P1: Named Find-Spot or “Almost Certainly” Therefrom

The first assumption has to do with the standard they set for their highest rating of p1. Some of the 1,396 objects they tabulated do indeed have a documented find-spot. The lion’s share of these appear in the Crossroads of Asia exhibition at the Fitzwilliam Museum, which drew heavily upon the stores in the British Museum filled with plunder from nineteenth-century colonial explorations. These include works such as an inscribed steatite casket excavated by Charles Masson in 1834 “at Stupa no. 2 at Bimaran, a village in the Darunta district, west of Jalalabad,” published in 1840. ^{Footnote 17} But the vast majority of the objects in these seven catalogues, including most of those classified as p1 by Chippindale and Gill, have nothing even close to this amount of specific archaeological documentation. Their rating of 145 objects as p1 is possible only because of their extremely capacious definition of the category. Chippindale and Gill’s p1 includes not just objects with specific, documented find-spots but all works for which a find-spot is named in the “tombstone” portion of the catalogue entry, regardless of how precise or imprecise the geographical information, what it is based on, and whether or not bibliography is cited to support it. The only criterion is that the information not be encumbered by qualifiers such as “said to be” (which would make it a p2), “probably” (p3), or “possibly” (p4). As the authors acknowledge, “[w]e accept findspot in the broadest terms: sometimes it is a specific site, sometimes a region or even a whole country.” ^{Footnote 18} Thus, a p1 rating is granted to a bronze statue of a man wearing a toga in the Fire of Hephaistos exhibition, which was, according to the tombstone, “found in France during the 19th century.” ^{Footnote 19} Also earning a p1 is a beautiful marble portrait head of a female figure in George Ortiz’s catalogue, whose provenance is given as “Asia Minor,” which is surely a hypothesis based on carving style rather than on a secure archaeological find-spot. ^{Footnote 20} Other geographically vague “find-spot” reports to which Chippindale and Gill award a p1 rating include “central Etruria”; ^{Footnote 21} “Etruria”; ^{Footnote 22} “Germany”; ^{Footnote 23} “Egypt”; ^{Footnote 24} “Iran”; ^{Footnote 25} and “Afghanistan.” ^{Footnote 26} A p1 is even awarded to a small, sitting, neolithic female figure in the Levy-White collection, about which we are told: “Some scholars believe that the type originated in Asia Minor, where this example almost certainly was found.” ^{Footnote 27}

In most of these cases and in many others granted a p1 by Chippindale and Gill, no bibliography was given in the catalogue to support the find-spot claim. While the assertion about the discovery of the togatus in France during the nineteenth century was presumably based on some preserved record (a letter? Diary entry? Sales receipt?), no such document is mentioned in the text nor cited in the bibliography. Given the fact that this was the statue’s first publication, one might have expected a full presentation of the data. Nor is there any mention of Tarquinia in the 150-word entry on a large bucchero olpe in pristine condition in the Etruscan exhibition, despite the tombstone’s confident assertion that the piece is “from Tarquinia” and Chippindale and Gill’s obliging rating of it as p1. ^{Footnote 28} The catalogue bibliography gives references only for the comparanda noted in the entry; the olpe itself, like most of the objects in this exhibit, appears to be surfacing for the first time. On what basis, then, is it assigned to Tarquinia? There is no way to know if that is second-hand information passed from the vendor to the collector to the catalogue author; if it is an inference based on style; or if it is data obtained by some other means. The specific find-spots posited for at least 25 percent of the 145 objects Chippindale and Gill rate as p1 appear to be unsupported by documentation of any kind.

A related matter is the primacy Chippindale and Gill’s rating system places on the tombstone portion of the labels. In many cases, the entry for objects they rate as a p1 will name a specific find-spot in the tombstone, but the text below (that is, the “chat”) will explain that the find-spot information is second-hand (in which case, it should be classed as p2). One example of this can be found in the Crossroads of Asia catalogue, where a tombstone gives “Afghanistan” as the provenance, but the last sentence of the entry informs us that this figure and another one like it “reputedly were found together in the vicinity of the Helmand river in Afghanistan.” ^{Footnote 29} In other cases, the chat makes it clear that a confidently asserted “find-spot” in the tombstone is actually an inference based on affinity (p6). So, again in Crossroads of Asia, entry no. 137 states simply “Taxila” for the provenance in the tombstone, but the last paragraph of the text explains that “[t]here are comparable figures at Taxila, which suggests a possible source for this one.” ^{Footnote 30}

My concern with these p1 ratings is how far most of the “find-spots” are from actual, documented archaeological origins. They tell us nothing about how, where, or when the objects were used or deposited in antiquity. They cannot provide even the most basic assurance that comes with a genuine find-spot, namely the very near certainty of the object’s antiquity (there are some known cases of archaeological sites being salted with precious objects in advance of an excavation, but for various reasons, this cannot be a common practice). The vagueness of many of these “find-spots” strongly suggests that they are nothing more than second-hand or third-hand reports or rumors (possibly calculated in their imprecision to avoid the attention of specific source countries) or inferences based on style.

Chippindale and Gill are aware of these problems and acknowledge them candidly and insightfully throughout the article. They admit to having “pervasive fears” that “inferred places of origin are equated to findspots.” They also stress that

Their explanation for their decision not to modify their criteria, despite “repeatedly encountering these difficulties in tabulating the collections,” is that

This is certainly true, but it seems to me that Chippindale and Gill here (and throughout the article) are perhaps claiming a more passive role—one of merely reflecting problems that are out there in the field already—than they have in fact taken. A numbered, six-tier scale codifies and reifies a hierarchy of values. Despite the authors’ repeated caveats, no reader could come away from the article doubting that p1 objects have the best provenance; that p2 objects have better provenance than p3 objects; that while the provenance of p4 objects is of only middling quality, it is still much better than that of p6 objects, and so on. Thus, to stay with the category of p1 for the moment, these highly regarded archaeologists are communicating to the whole field—to classicists, archaeology students, collectors, dealers, curators, amateur enthusiasts, and whoever else might be reading the American Journal of Archaeology—that the data “from Swat or Gandhara” is of equal value to the data “from Tomb 416 at Abydos, excavated in 1907” since both are rated p1. ^{Footnote 33} To grant their highest seal of approval to the former sanctions the curatorial practices and conventions that make such a label possible. It teaches students and buyers that such information represents as full an archaeological record as we can hope to have. And it might even influence dealers’ decisions about what words they use to describe the “provenance” of the archaeologically undocumented artworks they are selling.

P2, P3, and P4: Said to Be, Allegedly, Probably, Possibly

A related problem results from the fact that within their six-stepped scale, Chippindale and Gill identify the critical divide as that between p4 and p5, with objects rated p1–p4 seen as having “some indication of provenance” and those in p5 and p6 having “no indication of provenance.” Thus, all of their tables are laid out with a visible break between p4 and p5, and their concluding calculations combine the totals for all of the objects ranging from p1 to p4 (351 objects or 25 percent of the total) and contrast them with the combined totals for all of the p5–p6 objects (1045 or 75 percent). In my view, the problem with this framework is the legitimacy it grants to the various linguistic conventions that Chippindale and Gill rate p2, p3, and p4. P2 objects are those with an “alleged” or “said to be” find-spot. Unsubstantiated second-hand accounts of an object’s place of discovery, usually passed on by vendors, is information of a kind, and curators are right to report it in their catalogue entries. Whether it belongs in the tombstone, a discursive space widely understood as the realm of basic, empirical facts, is another matter.

It is important to remember just how uninformative and unreliable “said to be” reports can be, especially when, as is almost always the case, they are passed along with no indication of when the rumor was reported, by whom, or in what context. ^{Footnote 34} What exactly are we being told when we read that a bronze statuette in the Fire of Hephaistos exhibit is “reported to be from the eastern Mediterranean” or that three metal rhyta in the Ortiz collection are “allegedly from the region of the Black Sea”? ^{Footnote 35} From the point of view of both the buyer and the seller, these are very convenient factoids, as they would appear to assure all parties of the objects’ authenticity while simultaneously leaving the specific, aggrieved (that is, looted) source countries unidentified. Everyone can claim to be acting in “good faith.” When curators and scholars repeat such allegations without due diligence to identify and critically analyze the source of the claim, they protect the possibly conflicted interests that the report may have been crafted to hide. To pass possibly fraudulent claims along without drawing attention to their possible fraudulence is, in my view, to condone the possible fraud. Scholars and curators who do so are, arguably, complicit in the system that perpetuates the looting and destruction of historical sites.

Less common than the “said to be” formula is the potentially even more obfuscatory insertion of words such as “probably” or “possibly” before an alleged, but presumably undocumented, find-spot (p3 and p4, respectively). When there is no further discussion of the probable/possible provenance in the longer chat portion of the entry below, as is often the case, there is no way to know what the suggestion means (is the alleged find-spot probable/possible because it makes sense stylistically? Because a reliable source claimed that that is where the object came from? Because the object is similar to others found there?). Fourteen works in the catalogues surveyed by Chippindale and Gill are described in this way, such as the bronze bust of Demosthenes, which the Ortiz catalogue informs us is “probably from Greece.” ^{Footnote 36}

Here again, the narrow focus on the tombstone portion of the entry is problematic. Chippindale and Gill’s scale rewards authors for putting “almost certain” (p1), “alleged” (p2), “probable” (p3), and “possible” (p4) provenance claims in the tombstone and pays no attention to whether or not these claims are discussed or even mentioned again in the chat. But these assertions are not empirical facts. They do not belong in the tombstone but in the chat; not only because they lack the epistemological status of data about the object’s size, material, or current owner. They also belong in the chat because that is where there is room for the author to discuss the claims in detail, to name their sources, assess their merits, present supporting evidence, and so on. This method has been used in exemplary fashion in a few cases in these catalogues, particularly in the Ortiz publication. ^{Footnote 37} But in such cases—where hypotheses about undocumented find-spots are kept out of the tombstone and discussed in the chat instead—Chippindale and Gill rate the objects a much more ignominious p6.

The key point here is that, as far as our knowledge about the object’s archaeology is concerned, all of the tombstone phrases that earn a Chippindale and Gill rating of between p2 and p4 ultimately mean the same thing: the link between the work and its actual find-spot has been severed, so some vendor or owner, somewhere along the way, has filled in the missing information somehow. None of these objects, in my view, deserve to be counted in the “having some provenance” category. In a subsequent publication, Chippindale and Gill acknowledged some of these issues:

Such willingness to reconsider one’s own previous position is rare and admirable indeed. I see a need for further reconsideration. In my view, the 206 objects that Chippindale and Gill rate between a p2 and p4, and many of the 145 objects that they rate as a p1, are encumbered by too many epistemological uncertainties for them to serve as a foundation for scholarly theories or interpretation. To use them for this purpose would be to build houses upon sand. I would argue that, from an epistemological standpoint, the crucial divide is not between objects with “some” indication of provenance and “no” indication of provenance. Rather, it is between those with a specific, reliable, documented find-spot and those without one—the former a much smaller category than that encompassed by Chippindale and Gill’s p1–p4 rating or even just their p1 rating. I call such objects “grounded,” and those without a specific, reliable, documented find-spot “ungrounded.” The term “grounded” is meant to signal not only the fact that we can associate such objects with a particular place in the ground but also that interpretations of them are grounded or anchored by evidence external to the object itself. ^{Footnote 39} Their authenticity is all but certain, and we have at least one data point—the place of their final use or deposition in antiquity—about their ancient function. The terms can also serve as alternatives to the overly laden “provenanced” and “unprovenanced,” which, in general usage, encompass a number of different concepts that desperately need to be distinguished from one another. Museum labels often have a single line designated for “provenance,” where one may encounter a wide range of types of information, from a secure archaeological find-spot, to a hypothesized place of manufacture based upon style, to distinguished previous owners, to the source of the raw materials. Conversely, the term “unprovenanced” is routinely used to characterize objects with an unknown archaeological find-spot, an unknown ownership history, or both. Attempts by some scholars to introduce the term “provenience” to indicate archaeological find-spot have not been successful, as many have understood it merely as a “dialectical variant” of provenance. ^{Footnote 40}

Scholarly hypotheses about ancient material culture are strongest when they are anchored by grounded objects. Attributions of ungrounded objects to particular dates and places are strongest when they are based on comparisons to specific, grounded objects. By my reckoning, 40 objects out of the 1,396 surveyed by Chippindale and Gill (or 3 percent) meet this higher standard (see Appendices 1 and 3). That is to say, 40 of the entries offered a specific find-spot and cited the bibliography in which that find-spot was first reported. An additional 34 also indicated a specific find-spot but offered no supporting references. It is precisely because they are so rare and so epistemologically valuable that grounded antiquities deserve a much brighter spotlight than they usually receive.

P5 and P6: Silence and Affinity

And what should be done with the 1,356 objects in these seven catalogues that are ungrounded? Chippindale and Gill’s scale implies that the most valuable thing curators can do is to pass along unsubstantiated rumors about the find-spot of these objects—whether identified as rumors (p2) or stated as fact—for example “from Kush”—and leaving it up to the reader to notice the geographical vagueness and its lack of documentary support (p1). Next best is to make a good guess (“probably” or “possibly”), without explaining the basis of that guess (p3 and p4). Any of those practices would earn the object a classification of “having some provenance.” A bad option—p5—would be to honestly and accurately state “provenance unknown” in the tombstone; worst of all would be to hypothesize about where the object might have come from based on comparisons to other, similar objects—p6.

Chippindale and Gill condemn the p5 and p6 objects most harshly and suggest that they are most likely to have been looted. This may or may not be so. But from an epistemological standpoint, the “information” that earns a p2–p4 rating—and, in fact, most of their p1 ratings as well—is actually more damaging to our knowledge of the past than that graded as a p5 or p6, since the “evidence” behind it is rarely ever explicitly presented and thus cannot be refuted; and since such soft rumor often congeals through repetition into hard “fact.” ^{Footnote 41}

Chippindale and Gill disdain any attempt at contextualization based upon what they call “affinity.” I assume they mean “affinity with other objects” or what art historians call connoisseurship. Connoisseurship, in the case of an ungrounded ancient artwork, might entail looking closely at its formal traits, noting that they are similar to the formal traits of another object, and thereby hypothesizing that the two come from similar, or the same contexts or sites or times and places more generally. Whatever they know about the other object, they apply to the one in question. Often, connoisseurship is the only tool available to scholars and curators seeking to say something about an ungrounded ancient work’s use or manufacture in antiquity. At its best, connoisseurship yields good hypotheses, which is achieved when the comparanda (the works to which the one in question is being compared) are explicitly identified and are themselves grounded. Sadly, these two “best practices” are not always followed. For example, the Ortiz catalogue suggests that a fine marble portrait of the empress Sabina might be from Greece (Catalogue no. 232: “Provenance: no indication; Greece?”). Ortiz offers no explanation for his hypothesis. It is assumed that those in the field will understand that it is based on unstated affinities (probably concerning the very skillful carving) with other, unspecified portraits found in Greece. This particular attribution may or may not be accurate, and Ortiz rightly indicates its uncertainty with the question mark. But Chippindale and Gill are, in my view, correct in their claim that such generic, imprecise attributions are largely meaningless and often circular. ^{Footnote 42}

However, if the “best practices” noted above are observed—if the “affinities” are with objects that are named, specific, and grounded and if the scholar has taken care to identify the results as hypotheses and not firm facts—then connoisseurship deserves more credit than Chippindale and Gill’s p6 rating implies. Such meaningfully compared and stylistically analyzed objects are still of an entirely different epistemological order than their excavated kin, but they have greater evidentiary legitimacy than ungrounded pieces on which such careful, methodologically conscious work has not been done. In his discussion of a bronze aryballos with red and blue enamel decoration in the Fleischman collection, for example, John Herrmann notes that

A detailed bibliography at the end of the entry gives all of the relevant citations. This is important, careful, scholarly work. Students and scholars who are intrigued by the aryballos in the Fleischman collection, and see its implications for, say, their research on links between Celtic and Islamic art, can read about it in the catalogue and know where to go for grounded examples. The entries for Catalogue nos. 151 and 165, also by John Herrmann, are similarly conscientious in their use of grounded comparanda.

Best practices are demonstrated in a handful of other entries on ungrounded objects in some of the other catalogues under consideration in this article. ^{Footnote 44} The scrupulous reliance upon specific, epistemologically strong comparanda, however, is not the norm in the field. Indeed, for some classes of objects, such as Greek painted pottery, this practice is all but unheard of (for example, in the San Antonio show). ^{Footnote 45} By my count, only 43 percent of the entries on ungrounded objects in the seven catalogues even mention a grounded comparandum, much less make it the primary basis of their attribution. More typical is the treatment of the ungrounded “bust of an older man” in the Levy-White collection, about which it is stated:

However, no specific, grounded examples of works with comparably “excellent carving” and “subtle modeling of the surfaces,” securely attributed to the period of the emperor Claudius from the region of Rome or Naples, are adduced. My point is not that the hypothesis is incorrect but that as presented it is unsupported by evidence of any kind.

Greater methodological self-awareness and epistemological frankness in exhibition labels and catalogues are desirable for many reasons, not least of which is to mitigate the risk of being duped by forgeries. The demand for antiquities has long exceeded the supply, and we can never take the authenticity of ungrounded objects for granted. When exhibited in major museums and published in glossy, professional catalogues, forgeries have the power to radically distort our understanding of ancient material culture. A small number of entries in the catalogues surveyed here acknowledge the existence of forgeries in general. ^{Footnote 47} Most, however, admit of no uncertainty, even in cases of highly unusual objects. As Chippindale and Gill note as well, it is precisely in such entries that we often encounter the most untroubled and even rapturous descriptions of the object’s uniqueness, its lack of comparanda (grounded or ungrounded) and/or pristine condition rarely presented as cause for alarm. ^{Footnote 48} In the Levy-White catalogue, the complete chat for a 14-centimeter-high bronze figure mounted on a horse and holding his erect penis states: “This small but striking bronze sculpture of a nude, belted male figure mounted on the back of an equid—probably a horse—is unparalleled. It can be assigned to the Bronze Age in Central Asia both on the basis of its type and the treatment of the body of the figure.” ^{Footnote 49} No comparanda are cited, perhaps because it is “unparalleled,” although one is left to wonder how a figure can be both “unparalleled” and an iteration of a “type,” and whether the “treatment of the body” refers to the fact that the figure is holding his erection or some other aspect of the work.

Equally puzzling, in the same catalogue, is a Cycladic figurine with a smaller figure conjoined to the top of her forehead. ^{Footnote 50} The effect is that of a child standing on her mother’s head, even though both figures are otherwise in the standard pose for single Cycladic figures, with their arms across their midriffs and their knees slightly bent. The text notes:

How many is “very few”? What does it mean that four, geographically vague, “reputed findspots” are given for the two “related sculptures”? Most importantly, has the type been “clearly established” by grounded comparanda or only by ungrounded ones?

Particularly in cases of such extraordinary and puzzling iconography, the onus is on catalogue authors to find the grounded comparanda that would help audiences make sense of the work—or to discuss frankly the absence of such comparanda and what that might mean (this onus also extends to non-museum-based scholars who use ungrounded objects in their research). We need information like that provided by John Herrmann, quoted above, or that which is offered by George Ortiz in a noteworthy number of instances in his catalogue. Ortiz’s entry for Catalogue no. 79, a small bronze female figure seated on an ornamented crossbar (“bench”) raising an arm up to her head, offers a good model. There is a clear separation between fact and hypothesis in terms of the piece’s archaeological origins (“Provenance: no indication / Thessaly or Macedonia?”), some ownership history (although this is traced back only to the most recent previous owner; ideally, this would go back to the moment of the work’s first surfacing), and a careful report of its condition. The entry then continues with a lengthy text, which presents hypotheses about the work’s ancient significance, supported by numerous grounded comparanda from Thessaly and Macedonia (with bibliographic citations of their archaeological reports). It concludes with a frank discussion of remaining questions (“however appealing our female figure seated on a bench, we do not know what her gesture represents nor whether she is a divinity or a mourner”). Many additional examples of entries that demonstrate methodological self-awareness and epistemological clarity can be adduced in these catalogues: Catalogue nos. 146 and 212 in Crossroads of Asia; Catalogue nos. 97 and 30 in the Levy-White collection (“unfortunately, neither this vessel nor those like it have any controlled provenance, so their place of manufacture, function, and date have to be determined solely by stylistic and iconographic comparison with other objects that also float without provenances”) and Catalogue nos. 16, 23, 24, 30, 60, 61, 74, 81, 84, 109, 127, 137, 169, and 245 in the Ortiz collection. Chippindale and Gill’s scale, however, is not designed to recognize or encourage good work like this. Most of these careful, critically reflexive entries are rated as p2, p5, or p6, depending simply on whether or not a second-hand report of a find-spot is repeated in the tombstone.