Comparative analyses of extinction risk routinely apply methods that account for phylogenetic non-independence, but few analyses of extinction risk have addressed the possibility of spatial non-independence. We explored patterns of extinction risk in Banksia, a plant genus largely endemic to Australia's southwest biodiversity hotspot, using methods to partition the variance in two response variables (threat status and range size) into phylogenetic, spatial, and independent components. We then estimated the effects of a number of biological and external predictors on extinction risk independently of phylogeny and space. The models explained up to 34.2% of the variation in range size and up to 9.7% of the variation in threat status, nearly all of which was accounted for by the predictors, not by phylogeny or space. In the case of Banksia, therefore, high extinction risk can be clearly linked with biological syndromes (such as a brief flowering period) or geographic indicators of human impact (such as extensive habitat loss), but cannot be predicted from phylogenetic relatedness or geographic proximity.