A ‘composite seal' is defined by one or more examples of the impressions made with the same seal. The aim of the current exercise is to incorporate composite seals based on seal impressions on Ur III tablets into the CDLI database so that it is readily possible to call up all of the seals with a particular impression. The primary difficulty is the sheer volume of material. At present, there are c. 100,000 administrative & legal tablets from this period listed in CDLI and roughly 25% of this have seal impressions. There are approximately 5,150 different composites seals – excluding for our discussion physical artifacts – currently listed for this period. However, each of these numbers is certain to change as work is done to standardize this huge amount of data, while at the same time new material is being added regularly, given that there are still very large numbers of Ur III unpublished administrative texts remaining under the watchful eyes of occasionally protective museum curators.

When I started work on the Ur III composite seals, they had been electronically sorted and tagged with a seal number based on their transcribed legends (Englund 2014). There had been some fine-tuning (for example, legends that were so poorly preserved that they were ambiguous had not been given numbers) but essentially it was an electronic sort, and reduction of multiple instances of legends to one each. From my perspective, this made my job a lot easier because it created a simple structure and because very roughly 70% of tablets with seals were correctly slotted into that structure.

I set about the task of improving the referencing of seals in the database by using two large-scale studies of seals: the doctoral dissertations of Tsouparopoulou (for Drehem) and Mayr (for Umma). These pieces of work were different in their approach to cataloguing the seals.

Tsouparopoulou's Cambridge thesis, submitted in 2008 and approved in 2009, was clearly heavily reliant on CDLI and BDTNS, and, compared to Mayr, appears to have been less reliant on the examination of actual tablets. This made my task easier because Tsouparopoulou's catalogue was already aligned with CDLI. Furthermore, because Tsouparopoulou became part of the CDLI team and created that project's initial seals catalogue, the numbering of seals in CDLI corresponds very closely with that given in Tsouparopoulou's thesis. As a consequence I was able to work through the monograph, considering several tens of seals in a day, ensuring, where appropriate, that there was a level of consistency between CDLI and Tsouparopoulou's thesis. This work has been completed.

The main issue for the Drehem seals was the ‘standardization' of the data. We know that each scribe had a limited number of seals in his career and that many of the seal impressions are difficult to interpret. If the precise reading of each seal impression were transliterated there would be many times more different transliterations than the number of seals involved making those impressions. Hence the need to standardize the data. For a large proportion of published tablets, the readings of the seal legends have been ‘standardized' so that, for example, the publications of Sigrist usually imply perfect seal legends, whether or not there were difficulties reading those impressions. For the other cases, Tsouparopoulou was doing the standardizing. This was done without discussion within the thesis, though as I was working through the seal catalogue the rationale of the underlying decisions became evident. To a large extent I followed Tsouparopoulou, but it is worthwhile mentioning a couple of cases as examples.

   

In the first case (Fig. 1a; Snell and Lager, YOS 18, 33), the question was, should we take account of the two “stray” signs shown below the boxed legend, which Tsouparopoulou had ignored.

In this case, I hesitated and asked Bob Englund, and his decision was to follow Tsouparopoulou and ignore the two “stray” signs for the time being.

The second case concerns Ur-Šulpa'e, who made two groups of seal impressions of the following form, one dedicated to Amar-Suen and the other to Šu-Suen (Fig. 1b).

Dahl & Hebenstreit (2007) interpret the data literally and assumed that these were two near identical seals that were used concurrently. Tsouparopoulou takes the view that these are physically the same seal that was modified when Šu-Suen succeeded Amar-Suen. Here, I did not ‘standardize' the data.

Mayr's dissertation on Umma seals was submitted in 1997 and so did not have the benefit of CDLI and BDTNS. Instead, Mayr's catalogue of seals was based primarily on Yale tablets, with a relatively small use of tablets from the literature. Mayr seems to have examined more tablets than Tsouparopoulou. He frequently demonstrates that scribes who were active for a lengthy period of time had a number of seals, often with the same legend, so that worn seals could be replaced.

From my perspective, it is more time consuming to absorb Mayr's dissertation into CDLI. First, there is a larger body of data from Umma than Drehem (c. 6:1). But, second, because the CDLI database does not include unreferenced tablets from the Yale catalogue, many of the tablets that Mayr lists were not listed in CDLI and have to be introduced into the database, which can be time consuming.

For Drehem, much of the work was to eliminate duplicated variations of the same seal legend that had been generated by the electronic sorting procedure within CDLI. For Umma, this was less of an issue. Instead, I have to generate new seal numbers, for Yale tablets, quoted by Mayr, that I am newly introducing into the database. Since Mayr lists 900+ scribes and I process them at an average rate of 100 per week, then this is clearly a lengthy task.

In respect of the multiple seals per scribe: those with different legends are identified with separate seal numbers. However, where the seals have the same legend then (at least, at this stage) I have not identified the tablets with each separate variation of the seal, primarily because, although I could reproduce Mayr's lists based on the Yale tablets, I could not do this for the many other Umma tablets that Mayr did not consider.

One significant problem that I am having with Mayr's data is with differences between his listings and publications of those tablets. As an example, there are large numbers of tablets with the seal of the ensi₂ of Umma, A-kal-la/A-a-kal-la. All of the tablets in Mayr's listings have A-kal-a if the seal is dedicated to Amar-Suen and A-a-kal-la if it is dedicated to Šu-Suen. In particular, Mayr lists four TCNU tablets as A-kal-la even though I understand (from BDTNS and CDLI) that the published version gives A-a-kal-la. Furthermore, CDLI lists c. 20 similar tablets (27 in BDTNS) with varying authors. In fact, BDTNS includes four SAT 2 tablets with A-a-kal-la, even though Sigrist's published version has A-kal-la. There are numerous other examples: Mayr lists four Yale tablets as having an A-kal-la seal dedicated to Amar-Suen, whereas BPOA 6 publishes these four tablets with the seal of A-a-kal-la dedicated to Šu-Suen; Mayr also lists five Yale tablets as having the A-kal-la seal, whereas BPOA 6 states that these five tablets have the seal of Ur-^dLi₉-si₄. It is worth noting that all of the problems noted in this paragraph arise within one inch of text in Mayr's dissertation. Fortunately the density of problems arising in the remaining text is lower! In these cases, I favored the published versions over the dissertation.

Work on other proveniences

For the Girsu tablets, I have made a start, using two publications by Fischer (1992 & 1997), and added to the numbers of tablets with these impressions by including unpublished tablets listed in the British Museum catalogues. For Nippur, I made a start based on Owen's NATN list of duplicate seals. However, in both those cases, much more work needs to be done, and would benefit from comprehensive surveys if they exist. I have not done any work on the Garšana seals, but this should be straightforward because there is a preliminary listing of these seals by Mayr in CUSAS 3 with a promise of a fuller version in CUSAS 7. There is also a promised book of the Umma seals, which is stated as being a much-enhanced version of Mayr's dissertation, but my understanding is that this has not been published.

Future work

BDTNS has done some work to clean its listing of seals, but that this does not include standardizing that database against the Tsouparopoulou and Mayr dissertations. It would be worthwhile cross-checking the listing of seals in CDLI and BDTNS. I did this, to some extent, as I was going along with Drehem, but I had to stop doing it for Umma in order to make a practical rate of progress. Ideally, there should be crosschecking between the tablets listed with each of the seals in CDLI and BDTNS, but this would be a major task.

For Drehem, I started to check for seals included in CDLI but not included by Tsouparopoulou in her thesis, but since the latter was quite recent and was based around CDLI and BDTNS, I came to the conclusion that this check was unnecessary. However, it will be more important for Umma because Mayr's work is almost 20 years old, pre-dating CDLI, and because it was primarily based on the subset of tablets in Yale that have remained unavailable for research.

Ideally, there should be images of composite seals associated with all CDLI entries, but my attempts to produce acceptable images have been ill fated. It would be useful if CDLI implemented a facility to search exclusively in seal legends, similar to that already available in BDTNS. There should be some cross-linking of seals in the database that are different but belong to the same scribe, but I have not attempted to do this rigorously at this stage (cp. S000022a&b). This points to another weakness in the CDLI listing to which Englund (2014) already referred, namely the removal of the sort of iconographic tagging that would differentiate seals with the same legends, but based on imagery demonstrably different. How best to accomplish this capability would be an important discussion among organizers. Consideration should, as well, be given to whether Mayr's listing of tablets with differing seals with the same legend should be noted in the database. (I currently reference Mayr's seal numbers but I do not make explicit which tablets have the different variants.)

Mayr's listing of tablets with seals includes tablets where the text states that the tablet is sealed (kišib₃) even though the seal impression itself is stated (by Mayr) to be illegible. It is worth taking a view on whether such tablets should be given a seal number at all – likely, they should not. In like manner, Tsouparopoulou's thesis includes a significant number of ambiguous seals that were presumably included for completeness. Because of the way that CDLI seal numbering was done post Tsouparopoulou entry, these Drehem seals with ambiguous legends all have seal numbers. These are not useful and should also be reduced to unidentified status.