The recognition of geographic patterns in biodiversity is of considerable interest to biologists and natural resource managers. A key component is the mapping of biodiversity, which combines cartographic procedures for vegetation mapping with biological models to predict biodiversity. An important example is the gap analysis project of the National Biological Service (Scott et al., 1993), which seeks to map the spatial distribution of vertebrate species richness and to identify gaps in the protection of species richness. This paper describes cartographic and modelling features of gap analysis, and discusses some challenges that arise in assessing the accuracy of biodiversity maps. A conceptual model is given for assessment of vegetation and species richness maps produced in gap analysis. Several measures of map accuracy are identified, and their strengths and limitations are discussed. Assessment of map accuracy is described in terms of the influence of landscape heterogeneity, indistinctness of ecotones, scale dependencies and methods for atttribution of biological characteristics. Some sampling and statistical issues in the estimation of map reliability are highlighted, especially those concerning the selection of mapping units and collection of field data. Few universally applicable rules can be identified, and procedures for assessing map accuracy must be tailored to map resolution, the structure of the landscape being mapped and the intended use of the map. However, the inadvisability of validation based on patterns of occurrence for individual species is emphasized. Though difficult and expensive, the demonstration of reliability of biodiversity maps is an essential but largely unmet requirement for their scientific and management applications.