Bronze statuette of a horse

A bronze horse forty centimetres high, bought by the Metropolitan Museum from a Paris dealer on 16 February 1923 and for forty years described as the most important single object in its classical collection. In December 1967 the Museum's own Vice-Director for Administration, Joseph V. Noble, stood up at the close of a public seminar on art forgery and announced that it was a fake made in Paris between 1918 and 1923. He published the case in the Museum's Bulletin the following February under the title The Forgery of Our Greek Bronze Horse, and ended it with the sentence: it is famous, but it is a fraud.

He was wrong, and the way in which he was wrong is the reason this record exists. The seam that started the investigation lay in a layer of pigmented wax applied to the bronze in the 1920s or 1930s, and it disappeared when the wax was removed with xylene. The iron wire he read in a gamma radiograph was an armature of a kind found in the Piraeus Kouros and the Delphi Charioteer. Thermoluminescence measurements on zircon grains separated from the ceramic core, published in Archaeometry in 1974, gave an age of 2250 plus or minus 210 years before present.

The horse is back on the Museum's books as an ancient bronze. It is not, however, back where it started: acquired as an Early Classical masterpiece of about 480 to 470 BCE, it is now catalogued as Late Hellenistic, late 2nd to 1st century BCE. The attribution moved twice, and only one of those moves is famous.

The object is small, and that is the first surprise

The horse stands 40.2 centimetres high. Photographs of it, reproduced for decades in histories of Greek art, give no sense of that; it has the presence of a monument and the dimensions of a table ornament. It is hollow, cast in one piece, with no rivets and no welds. The head is turned very slightly, the mane is cropped into a stiff upright crest, and a hole through the crest and a second hole deep in the forelock indicate fittings that no longer survive.

Two small round holes about two millimetres across and ten millimetres deep sit ten millimetres behind each ear. They were invisible for decades because they were filled with a thick coloured wax coating, and were only found when the wax was taken off with xylene in 1968. The right hole was packed with sandy accretion; material in the left tested positive for iron and calcium carbonate, which indicates that a bridle fitting was once pinned there in iron. That detail matters more than it looks, because the placing of the harness was one of the arguments used to condemn the horse.

How a bronze like this was actually made

A hollow Greek bronze of this size was built around a core of refractory material, modelled in wax over that core, invested in clay, fired so that the wax ran out, and filled with molten bronze. The core has to be held in place while the wax is gone and the metal is flowing, and it is held by iron pins driven through it, called chaplets. When the casting has cooled the chaplet ends are cut off flush and the small holes left behind are stopped with rectangular bronze patches.

Every one of those features is present on this horse, and every one of them was read in 1967 as evidence of fraud. Petrographic analysis of the core by Frederick R. Matson found it to be roughly equal parts clear quartz sand and a calcareous paste best described as a marl: a crude calcareous sand of a kind available almost anywhere around the Mediterranean, and the obvious material for the job. The iron chaplets penetrate a short distance into the core and do not pass through it, which is customary in hollow lost-wax casting and would be useless in the loosely bound sand core of a sand casting. The wall thickness varies. There is interior flashing where metal ran. The legs are solid and carry armature.

This is the ordinary technology of the classical bronze foundry, and it is worth stating plainly, because the case against the horse rested on the claim that it was the technology of a nineteenth-century French one.

Forty years as the most important object in the collection

The Museum bought the horse on 16 February 1923 from a dealer in Paris, on the Fletcher Fund. It was dated between 480 and 470 BCE, placed among the highest achievements of Early Classical Greek sculpture, and reproduced in scholarly works as such.

It was also, in a way that is now hard to imagine for an object of this kind, popular. Thousands of plaster casts of it were made and sold over the years at the Museum's sales desk and at Brentano's bookstores. Noble bought one for his own house in 1946, twenty-one years before he condemned the original. Those casts are the reason the case began, and the reason it collapsed, though nobody understood that until 1968.

One line on the face, seen in the summer of 1961

Noble described the moment precisely. He had walked past the horse a thousand times. In the summer of 1961, approaching it, he saw a line running from the tip of the nose up through the forelock, down the mane and back, up under the belly and all the way around. He recognised it as a casting fin: the seam where the sections of a piece mould meet and the metal oozes out, filed off afterwards and leaving a trace.

A genuine Greek bronze of this size would have been cast by the lost-wax process, which leaves no such seam. A seam of this kind belonged to sand casting, a process Noble understood to be unknown before the fourteenth century AD. He turned in a report on 25 August 1961. Dietrich von Bothmer, curator of Greek and Roman Art, went to Greece to look at the bronze horse at Olympia, the marble horses on the Acropolis and other comparanda, and wrote back with three words: remove the horse from exhibition. It came off display on 15 September 1961 and the investigation ran privately for the next six years.

Along the way Noble added stylistic objections. The hole in the mane, he argued, was too low for harness straps, which should tie right behind the ears. The deep hole in the forelock, he thought, was a forger's misunderstanding of the sockets on Acropolis marbles that held a bronze spike, a meniskos, to keep birds off.

The foundry in Greenpoint

The most persuasive piece of the case was assembled a month or so after von Bothmer returned. Noble and von Bothmer took a plaster cast of the horse to the Bedi Rassy Foundry in Greenpoint, Brooklyn, one of the few firms in the area still working in the French sand piece-mould process, and asked how they would cast it.

Mr Rassy laid the cast on its side in a bed of sand mixed with clay and indicated the level to which he would fill the mould. The line he drew with his hand followed exactly the line running around the body of the horse. Then, unprompted, he added that he would not do it that way; he would cast it hollow, with a core of the same sand and clay, held with iron wire about the thickness of coat-hanger wire, the protruding ends cut off and the holes covered with little bronze plugs.

Noble and von Bothmer went back to the Museum and counted the plugs. They were exactly where Rassy said they would be, on the head and back and along the sides. They had always been read as ancient repairs of casting bubbles. A young assistant curator, Brian Cook, suggested testing for the iron with a magnet on a string; by the spring of 1962 there was a chart of the points of magnetic attraction, and they coincided with the plugs.

It is a beautiful piece of reasoning and it is wrong, because the identical features are produced by the lost-wax process. Chaplets are iron. Chaplet holes are patched. The demonstration proved that a modern sand-caster would make a horse that looked like this one; it did not prove that only a modern sand-caster would.

Gamma rays, and a museum condemning its own masterpiece in its own Bulletin

On 15 September 1967 a firm called Radiography Inspection Inc., which normally inspected nine-inch steel plate on submarines, arrived in the Museum's north parking lot with a truckload of equipment. Conventional x-rays would not give a clear image through that much bronze, so they used iridium-192 and exposed by gamma rays. The film, developed in the truck, showed the cavity, the sand core, a faint line of extra density running the length of the animal, and white spots at the ends of transverse wires.

Noble read the faint line as an iron wire armature and called the shadowgraph the definitive proof of the forgery. He dated the object to Paris between 1918 and 1923 on the grounds that the sophisticated French sand-casting technique had been used and that the neoclassical stylisation of archaic form would not have been possible earlier. He insisted on the word forgery rather than pastiche because the legs and tail had been deliberately mutilated to simulate damage.

He announced it in November or December 1967 at the third of a series of four public seminars on art forgery held at the Museum, and printed it in February 1968 in the Museum's own Bulletin, in an issue devoted to forgery which the director Thomas Hoving opened with an essay called The Game of Duplicity. It is one of the few occasions on which a great museum has published a full technical indictment of its own most celebrated object under its own imprint. That is to the Museum's credit, and it is also why the reversal took thirteen years and left a permanent mark.

What was under the wax

The horse was submitted to the Museum's Conservation Department on 3 January 1968, three weeks after the announcement, for a full technical examination. The first thing the conservators did was look at the mould lines.

The bronze turned out to be coated with thick opaque layers of black and green pigmented wax, of a type routinely applied to archaeological bronzes at the Museum and elsewhere in the 1920s and 1930s. The wax was removed with xylene. The mould lines went with it.

What remained was a single depressed line on the face, running up from the nose to between the eyes and stopping there, not continuing to the forelock as had been reported. The seam that Noble had seen running round the entire animal was in the wax, and it marked the edges of the piece mould made not for casting the bronze but for reproducing the plaster casts sold at the Museum's own sales desk. The horse was condemned on the evidence of its own souvenir industry.

The line that is genuinely on the metal is incised, with no flange and no step. Cyril S. Smith of MIT, the foundryman John Spring of the Modern Art Foundry and the Florentine master caster Bruno Bearzi all examined it and none considered it a mould mark. Bearzi pointed out that a piece which took months to make could have had such a line obliterated in minutes; the surfaces around it have been chased and filed, so it was deliberately kept. Comparable incised lines run down the faces of marble horse heads in the National Museum in Athens and the Agora Museum, where mould marks are impossible. It is a modelling line, put there by the sculptor.

Europe answers back

The counter-arguments began immediately and came mostly from people who had handled ancient bronzes. Carl Bluemel, retired director of the Berlin State Museums, published a note in Die Weltkunst in 1968 titled, flatly, Das Bronzepferd in New York ist keine Faelschung. He pointed to iron armatures in the core of the bronze Kouros found at Piraeus, in the Charioteer at Delphi and in castings in Berlin, and argued that the armature Noble described was a sign of authenticity rather than the reverse. Bruno Bearzi said he had found iron supports in the Piraeus Kouros and other ancient castings. Internal iron supports have since been observed in the serpent head from the Delphi tripod in Istanbul and in a mirror support in the Metropolitan's own collection.

The revealing detail is what Bluemel and Claude Rolley did with the seam. Both accepted Noble's assertion that a fin existed in the metal, because it had been published by the Museum, and both then tried to explain how an ancient object could have one: Bluemel suggested casting in two soldered halves, Rolley a wax model built in two halves over a core. Neither could have known that there was no fin. A false observation, printed under an institutional imprint, generated a literature of ingenious explanations for a thing that was not there.

The metal, the corrosion and the ore

The re-examination ran on ten collaborators and took the object apart analytically. Emission spectroscopy gave an alloy containing about 0.45 per cent lead and no zinc, a composition consistent with ancient practice and not with modern bronze. X-ray diffraction on the corrosion products identified paratacamite and atacamite, copper chlorides characteristic of long natural burial, with intergranular corrosion penetrating the metal to a depth of at most about a quarter of a millimetre. R. M. Organ at the Smithsonian's Conservation Analytical Laboratory found bornite and calcite in the mineral incrustation. Lead isotope work by Robert H. Brill and colleagues with the National Bureau of Standards placed the ore sources in the Mediterranean, most probably Italy and the Levant.

Radiography was repeated with iridium-192, cobalt-60, caesium-137 and high-energy x-rays. The internal structure it showed was an armature and chaplets, not coat-hanger wire.

None of this, on its own, settles the question, and the authors said so with unusual honesty in their opening paragraph: even where every traditionally understood property is consistent with an attribution, a sufficiently clever forger could have known and used the right materials in the right ways. Consistency with antiquity is not proof of antiquity. What was needed was an absolute date.

The sample that was taken before the gamma rays

Thermoluminescence dates fired ceramic. All ceramic contains traces of uranium, thorium and potassium; the energy released by their decay is trapped in the crystal lattice and released as light when the material is heated to around 500 degrees, and the accumulated dose divided by the dose rate gives an age. The ceramic core of a bronze is therefore datable, and its firing date is the date of the casting.

The horse should have been undatable. Gamma and x-ray radiography deliver an artificial radiation dose that is added to the natural one and cannot be distinguished from it, and this horse had been radiographed repeatedly. The examination was saved by an accident of sequence: a sizeable sample of core material had been extracted early, before most of the radiography was done. The radiographic dose that sample had received was calculated from the exposure records at about 15 rads, with the caveat that trial exposures may not all have been logged.

Pieter Meyers initiated the measurements and worked with David W. Zimmerman and Marjorie P. Yuhas at the Laboratory for Space Physics at Washington University in St Louis. They measured three different fractions separated from the core: zircon grains with very high internal radioactivity, large quartz particles with almost none, and fine-grained mineral particles in between. The accumulated doses came out at roughly 100,000 rads for the zircons, 625 rads for the quartz and 950 rads for the fine fraction. The inequality is itself the argument: the zircons had absorbed a dose several thousand times the estimated radiographic dose, which can only have come from their own internal radioactivity over a long period. The 325-rad gap between the fine fraction and the quartz is due to alpha radiation, and alpha radiation can only come from decay inside the core material itself.

Zimmerman's improved zircon technique gave an age of 2250 plus or minus 210 years before present, an average of measurements on three individual grains, with the uncertainty a reliable one-sigma estimate. His own statement of what that means is properly cautious: the results show only that the horse was probably made between about 500 BCE and 50 BCE, with, for practical purposes, equal probability for any time in that period. The results were published in Archaeometry in 1974 and set out in full in the Journal of the American Institute for Conservation in 1981, whose conclusion is that the thermoluminescence dating provides conclusive evidence that the horse cannot be a modern forgery.

Vindicated, and re-dated by seven centuries at the same time

The horse came back. What did not come back was the date. The Museum acquired it as Early Classical, about 480 to 470 BCE, and that is the attribution that made it the most important object in the classical collection. It is now catalogued as Greek, Late Hellenistic, late 2nd to 1st century BCE, with the Museum's machine-readable record giving a range of 115 BCE to 1 BCE.

The thermoluminescence result is the reason both statements can stand. A window of 500 to 50 BCE excludes a modern forger absolutely and excludes nothing else; within antiquity it is compatible with an Early Classical date and with a Hellenistic one, and it cannot choose between them. The choice was made on style, in the light of a better understanding of how thoroughly Hellenistic sculptors reused and quoted earlier manners. An object made in the first century BCE in a deliberate archaising idiom will look Early Classical to an eye trained to read style as chronology, which is exactly what happened in 1923.

There is a symmetry worth naming. In 1967 the horse was condemned because a technical observation was allowed to overrule the eye. In the 1970s it was re-dated because the eye, better informed, overruled a technical result that could not decide. Neither method is sovereign, and this object has been used to prove both propositions.

What the case actually establishes

Three things, and they are not the ones usually drawn from it.

The first is that a conservation coating is evidence about conservators, not about makers. Every observation Noble made about the seam was accurate as an observation. The seam existed; it simply belonged to a wax layer applied in the 1920s or 1930s for the purpose of taking moulds for the museum shop. A technical argument is only as good as its account of which layer it is looking at.

The second is that publication by an institution converts an assertion into a premise. Bluemel and Rolley, distinguished scholars both, spent their effort explaining a fin rather than asking whether it was there, because the Metropolitan Museum had printed that it was.

The third is that absolute dating answers one question and exactly one. Thermoluminescence told the Museum that the horse is ancient. It did not tell it, and cannot tell it, when in antiquity, or where, or by whom. The date on the label today rests on the same kind of stylistic judgement that produced the wrong date in 1923, and it is offered with the same confidence.

Joseph Noble's own standard, quoted by his director at the start of the very seminar series at which the horse was condemned, was Max Friedlaender's: it is an error to collect a forgery, but a sin to stamp a genuine piece with the seal of falsehood. The Museum did both to the same object, and then published the correction.