The limb proportions of 686 subjects (461 men and 225 women) from five ethnic groups (White, Inuit, Gurkha, Bantu, San) have been compared. Stature, limb and skeletal measurements were taken directly from the subjects by one observer. The brachial and crural indices of the Whites were markedly smaller (lower) than those of the other populations. The crural indices of the Inuit were similar to those of the two African populations, but this may be an artefact from relatively small numbers for the Inuit population. There is no sexual dimorphism for the brachial index, but men have larger (higher) crural indices than women, a finding which probably relates to the relatively broad pelvises and consequently long femurs of women. The two African populations have long limb lengths standardized for height compared to the Gurkha and Inuit populations, with the Whites intermediate. This finding is consistent with Bergmann's thermoregulatory rule. The correlations between distal abbreviation and limb abbreviation for both the upper and lower limbs are poor and negative. Relatively long limbs tend to have smaller distal segments than relatively short limbs and for the legs this may constitute a safeguard for the integrity of the medial and cruciate ligaments of the knee. For these five modern populations distal abbreviation cannot be used as a proxy for limb abbreviation and there is no justification for linking distal abbreviation with climatic selection. Skeletal data relating to nine Neanderthal and 25 early modern humans have also been analysed. The analysis confirms marked limb and distal abbreviation for the Neanderthals compared to early and contemporary modern humans, but this conclusion presupposes that the taxonomic classes are correct and that limb proportions were not used originally as a class discriminant. For these archaic populations there is a moderate positive correlation between lower limb abbreviation and distal abbreviation, but the numbers are small and the confidence intervals very wide. In view of the findings for modern populations, and until more relevant fossils are available, it is probably unwise to use the crural index as a proxy for limb abbreviation in archaic populations.