The claim

Barbegal was not a mill beside an aqueduct

The familiar description—“sixteen Roman watermills”—is true but incomplete. The mills only make engineering sense when the whole water route is treated as one system: spring capture in the Alpilles, gravity aqueduct, diversion, Vallon des Arcs crossing, rock cut, headworks, two parallel mill trains, wheel basins, flumes, wheels, gearing, millstones and tailwater.

The 2020 hydraulic study describes Barbegal as a unique arrangement of sixteen wheels in two parallel rows of eight. The aqueduct reached the ridge from the north and the mill complex occupied a natural slope below it.[2]

Landscape as machine

The Romans found stored energy in a hillside

The mills sit on a limestone outcrop with a natural slope of about 17°. A roughly nine-kilometre aqueduct brought water from the Alpilles. When the mill system was created around AD 120–130, the southern branch of the Arles water network was diverted to the mills while the northern system was extended so that Arelate retained an urban supply.[1]

That decision converted elevation into a productive resource. The same water could lose height repeatedly, doing useful work at each drop. Instead of spending water once, Barbegal spent head—the gravitational potential represented by the vertical fall.

Surveying

Before the first arch, someone had to solve the heights

No surveying instrument has been recovered from Barbegal, so we should not claim that a particular chorobates or dioptra was used there. But aqueduct construction required levelling, and Roman technical evidence shows that specialists possessed the means to do it.

Vitruvius lists the dioptra, libra aquaria and chorobates for levelling watercourses. Hero of Alexandria describes a dioptra being used to determine height differences for a watercourse. Most revealingly, the second-century military engineer Nonius Datus recorded that he surveyed the Saldae aqueduct, fixed its route and supplied the plan from which others attempted to build it.[11][12]

Aqueduct engineering

Changing the destination of water

Excavation upstream of the Vallon des Arcs revealed that the northern and southern branches of the earlier Arles system had once converged. When Barbegal was built, water from the southern branch was given a new destination: a channel led it towards the mills, while the arrangement supplying Arelate was altered at the same time.[4][5]

This is more consequential than adding a wheel to an existing stream. It is a reallocation of infrastructure. Water that had belonged to an urban supply system was made to perform industrial work before leaving the engineered network.

Masonry and construction

Expensive stone where it mattered; mass masonry where it did not

The two parallel bridges across the Vallon des Arcs were not identical. The earlier Arles crossing used large ashlar construction; the later “bridge of the mills” relied much more heavily on mortared small masonry and concrete-like rubble construction, with large blocks concentrated at structurally important parts such as pier bases and imposts. The departmental archaeology museum's teaching dossier gives the parallel crossings as about 325 m long, with 36 arches on the Arles bridge and 28 on the mill bridge.[13]

Archaeology also shows that the mill bridge belongs to a wider second-century rebuilding of the hydraulic system rather than an isolated architectural gesture.[4] Stone cutting studies demonstrate successive construction phases in the Vallon des Arcs and a very different technique between the early and later works.[6]

Hydraulics

Eight wheels in series create a new kind of problem

The surviving wheel pits are about 1.1 m wide and 4.9 m long. The upper pits are about 2.4 m deep and the lowest about 2.6 m. Modern reconstruction favours overshot wheels because the available head is approximately 2.4–2.6 m while the discharge available to each train was below about 0.13 m³/s.[2]

But serial operation means the tailwater of one wheel becomes the headwater of the next. Stopping one wheel suddenly releases the water held in its buckets into the downstream train. The 2020 model estimates a running wheel could contain roughly 0.08–0.12 m³. A shutdown therefore creates a transient pulse that the lower basins must absorb.[2]

Wheel-pit plan

about 1.1 × 4.9 m

Available head

about 2.4–2.6 m per mill position

Likely wheel diameter

about 1.8–2.0 m depending on pit depth

Maximum train supply

less than about 0.13 m³/s per row

Engineering consequence

each wheel affects every wheel downstream

The elbow flume

The lost timber survived as stone

Fernand Benoit's excavations preserved carbonate fragments that had formed where mineral-rich water flowed across wooden machinery. More than a century after the machinery decayed, those deposits retain impressions of planks, joints and hydraulic surfaces. Modern analysis identified fragments belonging to mill flumes and wheels.[1]

The most striking reconstruction is an approximately 2.1 m long elbow-shaped flume, originally around 0.30 m wide internally. Hydraulic modelling shows that it could establish critical flow at its entrance and then produce supercritical flow through the flume, while its changing slope helped deliver water to an overshot wheel at a useful velocity and impact angle.[2]

Sixteen machines

Standardised layout did not mean identical machinery

The symmetry of Barbegal is visually powerful: eight basins down the west side and eight down the east, divided by central buildings and access. Yet the carbonate record warns against imagining sixteen interchangeable copies.

The 2024 study found differences in flume slope associated with different wheel sizes. One flume appears to have been physically raised after several years, most plausibly because a larger replacement wheel changed the required relationship between flume tip, water trajectory and bucket entry.[3]

Maintenance

The carbonate preserves something close to a service log

Microstratigraphy in Group 1 carbonate records less than nine years of operation. The 2024 study correlates layers between flumes and probable wheel deposits, allowing parts of the western train's final working life to be reconstructed year by year.[3]

The result is a rare archaeology of maintenance: wheels wore out, timber was renewed, flumes were adjusted and parts of the system declined unevenly rather than failing in one dramatic moment.

Mechanical power

How much power did Barbegal actually concentrate?

Popular accounts often give a single kilowatt figure, but the archaeological inputs are ranges: discharge varied, wheel size varied and mechanical efficiency is not directly preserved. The safer conclusion is that the hydraulic parameters reconstructed for sixteen repeated drops imply a concentration of power in the tens of kilowatts, distributed across the mill train rather than generated at one shaft.[2]

Andrew Wilson uses Barbegal precisely as evidence that ancient investment in machinery could substitute natural power for human or animal labour. He stresses the considerable capital cost: building, machinery and the new aqueduct branch all had to be paid for before the first flour was sold.[7]

Production

The famous flour figure is not a measurement

Published estimates range from roughly 4.5 to 25 tonnes of flour per day. That wide range comes from different assumptions about wheel power, millstone performance, operating hours and downtime—not from excavated sacks of flour.[1][7]

Use the range, not a single headline figure. It better reflects the current evidence.

What was it for?

Arelate, ships—or a production pattern we have not yet recovered?

The obvious answer has long been the population of Arelate. The scale of investment strongly suggests a substantial market, and Barbegal's connection with Arles is not in doubt. Yet stable-isotope sequences in the carbonate show recurring interruptions in mill activity lasting months, which does not fit neatly with a factory expected to provide an even daily supply to a large urban population.[10]

The 2018 study therefore proposed that Barbegal could have produced durable ship provisions—hard bread or biscuit—for the ports of Arles and the Rhône delta. It is an ingenious explanation for seasonal operation, but it remains a hypothesis rather than a labelled ancient contract.

Ownership and imperial power

The empire did not have to own Barbegal to make Barbegal possible

Wilson's formulation is the safest starting point: Barbegal clearly had some connection with the needs of Arles, but its ownership and motives for construction remain unclear.[7] Older reconstructions favoured imperial or municipal management; private or collective investment has also been proposed.

The more defensible argument about imperial power is structural. The Roman world supplied the conditions in which such an investment could be conceived: legal control of water, monetised markets, an urban population, long-distance grain and shipping networks, specialist surveyors, standardised construction practices and the ability to mobilise large quantities of stone, lime and timber.

Evidence audit · what survives

What remains after the impressive claims are tested?