There is a limited choice of psychometric tests for Portuguese speaking people which have been evaluated in well defined groups. A Portuguese version of CERAD neuropsychological battery was applied to a control group of healthy elderly (CG) (mean age 75.1 years/ education 7.9 years), 31 Alzheimer disease (AD) patients classified by clinical dementia rating (CDR) as CDR1 (71.4/ 9.0) and 12 AD patients CDR 2 (74.1/ 9.3). Cut-off points were: verbal fluency-11; modified Boston naming-12; Mini-mental State Examination (MMSE) -26; word list memory-13; constructional praxis-9; word recall-3, word recognition-7; praxis recall-4. There was a significant difference between CG and AD-CDR1 (p<0.0001) for all tests. There was a less significant difference for constructional praxis and no difference for Boston naming. Comparison between AD-CDR1 and AD-CDR2 showed difference only for MMSE, verbal fluency, and Boston naming. The performance of CG was similar to that of a US control sample with comparable education level. These results indicate that this adaptation may be useful for the diagnosis of mild dementia but further studies are needed to define cut-offs for illiterates/low education people.