Thermal sterilization of canned viscous liquid foods using saturated steam is enabled by natural convective heat transfer. However, the governing equations for two-dimensional convective heat transfer may be only rigorously solved by numerical calculations. On the other hand, if conduction is assumed to be the only mode of heat transfer, the thermal sterilization problem has analytical solutions for simple boundary conditions. However, the conduction model may not be appropriate in describing thermal sterilization of even viscous liquid foods and may cause considerable error in the prediction of the important parameters such as slowest heating zone (SHZ) temperature and lethality. The longer time for sterilization recommended by the conduction model may lead to overprocessing and an unacceptable food product. The objective of this work is to quantify the faster temperature rise in the food can due to natural convection when compared to the temperature rise obtained by only conductive heating. The consequent enhancement in lethalities is also reported. In addition, this work's objective is to investigate how quickly the natural convective heat transfer effects begin to dominate over the solely conduction heating mode. The volume-averaged temperature as well as the SHZ temperature variations with time was calculated for the convection-augmented mode using computational fluid dynamics (CFD) simulations. Lethality values were then calculated based on volume-averaged temperature as well as the SHZ temperature. Food cans of different aspect ratios and food medium thermal conductivities are considered in this analysis. For the food system investigated, the critical Fourier number at which the transition to convection-augmented mode of heat transfer occurred is identified and explained from scaling considerations. In the conduction-dominated mode, it was possible to use analytical solutions to predict the volume-averaged and SHZ temperatures of the liquid food undergoing thermal sterilization. The Nusselt number correlation developed by Kannan and Gourisankar (2008) was used in the lumped parameter transient heat transfer model to predict the volume-averaged temperatures in the convection-dominated region. The volume-averaged temperatures from this approach were found to be in good agreement with the CFD simulation results. The time predicted for the SHZ to reach the minimum sterilization temperature was significantly lower when convective heating was also considered. The volume-averaged temperature and SHZ temperature enabled an estimation of overall sterility levels attained and minimum sterility levels prevalent inside the can, respectively. Even though the volume-averaged temperature increase due to convection was only about 10 K, the resulting accumulated lethality values were higher by an order of magnitude. The increase in SHZ temperatures was much higher in the convection-augmented mode, and consequently greater integrated lethalities were attained. The simple conduction model that is amenable to analytical solution cannot be used to approximate the heat-transfer-related phenomena even for "quick estimation" purposes when convection effects are significant. This precaution is found necessary even for the reasonably high viscous carboxy methyl cellulose system, whose average viscosity values ranged between 13 and 3 Pa s during the course of the sterilization process. © 2009 Springer Science + Business Media, LLC.