The study consisted of a survey of all new cases of Bell's palsy occurring between 1992 and 1996 in practices contributing data to the UK General Practice Research Database (GPRD). Data were extracted on age, sex, date of episode of Bell's palsy, household number, episodes of herpes simplex, treatment prescribed and referral to relevant hospital departments. A total of 2473 cases of Bell's palsy were identified. The overall incidence for the study period was 20.2 per 100 000 person years of follow-up (95% CI 19.4-21.0). Incidence increased with age. There was no difference in incidence according to sex or season but there were significant changes over time: incidence was higher in the first year of the study period than in subsequent years. There was no clustering of cases in households and no evidence of any tendency for herpes simplex infections to precede Bell's palsy. About 36% of cases were treated with oral steroids and 19% of episodes resulted in hospital referral. In conclusion, Bell's palsy is seen mainly in a primary care setting. The majority of cases are treated expectantly without drugs. Lack of household clustering and lack of a tendency of herpes simplex infections to precede Bell's palsy do not support a viral aetiology.