgotten. Graham Hancock, who says he has made more than 200 dives there, argues the geometry is too insistent, too repeated, to be accidental.

Then comes the skeptical reading, and it is a serious one. Geologist Robert Schoch, no reflexive debunker (he is the man who argued the Sphinx is older than Egyptology allows), dived Yonaguni and came away convinced it is primarily natural. The sandstone here fractures along bedding planes and along sets of parallel vertical joints. When those two directions intersect, the rock naturally calves off in rectangular blocks and flat steps. Schoch documented the same step-and-terrace pattern in the exposed cliffs on Yonaguni's shore, above water, where no lost civilization could be invoked. Nature, in other words, is demonstrably building 'staircases' on this island in plain sight.

The hard evidence cuts against the lost-city thesis in one crucial way: there are no portable artifacts. No tools, no pottery, no inscriptions that survive scrutiny, no bones, nothing you could carbon-date and pin to a maker. Kimura's 'characters' and 'animal carvings' have not been independently authenticated as worked stone. In archaeology, a wall without a single artifact is a very lonely wall. The parsimonious explanation, the one most geologists accept, is jointed sandstone weathering exactly as jointed sandstone does.

And yet the inversion will not quite die, because the timeline is genuinely uncomfortable. We know sea levels rose dramatically and fast as the Ice Age ended. We know coastlines that were dry land for millennia are now drowned, and that humans have always clustered on coastlines. Whatever the verdict on Yonaguni specifically, the larger logic holds: if any of our ancestors built in stone before 10,000 years ago, the evidence is most likely underwater, unsearched, and unfunded. The question Yonaguni leaves open is not really about one rock formation. It is whether we have been looking for the dawn of civilization in all the wrong places, on land, when the relevant real estate is now beneath the sea.