A category specific effect in naming tasks has been reported in patients with Alzheimer's dementia. Nonetheless, naming tasks are frequently affected by methodological problems, e.g., ceiling effects for controls and “nuisance variables” that may confound results. Semantic fluency tasks could help to address some of these methodological difficulties, because they are not prone to producing ceiling effects and are less influenced by nuisance variables. One hundred and thirty-three participants (61 patients with probable AD; and 72 controls: 36 young and 36 elderly) were evaluated with semantic fluency tasks in 14 semantic categories. Category fluency was affected both by dementia and by age: while in nonliving-thing categories there were differences among the three groups, in living thing categories larger lexical categories produced bigger differences among groups. Sex differences in fluency emerged, but these were moderated both by age and by pathology. In particular, fluency was smaller in female than male Alzheimer patients for almost every subcategory.