The thermal shock resistance of a brittle solid is analysed for an orthotropic plate suddenly exposed to a convective medium of different temperature. Two types of plate are considered: (i) a plate containing a distribution of flaws such as pores, for which a stress-based fracture criterion is appropriate, and (ii) a plate containing a single dominant crack aligned with the through-thickness direction, for which a critical stress intensity factor criterion is appropriate. First, the temperature and stress histories in the plate are given for the full range of Blot number. For the case of a cold shock, the stress held is tensile near the surface of the plate and gives rise to a mode I stress intensity factor for a pre-existing crack at the surface of the plate. Alternatively; for the case of hot shock, the stress field is tensile at the centre of the plate and gives rise to a mode I stress intensity factor for a pre-existing crack at the centre of the plate. Lower bound solutions are obtained for the maximum thermal shock that the plate can sustain without catastrophic failure according to the two distinct criteria: (i) maximum local tensile stress equals the tensile strength of the solid, and (ii) maximum stress intensity factor for the pre-existing representative crack equals the fracture toughness of the solid. Merit indices of material properties are deduced; and optimal materials are selected on the basis of these criteria, for the case of a high Blot number (high surface heat transfer) and a low Blot number (low surface heat transfer). The relative merit of candidate materials depends upon the magnitude of the Blot number, and upon the choice of failure criterion. The effect of porosity on thermal shock resistance is also explored: it is predicted that the presence of porosity is generally beneficial if the failure is dominated by a pre-existing crack. Finally, the analysis is used to develop merit indices for thermal fatigue. (C) 1998 Acta Metallurgica me. Published by Elsevier Science Ltd. All rights reserved.