Coins from Underwater Excavations. The Case of Thonis-Herakleion

1The organisers of the conference from which this volume springs invited me to contribute some thoughts on the methodological problems and possibilities arising out of working with coins from an underwater excavation. As will become clear, such coins are far from the most attractive to survive from antiquity, but they can certainly be among the most interesting. However, I should begin with a disclaimer: I do not personally dive. The coins under discussion here have been excavated by the extraordinary group of divers of the Institut européen d’archéologie sous-marine, under the direction of F. Goddio, and with the generous sponsorship of the Hilti Foundation¹. I cannot, therefore, offer any firsthand account of the process of excavation. My knowledge in this area comes solely from discussion with the excavation director and divers, and from my own inferences from the material I study. This material, once excavated, is first conserved at the laboratory of the Maritime Museum in Alexandria, before being registered and organised within the storerooms there². My procedure in examining these coins has been to work through them in a continuous sequence as they have been registered. This means that I have essentially worked through the coins in the order that they were discovered. There will, therefore, be a bias in the provisional picture of finds that I present here, as a result of the focuses of excavation and survey in the first six years of work. The exception to this pattern has been in the work on hoards from the site. When they were available to me for study, I have prioritized work on such assemblages. In presenting data regarding overall patterns of coin deposition across the site I have sought to negate this bias, as well as the tendency of hoards to skew proportions of coins, by excluding them from the figures quoted and distribution maps presented.

2A final caveat concerns the site itself. If there is such a thing as a typical underwater site or excavation, Thonis-Herakleion is most definitely not it. This statement may best be explained by consideration of the discovery of the site, and the picture that may be obtained so far of its nature. The site in question is Thonis Herakleion, which was located during the summer of 2000 by the archaeological expedition of the IEASM. It lies on the sea-bed c. 6,5 km off the coast of Egypt, to the East of the modern town of Aboukir. Survey and excavation of the site has proceeded on an annual basis – political and military circumstances permitting³. The main part of the city, with its great temple of Amun-Gereb was situated at the tip of a peninsula. To the east of this peninsula were a series of interconnecting basins that were separated from the main Canopic branch of the Nile by large sand dunes pierced by channels permitting communication, via the basins, with the city and its ports. To the west of the peninsula lay a lake, which was connected to the eastern basins by a large “Grand Canal”. The general effect of modern artistic reconstructions of this landscape is not so very different from the impression we receive from the famous 1^st c. BC Palestrina mosaic (fig. 1).

Fig. 1. Image of Palestrina Mosaic.

3Perched at the northern tip of the Canopic Branch of the Nile, Thonis-Herakleion constituted the entrance to Egypt from the Mediterranean⁴. The existence of the city of Herakleion was long known from Strabo and later writers⁵, along with its temple of Herakles, ancient, it seems, even in Herodotus’ day⁶. The identity of this city with the Thonis also mentioned by Strabo⁷ was strongly suggested by the Canopus Decree which refers in Greek to the temple of Herakleion⁸ and in hieroglyphics to (ll. 25-6) “the temple of Amun-Gereb in the mouth of the Hone⁹”. These names and identity of the city discovered by the IEASM were stunningly confirmed in 2001 by the discovery of the second copy of the Sais (Naucratis) stele, set up at Herakleion, which provides for its erection “in a town called the Hone-of-Sais¹⁰”.

4But in this image, we encounter one of the primary difficulties faced in working with material from such a site. This is an entire town underwater. It had temples, houses, canals, streets, harbours and all the variety of buildings and public spaces we might expect of an urban centre in antiquity. And all are completely invisible from the surface (fig. 2). It is impossible to gain a physical overview of the site as we might for a site on land. We can add a further layer of complexity to this by noting the maritime/riverine nature of this city means that there are archaeological contexts that were under water even in antiquity. Thus we are dealing sometimes with objects from areas that were above water when they were deposited, and sometimes with objects that have been recovered from locations that were under water when they were lost or deposited.

Fig. 2. Photo of site from boat (by the author).

5And indeed the process of identifying the basic topography of the town in what survives on the seabed today has proven challenging and has had to be built up from a combination of survey methods (sonar and magnetometry), as well as core-sampling, spot-cleaning and basic physical examination¹¹.

6Matters are still further complicated by the processes by which the city came to be under water. On the one hand there is evidence to suggest that the sea-level has been rising since antiquity with respect to this part of the coast at a rate of c. 2 mm per annum. But this by itself is not sufficient to account for the depth of the city at around 6 m below sea-level today. Geological investigation has suggested that the additional sinking must be the result of substrate failure, some of it probably sudden. The precise cause remains elusive: there may not have been one massive event such as an earthquake, but perhaps a series of smaller shocks or high Nile floods. But Herakleion was clearly a disaster waiting to happen: “This geoarchaeological analysis illustrates extreme consequences that occur when carefully implemented protection measures related to coastal sites and associated environmental conditions are overlooked. In the examples presented here, the Greeks and subsequent inhabitants built and maintained their sites on unstable wetland sediment, without foundations and pilings, at distributary mouths that experienced powerful annual Nile floods. Given the remarkable Greek and Roman building and engineering tradition, it is difficult to fully comprehend the decisions that led to emplacement of large, monumental structures directly on underconsolidated and unstable sediment prone to soil instability and associated geohazards¹²”. The city was thus doomed to disappear beneath the waves, and this may have happened in a series of stages. Moreover, the sheer instability of land on which the city was built should prepare us for the possibility of a number of possible minor or major disaster-events before the city finally vanished. The date of this final submergence remains unclear. The last writer to describe the landscape between Canopus and Thonis-Herakleion is St Sophronius, who lived in the region from the late 6^th to early 7^th c. AD. So far, the latest coins to be found at the similarly submerged site of East Canopus belong to the first half of the 8^th c. One possibility is that the final event that led to the complete submergence of this area was the abnormally high Nile flood attested for 741 or 742 AD¹³. It is probable, therefore, that Herakleion has been under water for more close to 1 300 years.

7This fact of long submergence brings with it one obvious benefit, of course. The site has been immune to casual gleaning or systematic plundering by treasure hunters. This should mean that we have the opportunity to explore the monetary history of the town at its various points of destruction (the plural is important here). But equally, the submerged nature, as for any underwater site, creates a very basic problem for the numismatist, and indeed for anyone working on or visiting the site, in that it is extremely difficult to obtain an overview of the contexts in which objects are found. However, the excavation’s response to this problem has created a huge opportunity, to which I will return.

8But first I should outline some of the difficulties in dealing with the numismatic material that has been found at Herakleion. Some of these difficulties are not specific to an underwater site, of course. First there is the sheer number of coins that are emerging. As we have noted, one obvious characteristic of a site that has been underwater for over a millennium is that it has not been metal-detected or otherwise robbed of its monetary finds. What we have from the site, potentially, is everything that was in the city when a particular phase of occupation was brought to an end by disaster and/or when it disappeared under the waves, apparently in the 8^th c. AD. Excavation continues on the site, so we do not yet have final numbers. The material I will drawn on for this paper consists of some 2 200 coins, of which approximately 1 400 have proven identifiable in some respect.

9Second there is the problem of context. The site of Herakleion preserves the fallen remains of a city, and beneath those remains lies a context that resembles that which we might find on a site of dry-land: an “archaeological layer” subject to excavation and a process of stratigraphic analysis. Within this layer, with a characteristic clay-like texture, hoards and other single finds have been excavated. However, on top of this intact context the entire site is covered by a layer of mobile silt (dubbed ‘tout venant’ by the archaeologists), the depth of which varies, but can be measured in the 10s of centimetres. While this might, at first sight, appear problematic, in fact it seems that the reverse might be true. Coins are never found in the tout venant. Rather they rest beneath it, and they rest there in large numbers. However, it is important to stress that the context of these coins is not securely that of their original deposition in antiquity. What seems to have happened is that at the point of substrate failure, coins buried or otherwise close to the ancient surface were dislodged from their original context and gently scattered. However, after this initial point of scattering, they have remained in place, and subsequently have been covered by the tout venant.

10An example may serve to illustrate the point. By good fortune we have evidence of one undoubted hoard that bears witness to the extent of this initial movement, while suggesting the relative immobility that follows. This find consists of a group of gold quarter staters of Ptolemy I (EH 1.153)¹⁴. Gold coins are not common, needless to say, among the finds from Thonis-Herakleion. Thus, to find five gold coins of identical type within a single area of the site is as sure an indication that there can be that these coins were deposited together. That said, these coins were not excavated from the clay-like layer beneath the tout venant, but were found on top of it during survey work around the main temple area. A map (fig. 3) of their distribution thus reveals not a single find, but rather a scatter.

Fig. 3. The Gold hoard from Thonis-Herakleion.

11This might, at first sight, appear alarming: the hoard has become dispersed. But it is at least clear that they seem to cluster around an original point of deposit, probably within the precinct of the main temple of Herakles (H1)¹⁵, and have not been carried from one side of the site to another. During a decade of prior and subsequent survey and excavation work no further specimen of this type has been found. This suggests that, while we should be aware that coins did move at the time of catastrophic substrate failure, this movement is limited, and that it is, therefore, worthwhile examining the distribution of coins from this layer for general patterns of deposit.

12But there is one important way in which this gold hoard is misleading: the state of preservation of these coins is wholly unrepresentative of the finds from the site in general. A few illustrations will make the key points about the nature of the finds (fig. 4, 1-4). The vast majority, over 95% of the total finds, are of bronze. Bronze, unlike gold, does not respond well to 1200+ years of immersion in sea-water. The result is that 36% of the coin finds cannot be identified beyond saying that they were definitely once coins. And when I claim 64% as identifiable in some way, by this I mean that they can securely be assigned to one of the following four categories: Greek (i.e. pre-Roman and non-Ptolemaic), Ptolemaic, Roman (exclusively so far from the mint of Alexandria) and Byzantine (this last category consists exclusively of the local production of dodecanummi).

Fig. 4.1. Bronze coin of ‘Kition’, 2^nd half of 4^th c. BC.

Fig. 4.2. Ptolemaic bronze diobol of Series 2, c. 301-261 BC.

Fig. 4.3. Ptolemaic bronze, uncertain denomination (37mm), Series 5, c. 220-197 BC.

Fig. 4.4. Byzantine bronze dodecanummium, c. 7^th c. AD.

13A serious deficiency resulting from the condition of coins from underwater excavation is the amount of weight-loss they have suffered as a result of long immersion. The effect of this can be most easily seen by comparing the average weights of some specimens from Herakleion with the average recently published by Wolf for coins that in all probability have not been so corroded. Table 1 presents the mean weights identified by Wolf for the diobols of seven subseries of Series 2¹⁶. The range is from 14,38 g to 15,99 g. The 28 coins of Series 2 recorded so far from Herakleion range from 3,21g-14,46 g, with a mean weight of 9,31 g (fig. 6). Herakleion coins weigh, on average, a little over half of their original weight, it seems.

Fig. 5. Series 2 Diobols from Herakleion.

14It is immediately clear that the weight loss is substantial. This has two corollaries. First it means that the Herakleion coins are not useful for metrological study, since their weights bear no clear relation to that at which they were struck. Second, it means that weight cannot be used as a reliable guide to denomination. Thus if a coin is illegible, or only partially legible, its weight cannot be used to deduce its identity.

15Illegibility, of course, is a significant problem. The coins illustrated in fig. 4 are entirely representative of the material at large. For the earlier coinages where types alone may suffice to attribute the coin to a mint, then such an attribution is possible so long as the types are clear. But when issues begin to be marked by variations in monogram, symbol or legend, it is almost impossible to subdivide identifications along these lines, for these subtleties of identifier are rarely if ever legible. So it might be possible to identify a Ptolemaic bronze by type and module, but it is generally impossible to read the letters that identify issues in the third century coinage. An issue of Roman Alexandria may be attributable to an emperor, and the reverse type may be discernable, but it is rare that the year of issue will be clear enough to complete the identification.

16Again, this will be a familiar problem to colleagues who work on dry-land sites, but I suspect that the problem is more severe for underwater material. Happily, however, this limitation of the evidence is not necessarily an impediment to what the numismatic material may offer at an interpretative level. In many contexts, it rarely matters to anyone other than the numismatist whether a coin was part of issue E or A, if that distinction in issue cannot be translated into one of clear chronological, economic or political implication. And it often cannot.

17In this respect, however, the location of the site in question, and the coinage it is likely to yield, is obviously a major factor in determining the usefulness of partially illegible material. Herakleion, of course, is in Egypt, and this fact brings with it a distinct set of challenges in terms of the material that is found; but also some obvious potential in terms of illuminating monetary practice in – dare one say – an exceptional part of the world. This potential has been very much enhanced recently by the work of C. Lorber, O. Picard and T. Faucher through their recategorization of Ptolemaic bronze, and the publication of an important body of excavation material from the CEA excavations in Alexandria¹⁷. This categorization, and the recognisability of Ptolemaic issues through a clear iconographic programme and differentiation by module, means that it is possible to place even badly corroded coins within the new Series into which these three scholars have now organized the bronze (table 2).

18This new categorization offers the prospect both of new forms of analysis of coin use and deposition within an individual site and of comparison of deposition across multiple sites. The latter is not specific to underwater sites, but it does make it possible to compare the coin finds from different types of excavation. I have discussed this possibility at greater length elsewhere¹⁸, so I confine myself to reproducing a chart (fig. 6) that summarises the coin finds from four sites in relatively close proximity to one another: Thonis-Herakleion, Naucratis, Athribis and Alexandria¹⁹.

Fig. 6. Comparison of sites by broad period.

19While the resulting pattern is interesting, and perhaps suggests different patterns of occupation at the different sites, the important point from our point of view is that the poorly preserved specimens from underwater excavation still have comparative value with better preserved coins from land-based excavations.

20Finally there is one last element of an underwater site, or at least of this particular underwater site, that makes a fundamental difference to the way in which we can use the numismatic evidence to enhance our interpretation. An underwater site, as we have noted, is invisible. All of our interpretation must take place through the use of maps and precisely recorded contexts. Here we are very fortunate that the working methods of the excavators at Herakleion have been precise and careful from the very beginning. On the one hand a picture of the topography of the site has been built up through electronic means. So we have base maps on which we plot the coins. But we also have the precise find-spots of the coins. Goddio and the IEASM have been pioneers in the use of underwater GPS systems. Every object found has its find-spot accurately recorded, and is therefore individually mappable. This creates enormous potential for tracing patterns of deposition over time, as the following series of maps (fig. 7-10) demonstrates.

Fig. 7. Distribution of pre-Ptolemaic Greek coins (©F. Goddio/IEASM).

Fig. 8. Distribution of Ptolemaic coins of Series 1-7 (©F. Goddio/IEASM).

Fig. 9. Distribution of Ptolemaic coins of Series 9-10 (©F. Goddio/IEASM).

Fig. 10. Distribution of Byzantine coins (©F. Goddio/IEASM).

21The distribution of finds appears to tell a story. Relatively intense deposition of coins centered predominantly on the Temple of Ammon-Gereb at the centre of the site in the late 4^th c. (fig. 7) gives way to a broader distribution and period of largest number of coin deposits in the Ptolemaic period down to the mid 2^nd c. BC (fig. 8). There then follows abrupt fall off in quantity of coins deposited, and a confinement of them across the site in the following 150 years down to the Roman period (fig. 9). Finally, the Byzantine finds attest to a relatively narrow zone of deposit based on the central island of the city (fig. 10). This deceptively cursory account of the coin finds across a site belies the vast amount of work that has gone into creating it. Hours of diving and careful recording of provenance combine with hours spent identifying coins in appalling condition. Sadly, few sites have this amount of care lavished on them. And fewer still are published in a way that the story can be told. But if it is possible under water, then surely it must be possible on land too.