The study area of approximately 2500 km2 is centered on Hosta Butte, a prominent geological feature located between Grants and Gallup, New Mexico (Figure 2). Previous investigations in this area have identified over 1,600 Anasazi sites and more than 20 sections of road of which all but one are closely associated with prehistoric communities (Kantner 1996; Nials et al. 1987). This information was assembled in a Geographic Information System database (GIS) using ESRI's Arc/Info software. Contextual information on elevation, water, and land use have been added to the GIS, and this is currently being augmented with additional data on soil characteristics.

The Anasazi sites in the study area are concentrated on the northern and southern sides of a large plateau known as Lobo Mesa. The majority of Anasazi occupation occurs during Pueblo II (A.D. 900 -1100), with most habitation sites found in about a dozen communities. Great Houses and Great Kivas similar to those found in Chaco Canyon are found in each community, which, together with the road segments associated with several communities, demonstrate that the area can be defined as part of the Chaco Anasazi system (Kantner 1996).

In order to evaluate the four general models of Chaco road function, a cost surface was generated for the entire study area using the GIS. First, Digital Elevation Models (DEMs) available through the U.S. Geological Service were acquired. Each DEM consists of a grid of points that record elevation at 30 m. or 100 m. resolutions. Based on these elevation data, a new grid of points covering the study area was generated in Arc/Info using the following formula:

T = D/(6 exp ( -3.5 * abs(S + 0.05)))

T = time to cross each cell of the DEM
D = distance across each cell
S = slope

This formula, known as the "hiking function," was developed by geographer Waldo Tobler (1993:4), and has been successfully evaluated using both archaeological and ethnographic data (Gorenflo and Bell 1991; Aldenderfer in press). Application of this formula produced a cost-surface measuring the amount of time it would take to cross each cell.

The study initially tested DEMs with both 30 m. and 100 m. resolutions. The results clearly demonstrated that cost-surface analyses are quite sensitive to data quality. Tests comparing the two resolutions showed that the use of 100 m. DEMs resulted in a cost-surface that did not compare well with real-world topography (Figure 3). For this reason, 30 m. DEMs were used in the final analyses.

The resulting cost-surface provided the basis for evaluating the models of road function. Based on expectations for each model, cost-paths could be generated and compared with the actual road segments. For example, a common interpretation of the roads is that they facilitated economic exchange and foot travel between Chacoan communities (e.g., Kane 1993; Wilcox 1993; Windes 1991). Using the cost-surface, the GIS could choose the paths between the communities that minimized travel time. The resulting cost-path network (Figure 2) could then be compared with the prehistoric road segments to see how closely the modeled paths fit the actual routes of the roads. The specific expectations and evaluations of each of the three models, as well as some unexpected patterns revealed in this study, are discussed in the remaining panels.