The causes of craniofacial variation among human populations have been the subject of controversy. In this work, we studied aboriginal populations from southern South America, the last continental region peopled by humans and with a wide range of ecological conditions. Because of these characteristics, southern South America provides a unique opportunity to study the relative importance of random and nonrandom factors in human diversification. Previous craniometric studies recognized remarkable differences among populations from this region, usually resorting to random factors as the main explanation. In contrast, here we suggest, using tests based on quantitative genetic models, that: (1) the rate of craniofacial divergence among these populations is too high and (2) the patterns of variation within and between populations are too different to be explained by genetic drift alone. In addition, the among-sample craniofacial variation is correlated with climate and diet but not with mtDNA variation. We suggest that the influence of nonrandom factors (e.g., plasticity, selection) on human craniofacial diversification in regions with large ecological variation is more important than generally acknowledged and capable to generate large craniofacial divergence in a short period of time. These results bring nonrandom factors into focus for the interpretation of human craniofacial variation.