The strength of two coarse-grained (grain size>200m) cubic ceramics, a magnesium aluminate spinel (MgAl2O4) and an AlON , along with a fine-grained (1.5m) MgAl2O4, was determined by conducting a series of four-point and equibiaxial flexure tests on specimens of different sizes. Weibull strength size scaling revealed a linear relationship on a log-log plot between average flexure strength and effective specimen area for the fine-grained spinel, but a nonlinear relationship for both coarse-grained materials. Initial fractography showed that each material had a single flaw population limiting the strength over the entire specimen size range, which does not account for the nonlinear size scaling relationship in the two coarse-grained materials. However, further fractography revealed that in both materials there was an initial flaw and a critical flaw. The former appears to be machining/polishing damage that started the fracture process while the latter was a cleaved grain in AlON or a cracked grain boundary in the HP/HIP spinel that lead to fracture of the specimen. The difference between the initial and critical flaw size coupled with a detailed analysis of the strength as a function of test specimen thickness accounted for the nonlinear strength size scaling relationship. As a result, strength values obtained using thin test specimens can lead to an erroneous strength prediction for large components made of these ceramics. The implication of these findings is that strength tests must be conducted using appropriately thick specimens to obtain a representative strength value. If appropriately thick specimens cannot be tested, then fractography must be conducted to determine the flaw size. If the flaw size is sufficiently large, compared with the specimen thickness, then the strength must be adjusted according to a stress field correction factor to obtain a more accurate strength value.