Internet Archaeol. 16. Rajala. Correlations and predictive modelling

5.0 Epistemology: How is knowledge acquired?

5.1 Correlations and predictive modelling

Crumley and Marquardt (1990) suggested that GIS results in archaeology are acquired by juxtaposing archaeological data and environmental and topographical attributes. Their own results showed how changing climate conditions affected settlement and land use in the Arroux valley. However, in the early phase most archaeologists thought that new knowledge could only be created through predictive modelling and the integration of data and statistical analysis (e.g. Warren 1990a; 1990b). Predictive modelling was based on two fundamental assumptions: 1) settlement choices in the past were strongly influenced by natural environments and 2) environmental characteristics could be drawn directly or indirectly from modern maps (Warren 1990b, 202-3). A further precondition was the necessity to sample the area adequately by archaeological survey.

Predictive modelling, like site catchment analysis and related GIS methods of buffering, cost surfaces and Thiessen polygons (cf. Wheatley 1993; Gaffney et al. 1995b; 1996; Wheatley and Gillings 2000), embedded optimising principles and positivistic thinking. By the mid-1990s the uncritical use of tools was criticised heavily and there was a move towards cognitive and interpretative analyses and the study of belief systems (cf. Wheatley 1993; Gaffney et al. 1995b; 1996). Price (1995) and Wheatley (2003) denounced correlative predictive modelling altogether. Woodman and Woodward (2002) pointed out that the different types of relationships between attributes had not been taken into account. However, there is a research community for predictive modelling in continental Europe (e.g. Kamermans and Wansleeben 1999). In North America optimised modelling has remained strong, even using evolutionary psychology in cognitive studies (Maschner 1996).