One of the nice things about a study season is that there is time to do little studies that can get lost in the bustle of a field season. There’s also time to catch up on some reading that all too easily gets set aside during the academic year.
This week, for example, I spent some time with a recent article by Elizabeth Murphy and collaborators in the most recent JFA (2019) (and here) Since we’re working on a preliminary report for WARP, it was useful to read how another project introduced their work, their methods, and their goals. More specifically, however, Murphy and her collaborators offered some insights into a collection strategy that was markedly different from ours at WARP. The LASS project in Sardinia counted 19,176 artifacts and collected 1371. In other words, 7.1% of the seen artifacts were diagnostic (7). This likely speaks to the relatively poor state of ceramics knowledge for Sardinia and what defines an artifact as “diagnostic” than any shortcomings of method on the part of the LASS team.
In contrast, WARP had the advantage of working in the general vicinity of a large number of well-know and thoroughly excavated sites (in varying degrees of publication) that helped us expand what we regarded as diagnostic. This also complemented a rather robust sampling strategy which provided us with a large assemblage of material for analysis. Here’s a short study that considers the significance of our collection strategy in the context of ceramic knowledge in the larger WARP region.
At WARP, over the course of standard transect walking, we counted 119,414 artifacts including sherds, tiles, lithics, and “other”. We collected 64,983 artifacts or 54.4% of all the artifacts seen. This was consistent with a sampling strategy that asked field walkers to collect all visible sherds, but to sample only unique examples of tiles on the basis of fabric and part of the tile (e.g. edge). In the end, we collected 39% of the tiles that we counted. Considering that 78% of the collected tile was either chronologically undiagnostic (what we called “Tile, Ancient Historic”), Modern, or Early Modern, this seems like a good decision.
A cursory statistical review of the collected artifacts likewise demonstrates that a more robust sampling strategy of each survey unit produced archaeologically significant results. While the definition of “diagnostic” pottery is largely in the eye of the field walker and ceramics experts, rims, bases, and handles (RBH) are almost universally recognized as potentially diagnostic and almost always collected. For WARP, these accounted for approximately 10% of the artifacts collected during typical field walking. The remaining 88.9% of the material were body sherds. During revisits to units where we collected more intensively over a more limited area, we produced assemblages that were 87.8% body sherds and 12.2% RBH. This suggests that the sample produced through regular field walking reflected the percentages of material in the plowzone. Interestingly, we also told our teams that they should collect any material from their units that they thought might be diagnostic, but did not fall into a swath, and designate this material as a “Grab” sample. We also designated as “grab” any material from units that we did not survey using our standard survey methods including amid the collapse of Early Modern houses and the overgrown and rocky acropolis of Orneai. These grab samples were 83.6% body sherds. When we filtered out tiles from this assemblages, which produce a large number of body sherds and appeared in an almost continuous carpet across our survey area, the percentages remained more or less the same. For standard survey, 78.9% of our sherds were body sherds, for more intensive resurvey, 81.7%, and for grabs, 77.7%
Of course, most surveys – including LASS – recognize that decorated body sherds are diagnostic. Like most surveys, WARP found that only about 9.3% of our body sherds preserved decoration. Of these decorated sherds, about 45% of them could be dated to a period more narrow than 1000 years. What’s more remarkable, however, is that of the 90% of our body sherds that lacked decoration, 33% of these sherds could also be dated to a narrow period on the basis of preserved shape or fabric. For grab samples, which are by definition more diagnostic, our ability to data undecorated pottery to less than 1000 years improves to 67.1% which is not far from the 72.6% for decorated sherds.
For some periods, like EHIII, Classical-Hellenistic, Hellenistic, and Roman and Late Roman, undecorated body sherds far outnumber decorated body sherds. As a result, a collection strategy that overlooked these sherds would have significantly biased our view of the landscape.