Friday, March 16, 2012

I thought that was a WHAT? Inter-analyst variability and yet more ceramic seriation

One of those little problems that archaeologists politely ignore most of the time is the issue of inter-analyst or inter-annual classification variability. A couple weeks ago, I was skimming Ian Robertson’s dissertation and realized that many of the issues he describes for the Teotihuacan mapping project ceramic data could well apply to most archaeological projects fifty years after the fact.

For the Calixtlahuaca project, our excavated ceramics have been classified by a whole bunch of different people over the last five years. In an effort at standardization, our lab procedures have always required that a second person check the classification of a bag of sherds before it is finalized, and all previously classified examples of a type are reclassified if we make a major change in the sorting rules. Despite this, I know from personal experience that our sorting rules have drifted a bit over the years. (This also means that our classification manual doesn’t quite match the way things actually work, much to the frustration of anyone new trying to learn the system.)

Because of this cluster of potential issues, I (with lots of help from Judith and Janeth) did a quick classification check/ reclassification of all the lots that I was pulling attribute samples from last year, since the bags were out anyways. We went through probably two thirds of the domestic context sample this way, which should cut down quite a bit on any inter-observer variability within the DCS as a whole. My gut impression was that we made quite a few changes to lots classified during the 2007 and 2008 seasons, with things stabilizing after that. During the first time period we were still apparently working out some of the basic dividing lines (like how to tell the difference between a plain olla body and a plain bowl body) for this particular ceramic assemblage. For lots classified in later seasons, my impression was that the reclassification accounted for some subtle variation due to gradually shifting dividing lines between types, with most changes being from one sub-variant to another within a single decorative family.

As of earlier this week, this semester’s undergraduate volunteers here at ASU finished entering all of the reclassification changes into the ceramic database, which allowed me to do a final ceramic seriation run to see if all the reclassification had effected the dating of particular contexts. (At least, I hope it’s the last run!) The run seems to have produced the usual handful of changes in the phase assignment of particular stratigraphic capas, so if you are analyzing any sort of material from the project by phase, you should probably check in for the updated list in about another week.

Now, while I’m thinking about it, I really ought to go update our ceramic classification manual/type description book, so that it does match how we actually sort sherds…