A study was conducted, which simultaneously used GPs and epidemiological assessment procedures (MMSE, CAMDEX), to identify individuals with DSM-III-R dementia in the same population. In addition, a 1-year follow-up assessment was conducted in patients with a CAMDEX-diagnosis of dementia. In the non-institutionalized group, which was assessed both by the GP and with the epidemiological test battery, the prevalence of dementia according to the GP was 2.2%, whereas the prevalence based on the epidemiological approach amounted to 5.2%. In general CAMDEX-diagnoses of dementia were confirmed at 1-year follow-up, and thus the discrepancy between the two prevalence estimates must be attributed to the low sensitivity of the GPs. Sensitivity of the GP was related to help-seeking behaviour, with low sensitivity in patients with a low contact rate. It was also related to the use of less specific diagnostic labels by the GP (cognitive impairment), and to poor recognition of cognitive impairment in patients who visited their GP.