The paper presentsa method for terminological reasoningabout spatial objects on the basis of a KL-ONE-like framework (LOOM). We apply this method to the domain of deductive geographic information systems and parsing of visual languages.In contrast to existing work, which mainly focusseson reasoningabout qualitative spatial relations, we integrate quantitative information with conceptual or terminological reasoningby the use of "generative"qualitative relations. These relations allow a modularization of systems for terminological reasoning and domain-specificstorageand indexingof, e.g., spatial data. Qualitative relationsare computed on demand from quantitative data during forward-chaining assertional reasoning.