The Antikythera Mechanism

The gearing is understood in its main lines. An input turned a crown wheel that drove the main train; the front dial showed the sun and moon in the zodiac and the moon's phase on a rotating ball; a pin-and-slot coupling on epicyclic gears varied the moon's speed according to Hipparchus's lunar theory, the earliest known mechanical model of an astronomical anomaly⁠[4]. The back had two spiral dials: an upper one of 235 months for the Metonic cycle, with a subsidiary games dial and a Callippic dial, and a lower one of 223 months for the Saros cycle, with glyphs predicting lunar and solar eclipses and their times, and a subsidiary exeligmos dial⁠[4]⁠[5]. The Metonic month names belong to the Corinthian family of calendars, which pointed the 2008 authors toward Corinth's colonies and Iversen's 2017 study toward Epirus⁠[5]⁠[14]. The back-cover inscription names the planets and their displays, which the 2021 reconstruction used to propose a front display of all five with their synodic cycles, a reconstruction that is mechanically plausible but not attested by surviving gears⁠[6].

On dating, Carman and Evans showed in 2014 that the eclipse table's starting point is best explained by a full-moon epoch in 205 BC, which puts the astronomical scheme earlier than the 150 to 100 BC range assumed in 2006; Jones's 2020 study confirms the eclipse-sequence dating, reaffirms the calendrical epoch and argues that the design decisions could still have been made by a maker working as late as the shipwreck⁠[12]⁠[13]. The wreck itself is placed in the first century BC by its cargo⁠[1].

The calendar ring. The front dial's outer ring carried the 12 Egyptian months and could be rotated to compensate for the Egyptian year's drift against the seasons; it was fixed by pins in a ring of holes beneath. The 2006 and later reconstructions assumed 365 holes. In 2020 Budiselic, Thoeni, Dubno and Ramsey measured the 81 surviving holes in the CT data and, from their spacing, estimated a full count near 354, a lunar year⁠[7]. In 2024 Woan and Bayley applied Bayesian techniques from gravitational-wave astronomy to the same coordinates and found 354 or 355 holes, with 360 strongly disfavoured and 365 not plausible under their model⁠[8]. Dickens and Malin, re-analysing the same data with least squares and criticising the 2020 statistics, found 352.3 plus or minus 1.5 and put the chance of 365 below one in ten thousand⁠[9]. Specialists objected that the mechanism already had a lunar calendar on the Metonic dial, which would make a second, cruder one redundant⁠[16]. In 2025 Voulgaris, Mouratidis and Vossinakis published a reconstruction of the front dial that keeps 365 holes and argues that any other count would make the Egyptian ring malfunction⁠[10]. The dispute turns on the statistics of 81 hole positions in corroded metal and has not been resolved.

Whether it worked. Szigety and Arenas built a computational model of the gear train with triangular teeth and the manufacturing errors measured by Edmunds in 2011, and found that the train jammed in 90 per cent of simulations, typically before the solar pointer completed a third of a turn; they conclude that either the device never functioned or the true errors were smaller than measured, and note that corrosion has deformed the gears the measurements rest on⁠[11]⁠[17]. The 2018 reprocessing of the Fragment A scan tested and rejected an earlier suggestion that the fragment had moved during scanning⁠[15].