In this paper, a logarithmic model and Giner's differential model are compared in the process of simulating the stationary deep-bed drying of a product. The logarithmic model can be interpreted as the propagation of boundary conditions along characteristic lines throughout the bed. This propagation depends on drying air velocity and becomes less pronounced with respect to the depth of the bed. Therefore, transfer time, as defined by the ratio of the depth of the bed to the inlet velocity of drying air is a relevant parameter that can strongly influence the behaviour of the logarithmic model. The logarithmic model is formulated using dimensionless numbers derived from the theory of Transfer Units. The logarithmic model developed in this paper is compared to numerical approximations of the differential model. This comparison has been performed in terms of accuracy (with respect to Giner's differential model). parsimony (with respect to the number of variables and parameters of the models) and specialization (with respect to the scope of the application fields of the models). Drying conditions ensuring that the logarithmic model describes drying phenomena as well as Giner's partial differential equation model are determined. The application field of the logarithmic model appears to be more restrictive than Giner's model. however. it is much more parsimonious than the numerical approximations of the differential model and it may be used by decision-support systems for the design of driers. As an illustration, numerical simulations based on wheat drying are performed. (c) 2005 Elsevier Ltd. All rights reserved.