An Expert System Development Methodology is proposed, based on experimentation in developing a soils expert system (SES) which identifies a soil from incomplete field data. Tools for conceptual modeling of the soils domain are examined. The tools developed provide a means of recording the conceptual model of the knowledge from three different view points: the inference structures, the domain objects and the functional aspects. A review of the structures used in the knowledge bases of existing classification problems identifies eleven categories for grouping these structures. Using this information with the conceptual model, a detailed design of the knowledge base for SES is created. This design closely models those structures identified as important in the soils domain ensuring that important knowledge is represented explicitly.