Water Clocks: Measuring Time Through Flow

Why use water at all?

People needed a stretch of time they could share: a watch of the night, a ritual, a speech, a stretch of work. Sunlight will not give you that after sunset. A vessel with a small opening will, if you are willing to fill it and trust the reading only as far as the instrument deserves.

The simple outflow clock is easy to picture and easy to get wrong. Water leaves through a hole near the bottom. As the bowl empties, there is less weight of water pressing on that hole, so the flow slows. A straight-sided tank does not lose height at a steady rate. Makers answered that in more than one way: a vessel that narrows toward the base, a supply held at a steadier level, or marks spaced to match the way that particular bowl actually drained. None of those fixes is a computer. All of them are somebody noticing that the water will not behave unless the instrument is built for it.

The Karnak bowl is an object, not a birthday

The earliest Egyptian outflow water clock that still exists is a calcite vessel from the reign of Amenhotep III, in the 18th Dynasty. Georges Legrain found it in pieces in 1904, in a burial cachette in the court of the seventh pylon at Karnak. It was restored. The McMaster Ancient Egyptian Astronomy catalogue, which tracks these objects, calls it the earliest surviving example of its kind and lists it as JE 37525. As of that catalogue’s 2024 note, it is at the National Museum of Egyptian Civilization in Cairo. The Science Museum in London holds a plaster cast made in 1922, not the original.

Inside are twelve vertical scales, one for each month named on the rim. Egyptian “hours” of the night were not our equal hours. Daylight and darkness were each split into twelve parts, so an hour grew and shrank with the season. A monthly scale is there so the reader can use the scale that matches the time of year. The Science Museum describes the bowl as shaped so the water level would fall at a more even rate, and says it was probably used for the hours of the night. That is a reading of the object’s form and its scales. It is not a laboratory slip saying the clock kept modern minutes.

Egyptologists do not all print the same first and last year for Amenhotep III. I will not pretend a two-year argument is settled by picking one pair. The reign is the date we have. What we do not have is the name of a first inventor, or a year when water timekeeping “began.” This bowl is the oldest Egyptian outflow clock we can put a hand on. Older practice may have existed. This object does not prove the day it started.

Ctesibius, as far as the writing goes

Ctesibius worked in Alexandria. Later technical writers put him in the third century BCE. We do not have his book. What we have is Vitruvius, On Architecture, book 9, chapter 8, written in the Roman period, long after the man was dead.

Vitruvius says Ctesibius made the outlet in gold, or drilled it through a gem, because those are not worn away by water and do not collect dirt and choke. Water passing that opening raises an inverted bowl — the float. A toothed bar on the float turns a drum, and further gears can move figures or trip little displays. He also describes two cones, one solid and one hollow, fitted so they can be drawn together or apart and the inflow made stronger or weaker.

That is a real description of a more elaborate clock: a durable orifice, a float, geared indication, and a way to set the feed. It is not a patent, and it is not proof that Ctesibius invented every water clock. Simple outflow bowls were already old by his century. Some modern engineering papers call his regulating float “the invention of feedback control.” That is their label, laid on a text that never uses it. I will not teach it as an ancient fact. If you work controls, you already know the difference between a float that lifts a pointer and a loop that measures a variable and drives an output. Do not let a headline erase that difference.

How a more careful clock could be arranged

Clocks were not built from one drawing. When a later maker went past the plain bowl, the pieces tend to fall in an order like this. Treat it as a way to read the descriptions, not as a kit.

1. A supply vessel feeds the timing vessel.
2. An orifice, and sometimes a fitting that sets the feed, limits how fast the water moves.
3. The level in the timing vessel rises or falls.
4. Marks on the wall, or a float tied to a pointer, give a person something to read.
5. On some elaborate clocks, that same movement trips a figure, a sound, or another sign. Vitruvius mentions displays of that kind in the tradition he is describing. Not every clock had them.

A plain Egyptian outflow bowl stops at the marks inside the vessel. If a diagram shows gears, puppets, and a regulator all at once, ask which century and which text it came from. A lot of popular drawings have glued several centuries into one picture.

Other countries kept their own clocks

Greek and Roman writers describe water clocks. That does not make Alexandria the source of every later one. China has a long written tradition of clepsydras. Some of those books credit the Yellow Emperor, thousands of years before any clock we can date. I will not put that on a timeline. A legend in a later history is not an artifact.

What we can stand on is later and written by the maker’s own circle. Su Song, a Northern Song official, completed a treatise in 1092, the Xinyi Xiangfayao, describing a water-driven astronomical tower: an armillary sphere, a celestial globe, and jack-work for telling time. Joseph Needham’s study of that book is the place to start. The tower itself does not survive. After the capital moved in 1127 it was taken apart, and it was not successfully put back into service. So we have the book. We do not have the machine to go and inspect.

In 1206, al-Jazari finished a book of machines for his Artuqid patron. Donald Hill’s English translation, The Book of Knowledge of Ingenious Mechanical Devices, includes water clocks he designed and drew, among them the elephant clock. That is a different kind of evidence from Ctesibius. Al-Jazari’s descriptions are his. We are not depending on a Roman author writing centuries later.

Ideas did travel. They were also rebuilt for local use. Do not draw one arrow from Egypt to Greece to China to Islam and call it the history. Each of these clocks sits in its own workshop.

What this has to do with automation

If you run equipment, the familiar part is this: a physical change was made into information a person could use, and on the fancier clocks that change was made to move a sign or trip an action. Makers worried about the same practical messes you would worry about. The hole wears. The head of water changes. The reading has to be checked against something else, or it drifts.

That is the family resemblance. It stops there. These clocks did not sense a process and correct an output the way a controller does. Somebody filled them, chose the month’s scale or the setting of the cones, and read them. A waterline is not a sensor. A float that lifts a hand is not a program.

What I want you to keep

• A draining or filling vessel can stand in for time. It cannot check itself.
• As a straight tank empties, flow slows. Shape, supply, and spacing of the marks were the old answers. They were not equally good.
• The Karnak bowl is the earliest surviving Egyptian outflow clock. It is not the birth date of the idea, and it did not keep equal modern hours.
• Vitruvius, not Ctesibius, is the text we have for the Alexandrian geared clock. Do not promote that description into “he invented feedback.”
• A Yellow Emperor origin year is a tradition. Su Song’s 1092 book and al-Jazari’s 1206 book are documents. Treat them differently.
• Electronic clocks and feedback loops are a later trade. They answer a cousin of the old problem. They are not small clepsydras.

Water clocks belong in this history because people made a natural process into a reading they could share, and then spent centuries fighting the ways that reading went wrong. They do not belong here as ancient robots.