Abstract The islands of Bocas del Toro, Panama, were sequentially separated from the adjacent mainland by rising sea levels during the past 10,000 years. Three-toed sloths (Bradypus) from five islands are smaller than their mainland counterparts, and the insular populations themselves vary in mean body size. We first examine relationships between body size and physical characteristics of the islands, testing hypotheses regarding optimal body size, evolutionary equilibria, and the presence of dispersal in this system. To do so, we conduct linear regressions of body size onto island area, distance from the mainland, and island age. Second, we retroactively calculate two measures of the evolutionary rate of change in body size (haldanes and darwins) and the standardized linear selection differential, or selection intensity (i). We also test the observed morphological changes against models of evolution by genetic drift. The results indicate that mean body size decreases linearly with island age, explaining up to 97% of the variation among population means. Neither island area nor distance from the mainland is significant in multiple regressions that include island age. Thus, we find no evidence for differential optimal body size among islands, or for dispersal in the system. In contrast, the dependence of body size on island age suggests uniform directional selection for small body size in the insular populations. Although genetic drift cannot be discounted as the cause for this evolution in body size, the probability is small given the consistent direction of evolution (repeated dwarfism). The insular sloths show a sustained rate of evolution similar to those measured in haldanes over tens of generations, appearing to unite micro- and macroevolutionary time scales. Furthermore, the magnitude and rate of this example of rapid differentiation fall within predictions of theoretical models from population genetics. However, the linearity of the relationship between body size and island age is not predicted, suggesting that either more factors are involved than those considered here, or that theoretical advances are necessary to explain constant evolutionary rates over long time spans in new selective environments.