Vitruvius’s hodometer · On Architecture 10.9
The Roman cart designed to count miles with pebbles, and the argument over its wheel
Writing under Augustus, Vitruvius described a cart that counts the miles of a journey: a wheel that rolls 12½ Roman feet per turn, a drum with 400 teeth, and one pebble falling into a bronze bowl every 5,000 feet. The diameter in the surviving text does not match that 12½. Pick a Roman road or your own distance, choose a wheel, and watch where the pebbles fall.
What Vitruvius wrote
The description takes four short paragraphs (10.9.1–4). The wheels of the carriage are to have a stated diameter, “so that, when a point is marked on the wheel and it begins to turn on the road from that point, by the time it comes back round to where it began it has covered exactly 12½ feet.” A drum fixed inside the hub carries one projecting tooth. Above it, in a box on the carriage, a drum set on its edge has 400 teeth, which that one tooth pushes on by one tooth for every turn of the wheel. A second tooth on the side of this drum moves a flat plate above it by one tooth for every turn of the drum.
The plate has holes “as many as the miles a carriage can travel in a day; more or fewer makes no difference”, each holding a round pebble. When a hole comes over the single opening in the box, its pebble drops through a channel into a bronze bowl. Four hundred turns of 12½ feet make 5,000 feet, “that is, a thousand paces”: one Roman mile. The sound of each pebble announces a mile, and the pebbles in the bowl at the end of the day give the day’s mileage.
pedes XII S · numero quadringenti · pedum milia quinque, id est passus mille
The arithmetic is tight: 400 × 12½ = 5,000. Everything depends on the wheel really rolling 12½ feet per turn, and the diameter in the surviving text does not give that.
One sentence, three wheels
All five manuscripts that preserve the passage write “four feet” followed by wording for a fraction, and they disagree on its form. Panu Hyppönen checked the witnesses for a 2014 paper in Arctos:
| Manuscript | Library, date | Diameter | Distance per turn |
|---|---|---|---|
| H | British Library, Harley 2767, c. 800 | pedum quaternú & sextantes | pedes · XII · S · |
| W | Vatican, Reg. lat. 2079, 12th century | pedum quaternum et sextantes | pedes XII S |
| V | Vatican, Reg. lat. 1328, 15th century | quaternum et sextantis | XII (the ½ dropped) |
| E | Wolfenbüttel, Gud. lat. 132, 9th century | pedum quaternum & sextante (final s erased) | pedes · XV · S · |
| G | Wolfenbüttel, Gud. lat. 69, 11th century | pedu quaternu et sextante (s added later) | pedes · XII · S · |
The Sélestat manuscript (S) has a gap from 10.6 to 10.10. Readings as reported by Hyppönen (2014), excerpted; V and W from the Budé edition’s apparatus. The Latin line above follows the Budé text he quotes; LacusCurtius prints pedum XII S.
Four and a sixth feet. Only manuscript V has the grammatical form, “four feet and a sixth”. Pair that 4⅙ with the usual 12½ feet and π comes out at exactly 3, as in the Bible’s bronze sea (1 Kings 7:23). John Pottage (1968) thought a pious scribe made the text agree with scripture; George Hollenback (2018) adds that a copyist may simply have restored damaged wording with that ratio in mind.
Four feet. Claude Perrault, one of the architects of the Louvre, deleted the fraction in his 1684 edition to make the sum work, and many editors have followed him. It still does not quite work: it makes π equal to 3⅛.
Three feet and 47⁄48. Hyppönen proposes the number that gives almost exactly 12½ feet with the true π: 3 feet 11¾ inches, in Roman fractions pedum trium deuncis semunciae sicilici. He suggests early copyists misread or lost fraction signs they no longer understood. Hollenback accepts the value and notes it implies π ≈ 3.141361, closer than Archimedes’ upper bound of 3 1⁄7.
What each wheel does to a mile
Any error in the distance a wheel covers per turn is added up 400 times between two pebbles.
| Wheel | Rolls per turn | One pebble every | Error per mile | Implied π |
|---|---|---|---|---|
| 4 feetPerrault’s reading (1684), adopted by many editors | 12.5664 ft | 5,026.5 ft (1,486.4 m) | +26.55 ft (7.85 m) | 3.125000 |
| 4⅙ feetManuscript V, “and a sixth” (Pottage 1968; Hollenback 2018) | 13.0900 ft | 5,236.0 ft (1,548.3 m) | +235.99 ft (69.78 m) | 3.000000 |
| 3 47⁄48 feetReconstruction (Hyppönen 2014) | 12.5009 ft | 5,000.4 ft (1,478.6 m) | +0.37 ft (0.11 m) | 3.141361 |
Check: 400 turns × 12½ feet = 5,000 feet (Vitruvius 10.9.4). 400 × π × 4 feet − 5,000 ≈ 26.55 feet (Hyppönen: 26.55). 400 × π × 4⅙ − 5,000 ≈ 236.0 feet (Hollenback: about 236). 12½ ÷ 3 47/48 ≈ 3.141361 (Hollenback: 3.141361).
The wheels read any road short by a fixed share: 0.53% for the 4-foot wheel, 4.5% for the 4⅙-foot wheel. Metres here use a Roman foot of 29.57 cm (a mile of 1,478.5 m); Roads to Rome uses a mile of 1,481.5 m, which changes the metres by a fifth of a percent. A whole-pebble count can shift by one when a route’s end sits close to a drop.
Do the milestones show the error?
If carts built to the printed text had surveyed Roman roads, the milestones would sit too far apart: 7.85 m more per mile with the 4-foot wheel. Over the Via Salaria’s 139 or so miles, Hyppönen’s rounded calculation gives about 3,670 feet (1,085 m) of accumulated overshoot.
Most of the odd miles he cites run the other way. On the Via Appia, between the 42nd and 46th milestones, a mile works out at 1,471.23 m, shorter than the standard (Letronne, 1851). On the Via Salaria, Hyppönen’s own reconstruction of the road from the Porta Collina to the 18th milestone put its apparently short miles down to gaps in what is known of the road’s exact line.
Roman engineers, he suggests, probably tested a hodometer before using it, and would have noticed a wheel that rolled too far.
Was it ever built?
Whether a cart was ever built to these specifications is not known. Vitruvius presents the design as one “handed down by our ancestors” and gives the building instructions in the subjunctive (“the wheels are to be…”), a point Drachmann had noticed; André Sleeswyk took these clues to mean Vitruvius may never have seen one. Sleeswyk’s reconstructions, built from the text and from Leonardo da Vinci’s unsuccessful attempts and published in Scientific American in 1981, showed that it could be made to work. He also argued that the device goes back to Archimedes; that origin is his hypothesis. Heron of Alexandria describes a hodometer of his own in his Dioptra, in the first century AD.
Vitruvius adds a version for ships: paddle wheels on an axle through the hull, the same 400-tooth drum, the same pebbles. In that passage the Latin text gives the paddle wheels four feet “and a sixth”.
Vitruvius does not give the number of holes in the plate; the drawing on this page shows twenty.
Sources
Road lengths are estimates along Itiner-e, the 2025 dataset of Roman roads, between the mapped junctions nearest each place, favouring the named road where one is given. About nine tenths of its length is conjectured and often follows modern roads, so the ancient distances may differ. That changes the number of miles, not the share a given wheel loses.
- Vitruvius, De architectura 10.9.1–7. Latin text: LacusCurtius (Bill Thayer); translations on this page are by Marius Comper.
- Panu Hyppönen, “4π = 12.5? – The Problems in the Vitruvian Hodometer”, Arctos 48 (2014) 185–204, journal.fi.
- George Hollenback, “The Problems in the Vitruvian Hodometer Revisited”, Arctos 52 (2018) 95–98, journal.fi.
- J. Pottage, “The Vitruvian Value of π”, Isis 59 (1968) 190–197; A. Sleeswyk, “Vitruvius’ odometer”, Scientific American 245 (1981); A.-J. Letronne, Recherches critiques… sur les fragments d’Héron d’Alexandrie (Paris 1851) — cited through Hyppönen.
- P. de Soto, A. Pažout, T. Brughmans et al., “Itiner-e: A high-resolution dataset of roads of the Roman Empire”, Scientific Data 12, 1731 (2025), doi:10.1038/s41597-025-06140-z; data Zenodo 17122148, CC BY 4.0.
- Place coordinates: OmnesViae (MIT), after Pleiades and Talbert 2010.