Mean daily temperature in natural nests of freshwater turtles with temperature dependent sex determination is a poor predictor of hatchling sex ratios when nest temperatures fluctuate. To account for this, a mathematical model has been developed on the assumption that hatchling sex depends on the daily proportion of embryonic development that occurs above the threshold temperature for sex determination rather than the proportion of time spent above the threshold. The model predictions are borne out by experiments using the marine turtle Caretta caretta. Average developmental rates, both overall and during the period that sexual differentiation is sensitive to temperature, are unaffected by diel fluctuations about the mean incubation temperature. Sex ratios, on the other hand, were affected by diel fluctuations and ranged from ca. 100% males under regimes 26 ± 0°C and 26 ± 3°C to 100% females for regimes 26 ± 7°C and 26 ± 8°C. These and intermediate sex ratios were in close agreement with model predictions. Demonstration of an impact of temperature on sex, while holding overall developmental rate constant, gives support to hypotheses invoking a direct role for temperature rather than alternative hypotheses invoking overall developmental rate as a more proximal influence on sex. The model explains why mean temperature is a poor predictor of hatchling sex ratios. It urges caution in using “hours above the threshold” for predicting sex ratios, because 1 hr at 1°C above the threshold will not be equivalent to 1 hr at 4°C above the threshold. It provides a general framework for integrating experiments at constant temperatures with those in the field or laboratory using fluctuating regimes. It provides greater scope for exploring how reptiles with temperature dependent sex determination might respond to climatic change or other disturbances to the incubation environment. And it provides an explanation of why secondary factors such a hydric conditions and oxygen potentials might influence hatchling sex, even if temperature acts directly to influence sex ratios rather than through its influence on overall developmental rate. © Wiley-Liss, Inc.