Wesley Salmon has recently proposed a new theory of scientific explanation based on a model which he calls the statistical-relevance model. It is intended primarily as an account of the structure of explanations of particular events—explanations which, according to Salmon, are very often motivated largely by practical concerns. Two important features of this account are the concepts of homogeneity and screening off. In this paper we argue that the employment of these two concepts (which, in fact, are intimately connected) is fundamentally at odds with the desire to capture everyday explanations of the practical sort, and that it leads instead to an account in which the only satisfactory explanations are those expressed in the very deepest physical terms.