The Machine in the Wreck

In 1901 sponge divers sheltering from a storm off the island of Antikythera found a Roman-era shipwreck. Among the bronzes and statuary they brought up was an unremarkable corroded lump, which was catalogued and left largely alone while the more obviously valuable material was studied.

It split. Inside were gear wheels with triangular teeth, cut to a precision that had no business existing in the second century before the common era.

The object is now understood, after a century of work culminating in x-ray computed tomography of the fragments published in 2006, to be a geared calculating device of considerable sophistication. It carried at least thirty bronze gears, and probably more in the portions that have not survived. It modelled the positions of the sun and moon, tracked the Metonic nineteen-year lunar cycle, displayed the four-year cycle of the panhellenic games, and predicted eclipses using the Saros cycle.

Two details raise it above the level of an impressive curiosity. The first is that it modelled the moon's variable speed across the sky, the anomaly later described by Hipparchus, using a pin-and-slot arrangement between two gears mounted on slightly offset axes. That is a mechanical solution to a problem in observational astronomy, executed in bronze. The second is that the workmanship shows no sign of being a first attempt.

That second point is the one that matters and it is the one usually skipped. Sophisticated mechanisms do not appear fully formed. A device of this complexity implies a tradition: predecessors, failures, an accumulated body of technique, and craftsmen trained by other craftsmen. None of that tradition survives. We have one instrument, badly damaged, from a shipwreck.

Nothing of comparable gearing complexity is known in the surviving European record until the astronomical clocks of the fourteenth century. That is a gap of well over a thousand years, and it is a gap in what survives rather than a demonstrated gap in what existed. Cicero describes a device made by Archimedes that showed the motions of the sun, moon and planets, and another owned by Posidonius. Those passages were read for centuries as literary exaggeration. They are now read differently.

The lesson here is not that the ancients possessed hidden technology in the sense the phrase usually carries. It is about the reliability of arguments from absence. Before 1901 the confident position, held by competent people on the available evidence, was that the Greeks did not build complex geared mechanisms. That position was not unreasonable. It was simply wrong, and it was overturned by a single object that survived by accident because a ship sank in the right place.

Which raises a question about method rather than about antiquity. How many other confident negatives currently rest on the same footing: not on evidence of absence, but on the absence of evidence, in a domain where the survival rate of physical objects across two thousand years is close to zero.

Bronze was routinely melted down and reused. Papyrus rots outside the dry conditions of Egypt. The survival of any given ancient object is close to a lottery, and we are reasoning about the capabilities of entire civilisations from the winning tickets.

One machine, in one wreck, moved the accepted date for a whole class of technology by more than a millennium. It is worth holding the current dates loosely.