ABSTRACT
Many of the case-studies in Interpreting Space were concerned with predicting the location of archaeological sites using formal methods generally known as Predictive Modelling, techniques that were a major area of interest predating GIS with a survey in 1986 citing over seventy papers on the topic (Kohler and Parker 1986). This level of interest is rooted within the CRM concerns of North American archaeology that deal with large tracts of land often with only small-scale survey results on which to base an understanding of site distributions. The methodology of predictive modelling is ideally suited to raster-based GIS where complete landscape coverages representing different variables can be quantified. One underlying assumption is that past peoples did not locate their sites at random but employed logical decision-making processes. It follows, therefore, that if the variables involved in that decision making can be identified and measured then predictions of site location can be made based on the characteristics of known sites. Two classic papers provide a detailed introduction to predictive modelling within a GIS environment, the underlying assumptions, the different types of statistical models and their testing (Kvamme 19901; Warren 1990).
