The purpose of the study was to investigate the risk of stroke with typical and atypical anti-psychotics in elderly subjects, weighting for a number of known risk factors, including dementia. Data were retrospectively drawn from the primary care setting from the Health Search Database, which stores information on about 1.5% of the total Italian population served by general practitioners. All elderly patients (65+ years) prescribed an anti-psychotic in monotherapy from January 2000 to June 2003 were selected for the study. A cohort of patients not exposed to anti-psychotics was taken from the same database. Subjects who had previously had a stroke were excluded. The main outcome measure was the incidence of first-ever stroke during exposure to an anti-psychotic.The sample included non-users (69,939), users of atypicals (599), butyrophenones (749), phenotiazines (907) and substituted benzamides (1,968). The crude incidence of stroke in subjects not exposed to anti-psychotics was 12.0/1000 person-years. Risk was significantly higher for those on butyrophenones (47.1/1000), phenotiazines (72.7/1000) and in the atypical anti-psychotic group (47.4/1000). Substituted benzamides had an almost significant higher risk (25.0/1000). Cox regression modelling, weighting for demographic and clinical variables with non-users as the reference group, showed that the risk for stroke was 5.79 times for phenotiazines, 3.55 times for butyrophenones, and 2.46 times for atypicals. Clinicians should be cautious in prescribing phenotiazines and butyrophenones in elderly patients, since the risk for stroke would seem comparable or even greater than with atypicals.