The Water Thief and the Uneven Hour

AThe problem that water clocks were built to solve was not telling the time but dividing the dark. A sundial fails at dusk, and for temples that owed particular rites to particular hours of the night, the failure mattered. The oldest surviving answer is a flared alabaster vessel recovered at Karnak and dated to roughly 1400 BCE, which drains through a small hole near its base while twelve columns of marks cut inside the wall record how far the water should have fallen. Greek writers later called such vessels klepsydrai, water thieves. The name has done quiet damage. It invites the reader to imagine an instrument whose business was to steal away equal portions of time, which is what a clock does for us; whether that is what the vessel did for its owners is precisely what the past decade of experimental work has reopened.

BThe physics is unforgiving. Water leaves an orifice more slowly as the level above it drops, so a straight-sided container empties at a falling rate and its hours grow longer as the night wears on. The Karnak vessel is not straight-sided. Its wall leans outwards at roughly 105 degrees to the base, so that the surface shrinks as the level falls and the level therefore falls faster, offsetting the loss of pressure. Marisa Venn, who cast fourteen replicas in gypsum and ran them through 340 timed drainings between 2019 and 2023, reports that the sloping profile held the error across a twelve-hour night to under six minutes, whereas straight-sided copies of the same capacity were out by nearly forty minutes. A shape that difficult to hit by accident, she argues, records an intention: someone tested, adjusted and tested again.

CTobias Grell accepts the measurements and rejects the inference. Potters, he points out, lean a wall outwards for reasons that have nothing to do with hydraulics, since it lets a mould be withdrawn cleanly and it is simply what a scaled-up drinking bowl looks like. Accuracy, on this reading, is a windfall rather than a target. His stronger evidence is the marking itself. The twelve columns inside the Karnak vessel are cut to twelve different depths, one for each month, because an Egyptian night hour was a twelfth part of an actual night and so ran from about fifty minutes in June to about seventy in December. A device calibrated to a quantity that changes with the season, Grell writes, is not an instrument for measuring equal intervals; it is a machine for finding your place in a ritual sequence, and its makers would not have thanked us for the compliment we keep paying them.

DSomething did change in Hellenistic Alexandria. Ctesibius, working around 270 BCE, inverted the arrangement: water ran into a receiving vessel carrying a float, and a pointer on the float rose against a scale. Inflow can be regulated in a way that outflow cannot, and the surviving descriptions credit him with a conical valve that admitted more water as the level of the supply fell, so that the feed stayed constant. The Athenian courts had already put plain outflow vessels to a use that needed no such refinement: an advocate spoke while a measured jar emptied, and fell silent once it was dry. Even there the arrangement was narrower than it sounds, because the flow was stopped by a clerk whenever a law or a deposition was being read aloud, so what the jar limited was argument rather than proceedings.

EThe most accurate water clocks were built where the tradition ran longest. Chinese engineers abandoned the shaped wall for a different remedy, dividing the supply into a cascade of tanks, each of which held the one below it at a constant level. The tower completed by Su Song in 1088 CE coupled such a supply to a wheel released one scoop at a time, an escapement in everything but name, three centuries before the earliest European mechanical clock. Al-Jazari's water-driven devices of 1206 CE went further in showmanship, if not in accuracy. It is tempting to read this sequence as a slow climb towards precision, but the Chinese instruments stood inside imperial observatories whose business was to keep the calendar in the hands of the state; better timekeeping was one means to that end, and the end was not itself a technical one.

FWhere the two readings can be tested against each other is in the residues. Venn's team found calcite scaling inside three temple vessels thick enough to have closed a quarter of the orifice, and she notes, awkwardly for her own case, that a vessel in that condition would keep no better time than the straight-sided copies she uses as a control, so the fine shaping of the wall was plainly not being maintained. Grell concedes in turn that two orifices had been reworked to a wider bore, a change visible under magnification and hard to explain unless somebody was dissatisfied with a rate. Both concessions point the same way. The Egyptians were capable of caring about a rate without holding the modern conviction that an hour must be the same length in June as in December.