The Mini-Mental-State Examination (MMSE) is widely used as a screening tool for dementia in epidemiological studies. Its applicability in population-based studies is nevertheless limited by its low specificity. The effect of age and educational level have been usually ignored when cut-off scores have been selected. The aim of this study was to evaluate the effect of age and educational level on the MMSE scores in a representative sample of subjects older than 70 and provide adjusted normalised data according to these two variables, after excluding subjects with dementia or cognitive decline. Population-based, cross-sectional and longitudinal study of a representative cohort of 1367 subjects older than 70. All subjects with suspected dementia or cognitive decline received a neurological evaluation where clinical and etiological diagnosis were established. Normal MMSE scores, as defined by the 10th percentile, varied significantly across age and educational level groups. Exclusion of demented or cognitively declined patients from the reference population reduced the variability and "range of normality", but this remained excessively high in the older and less educated groups. The use of different cut-off points for each age and educational level groups may improve the specificity and applicability of the MMSE in population-based epidemiological studies. However, the wide amplitude of the range of normality suggests that different approaches, other than this vibariate analysis, may prove more adequate in the selection of cut-off scores for the MMSE.