ABSTRACT
Predictive modeling-the practice of building models that in some way indicate the likelihood of archaeological sites, cultural resources, or past landscape use across a region-has its roots in the 1960s and earlier. Such models were implicit in the earliest expressions of settlement archaeology (e.g., Willey 1953) and in later work that actually formulated explicit statements about prehistoric location (e.g., Williams et al. 1973). The First Age of Modeling, in the early to mid-1980s, saw many stumbling blocks to be overcome: ways of thinking that concentrated more on difficulties and sources of variation that seemed to dictate why archaeological models could not be developed, the “processualist school” that advocated deductive or “lawlike” behavioral statements as a basis for modeling and decried uses of statistical methodologies based on simple correlations, and a lack of effective computer technology for the application of models across regions. Yet, despite these disadvantages, real progress was made, largely in university research settings made possible by cultural resource management (CRM)funded projects. Some of these advances included recognition of sampling biases in archaeological databases, procedures for characterization of background environments, applications of univariate and multivariate statistical tests and models, the use of independent test samples for model performance assessments, and the pioneering applications of geographic information system (GIS) technology in the discipline (see Judge and Sebastian 1988; for historical overviews see Kvamme 1995; Wheatley and Gillings 2002:165-181).
