Twenty-Six Metres Down: The Yonaguni Terraces and the Argument That Will Not End
In the winter of 1986 a diver named Kihachiro Aratake went into the water off the southern tip of Yonaguni looking for hammerhead sharks.
Yonaguni is the westernmost inhabited island of Japan, closer to Taiwan than to Okinawa, a place of about seventeen hundred people, strong currents and famously difficult diving. Aratake ran the local tourism association and was scouting new sites. Below the Arakawabana cliff, at a depth of around twenty-six metres, he swam over something that stopped him: a mass of rock rising from the seabed in flat terraces, with faces that met at right angles and ledges stacked like a staircase.
He surfaced, and within a few years the argument that followed had reached the international press. It has not been settled since.
The formation is roughly fifty metres long and about twenty-seven metres high at its tallest point. Divers have given its features nicknames that stuck and that quietly beg the question: the steps, the terrace, the loop road, the twin megaliths, the turtle, the stadium. The top is broad and flat. Several of the vertical faces are strikingly clean. In the underwater light, with the surge moving and visibility limited, it looks like the corner of a building.
The rock itself is not in dispute. It belongs to the Yaeyama Group, sandstones and mudstones laid down in the Early Miocene, roughly twenty million years ago. That kind of rock is deposited in flat layers and, when it fractures, tends to break along two directions at once: horizontally along the bedding planes between layers, and vertically along joints. The result, in the right conditions, is a stack of clean rectangular blocks. Yonaguni also sits in one of the most seismically active regions on earth.
The first scientist to make the site his life's work was Masaaki Kimura, a marine geologist at the University of the Ryukyus in Okinawa. Kimura began diving the formation shortly after Aratake reported it and returned for decades, mapping it dive by dive. He did not think it was rock shaped by chance. He identified what he read as a pyramid, a castle, roads, a stadium and a series of monuments. He pointed to drainage channels cut into the terraces, to holes he interpreted as post sockets, to a pair of standing blocks that appear to have been positioned, and to surfaces he described as carvings, including one he read as a face and another as an animal.
Kimura's early dating rested on the sea. At the height of the last ice age global sea level stood well over a hundred metres below the present shoreline, and the shelf around Yonaguni was dry land. The water rose fast after the ice melted. A surface now twenty-six metres down would last have stood in open air roughly ten thousand years ago. If people cut those terraces, Kimura argued at first, they cut them before that date, which would place organised stone construction in East Asia thousands of years earlier than any accepted record. In his earliest public statements he linked the site to Mu, the lost Pacific continent of nineteenth-century speculation, a claim that mainstream archaeology treats as fiction and that did his argument no favours with colleagues.
The story travelled. Television crews came, then popular authors, and Yonaguni entered the standard catalogue of lost-civilisation sites alongside the ones that are genuinely made of ruins. The island gained a dive industry. Almost none of the visitors were archaeologists.
In 1997 and 1998 the geologist Robert Schoch of Boston University dived the formation himself. Schoch is not a conventional sceptic; he had already made his name arguing for a much older date for the Great Sphinx, which put him at odds with Egyptologists. On Yonaguni he came up with the opposite verdict to the one his reputation might have predicted. The structure, he concluded, is most likely natural. What looks like masonry is bedding planes and vertical jointing in a layered sandstone that has been shaken by earthquakes and scoured by current. He noted that the steps are not uniform, that the terraces continue into the surrounding bedrock rather than sitting on it, and that nothing there requires a builder. He did leave a door open: a natural formation that people later used, cleaned up or modified is possible, and he has said so repeatedly.
Others reached the same place. In 2001 the German geologist Wolf Wichmann dived the site for the magazine Der Spiegel and reported finding no evidence of tool marks or quarrying. Patrick Nunn, a geologist who has spent his career on Pacific coastal change, has described the formations as purely natural. A 2019 study by Ogata and colleagues attributed the shapes to weathering and erosional processes acting on the local geology.
Then, in 2007, Kimura moved. He revised his dating from ten thousand years to roughly two to three thousand, and suggested the site was the work of a known culture rather than an unknown one, sinking not because the sea rose but because the land dropped in an earthquake. The revision made the archaeology far more plausible and the geology far harder, because it now required a specific and large tectonic event within recorded prehistory to put a coastal site twenty-six metres under.
What has never happened is the thing that would settle it. The Japanese Agency for Cultural Affairs does not list the formation as a cultural property. Okinawa Prefecture has conducted no research or preservation work there. No excavation has been carried out, no systematic underwater archaeological survey has been published, and in forty years not one artefact has been recovered from the site: no tool, no pottery, no worked fragment, nothing that a quarry or a settlement normally leaves behind. There is also no local tradition, no story on Yonaguni or in the wider Ryukyu chain, about a city that went under.
Conclusions and Open Questions
Three readings of the same rock have been on the table for nearly four decades.
The first is that the formation is entirely natural. Its strength is that the mechanism is ordinary and observable: layered sandstone in a seismic zone fractures into rectangles, and comparable step-like structures occur above water elsewhere on Yonaguni and along that coast, where nobody claims a builder. It also explains the total absence of artefacts without needing to explain it away. Its weakness is that natural does not mean self-evident. Very few of the geologists who have dismissed the site have published a detailed survey of it, and the case is often made by analogy rather than by measurement of these specific surfaces.
The second reading is Kimura's: the formation is substantially artificial, quarried and shaped by people whose coastline later went under. Its strength is that it is the only position built on sustained fieldwork. Kimura dived the site over decades and produced maps, and his identification of channels, sockets and paired blocks is at least a set of testable claims. Its weakness is severe. He has revised his own date by an order of magnitude, his early appeal to the lost continent of Mu damaged his standing, and after decades of diving nothing portable has come up. Faces and animals seen in eroded rock are also the oldest trap in the field, and the human eye finds them reliably in stone that no one carved.
The third is Schoch's middle position, and it is the one hardest either to prove or to dismiss: a natural formation that human beings found, used and altered. Its strength is that the two sides are not actually arguing about the same claim. Bedding planes can be real and someone can still have cut a step into them. Cultures around the world have adapted natural rock rather than building from nothing. Its weakness is that it is close to unfalsifiable underwater. Tool marks on sandstone that has spent ten thousand years in a current-swept channel are unlikely to survive, so absence of evidence proves very little in either direction.
What remains genuinely open is smaller than the popular version of the mystery and more interesting. Nobody has published a high-resolution photogrammetric model of the whole formation, the kind of survey that would show whether the terraces are continuous with the bedrock everywhere or interrupted anywhere. Nobody has cored the sediment at its base to establish when it was last exposed. No one has explained why a site claimed as an ancient quarry or town has produced no debris field, and no one has explained on the other side why so little formal geological work has been published on a feature that has been argued about internationally since 1986.
The most likely answer remains the plain one: that the sea off Yonaguni has done something that looks, to an animal built to recognise architecture, exactly like architecture. But the disagreement is not between science and fantasy. It is between two credentialled geologists who dived the same rock and came up with different answers, and it has stayed unresolved for forty years mainly because the work that would resolve it has never been funded.