Generalization of spatial databases consists of complicated operations including not only geometric transformations but also topological changes. The changes often result in an inconsistency between the original and derived databases. In order to control the quality of derived databases, we must assess the topological consistency. In this paper, we propose a set of rules to assess topological consistency for the collapse operation. The rules are based on a rigorous classification of topological properties in a collapse operation. By these rules, we can detect inconsistent topological changes in the generalization process and improve the quality of derived databases.