The sensitivity of bat species to stochastic disturbance was investigated by exploiting the natural experiment provided by Hurricane Georges, which struck the island of Puerto Rico (Caribbean) in September 1998. Six forest habitats and three cave roost sites sampled for bats prior to the hurricane were sampled in the same way after the hurricane. Populations showed significant declines in abundance and species richness across all forest habitats sampled. Species' sensitivity to disturbance were not equal: larger species were significantly more affected by disturbance than smaller species, once the effects of phylogenetic non-independence were removed. There was some evidence that frugivorous and nectarivorous species are more affected by hurricane disturbance than insectivorous species. These findings have important implications for maintaining viable populations of species in areas that experience a high degree of environmental fluctuation.