Previous studies have shown that the verbal fluency test (VFT) is a sensitive measure of cognitive dysfunction in Alzheimer's disease (AD). However, other studies have shown that the performances were significantly influenced by education in the normal elderly population. In order to examine the utility of the VFT as a tool for screening for AD, it is necessary to study the effect of education not only in the cognitively intact population but also in the population of early AD patients. Patients with AD (n = 345) and individuals with amnestic type of mild cognitive impairment (MCI) (n = 123) were asked to generate as many words as possible belonging to a category "animal" and beginning with " [ka]" in syllabic Japanese "kana" script. In order to determine the education effect after adjusting for age and cognitive state on the VFT performance in early stage of AD, we performed multiple regression analysis with 396 individuals including both amnestic MCI and AD. After adjusting for patients' age, sex, and cognitive state, the years of education were significantly related to category fluency test scores, but not significantly related to letter fluency test scores. Our results demonstrated that a category fluency performance reflected not only AD-specific changes but also educational background. These results suggest the limitation of using the category fluency task for screening subjects at risk for developing AD without taking subjects' educational background into consideration.