Bioclimatic models (also known as envelope models or, more broadly, ecological niche models or species distribution models) are used to predict geographic ranges of organisms as a function of climate. They are widely used to forecast range shifts of organisms due to climate change, predict the eventual ranges of invasive species, infer paleoclimate from data on species occurrences, and so forth. Several statistical techniques (including general linear models, general additive models, climate envelope models, classification and regression trees, and genetic algorithms) have been used in bioclimatic modeling. Recently developed techniques tend to perform better than older techniques, although it is unlikely that any single statistical approach will be optimal for all applications and species. Proponents of bioclimatic models have stressed their apparent predictive power, whereas opponents have identified the following unreasonable model assumptions: biotic interactions are unimportant in determining geographic ranges or are constant over space and time; the genetic and phenotypic composition of species is constant over space and time; and species are unlimited in their dispersal. In spite of these problematic assumptions, bioclimatic models often successfully fit present-day ranges of species. Their ability to forecast the effects of climate change or the spread of invaders has rarely been tested adequately, however, and we urge researchers to tie the evaluation of bioclimatic models more closely to their intended uses.