Abstract. Although maps and partitions are ubiquitous in geographical information systems and spatial databases, there is only little work investigating their foundations. We give a rigorous definition for spatial partitions and propose partitions as a generic spatial data type that can be used to model arbitrary maps and to support spatial analysis. We identify a set of three powerful operations on partitions and show that the type of partitions is closed under them. These basic operators are sufficient to express all known application-specific operations. Moreover, many map operations will be considerably generalized in our framework. We also indicate that partitions can be effectively used as a meta-model to describe other spatial data types.