Heuristics (rules of thumb) are proposed for extending the chilling time prediction method proposed in Parr I and tested for model substances in Part 2 to real foods with non-unity water activity. Guidance is given for selecting three water activity values-one representing the maximally wetted starting condition, one representing the mean value during the active chilling phase, and the third describing the surface condition in the quasi-equilibrium state reached at the end of chilling. Chilling rimes of a product retaining a well-wetted surface during chilling (peeled carrots) were predicted to within approx. -10 to +15% of measured values. At least part of this difference can be attributed to experimental error: For a product not retaining a well-wetted surface due to skirt resistance (unpeeled carrots) predictions of only slightly lower accuracy were achieved. Accurate prediction of chilling time across a wide range of conditions by a simple algebraic prediction method is possible in spite of the complexity introduced by evaporative cooling at the product surface with water activity less than I.