The First Number in a Hoard Is a Counting Rule

A hoard record begins with a number that is not a fact

Every record of a group of objects opens with a count, and the count is a rule applied to a pile rather than a property of the pile. Fourteen records in this archive describe a group: eight hoards and treasures from the ground, three sets made as sets, a dispersed manuscript, a temple's sculpture and a monument's carved facing. We read their counts together because the same thing happens in all of them. The number announced at the find is revised, usually more than once, and every revision comes from a decision about what to count as one, whether to count what was found or what survives or what one museum holds, and whether a fragment is an object.

None of that is carelessness. A helmet broken into a thousand pieces is a thousand finds and one object, and both counts are true. What the record has to do is say which rule produced which figure, because a reader comparing two hoards by their headline numbers is almost always comparing two rules. This piece sets the fourteen side by side, gives the rule behind each figure and shows where the same institution publishes two counts of the same material without saying so.

Four official counts of one hoard

The Staffordshire Hoard is the clearest case because its holders have published four figures and all of them are right. Birmingham Museums' own resources page says the hoard contains just under 4,000 objects and broken fragments, and its conservation page calls the work a giant jigsaw puzzle with 4,000 fragments. Historic England says the hoard is composed of more than 4,000 fragments, which together make up around 700 individual items. The announcement of the Society of Antiquaries' 2019 catalogue puts it as more than 600 significant objects found in 4,600 fragments. Our own record carries approximately 4,600 individual pieces and fragments across the 2009 and 2012 finds, and notes that the press figures of autumn 2009 ran from about 1,500 pieces to 3,500 to 3,940 as recovery went on.

The rule behind each number is visible once you look for it. The 4,000 and 4,600 figures count what came out of the field, and they differ by whether the 81 Treasure-declared pieces of the November 2012 return to the same field, and the smallest fragments, are in. The 600 and 700 figures count what the fragments belong to, after a decade of conservation joined some objects from as many as 27 pieces. The gold weight moves with the rule too. Our record carries about 5.1 kilograms of gold and 1.4 of silver, the figures that circulated from the find onward, and the 2019 announcement gives nearly 4 kilograms of gold and 1.7 of silver. The two sets of figures have not been reconciled in anything we read. A reader who quotes 4,600 objects has counted fragments as objects; a reader who quotes 700 objects has counted a dismantled helmet, which is about a third of all the fragments, as one. Both are honest, and neither is the other.

The count as found and the count a museum holds

The second thing a hoard count hides is whether it describes the find or the holding. The Kaper Koraon treasure is the clean example. The Walters Art Museum holds twenty-three vessels from the find, by its own repeated description across the object records, and our record says plainly that this is not, on the evidence available, the whole of what was found. Dumbarton Oaks holds a paten, a chalice and a fan from a nearby village that its own catalogue describes on their own terms, and the scholarly title that covers the material, the Kaper Koraon and Related Treasures, groups several find-labels under one name precisely because the original division into market lots cannot be undone from the museum side.

The Cheapside Hoard shows what happens when the holders count each other. The London Museum's published account says five pieces are at the Victoria and Albert Museum and twenty-five at the British Museum. Hazel Forsyth, the curator who reassembled the hoard for its 2013 exhibition, gave twenty to the British Museum in 1914 and did not mention the V and A in that context. Our record carries both figures and does not average them. The bulk was divided in 1916 between the London Museum and the Guildhall Museum, which merged in 1976, so the Cheapside count is a sum across three institutions and two eras of the same one, and the roughly five hundred pieces in every summary is a total nobody has published a working for.

Sets that were made as a number and lost it

A set made to a number can lose the number after the fact, and three records here show three ways. The Unicorn Tapestries were seven hangings woven around 1500. They are eight objects at the Metropolitan Museum, accessioned as 37.80.1 to 37.80.6 for the six intact hangings and 38.51.1 and 38.51.2 for the two narrow strips that are all that survives of the fifth, given a year later than the rest. Seven is the set, eight is the inventory, six is what a visitor who counts the big ones sees. Our record uses all three and says which is which.

The Aldobrandini Tazze were twelve silver standing cups, each with a statuette of one of the Twelve Caesars above a dish chased with four scenes from that emperor's life, forty-eight scenes in all. Because the emperors unscrew, the set was broken up in the nineteenth century and the figures went back onto the wrong dishes, so that for more than a century a count of twelve tazze was correct and a count of twelve matched emperor-and-dish pairs was not. The Beauvais Missal was a book of 309 leaves for about 450 years. It has been 309 separate objects since 1942, when a dealer bought it for a thousand dollars and sold the leaves at 25 to 40 dollars each; 122 have been located in more than a hundred collections. The count went from one to 309 by a sale and is coming back to one, leaf by leaf, through a virtual reconstruction that will not be finished.

Two figures that count different things

Priam's Treasure carries two numbers that reporting routinely swaps, and they are not two estimates of one thing. Schliemann's own itemised list for Treasure A, the hoard he announced in 1873, comes to roughly 8,750 small gold rings, buttons and other small objects on top of the named larger pieces. A separate figure of about 259 to 260 catalogued items describes the entire Schliemann Trojan collection at the Pushkin Museum, drawn from several find-groups across his campaigns at Hisarlik between 1871 and 1890. The first counts beads; the second counts catalogue entries for a collection of which Treasure A is one part. Our record keeps them apart and says they are not interchangeable.

The Morgantina Treasure has a smaller version of the same problem. It is sixteen pieces by the Metropolitan Museum's own published count, and fifteen in a 2003 letter from the museum's director and in at least one later account. The material reached the Met in two purchases, in 1981 and 1982, and whether a lid and its box are one piece or two is the kind of question that produces exactly that discrepancy. The record says it cannot resolve it, which is the right thing to say about a count that depends on a rule nobody wrote down.

A monument's count is a fraction, and the denominator moves

The count of a monument's sculpture is a fraction, and the record has to give both halves. The British Museum's Parthenon holdings are seventeen figures from the two pediments, fifteen of the temple's 92 carved metopes and 56 blocks of the frieze, about 75 metres of an original 160. Each of those is a count of what survives in one building against what the temple carried in 432 BC, and the denominator is itself an estimate for the parts that were lost in antiquity, in the explosion of 1687 and in the years between. The Amaravati collection is over 120 carved limestone pieces from the great stupa, about 70 on display, out of a monument whose facing slabs were turned round and recarved on their plain backs in the late building phase, so that one slab can carry reliefs of two dates on its two faces.

The Devonshire Hunting Tapestries are four hangings, and the museum's own catalogue cautions that the group may be the remains of three or four separate weaving campaigns rather than one commission. Four is the count of objects in the room. The count of sets they came from is unknown, and it matters, because a set of four and four survivors of four sets are different provenance problems with the same number on the label.

The fourteen group records and the rule behind each count

Set side by side, the fourteen counts are fourteen rules, and the column that matters is the last one.

Two of fourteen have no competing count, and both are hoards that were declared whole to one authority at one time, Mildenhall in 1946 and the Lydian group by a court in 1993. Every group that went through a market, a division between museums, a conservation campaign or a sale by the leaf has more than one number, and the numbers are not errors. They are the history of the group written as arithmetic.

How to read the first number in a group record

The first number in a group record should be read with three questions beside it. Is this a count of what was found, of what survives, or of what one institution holds; is a fragment an object under this rule; and has anyone published the working. A count with the working published, like the Bodleian's lot price or the Met's eight accession numbers for seven tapestries, is a fact about the record. A count without it, like the five hundred pieces of the Cheapside Hoard, is a tradition, and the record should say so.

What we have changed in our own practice as a result of reading these fourteen together is small and specific. Where a record describes a group, the count now carries its rule in the same sentence, and where two official figures exist we print both with their sources rather than choosing. A hoard's number is the first thing a reader remembers and the last thing anyone checks, and the fourteen records here show that it is the figure most likely to be wrong by a factor of six, and most likely to be wrong because two true things were counted as one.