Named urban neighborhoods (localities) are often examples of vague place extents. These are compared with current knowledge of vagueness in concepts and categories within semantic memory, implying graded membership and typicality. If places are mentally constructed and used like concepts, this might account for their cognitive variability, and help us choose suitable geospatial (GIS) data models. An initial within-subjects study with expert geographic surveyors tested specific predictions about the role of central tendency, ideals, context specificity, familiarity and expertise in location judgements – theoretically equivalent to categorization. Implications for spatial data models and a further research agenda are suggested.