Abstract: Widespread and rapid losses of natural habitats and biodiversity have made the identification of extinction-prone species a major challenge in conservation biology. We assessed the relative importance of biologically relevant species traits (e.g., body size, ecological specialization) obtained from published records to determine the extinction probability of butterflies in a highly disturbed tropical landscape (i.e., Singapore). We also developed a taxon-specific model to estimate the extinction of butterflies in Southeast Asia. Logistic regression analysis showed that adult habitat specialization, larval host plant specificity, geographical distribution, sexual dichromatism, and congener density were significant and independent determinants of butterfly extinctions in Singapore. Among these traits, specificity of larval host plant and adult habitat specialization were the best correlates of extinction risks. We used this phenomenological extinction-regression model to estimate the relative extinction of 416 butterfly species in Southeast Asia. Our results illustrate the utility of available taxon-specific data for a localized area in estimating the extinction of closely related species on a regional scale. (Eg, Southeast Asia), similar approaches could be used to estimate extinction threats for other taxonomic groups.