Ecological niche models are routinely used to not only generate predicted distributions for species at the present time, but they are also used to generate past species distributions (e.g., those at the last glacial maximum). There has been a disturbing absence of testing to determine the degree to which these models accurately predict the past distribution of species. We are using the bat species of the Bahamas as a test case whereby we are generating predictive models of species distribution using various methods and testing their accuracy based on the fossil record of bats from the Caribbean. This study involves considerable field collection, computational effort (niche modeling), and molecular data analysis (to define taxonomic boundaries for niche modeling).