Thermal properties of foods are vital inputs for many food process models. With the recent advances in mathematical modelling and significant reduction in the cost of computational power, uncertainties in model inputs are more and more becoming the limiting factor in model accuracy rather than the model formulation or solution process. In this chapter methods and models are presented for predicting thermal properties based solely on data for the composition of the food in terms of its basic components (liquid water, ice, protein, fat, carbohydrate, ash and air). This type of model provides genuine predictions of thermal properties since no thermal property measurements are required, as is necessary with some effective property models that may be found in the literature. Foods are divided into different classes, depending on the difficulties they pose for thermal properties prediction. Simple guidelines for thermal property prediction are presented along with worked examples to serve as illustrations.