The fact that climate influences the continental-scale distributions of species is one of the central tenets of ecology and biogeography. Equally elemental is that species exhibit enormous variation in geographic range size, with most occupying comparatively small areas. The degree to which climate can account for this variation remains unclear. Here, I test three alternative climate-based hypotheses for variation in range size using a large sample of tree and shrub species native to North America north of Mexico. I show that the lowest value of January average daily minimum temperature encompassed by a species' geographic range (T(MIN)), representing the 'climate extremes hypothesis', explains almost 80% of the variation in range size. Hypotheses based on seasonality and climate optima find substantially less support. The relationship between range size and T(MIN) does not change across the breadth of latitudes examined, and is general for conifers and hardwoods, and growth form (tree versus shrub). Differential freezing resistance gains support as the mechanism underlying interspecific variation in range size: using 35 species for which data were available, both T(MIN) and range size are shown to be strongly correlated with measures of freezing resistance.