Although doubts remain regarding the distributions of thesepocket mice, our analyses illustrate the kind of research thatcan now be accomplished with a large-scale geographicalperspective using reliable species occurrence records andelectronic maps of environmental variables. In particular, theobjective criterion of suitable environmental conditions holdsapplication both to studies assessing connectivity of knownpopulations as well as research examining the ecological andhistorical factors affecting species’ distributions. The utilityof predictive modelling will vary according to the localityrecords and environmental data available, however (for datarequirements see Stockwell & Peterson, in press; Peterson& Cohoon, 1999). In temperate regions with relatively well-documented species distributions and higher-quality environmental coverages, more accurate predictive models withgreater spatial resolution should be possible. Nevertheless,modelling may actually prove more useful in tropical areaswhere species’ distributions are poorly known due to inadequate sampling. There, these approaches allow evolutionary,biogeographic and conservation research based on data fromspecimens present in natural history collections when fieldwork is not possible or timely