The stroke classification developed for use in the Oxfordshire Community Stroke Project (OCSP) is simple, clinically meaningful and requires no investigations. However, its inter-rater reliability in a community setting is unknown. We aimed to assess the inter-rater reliability of this classification among neurologists and nurses within a community-based stroke incidence study. Fifty-four stroke patients with cerebral infarction who were registered in the North East Melbourne Stroke Incidence Study (NEMESIS) were assigned OCSP classifications by two neurologists and one of seven research nurses. There was moderate agreement between neurologists (kappa = 0.53), fair agreement between neurologist 1 and nurse (kappa = 0.31) and moderate agreement between neurologist 2 and nurse (kappa = 0.45). Disagreement about the neurological signs was an important reason for classification differences. The OCSP classification can be easily applied in a community setting with moderate inter-rater reliability and is thus a useful instrument for commun ity-based epidemiological studies.