Microwave heating of prepared meals in sealed retail containers can be used to extend their shelf life. A numerical model to predict food temperatures after heating was developed and tested for use in the design of microwave cavities and foods. The model uses a finite difference scheme to predict temperatures by solving Maxwell＇s equations for electromagnetic fields and the heat conduction equation. Complex-shaped trays containing mashed potato were heated in a microwave tunnel operating at 2.3 kW for 60 s. The temperature distribution in the food was then measured over a 120 s period during which the food partially cooled. The heat capacity and dielectric properties of the potato, and the power dissipated in the food during heating, were measured. Intricate temperature profiles, with large temperature gradients, were measured and predicted. The model predicted the positions of regions of highest and lowest temperature but large differences of up to 30 degrees C were found between measurements and predictions. The differences were largest near to regions where the finite difference mesh, which consisted of 4 mm cubes, was least able to simulate accurately the shape of the tray Predictions carried out on a high performance workstation required 43 h of cpu. Further studies using a finer mesh on a more powerful computer along with experiments using simple regular shapes of food, are required to test the model further.