Distributional Analysis

[Originally posted on Bill’s blog]

One of the challenges of siteless survey is shifting our intention from a focus on sites to the distribution of artifacts across a landscape. Over the last four years at the Western Argolid Regional Project we have collected artifact level data from over 7000 survey units that cover a significant percentage of our 30 sq km survey area.

The material includes several clear clusters of high density units some of which are associated with known sites as well as a wide scatter of material clustered in different ways across the modern countryside. The temptation is to focus on the larger and higher density clusters which have produced more robust assemblages of material and are more susceptible to analysis on the basis of function, chronology, and settlement structure. In fact, there is no escaping from the fact that the more material an area produces, the more we are able to say about the areas history, use, and regional context. What is harder to understand is how areas or even single survey units that produce small assemblages can contribute to the greater understanding of the landscape and region.

I’ve spent the last two weeks attempting to figure out how to describe the contours of the artifactual landscape of our survey area as a whole and to pull apart the high and low density clusters that constitute the artifact distribution. Some of the things that I had to consider are how to define a cluster: is it related to the number of objects? do the units that produced artifacts have to be contiguous or can they be interrupted? how do we control for surface visibility, background disturbance, and other variables that impact recovery rates on individual units?

Even when I was able to use various kinds of buffering and neighborhood analysis to create archaeologically plausible clusters of units with material from various periods, we then had to determine the arrangement of these clusters across the landscapes. The distance of one group of cluster from another (and the impact of the vagaries of our survey area on this kind of distribution) would appear to offer at least one indication of connectivity in our survey area and perhaps an indicator of density or intensity of human activity in the landscape. At the same time, factors such as period length and recovery rates associated with particular classes (or types) or artifacts likewise shape the visibility of periods and functions in the landscape.

Developing a template or a lens through which we define and construct assemblages for analysis is among the most challenging aspect of siteless survey and one that will likely occupy my time and energy for a quite some time to come!