HISTORY OF AUTOMATION • ANCIENT ENGINEERING

Ancient Water Mills: Water Power and Early Mechanical Grinding

A wheel in a stream, a pair of stones, and a miller who still had to show up

Grinding grain is dull, heavy, and daily. For a long time a person did it at a quern, or an animal walked in circles and turned the stone. An ancient water mill moves that turn onto a wheel the stream pushes. The river does not get tired. That is the real change, and it is easy to oversell.

The mill still has to be built, fed, and repaired. The stone still has to be dressed. If the water falls off, or a neighbor takes the race, the wheel stops. People kept using hand mills and donkey mills in the same centuries. Water power was a choice where the stream, the money, and the grain all met. It was not a switch that turned antiquity into a factory age.

If a title on this site calls the water mill the first automated industrial machine, do not believe the title. A mill is powered machinery. It is not a machine that runs itself, and it is not an industry in the later sense of that word. I want the mechanism, and I want the limit, in the same paragraph.

Illustration of a water mill using flowing water to turn grinding machinery
A teaching drawing of one kind of wheel. Undershot, overshot, and horizontal wheels are different machines. This picture is not a plan of Barbegal.

Two wheels, not one invention

There is no single ancient mill. The kind Vitruvius describes, late in the first century BCE, sits in a river. Boards on a vertical wheel are struck by the current. On the end of that axle is a toothed drum, standing upright. It turns a smaller toothed drum lying flat, and that flat drum turns the millstone. A hopper drops grain in. Meal comes out the same revolution. That is a right-angle gear: the wheel spins one way, the stone another.

The other common kind uses a horizontal wheel, lying in the race. The shaft goes straight up to the stone. No gear is required. It is simpler, and it is not, on that account, the older machine. Örjan Wikander’s surveys of the finds have more secure early evidence for the vertical wheel than for the horizontal one. Simpler on a napkin is not the same as earlier in the ground.

Among vertical wheels, the water can hit the bottom or fall on the top. An undershot wheel can sit in a current, which is the arrangement Vitruvius is talking about. An overshot wheel needs a head of water and a flume, and it uses the weight of the water as well as the push. Which one a builder chose depended on the fall he had, not on a chart of “efficiency” hung in a Roman shop. I will not give you a percentage. A wheel in a dry August is a wheel in a dry August, whatever the diagram says.

The split to keep: the stream supplies the turn. A person supplies the grain, the stone setting, and the decision to open or shut the water. That second part is the miller’s job. It is not a feedback loop.

What has to be true before the stone turns

A mill is a ditch as much as it is a wheel. Somebody has to bring the water to the wheel and let the spent water leave. A weir, a race, and a gate do that. Lift the gate by hand and you have changed the push. That is a person regulating a flow. It is not a controller comparing a measurement to a setpoint.

THE CHAIN, IN SHOP LANGUAGE

Water in the race, then the wheel, then the gear or the shaft, then the stone

Later machines swap the first step. Steam, then electricity, push instead of the stream. The rest of the question stays: get the motion to the tool, and have a person who knows when the tool is lying. A motor does not retire that person. It moves him.

Grain goes into a hopper. The gap between the stones decides whether you get meal or heated grit. Stones wear out of true and have to be dressed. Axles, cogs, and buckets rot because they live in water. None of that is in the sentence “the mill ran by itself.”

When we can first prove it, and when we cannot

Three writings from around the end of the first century BCE are the earliest clear literary cluster. Vitruvius gives the geared undershot mill. Strabo says that at Cabeira in Pontus the palace of Mithridates was built, and the water mill with it. One mill, not a row of them. A poem in the Greek Anthology, usually given to Antipater of Thessalonica, is read by many scholars as women at a mill being relieved by the nymphs of the water. Read together, they put the machine in the eastern Mediterranean by that date. They do not name an inventor.

M. J. T. Lewis, in Millstone and Hammer (1997), argues for something earlier: a vertical mill in Alexandria in the 240s BCE, and a horizontal mill a little before that. It is a serious argument from texts and mechanisms. It is not a wheel we have dug up from those years. Wikander, who has lived with this evidence longer than most, treats the whole early history as a working hypothesis that new finds keep revising. I will stand on the late first century BCE texts. I will not sell Lewis’s date as a discovery.

How common the mills became is a different argument, and Andrew Wilson has made it twice. In the Journal of Roman Studies in 2002 he set water power inside a wider case: the Roman economy used machines more than an older generation of historians allowed. In a 2007 paper for the Oxford Roman Economy Project he counted sites. More mills per century show up in the first through fourth centuries CE than in the fifth through ninth, with the thickest evidence in the second and third. That is uptake. It is not an industrial revolution, and it does not mean the hand quern or the donkey mill disappeared. The hourglass mills in the bakeries buried at Pompeii in 79 CE were still turned by animals walking in a circle.

Barbegal is sixteen wheels. It is not a factory that never stopped

Near Arles, in the second century CE, Roman builders set two rows of eight wheels down a slope and fed them from an aqueduct. The wood is gone. What survives is the masonry and, in the museum at Arles, pieces of the lime crust that grew on the wood while the wheels turned. Sixteen wheels on one channel is a rare concentration. Call it that. The headline “the first industrial complex in Europe” is the researchers’ frame. It is not a second body of evidence.

The 2018 study of those crusts, by Gül Sürmelihindi, Cees Passchier, and their colleagues, is the part I do not want skipped. Stable isotopes in the carbonate show the water was shut off, regularly, for several months at a stretch, often in the late summer and autumn. The authors say this cuts against the old idea that Barbegal ran steadily to feed a city. They suggest, as a reading, that the mills may have made a hard biscuit for the nearby ports, something that could sit in a warehouse. “May have” is their word, and it should stay their word. We do not have the shipment list.

That same paper reads the crusts as a maintenance record. The carbonate was left on the wood during use, and the authors judge that the wooden parts were replaced about every five to ten years, from mold and rot. That is their reading of the deposits. It is not a workshop log, and a later study of the same kind of crust has offered a different span. I will not average the two numbers into a third one nobody published.

A 2020 reconstruction of the water path, led by Passchier, argues for overshot wheels in pits 1.1 meters wide, with a flume shaped to throw the stream onto the wheel and to soften the surge when an upstream wheel was stopped. That flume, if the reconstruction holds, is a passive way to keep the next wheel from being starved or flooded. It is not a sensor. Wood that has to be renewed on a cycle of years means a crew. The flume did not replace one.

Why Barbegal still matters: somebody stacked sixteen machines on one managed supply. That is organized power. The isotope record says the supply was not continuous. Both sentences are the site. A daily tonnage you may have read online is not in this account, because the steady-output figure does not survive the shutdowns.

The mill takes away one kind of work and hands you another

You no longer have to push the stone for every turn. You do have to keep grain in the hopper, listen for a stone that is glazing, open a gate when the mill above you changes the flow, and dress the faces when the grind goes coarse. Wheels, buckets, and wooden cogs live wet and die wet.

That trade is the pattern in the rest of this history. A later machine takes the repetitive push off a person’s back and puts a different job in his hands: setup, judgment, and repair. Anyone who tells you the mill “eliminated labor” has not stood in a mill race. It moved the labor, and it left the places without a good stream on the old quern.

What a plant inherited, and what it did not

A modern line still has to answer the mill’s questions. Where does the push come from? How does it reach the tool? Who notices when the tool is wrong? Electricity and a drive replace the race. A sensor and a controller can now compare the result with what was asked and change the push. The Roman wheel could not.

I am describing equipment the trade uses now, not a rule that every shop owns it. The kinship is the powered, repeating motion. The break is the closed loop. Barbegal’s flume may have softened a surge. It did not measure the flour and correct itself. Do not promote a waterwheel into a PLC because sixteen of them sat on one hill.

What I want you to keep: Vitruvius, late first century BCE, describes a geared undershot mill. That is a text, not a birth certificate. Strabo names one mill at Cabeira. Lewis’s earlier date is an argument. Wilson’s counts show Roman mills were more common than the old textbooks said, and still not an industrial revolution. Barbegal had sixteen wheels, regular months of shutdown, and a five-to-ten-year wood reading that belongs to the 2018 paper. A gate you lift by hand is not automatic control.

The long use, without the straight line

Water kept turning stones, and later hammers and saws, for many centuries after Rome. Those later mills are their own history. They are not proof that the Roman wheel was waiting to become a steam engine. Steam replaced a limit of the stream: you could put the power where the fuel was, not only where the fall was. Electricity then moved the power farther still. Digital control came after that. Each step answered a problem the previous machine actually had. None of them was hiding inside Vitruvius’s hopper.

The mill earns its place as the earliest widely attested way we can show for taking the repeating heavy turn of grinding off muscle and putting it on a stream, in a machine a crew could maintain. Water-lifting wheels are a different job, and an older one. The grain mill does not earn the title “automated industry.” The stream does not care what we call it. It only turns what the race will feed.

References & Further Reading

  1. Vitruvius, On Architecture 10.5.2. The geared vertical wheel driven by a current, the hopper, and the meal. Late first century BCE. A description of a type, not a census.
  2. Strabo, Geography 12.3.30. At Cabeira, the palace of Mithridates and the water mill. Singular in the text. With Vitruvius and the epigram usually assigned to Antipater of Thessalonica, Greek Anthology 9.418, the late first-century BCE cluster. The poem is read as a mill by many scholars; it is not a technical drawing.
  3. M. J. T. Lewis, Millstone and Hammer: The Origins of Water Power (Hull: University of Hull Press, 1997). The argument for a third-century BCE origin. Cited as an argument, not as an excavated wheel.
  4. Örjan Wikander, “The Water-Mill,” in Handbook of Ancient Water Technology, ed. Örjan Wikander, Technology and Change in History 2 (Leiden: Brill, 2000), and his chapter in John Peter Oleson, ed., The Oxford Handbook of Engineering and Technology in the Classical World (Oxford: Oxford University Press, 2008). Vertical and horizontal wheels, and the warning that the early history is a working hypothesis.
  5. Andrew Wilson, “Machines, Power and the Ancient Economy,” Journal of Roman Studies 92 (2002): 1–32, Cambridge Core. Roman use of machines was greater than the older primitivist picture. Not an industrial revolution.
  6. Andrew Wilson, “The Uptake of Mechanical Technology in the Ancient World: The Water-Mill,” Oxford Roman Economy Project working paper, 28 February 2007. More mills per century in the first through fourth centuries CE than in the fifth through ninth.
  7. Gül Sürmelihindi, Cees W. Passchier, et al., “The Second Century CE Roman Watermills of Barbegal: Unraveling the Enigma of One of the Oldest Industrial Complexes,” Science Advances 4 (2018), doi:10.1126/sciadv.aar3620. Sixteen wheels; carbonate crusts; regular multi-month shutdowns; their reading that wooden parts were replaced about every five to ten years. This paper is in Science Advances, not Scientific Reports. The hardtack reading is the authors’ suggestion. A later study gives a different replacement span; the two are not merged here.
  8. Cees W. Passchier, Marcel Bourgeois, et al., “Reconstructing the Hydraulics of the World’s First Industrial Complex, the Second Century CE Barbegal Watermills, France,” Scientific Reports 10 (2020): 17917, doi:10.1038/s41598-020-74900-5. Two rows of eight, pits 1.1 meters wide, and the flume reconstruction. Their title’s “first industrial complex” is not adopted here as a finding.

No daily flour tonnage is given. The isotope record of shutdowns will not carry one. A page title that says “automated” is a title, not a description of the machine.