The thermal inactivation of Salmonella enteritidis phage type 4 and Escherichia coli O157:H7 as affected by temperature (54.5-64.5 degrees C), pH (4.2-9.6 with HCl or NaOH) and NaCl concentration (0.5-8.5% w/w) was studied. Cell suspensions in modified tryptone soya broth were heated in a submerged-coil heating apparatus and survivors were enumerated on tryptone soya agar incubated aerobically. For most thermal inactivation data there was a logarithmic decrease in the viable cell concentration over the initial 4-6 log10 reduction and D-values were fitted. In some cases, tailing of the survivor curves was observed with cells surviving longer than the D-values predicted. Models describing the effect of temperature, pH and NaCl concentration on the thermal inactivation of S. enteritidis and E. coli O157:H7 were produced. For both organisms, predicted z-values of 4.6-7.0 C degrees were obtained depending on conditions, with larger z-values at higher levels of NaCl. Optimum survival occurred between pH 5 and pH 7 and increasing acidity or alkalinity caused a decrease in the predicted D-values. At equivalent pH, acetic acid and lactic acid (at 0.5, 1 and 2% w/w) generally had a similar, or increased, lethal effect compared with HCl, whereas in most cases citric acid had a less lethal effect. For E. coli O157:H7, increasing NaCl concentration had a protective effect up to the maximum tested (8.5% w/w), while for S. enteritidis optimal survival at a NaCl concentration of 5-7% w/w was predicted. The models were validated in foods by comparing predictions with published data. Most (80%) of the predicted D-values from the S. enteritidis model were within the 95% confidence interval (within 2.45-fold of the published data) for different Salmonella serotypes in whole egg, egg albumen, egg yolk, beef and milk. Most (93%) of the predicted D-values from the E. coli O157:H7 model were larger than the limited published data for this organism in meat, poultry, milk and apple juice with 42% within the 95% confidence interval (within 2.05-fold of the published data). The D-value models were incorporated into Version 1, and subsequent versions, of the predictive microbiology software program, Food MicroModel.