Previous research has shown that sub-lethal heating can increase subsequent thermal resistance of bacteria. If this phenomenon occurs during slow roasting of meat products, it might compromise the validity of thermal process validations. Therefore, this research evaluated the accuracy of a traditional log-linear (Bigelow) inactivation model, developed via prior laboratory-scale, isothermal tests, applied to pilot-scale, slow cooking of whole muscle roasts. Irradiated turkey breast, beef rounds, and pork loin were inoculated via vacuum tumble marination. The inoculum consisted of an 8-servovar Salmonella cocktail in a salt/phosphate marinade. The resulting initial core Salmonella concentration was 7.0, 6.3, 6.3 log CFU/g for turkey, beef and pork respectively. The experimental design consisted of seven different cooking combinations representing industry practices, in a pilot-scale, moist-air convection oven. Core temperature was recorded during cooking, and was used to calculate lethality real-time via the log-linear model. Calculated lethality, using the log-linear model, was greater (P<0.05) than the actual lethality for turkey and beef. A path-dependent model that accounted for the sublethal history of Salmonella resulted in a safer end product as compared with the state-dependent model. Results demonstrate that if slow-cooked roasts are processed according to state-dependent model lethality at or near that required by the regulatory performance standards, Salmonella might survive processing.