A semantic battery comprising items from living and nonliving categories was administered to a sample of subjects suffering from probable Alzheimer's type dementia (n=20). The results were analyzed to test the role of semantic distance in predicting group accuracy and its possible causal link with the phenomenon of a category-specific deficit for living things in the experimental population. Our findings confirm a category effect favoring nonliving items (over and above the role of confounding variables) in patients with Alzheimer's disease and support the major role of an imbalance of semantic distance in the testing material in the genesis of this phenomenon.