The Model That Had To Be Destroyed

AA hollow bronze figure is, in one sense, the record of something that no longer exists. In the process known as lost-wax casting, or cire perdue, the object is first modelled in wax around a core of clay, encased in a mould, and then heated until the wax runs out through channels cut for the purpose. Molten metal is poured into the space the wax has vacated, and the mould is broken open to release the casting. Nothing of the original model survives the operation that reproduces it. The technique appears at widely separated places and periods, from the Levant in the fourth millennium BC to Shang China, to the courts of West Africa and to Renaissance Florence, and the objects it produced are among the most closely studied artefacts in any museum. The process that made them, by contrast, has been reconstructed largely from what workshops threw away: broken moulds, spilt metal, and the clay cores that casters left sitting inside their finished figures.

BTwo variants must be distinguished, and that distinction carries most of the argument which follows. In the direct method the wax model is unique, and the mould which destroys it can yield one casting only. In the indirect method the sculptor first makes a durable master model, takes piece-moulds from it, and presses wax into those moulds to produce as many wax copies as are required; a workshop can then cast a series of near-identical figures without ever risking the master. The indirect route demands an additional and distinctly unobvious step, the deliberate manufacture of a negative which is itself never cast, and it leaves traces that survive burial: seams where piece-moulds met, wax layers of unusually even thickness, and core surfaces bearing the imprint of a mould rather than of a modelling tool. In a 2019 survey of 340 hollow bronzes held in European collections, Marta Belisario of the Institute of Archaeometallurgy at Padua classified 71 per cent as indirect.

CBelisario uses that corpus to argue that lost-wax casting was invented more than once. Her group analysed the mineral composition of cores taken from 96 of the figures and found that the clays, the organic tempers and the proportions of ground quartz differed systematically by region: casters in each area had solved the problem of building a core that would survive the pour using whatever the nearest riverbank supplied. Had the method travelled as a package from a single source, she reasons, some element of the recipe should have travelled with it, because the core is the part of the process most likely to be taught and least likely to be worked out twice. No such shared recipe appears anywhere in her sample. Belisario is explicit that her evidence bears on the transmission of technical knowledge and not on the movement of the finished objects, which she accepts circulated widely.

DTobias Renner, a metallurgist at the Freiburg Centre for Historical Technology, accepts the core measurements and rejects the conclusion drawn from them. Recipes for refractory material, he observes, are constrained by what will not crack at 1,100 degrees, and a caster arriving in unfamiliar territory would in any case be compelled to work with local clay within a season. What he regards as diagnostic is geometry rather than chemistry. In 2021 he measured the angle at which the pouring channel meets the mould cavity on 58 indirect castings from four regions, and reported that 49 of them fell within a band of eleven degrees. That value is not dictated by physics, since castings succeed across a much wider range, and Renner treats its recurrence as the signature of instruction passed from hand to hand. Critics note that his four regions were not chosen at random but were simply the four for which intact gating survives.

EThe disagreement has narrowed in one respect. Both researchers now accept that the direct and the indirect variant must be dated separately, and that the question of origins may have different answers for each of them. Renner's own figures show the indirect method appearing across his four regions within a span of roughly two centuries, a compression he regards as too tight for repeated independent discovery; Belisario replies that those dates rest on a handful of securely stratified pieces and that the span would widen considerably if the sample grew. Neither claims that the direct method, which is after all little more than the modelling of an object in a substance that melts, required transmission at all. Belisario has described it as the kind of solution a workshop already casting in open moulds would arrive at unprompted.

FSince 2015 the argument has been fed by neutron tomography, which images the interior of a sealed bronze without cutting into it and reveals the iron pins, or chaplets, that held the core in place during the pour. Chaplet patterns are workshop habits rather than physical necessities, and the method has already divided into two a set of figures that earlier cataloguers had grouped together on the basis of style. What tomography cannot supply is a date, and the shortage of well-dated pieces remains the binding constraint on both cases. Until the corpus of securely stratified castings is several times larger, the question of how many times the process was invented will go on being answered by the choice of evidence rather than by the evidence itself.