Reduced N-acetyl aspartate (NAA) and increased myo-inositol (MI) levels have been reported in patients with dementia of Alzheimer type (DAT) in comparison with controls. We wished to assess the validity of these findings and to evaluate possible correlations of metabolite proportions with cognitive dysfunction in DAT. Twelve patients with DAT and 10 healthy age-matched controls were included. The severity of dementia was assessed using different scales including the Mini-Mental State Examination. MRS was performed with a conventional 1.5 Tesla scanner in a single voxel in the centrum semi-ovale (TE = 30 ms or TE = 136 ms; TR = 1500 ms). The evaluation of MRS results was limited by low interrater, intermeasurement (different echo times) and test-retest reliabilities, by a high interindividual variance and by the failure to measure absolute metabolite concentrations. These problems in mind, it was remarkable that previously reported reductions of NAA levels in patients with DAT could be reproduced in the present sample. The proportion of NAA was diminished in demented subjects in comparison with controls (37% vs 44.90%, short TE). A non-significant trend towards minor reductions of creatine, choline and MI proportions in these subjects might indicate that proportions of other metabolites necessarily increase when NAA is reduced. Cognitive dysfunction of demented subjects was significantly correlated with reductions of NAA, but not with increases of MI. Due to the present technical and methodological problems and to the non-specificity of findings, proton MRS cannot be applied to support the diagnosis of DAT in a clinical setting.