The integration of spatial datasets from different sources is becoming an increasingly important issue. It is very desirable to have a rigorous approach to integration, as an ad hoc approach can easily lead to incorrect inferences. This paper takes a formal approach, giving the beginnings of a logical semantics framework which allows meaning to be defined mathematically for spatial datasets which represent certain types of thematic map data. The basic idea is to interpret the datasets as summarizations of spatial variables, which are in turn interpreted as sets of possible worlds ( ways the world could be). This semantics approach can give a formal meaning to pairs ( or sets) of such datasets, which can then be used to determine the valid inferences from an integrated dataset.