We kept 48 gravid short-tailed pit vipers (Gloydius brevicaudus) under four laboratory thermal conditions during gestation and collected 10 females from the field soon before they gave birth to test whether Shine's (1995) maternal manipulation hypothesis applies to temperate reptiles. Females thermoregulated more precisely but did not shift their selected body temperatures during pregnancy, with females at high body temperatures giving birth early in the breeding season. The lowest (22 degrees C) and highest (32 degrees C) temperature treatments Increased maternal mortality and resulted in production of offspring with smaller body dimensions. More deformed offspring were produced at 32 degrees C, and more poorly performing offspring were produced at 22 degrees C. In the field, air temperatures lower than 22 degrees C and higher than 32 degrees C accounted for about 9% and 33% of total temperature readings, respectively. However, offspring produced by field-caught females did not differ from those produced by laboratory-kept females with body temperatures optimal for embryonic development in nearly all traits examined. This suggests that in nature, gravid females avoid exposure of their embryos to temperature extremes through thermoregulation. Our study validates the key prediction of the maternal manipulation hypothesis that maternal thermoregulation should enhance fitness-related offspring traits, and demonstrates that viviparity evolves in temperate reptiles because internal development shields offspring from temperature extremes.