The Roman Dodecahedron

Can a measuring function be tested?

Yes, as a conditional geometric model. A successful calculation alone cannot identify the Roman object's purpose. English Heritage describes hollow copper-alloy objects with twelve perforated faces and corner knobs, made by skilled craftspeople. Its account leaves their function unresolved.¹

Begin with one object, its measurement points and a prediction. A rangefinder proposal must explain how the observer aligns the openings and how a known target size becomes a distance. A model that needs different assumptions for every specimen may still demonstrate geometry, but it has not established a shared Roman instrument.

One object: Mainz 3

Guggenberger and Leach identify Mainz 3 as catalogue no. 134, formerly in the Henryk Klunder collection and now in a private collection in Winterthur. Its published width is 55 mm from face to opposite face, excluding knobs. This is neither a vertex diameter nor an excavation location.²

Guggenberger and Leach (2025), p. 36, fig. 4.²

The paired values matter. Replacing each opening with a single circle discards its shape. The experiment below uses selected widths from that description only as conditional circular substitutes. It does not recover a calibrated sightline from the artefact.

What the calculation assumes

Amelia Carolina Sparavigna proposed a distance-measuring experiment in 2012: look through a smaller near hole and larger far hole until their apparent outlines coincide. Similar triangles then relate a known target height to its distance. Her worked example assumes a 50 mm separation; that assumption is not a measured dimension of every dodecahedron.³

Here the eye, both aperture centres and the target centre lie on one axis. The two aperture planes are parallel, their circular outlines coincide in the view, and the target fills that view. Let d be the near diameter, D the far diameter and B the separation, all in millimetres. Let H be the target height in metres.

These equations follow by equating d / x with D / (x + B), where x is the eye-to-near-face distance. The target equation gives distance from the eye. To measure from the near face instead, subtract x after converting it to metres. Keeping those reference points distinct avoids a small but systematic mismatch.

Small aperture differences change the answer

The initial conditional example uses B = 55, d = 14, D = 17 and an assumed 2 m target. Its difference is 3 mm, so the model gives about 36.7 m. Selecting 14 and 15 mm reduces the difference to 1 mm and raises the result to 110 m. Selecting equal 15 mm circles leaves no finite position at which their rims coincide exactly. These are calculated consequences of the chosen substitutes, not three measured performance claims.

A smaller denominator magnifies changes in the input. Halving the diameter difference doubles the predicted distance; doubling the assumed target height also doubles it. If an opening is not circular, choosing another direction across it changes the width being compared. An effective instrument would need a repeatable alignment rule, a way to know target size, and a usable account of measurement error.

No tolerance has been assigned to the catalogue measurements. The selectable cases are not confidence limits. They show why an attractive numerical result needs testing against the actual openings, their orientation, wear and viewing conditions. Decimal output cannot supply that missing calibration.

Competing proposals make different predictions

Rangefinding, Sparavigna: the proposal predicts useful, repeatable viewing angles. Testing should use independently measured aperture pairs and targets, with errors recorded before adjusting the model.³

Seasonal observation, Sjra Wagemans: as described by Sparavigna, the proposal links observations through a replica to the sowing season. A test would need a stated observing procedure and predictions across locations and dates.³

Cosmological or divinatory interpretation, Guggenberger and Leach: the authors connect the form with Platonic and Pythagorean traditions. That is an attributed interpretation; its test concerns historical context and transmission, rather than a range calculation.²

Different sizes need not rule out every measuring proposal: a family of instruments could have different calibrations. But a universal rule must either work across those differences or explain them in advance. Showing that one replica performs a task establishes capability under those conditions. It does not establish that Roman users chose that task. No proposed function is established by this experiment.

Keep the regular solid separate from the object

The geometric calculator below uses a regular dodecahedron with twelve identical pentagonal faces, thirty edges and twenty vertices. Its overall diameter joins opposite vertices. For edge length a, that diameter is a × √3 × (1 + √5) / 2. The face-to-face distance is a × √(250 + 110√5) / 10.

Under that idealisation, a 55 mm face spacing corresponds to an edge of about 24.7 mm and a vertex diameter of about 69.2 mm. Those are derived model dimensions, not additional measurements of Mainz 3. Entering 55 as the vertex diameter would describe a different, smaller solid.

The radial inset reduces the vertex diameter by twice the entered amount and creates a smaller, similar solid. It is not a perpendicular wall-thickness measurement. The volume difference excludes neither apertures nor casting details and therefore does not estimate the object's metal content. The aperture output treats just one opening as a circle; it is not a ranging calibration.

From a possible use to a historical explanation

Keep three records together: the catalogued object, the assumptions used to simplify it, and the result of the test. A useful model makes those boundaries inspectable. The surviving object remains the check on the model.

Keep precision proportionate · Geometry and interpretation

Last verified: 29 September 2026