Brief cognitive tests are widely used for dementia screening, but are usually influenced by education. The present work aimed to determine education-adjusted cut-off scores and correspondent sensitivity (S) and specificity (Sp) values of the category fluency (CF) test (animals/min) as a screening tool for Alzheimer disease (AD). Eighty-eight patients with mild AD and 117 normal matched controls were evaluated. Patients and controls were divided into 4 groups according to educational level (illiterates, 1 to 3, 4 to 7, and > or =8 y) and were administered the CF test. In each group, cut-off values were determined using Receiver Operator Characteristic analysis. The areas under Receiver Operator Characteristic curves were 0.922/0.914/0.963/0.954, for the identification of AD among the groups of illiterates, 1 to 3, 4 to 7, and > or =8 years of education, respectively. The cut-off points for each group were 9 (S=90.5% and Sp=80.6%) for illiterates; 12 (S=95.2%% and Sp=80.0%) for 1 to 3 years; 12 (S=91.3% and Sp=91.9%) for 4 to 7 years, and 13 for those with > or =8 years (S=82.6% and Sp=100.0%). These results suggest that the CF may be a useful screening test for mild AD in different educational levels, with the need of using specific cut-off scores adjusted for each range of schooling.