BACKGROUND Factors that determine geographical differences in incidence rates of 'probable' presenile Alzheimer's disease (AD PSD) may help to clarify the possible role of the environment in its aetiology. METHOD We have ascertained the treated incidence of AD PSD in Scotland by scrutiny of hospital cassenotes and searched for cases outside hospital settings. Small area geographical analysis compared the observed distribution of cases (each allocated to one of Scotland's 898 postcode sectors) with the estimated random distribution of cases. RESULTS There was non-random geographical distribution of AD PSD but not of the comparison conditions (vascular dementia (VaD), motor neurone disease, prostatic or ovarian cancers). Substantial differences between Scottish regions were probably not attributable to methodological artefact, as other techniques of case finding showed the same regional differences. CONCLUSIONS The observed differences in incidence of AD PSD between Scotland's regions are real and some localities have a higher incidence, mostly in central Scotland.