We revisited a classic study of morphological variation in the oldfield mouse (Peromyscus polionotus) to estimate the strength of selection acting on pigmentation patterns and to identify the underlying genes. We measured 215 specimens collected by Francis Sumner in the 1920s from eight populations across a 155-km, environmentally variable transect from the white sands of Florida's Gulf coast to the dark, loamy soil of southeastern Alabama. Like Sumner, we found significant variation among populations: mice inhabiting coastal sand dunes had larger feet, longer tails, and lighter pigmentation than inland populations. Most striking, all seven pigmentation traits examined showed a sharp decrease in reflectance about 55 km from the coast, with most of the phenotypic change occurring over less than 10 km. The largest change in soil reflectance occurred just south of this break in pigmentation. Geographic analysis of microsatellite markers shows little interpopulation differentiation, so the abrupt change in pigmentation is not associated with recent secondary contact or reduced gene flow between adjacent populations. Using these genetic data, we estimated that the strength of selection needed to maintain the observed distribution of pigment traits ranged from 0.0004 to 21%, depending on the trait and model used. We also examined changes in allele frequency of SNPs in two pigmentation genes, Mc1r and Agouti, and show that mutations in the cis-regulatory region of Agouti may contribute to this cline in pigmentation. The concordance between environmental variation and pigmentation in the face of high levels of interpopulation gene flow strongly implies that natural selection is maintaining a steep cline in pigmentation and the genes underlying it.