The Greek Computer at the Bottom of the Sea

In the spring of 1900, a small fleet of Greek sponge divers from the island of Symi, returning home from a working trip to the North African coast, was forced by a storm to take shelter behind the uninhabited island of Antikythera, between Crete and the Peloponnese. While waiting out the weather, one of the captains, Dimitrios Kondos, suggested they dive on the local seabed to see whether there were any sponges worth taking on the way home.

The first diver down, a Symiot named Elias Stadiatis, surfaced after about ten minutes in a state of high agitation. He reported, in some version of words, that the seabed at sixty meters was littered with the bodies of horses and beautiful women. His captain, suspecting either nitrogen narcosis or a practical joke, went down himself in the next round. He came back with a bronze arm.

The Symiots had stumbled onto the wreck of a Roman merchant ship that had sunk in the first century BC, en route from Greece or Asia Minor to Italy, carrying a substantial cargo of Hellenistic art. They reported it to the Greek government. The Hellenic Navy sent two warships to assist with salvage. Over the next ten months, divers brought up bronze and marble statues, pottery, jewelry, glassware, and the human bones of what had been the ship’s crew. Many of the artifacts went directly to the National Archaeological Museum in Athens, where they were catalogued and put on display by the museum’s director, Valerios Stais.

In May 1902, while sorting through the corroded fragments that had been initially classified as ship hardware, Stais picked up a particular bronze object — a lump roughly the size of a large book, encrusted with calcified marine deposits — and noticed, faintly visible on its surface, what appeared to be a precision gear wheel.

It would take the following one hundred and twenty years to figure out what it was.

What everyone thought it was

The standard assumption, from 1902 through the 1950s, was that the gear-bearing object — by then known as the Antikythera mechanism — was an astronomical clock of medieval origin that had somehow ended up on the ancient seabed. This was not because the wreck’s other contents looked medieval (they didn’t; they were unmistakably Hellenistic), but because no one believed that Greek artisans in the first century BC were capable of producing precision-geared machinery. The first known European clock with comparable complexity was the astronomical clock at Strasbourg Cathedral, completed in 1354. Geared mechanisms of that sophistication, the consensus held, did not exist in the ancient world.

The mechanism was put on display, briefly studied, and largely ignored. Periodic papers in the 1920s and 1930s suggested it might be authentic and Greek, but the evidence — partly visible gears, partly visible inscriptions in Greek — was not enough to overturn the consensus.

The person who first took the problem seriously was Derek de Solla Price, a British-American historian of science at Yale, who began studying the mechanism in 1951 and spent twenty-three years on it. Price, working with X-ray images from the 1970s, was the first to identify and count the gears, the first to publish a detailed reconstruction (in his 1974 paper Gears from the Greeks), and the first to argue, on the basis of the inscriptions and the gear ratios, that the mechanism was Greek, that it dated to roughly 100 BC, and that it had been used to predict the positions of the Sun, the Moon, and some of the planets.

The Price reconstruction was correct in outline and wrong in many details. He undercounted the gears (he found about thirty; the actual number is closer to thirty-seven). He misidentified some of the gear functions. He did not understand the back-face dial mechanisms, which he treated as decorative. His reconstruction was nevertheless the first proof of concept: a complex, multi-function geared machine had been built in the Greek world more than a thousand years before anything comparable in medieval Europe.

This was a substantial enough claim that Price spent the rest of his career defending it. He died in 1983 with most of his colleagues still skeptical.

What it actually does

In 2005 a team led by the British mathematician Tony Freeth and the astronomer Mike Edmunds, in collaboration with X-Tek Systems (now part of Nikon), built a custom 12-ton X-ray tomography rig and used it to take three-dimensional X-rays of the mechanism’s surviving fragments at sub-millimeter resolution. They could see, for the first time, the gear teeth that had been buried inside the corroded mass for two thousand years. They could also read inscriptions on the inner surfaces that no one had ever seen.

The 2006 Nature paper that resulted from this work — and the subsequent papers by the Antikythera Mechanism Research Project through the 2010s — settled the question of what the mechanism did, comprehensively.

It was a hand-cranked analog computer for predicting astronomy. The front face showed the position of the Sun, the Moon, and probably the five planets known to the Greeks (Mercury, Venus, Mars, Jupiter, Saturn) against a zodiac dial. The Moon’s position was shown using a clever pin-and-slot mechanism that correctly modelled the lunar orbit’s variable speed — the same kind of correction that wouldn’t appear in European clockwork for another fifteen hundred years.

The back face was even more elaborate. The upper rear dial was a 235-month Metonic cycle, used to coordinate the lunar and solar years. Inside that was a 76-year Callippic cycle. Below was a 223-month Saros cycle, used to predict solar and lunar eclipses, with engraved glyphs at the predicted eclipse months indicating the eclipse type and time. Inside the Saros was a 54-year Exeligmos cycle that adjusted the Saros for the rotation of the Earth.

And in a separate small dial, a four-year Olympiad counter that tracked the schedule of the Olympic Games. The 2008 paper that identified this counter is one of the rare scholarly publications to contain the sentence “The mechanism could be used to plan attendance at the Games.”

There is also, on the front of the device, a so-called parapegma — a star-rising calendar showing which constellations would be visible on which dates. Latitude-specific. The latitude the parapegma is calculated for is approximately 35° north — which corresponds to Rhodes, the most plausible candidate for where the mechanism was made.

Where it came from

The current best guess is that the mechanism was built on Rhodes between 150 and 100 BC, probably in the school of the astronomer and engineer Hipparchus or one of his immediate successors. The gear ratios used in the lunar mechanism match Hipparchus’s calculations of the lunar orbit. The geographic latitude of the parapegma matches Rhodes. The shipwreck dates to roughly 70-50 BC, and the cargo of art objects is consistent with a Roman buyer purchasing Greek luxury goods, possibly in Rhodes itself, possibly via an intermediary.

The mechanism was, in other words, a high-end consumer good in the late Hellenistic world. It would have been built by a specialist workshop, sold to a wealthy individual interested in astronomy, and used as a combination scientific instrument, calendar, and conversation piece. The shipwreck was probably the moment of its delivery.

There may have been others. A passage in Cicero’s De Re Publica, written around 54 BC, describes a planetary-motion device built by Archimedes a century earlier and brought back to Rome by the general Marcellus after the siege of Syracuse in 212 BC. The device, Cicero says, modelled the motions of the Sun, Moon, and planets accurately enough to predict eclipses. He saw it personally. He says it was famous. There are several other ancient references to similar devices — a passage in Pappus of Alexandria, a passage in Vitruvius — that suggest geared astronomical mechanisms were a known, if rare and expensive, category of object in the Hellenistic and early Roman world.

We have one of them. Everything else was scrapped for the bronze.

Where it is now

The surviving fragments — there are 82 of them, in three main pieces and many smaller chips — are in the National Archaeological Museum in Athens, in a case in a small side gallery. The display includes a working modern reconstruction next to the original, so visitors can see what the gear train would have done if it had been complete. A small placard explains the function.

The shipwreck site is still there, off the cliffs at Point Glyphadia on Antikythera’s northeast coast. It has been re-excavated three times — most recently by a Greek-American team in 2014–2024 — and continues to produce artifacts. In 2023 the team brought up a marble head, possibly belonging to one of the statues whose decapitated bodies the sponge divers found in 1900. The face is intact. It is, the report notes carefully, that of a beautiful woman.