[Objective] To investigate the growth of Staphylococcus aureus in cooked chicken at different incubation temperature and initial inoculating level, three common predictive models were compared and the best-fit one was chosen to build the primary and secondary model, the result of this study could be applied to develop the tertiary model, from which the density of the S. aureus at any time in cooked chicken can be calculated from any combination of temperature and initial inoculating level. [Methods] The S. aureus strain was inoculated into cooked chicken under various initial concentrations of 102, 103, 104 CFU/g and stored at 15, 22, 29, 36 °C. The number of colonies was counted by 3M Petrifilm? Staph Express Count Plate. The modified Gompertz model, modified Logistic model and Baranyi model for describing the growth of S. aureus were established by using Matlab software. The best model was chosen by comparing the residuals and the goodness-of-fit [Residual Sum of Squares (RSS), Akallke Information Crlterlon (AIC), Residual Standard Error (RSE)]. The growth parameter (lag phase, maximum specific growth rate and maximum population density) were then obtained from the best model. The secondary model was set up by using response surface equation. Finally, the reliability of the model was verified by internal and external validation. [results] At 36 and 29 °C, the best choice to describe the growth of S. aureus was modified Gompertz model at all initial inoculating level. At 22 and 15 °C, the most suitable model is modified Gompertz model, modified Logistic model and Baranyi model successively according to the initial inoculation level. By comprehensive considerations, the modified Gompertz model was thought of the optimal primary model. The secondary model was verified by calculating the standard error of prediction (%SEP), Root-Mean-Squares (RMSE), Accuracy factor (Af) and Bias factor (Bf), the results of verification were all within acceptable range. [Conclusion] The modified Gompertz model coupled with response surface model could provide a useful and accurate basis to develop the tertiary model.