We evaluated the Addenbrooke's cognitive examination (ACE), a simple instrument to differentiate frontotemporal dementia (FTD) from Alzheimer's disease (AD), in our dementia patients clinic population. The Verbal-Language/Orientation-Memory (VLOM) ratio, which compares its language and memory scores, determines whether FTD or AD is more likely. The ACE was translated into French with adaptation maintaining the number of words in the name and address learning and delayed recall test, and with cultural adaptation for the semantic memory. The 85 included subjects had no evidence of two or more organic pathologies, after at least six months of follow-up, and an MMSE score>or=20/30. Patients with cognitive impairment due to alcohol intake were excluded. The diagnosis of a specific dementing illness was based on the consensus of the neurologist and neuropsychologists in the team. Thereafter, another neurologist expert in dementia, blinded to the ACE result and to the diagnosis and treatment, reviewed all cases files and proposed a diagnosis. A diagnostic agreement was reached for 79 cases (92.9%) with 40 (50.6%) dementia: 25 AD (62.5 %), 9 FTD (22.5 %). We estimated that the sensitivity for detecting dementia of an ACE score<or=83/100 was 90% with a specificity of 64.1%. When the ACE score was<or=88/100, the sensitivity for diagnosing FTD of a VLOM ratio<2.2 was 11.1% with a specificity of 88 % and the sensitivity for diagnosing AD of a VLOM ratio>3.2 was 72%,with a specificity of 69.4%. We conclude that, when used as originally proposed, ACE is very accurate for the detection of dementia, but much less effective in discriminating the most common frontal variant of FTD.