Conclusion Using only composition and initial freezing temperature data and knowledge of the food's temperature, a multi-step thermal conductivity prediction procedure provided the most accurate thermal conductivity predictions for the range of foods considered. However, the single-step Co-Continuous model also provided predictions within ±20% other than for food containing both ice and air voids. On balance, however, the multi-step procedure is recommended for general use, since it provided the most accurate predictions over the widest range of foods, and also because there is scope for enhancements to be made within its framework, unlike the single-step method. It is emphasised that this methodology is intended for first approximations based on the minimum of input data, rather than as a rigorous modelling framework.