Roman bronze dodecahedra remain one of the most debated classes of artefacts from the north-western provinces of the Roman Empire. This article proposes a new reading of the Nouwen corpus of 77 finds: not as a system of precise gauges governed by a rigid metrological standard, but as a possible group of hand-operated devices for the coarse sorting of mixed coin flows, or other small valuable objects, by size classes. The article is based primarily on the metrically tractable Nouwen corpus (1–77), supplemented by a small number of later, context-rich cases used qualitatively rather than as part of the main quantitative dataset. The analysis is based on published pairs of opposite apertures and on the identification of internal size bands within each object. A number of dodecahedra display not random diameter distributions, but recognizable “steps” and “shelves” that broadly overlap with Roman monetary regimes, including small fractional bands, a denarius-sized band, an antoninianus-sized band, and a large-bronze band. The article also allows for chronological drift in the internal “steps” and “shelves” of dodecahedra over time, broadly in parallel with changes in coin-size regimes, including the emergence of an antoninianus-sized band after the early third century CE. Particular attention is given to military sites, administrative-fiscal centres, urban nodes, coin-rich contexts, and selected cultic settings. The article does not argue that all dodecahedra had a single function, nor that they formed an exact empire-wide metrological standard in the strict sense. It does, however, allow for a broader imperial standard of function: a shared manufacturing logic for devices intended to operate within Roman systems of monetary handling, control, and distribution, without requiring exact uniformity. Atypical cases are treated separately as material, formal, weight, or metric anomalies and are interpreted as exceptions, secondary variants, or objects belonging to adjacent traditions rather than the norm of the corpus.
Leonid Volzon (Thu,) studied this question.