The estimation of body size among the earliest members of the genus Homo (2.4-1.5Myr [millions of years ago]) is central to interpretations of their biology. It is widely accepted that Homo ergaster possessed increased body size compared with Homo habilis and Homo rudolfensis, and that this may have been a factor involved with the dispersal of Homo out of Africa. The study of taxonomic differences in body size, however, is problematic. Postcranial remains are rarely associated with craniodental fossils, and taxonomic attributions frequently rest upon the size of skeletal elements. Previous body size estimates have been based upon well-preserved specimens with a more reliable species assessment. Since these samples are small (n < 5) and disparate in space and time, little is known about geographical and chronological variation in body size within early Homo. We investigate temporal and spatial variation in body size among fossils of early Homo using a 'taxon-free' approach, considering evidence for size variation from isolated and fragmentary postcranial remains (n = 39). To render the size of disparate fossil elements comparable, we derived new regression equations for common parameters of body size from a globally representative sample of hunter-gatherers and applied them to available postcranial measurements from the fossils. The results demonstrate chronological and spatial variation but no simple temporal or geographical trends for the evolution of body size among early Homo. Pronounced body size increases within Africa take place only after hominin populations were established at Dmanisi, suggesting that migrations into Eurasia were not contingent on larger body sizes. The primary evidence for these marked changes among early Homo is based upon material from Koobi Fora after 1.7Myr, indicating regional size variation. The significant body size differences between specimens from Koobi Fora and Olduvai support the cranial evidence for at least two co-existing morphotypes in the Early Pleistocene of eastern Africa.