Some brain-damaged patients seem to have more difficulty retrieving information about living things than about nonliving things. Does this reflect a distinction between two different underlying brain systems specialised for knowledge of living and nonliving things, or merely a difference in the diffculty of retrieving these two kinds of knowledge from a single semantic memory system ? Two recent articles (Funnell & Sheridan, 1992; Stewart, Parkin, & Hunkin, 1992) have concluded the latter, on the basis of experiments in which various determinants of naming difficulty were matched for living and nonliving things and the previously observed dissociation was found to vanish. We argue that these null effects are due to insufficient power, and that knowledge of living things can be selectively impaired. In support of this, we use the same stimulus materials, design, and data analysis as did Funnell and Sheridan (1992), with two different subjects having the same aetiology and general behaviour in the domain of semantic memory, and show that: (1) when, like the authors of these articles, we use only a single replication of each item, no effect is found, and (2) when we use more replications of the same items, highly significant differences between