Morphological variation in natural populations is a genomic test bed for studying the interface between molecular evolution and population genetics, but some of the most interesting questions involve non-model organisms that lack well annotated reference genomes. Many felid species exhibit polymorphism for melanism but the relative roles played by genetic drift, natural selection, and interspecies hybridization remain uncertain. We identify mutations of Agouti signaling protein (ASIP) or the Melanocortin 1 receptor (MC1R) as independent causes of melanism in three closely related South American species: the pampas cat (Leopardus colocolo), the kodkod (Leopardus guigna), and Geoffroy’s cat (Leopardus geoffroyi). To assess population level variation in the regions surrounding the causative mutations we apply genomic resources from the domestic cat to carry out clone-based capture and targeted resequencing of 299 kb and 251 kb segments that contain ASIP and MC1R, respectively, from 54 individuals (13–21 per species), achieving enrichment of ~500–2500-fold and ~150x coverage. Our analysis points to unique evolutionary histories for each of the three species, with a strong selective sweep in the pampas cat, a distinctive but short melanism-specific haplotype in the Geoffroy’s cat, and reduced nucleotide diversity for both ancestral and melanism-bearing chromosomes in the kodkod. These results reveal an important role for natural selection in a trait of longstanding interest to ecologists, geneticists, and the lay community, and provide a platform for comparative studies of morphological variation in other natural populations. Color polymorphism in closely related animal species provides an opportunity to study how the balance between natural selection and genetic drift shapes the evolution of appearance and form. The cat family, Felidae, is especially interesting; 13 of 37 extant species exhibit polymorphism for melanism, but evidence for any adaptive role is lacking, in part because the potential benefits of melanism to felid predators are not clear, and in part because the tools for genomic analysis of natural populations are limited. We identify the mutations responsible for melanism in three closely related South American wild felids, the pampas cat, the kodkod, and Geoffroy’s cat, then adapt a new approach for targeted genome sequencing to characterize molecular variation in the region surrounding each melanism mutation. We find that each mutation has developed independently, with strong evidence for natural selection in the black pampas cat, and reduced genetic variation in the entire population of kodkods. Our results demonstrate that some “black cats” are black not by chance, but by selection for a mutation that provides increased fitness.