Seventeen populations of T. hermanni in Greece differed substantially in mean adult body size, over a range of about one and a half times in length and three times in mass. Females were the larger sex in all populations; the degree of sexual size dimorphism did not vary with mean body size. The proximate cause of the variation of body size among populations was differences in the duration of growth, rather than egg or hatchling size or the growth rates of juveniles. The age at maturity (a) increased with body size in the 17 populations, while the Bertalanffy growth constant (k) decreased with body size. The quantities aM and M/k (where M is the instantaneous mortality rate) were invariant with body size, suggesting that differences between populations were adaptive rather than the result of short-term disturbance. Body size was greater in cooler areas, and increased with both latitude and altitude. This pattern is opposite to that found in most ectotherms (the reverse Bergmann's rule), and to that which occurs between tortoise species. Several hypotheses about the possible ultimate causes of variation of body size were rejected, including adaptation to long-term habitat disturbance (land use), character displacement, social factors, energetics, thermoregulation, r-K selection, the length of the season available for incubation, or differences in juvenile mortality. The most likely ultimate cause of size variation between sites is differences in adult mortality, the correlation with environmental temperature being through the frequency of ®res.