As large spatial databases become increasingly available to researchers in the social and physical sciences, new tools are needed for the analysis of this information that match the sophistication in storage, retrieval and display provided by the rapidly evolving technology of geographic information systems (GIS). In many instances, the context is data rich but theory poor (Openshaw, 1991, 1993) and techniques are needed to ‘let the data speak for themselves’ (Gould, 1981), that is, to aid in discovering patterns, and to suggest potential relationships and hypotheses. A large battery of such methods now exists, following the pioneering ideas of Tukey (1977) on exploratory data analysis (EDA), which stress the interaction between the individual and the data by means of summarising displays, innovative graphics and other highly computational tools (see for example, the overview in Cleveland and McGill, 1988). EDA techniques such as box plots, Chernoff faces, Tukey star diagrams, and scatter-plot matrices are commonly used in studies that combine GIS and spatial analysis, for example, as illustrated in the applications of a so-called archaeologist’s workbench in Farley et al. (1990) and Williams et al. (1990). However, such applications are aspatial in that they ignore the special characteristics of spatial data, such as spatial dependence and spatial heterogeneity (Anselin, 1990). As is well known, such properties will affect the validity of standard statistical techniques, and a special set of spatial statistical methods or spatial econometric methods are needed (for overviews, see Cliff and Ord, 1973, 1981; Anselin, 1988a; Cressie, 1991; Haining, 1990).