The diagnosis of dementia can be difficult, yet diagnostic accuracy has important prognostic and therapeutic implications. Nevertheless, conventional electroencephalography (EEG) has always played a secondary role in dementia investigation. More recently quantitative EEG (qEEG) has allowed more detailed and objective analysis of EEG data, but there is still no clearly defined clinical role for qEEG. We have used relative power qEEG measures made during resting and active brain conditions (serial subtraction and odour detection tasks) to differentiate between demented and non-demented subjects, and between subjects with different forms of dementia. Electroencephalograms were obtained from 15 subjects with clinically diagnosed Alzheimer's disease (AD), 16 with a clinical diagnosis of vascular dementia (VaD), and 16 non-demented control subjects. Discriminate function analyses were used to differentiate groups according to task, electrode site, and frequency bandwidth. Correct classification, as demented or non-demented, was made for 93% of cases using qEEG comparisons of resting states with eyes closed and eyes opened. Almost all subjects with AD and VaD were correctly classified with qEEG recorded during odour detection (95%). qEEG for serial subtraction correctly classified AD and VaD in 91% of the dementia group. These results have important implications for future qEEG research, and may be pertinent to the precision of diagnosis in patients with dementia.