The Antikythera Mechanism

Start here, because it is the single most common error made about this object. The Antikythera mechanism is a set of corroded bronze fragments in the National Archaeological Museum in Athens. The museum's own gallery label states the count precisely: seven large fragments, labelled A to G, and 75 minor pieces numbered 1 to 75, made of bronze, and it adds that it is not certain that all 75 small fragments belong to the mechanism at all. Seven and seventy-five is the eighty-two that is always quoted, and the museum's caveat about the seventy-five is almost never quoted with it.

What circulates instead are reconstructions. There are working models in Athens, in Thessaloniki, in Bozeman, in Sao Paulo, in Kassel; there are Lego versions, transparent acrylic versions, brass versions turned by a crank. They are photographed constantly and captioned as the Antikythera mechanism. They are hypotheses in metal, and several of them embody gear trains that later research has abandoned.

The distinction matters more here than for almost any object in this archive, because the reconstructions disagree with each other and with the fragments. A reader looking at a polished brass model with a hand crank is looking at an argument, not at evidence.

The National Archaeological Museum's own account, printed on the gallery wall beside the finds, is worth following because it is the holding institution's position rather than a retelling. About Easter of 1900, near the island of Antikythera, sponge-fishers from Symi located by chance a very important ancient shipwreck. They were on their way home from a voyage to the north coast of Africa and attempted to anchor on the east coast of the island, near the harbour of Potamos. The wreck lay a mere 25 metres from the coast, at a depth of about 50 metres. The work of raising it began at the end of the same year, under difficult conditions, and lasted for many months.

The captain was Dimitrios Kontos and the diver usually named as the first to see the site is Elias Stadiatis. The salvage was carried out by the sponge divers themselves, in standard diving dress, working at a depth that was at the outer limit of what that equipment allowed and well inside the range at which decompression illness kills. The Hellenic Royal Navy supplied a vessel and supervision; a Ministry of Naval Affairs document of 26 February 1901 concerning the despatch of the vessel Mykali to Kythera for the Antikythera works survives and is reproduced by the museum.

The first object retrieved from the bottom of the sea, according to the same gallery text, was the right arm of a bronze statue of the third century BC depicting a philosopher. Nobody was looking for a machine.

The mechanism came up with everything else and went into the museum as corroded bronze. It was recognised on 17 May 1902, when Valerios Stais, then at the National Archaeological Museum, saw a gear wheel in one of the lumps. His cousin Spyridon Stais is named alongside him in some accounts.

The two-year gap is the most instructive fact in the object's modern history. It was not found by an archaeologist, it was not recognised on the seabed, and it survived the salvage, the transport and two years of museum storage as an unremarkable mass of corrosion. The bronze statues from the same wreck were understood immediately, because a statue looks like a statue. The single most important object recovered from any ancient wreck sat unread because nobody had a category for it.

The museum's gallery text is the conservative statement and is worth taking as the baseline: the mechanism contained at least 30 gears, dials, scales, axles and pointers; the Greek inscriptions on the surface of many fragments refer to astronomical and calendar calculations and to instructions for its use; and it was contained in a wooden-framed case, probably with bronze plates on its front and back.

Beyond that baseline, and on the strength of the imaging campaigns rather than of the naked eye, the readings that have gained acceptance are these. The front carried a zodiac and calendar display showing the positions of the sun and moon. The back carried two large spiral dials, one running on the 19-year Metonic cycle that reconciles lunar months with solar years, the other on the Saros eclipse cycle of 223 lunar months, with subsidiary dials for the Callippic and Exeligmos corrections. A dial giving the four-year cycle of the games, including the Olympiad, was identified in 2008. The month names on the Metonic dial are Corinthian.

The lunar display incorporates a pin-and-slot arrangement that reproduces the variable speed of the moon around its orbit, which is a mechanical solution to a problem in observational astronomy and is the detail that most persuaded specialists that this is an instrument built by somebody who understood the theory rather than a curiosity.

In 2024, University of Glasgow physicists Graham Woan and Joseph Bayley applied statistical methods developed for gravitational-wave detection to the calendar ring's surviving hole positions, most of which are damaged or missing outright. Their Markov Chain Monte Carlo analysis put the original hole count at 354 or 355 (68% credible interval 354.08 to 355.24), matching the 354-day lunar year rather than the 365-day solar year some earlier reconstructions assumed, with the holes spaced to an average radial precision of 0.028mm. Published in the Horological Journal in July 2024, it is a statistical inference from damaged physical evidence, not a new fragment, and it corroborates rather than overturns the reading above.

It is worth being precise about the claim that this object is a computer or a clock, because both words are doing work they should not.

It has no escapement and it keeps no time. It is turned by hand, and what it computes it computes as a function of how far the input has been turned. It models cycles, not hours. The proper description is a geared calculating instrument that mechanises astronomical cycles, and the ancient category it belongs to is the sphere or planetarium that Cicero describes rather than anything horological.

Its importance to the history of technology rests on the gearing itself: differential and epicyclic arrangements, triangular teeth cut by hand, a largest wheel of about 13 centimetres, and a level of miniaturisation that has no surviving parallel until medieval Islamic geared calendars and European clockwork. Whether it was a unique object or the survivor of a tradition of such instruments is unresolved and is one of the real open questions about it. The Fragment G inscription analysis has been read as showing that the maker built the instrument for other people to use, perhaps students, astronomers or wealthy individuals, and not only for himself, which is an argument for a tradition rather than a one-off.

Albert Rehm, a German philologist, was the first to propose that the object was an astronomical calculator, and in 1905 he read the word sterigmos, stationary point, on a small piece later reassembled into Fragment G. Stationary points are where a planet appears to reverse direction, so the word is direct evidence for planetary modelling, read in 1905 and not built into any reconstruction for decades.

Derek J. de Solla Price, a historian of science at Yale, became interested in 1951 and published a popular account, An Ancient Greek Computer, in 1959. His decisive step was not textual. In 1971 he and the Greek nuclear physicist Charalampos Karakalos made X-ray and gamma-ray images of the fragments, and it was those images that first showed the gear counts inside the corrosion. Price published the results in 1974.

The pattern holds for every subsequent advance. What moves the reading forward is a new way of seeing inside the metal, not a new excavation, because the object has been complete, in the sense of being all that survives, since 1901.

The campaign that produced the current reading was mounted in 2005 by a collaboration of Cardiff University, the National Archaeological Museum, the University of Athens and the Aristotle University of Thessaloniki with two industrial partners.

X-Tek Systems built a 400 kilovolt microfocus computed tomography machine, weighing around eight tonnes, and moved it to Athens, because the fragments could not travel. Hewlett-Packard supplied a reflectance imaging dome implementing polynomial texture mapping, which lights a surface from many directions in turn and combines the images so that a surface can afterwards be relit at will, bringing up inscriptions that are invisible under any single fixed light.

The results were published in Nature in 2006 and again in 2008. The 2006 paper reported the gear train and functions; the 2008 paper reported the Olympiad dial and the Corinthian month names. The 2016 publication of the inscriptions and a 2021 model of the front dial by the University College London team continued the work. The imaging data have supported readings of 1,380 letters on Fragment G alone.

For a reader wanting to know how far the object has been made available: this is one of the best-documented objects in the ancient world in terms of published technical imaging, and among the worst-documented in terms of freely licensed photography of the fragments themselves.

Three positions are in circulation and each rests on a different kind of evidence.

The National Archaeological Museum's own gallery label dates it to about 150 to 100 BC. That is the institutional position and it is the one this record leads with.

A date of 205 BC has been argued from the eclipse predictions on the Saros dial, on the reasoning that the sequence has a start point and that the start point can be calculated. If that is right, the instrument was made before Hipparchus did the work usually credited with making such modelling possible, which has consequences for the history of Greek astronomy well beyond this object.

The wreck itself is dated by its cargo and coins to the second quarter of the first century BC. A ship's date is a terminus for when the object went to the seabed, not for when it was made, and an instrument of this cost would plausibly have been decades old when it was loaded. The dating dispute is therefore not resolvable from the wreck, and this record does not resolve it.

As a provenance chain this is a short document and an unusually clean one, and it is worth setting out because it is the opposite of the usual antiquities problem.

There is no collector, no dealer, no export licence and no restitution claim. The object was recovered inside Greek territorial waters by Greek divers, under the supervision of the Greek navy, and entered the Greek national collection immediately. It has never been on the market, has never been offered at auction, and there is no price of any kind in its history. The Ministry of Naval Affairs document of 26 February 1901 is a piece of the state's own paperwork for the operation.

What is not documented is everything before 1901. Who made it, who owned it, who was shipping it and where to are all unknown. The cargo is generally read as a shipment of Greek goods travelling west, probably from Delos or from a city on the Asia Minor coast, towards Italy, and the object was therefore probably in transit as property when it sank. Whether it was cargo, or a passenger's instrument, or ship's equipment, is not established.

The divers were, by most accounts, compensated by the Greek state for the finds. The terms and the sums are not established in the sources consulted here, and at least one diver died and others were paralysed in the course of the salvage. A record that lists the fragments and omits that has left out the price actually paid for them.

The fragments are on display in the bronze collection of the National Archaeological Museum in Athens, with the mechanism catalogued on the museum's own gallery label as inventory 15087, alongside other bronzes from the same wreck: the lyre, 15104, and parts of couches, 15098, 15099 and 15101. The marble statues from the wreck are displayed separately in the museum atrium.

The wreck site itself has been worked again. Jacques Cousteau dived it in the 1950s and again in 1976, and a series of modern campaigns from 2012 has recovered further material including human skeletal remains, which is exceptional for an ancient wreck and opens the possibility of identifying the people aboard. No further piece of the mechanism has been recovered in any modern campaign, which is itself a finding: the 82 fragments are very probably all there is.

The reading of the object is not closed. Every major imaging campaign since 1971 has changed the reconstruction, and the 2021 model of the front dial is a proposal rather than a settlement.