Adult annual survival rates estimated by mark–recapture for eight Greek populations of the tortoise Testudo hermanni were in the range 0.80–1.0 (mean 0.886). Estimates from growth rings for the same populations were in a similar range (0.84–0.95, mean 0.899). The average survival rate from the two methods was significantly positively related to mean adult body size (r=0.815) and negatively related to effective temperature (r=−0.880). A life table model was used to investigate the consequences of adult survival rate for optimum age at maturity and mean adult size of females. The basis of the model was the trade-off between increased reproduction from large adult size, and decreased survival to maturity from delayed age at maturity. The model showed that the observed variation in adult survival rates was more than sufficient to account for the variation of adult size among populations. The model thus predicted an even greater range of mean adult size than that observed. A negative correlation between growth rate and adult survival rate or age at maturity, as observed, was needed to fit the size data. The model supports the hypothesis that variation in adult size of T. hermanni in Greece is adaptive (rather than a non-adaptive environmental or phylogenetic effect) and due to variation in adult survival rates (rather than other factors such as juvenile survival).