Special-purpose land use maps based on slope attributes form digital elevation models (DEMs) often are produced with geographic information system technology. Field verification is lacking for accuracy and precision of computer-generated slope class maps from 10- and 30-m DEMs, a standard USDA (1:24 000) soil survey, and field-measured slopes. Two 16-ha sites in Atchison County, Missouri, were selected, flagged on 10-m grids, and the slope was field measured for each grid cell. The field-measured slope class maps were templates for cell-by-cell comparisons with the computer-generated slope classes. More than 50% of the areas were classified into correct slope classes with 10-m DEM maps. The 30-m DEM maps were 30 and 21% correctly classified in the two study areas. Two iterations of low-pass filters increased the accuracy of the 10-m derived maps, but decreased the accuracy of one 30-m derived map. Soil survey techniques correctly classified >30% of each area, but did not capture landscape heterogeneity. The GIS-produced maps underestimated slopes on convexities and overestimated slopes on concavities. Slope class maps from 10-m DEMs have potential use for soil survey and land use planning