Earlier in the week, I offered some comments on Kristina Winther-Jacobsen’s new book on ceramics from intensive survey on Cyprus. One of this book’s strengths is its openness to reflecting on the role of the ceramicist in the process of producing archaeological knowledge from assemblages. Even such seemingly mundane activities as preparing sherds for analysis, sorting them into lots, and adjusting typologies in small ways on the fly influence the kinds of conclusions one tends to reach from the material collected from the field. For example, our ceramicist, Scott Moore, tends to read units in no particular order in order to avoid allowing the location of the units or their proximity to other known units from biasing his analysis of the material. Winther-Jacobsen, like many ceramicist, prefers to read site and units with some attention to their spatial arrangement allowing them to observe any natural connections between units from the same area. Neither procedure produces inherently “better” results, although one can easily imagine benefits and drawbacks to either process.
This past week, John Wallrodt has been laying out how he organizes data for the PARP:PS project at Pompeii. This is part of a larger study on “Paperless Archaeology”. He made available as a pre-print an article that he co-authored with Sebastian Heath and Billur Tekkok on the structure of the ceramic database from the site of Troy. One thing that stood out to me in the database described in this article is that it included no field linking the ceramic object to the ceramicist. In other words, when the ceramic object becomes a database object it acquires an identity that preserves no record of the role of the ceramicist in this change.
In our first version our PKAP finds database, we attempted to include a reader number field in the our basic data structure. Each ceramicist had a unique number. The reader number would become part of the unique number for each digital object and make it clear that each digital object was the product of an individual’s interpretative decision. The physical, archaeological, object would not include the reader number as a concession to its existence outside of the ceramicist’s analysis. So each object in the database had a number based on a unit of space, an arbitrary unique number within that unit of space (a lot or a batch), and a number associated with a particular reader.
It was our how that with this design, if the interpretation or identification of the object changed, the object would get a new digital record with a new reader number. As a result, the new identification of an object would not overwrite an older identification of an object, and the digital record would stand as record of analysis rather than as a digital stand-in for the physical object. In practice, the design was hard to manage. We did not have a method in place for new identifications of objects by the same ceramicist. While this was not insurmountable, we also did not quite know how to separate what qualified as a “new” analysis by a ceramicist from an “old” analysis of an object. Was the production of a digital record complete when we entered the object into our database? Or was it complete at the moment the artifact received a clear identification? More problematic still was the prospect of a ceramicist later dividing a lot of artifacts that he or she originally grouped together into separate lots based on a new identification. How could we manage to keep a record of one large lot produced by one ceramicist and two or more separate lots of the same artifacts produced by another? (This was most likely to occur in the case of coarse and utility wares). Winther-Jacobsen’s typologies based on wall-thickness, for example, would cut across typologies that our project used for grouping coarse wares, utility wares, and amphora. How could such a system be integrated into our existing data structure?
Projects often tell excavators to maintain working interpretations in their notebooks and not on scratch paper because their working hypotheses, no matter how incorrect or problematic they ultimately become, form a key context for decisions made during excavation.
Of course, there are technical solutions to these issues but they all add a layer of additional complexity to our finds database and make using the finds database to understand larger patterns of ceramic distribution across the landscape or stratigraphy even more complex. Our effort to implement such a system was eventually scrapped, but we did feel that our work acknowledged a key (if somewhat unrealized) aspect of Winther-Jacobsen’s study: archaeologists produce archaeological data.