The thermal inactivation kinetics of Salmonella enteritidis PT4 between 49 and 60°C were investigated. Using procedures designed to eliminate methodological artifacts, we found that the death kinetics deviated from the accepted model of first-order inactivation. When we used high-density stationary-phase populations and sensitive enumeration, the survivor curves at 60°C were reproducibly biphasic. The decimal reduction time at 60°C (D60°C) of the tail subpopulation was more than four times that of the majority population. This difference decreased with decreasing temperature; i.e., the survivor curves became more linear, but the proportion of tail cells remained a constant proportion of the initial population, about 1 in 10^4 to 10^5. Z plots (log D versus temperature) for the two populations showed that the D values coincided at 51°C, indicating that the survivor curves should be linear at this temperature, and this was confirmed experimentally. Investigations into the nature of the tails ruled out genotypic differences between the populations and protection due to leakage from early heat casualties. Heating of cells at 59°C in the presence of 5 or 100 mg of chloramphenicol per ml resulted in reductions in the levels of tailing. These reductions were greatest at the higher chloramphenicol concentration. Our results indicate that de novo protein synthesis of heat shock proteins is responsible for the observed tailing. Chemostat-cultured cells heated at 60°C also produced biphasic survivor curves in all but one instance. Cells with higher growth rates were more heat sensitive, but tailing was comparable with batch cultures. Starved cells (no dilution input) displayed linear inactivation kinetics, suggesting that during starvation a rapid heat shock response cannot be initiated.