This paper presents our recent efforts on the two key issues in digital terrain analysis for digital soil mapping: quantification of spatial gradation of geomorphic objects and scale of digital terrain analysis. Geomorphic objects (such as slope positions, landform types) are rarely used as continuous variables in digital soil mapping. We developed a framework for characterizing and representing the spatial gradation of geomorphic objects. This framework consists of two major components: a fuzzy logic-based scheme for representing the spatial gradation and a prototype-based technique for quantifying the spatial gradation. Case studies in digital soil mapping have shown that the framework is effective in quantifying the spatial gradation of geomorphic objects. Our examination of scale in digital terrain analysis for digital soil mapping focuses on the following aspects: sensitivity analysis of computed terrain attributes and digital soil mapping to neighborhood size, and the relationship between neighborhood size and scale. Results show: 1) curvature measures are much more sensitive to neighborhood size than slope gradient and that terrain variables are more sensitive to neighborhood size at small neighborhood sizes than at large neighborhood sizes; 2) neighborhood size of digital terrain analysis has profound impact on the accuracy of digital soil mapping and the most accurate soil map is not obtained at the smallest neighborhood size; 3) There is no single spatial scale that can be unambiguously associated with a given neighborhood size and neighborhood size cannot be an adequate indicator of spatial scale.