Common thematic map accuracy measures (e.g., error matrix and kappa coefficient) are inadequate for analyzing the spatial variation of thematic map accuracy. This paper presents a method of mapping spatial accuracy of thematic land cover maps using fuzzy sets. First, we define a multi-level agreement between reference data and corresponding pixels on the map. Next, we formulate a discrete, multi-level agreement fuzzy set regarding a particular land-cover type for each pixel on the map. Then, we construct spatial accuracy maps based on information from the multi-level agreement fuzzy sets. We illustrate the method by constructing accuracy maps of several land cover types for the Mid-Atlantic Regionregion. Results show that thematic map accuracy is not spatially homogenous, but varies across a landscape and different land cover types having different spatial patterns of accuracy. The method provides valuable information about thematic map accuracy for both scientists and decision makers, such as in designing adaptive sampling schemes, deriving landscape indicators, or thematic map-based modeling.