The significance of the statistical character of wall-cooled packed-bed reactors was studied by measuring angular temperature variations, which result from the random nature of the packing. These are neglected in present-day mathematical models designed to describe the reactor behavior. The amplitude of these variations was determined experimentally as a function of the mass flux and the position in the tube, under conditions of heat transfer as well as of reaction. Angular variations are significant whenever radial temperature gradients are large. Different methods used to average observed variations were surveyed. For the system presented, the angular variations have only a limited influence on the reaction rates if the angular averaged temperature is used. Thus, a two-dimensional deterministic continuum model can be used, notwithstanding the statistical character of the packed bed.