To determine the prevalence of multiple sclerosis (MS) in the Tayside Health Board Area, Scotland, we carried out a population-based survey using four intersecting sources (Neurology Department records, a survey of general practitioners, Scottish Morbidity Records of discharges from hospitals and visual evoked response requests). A two-source capture-recapture model estimated survey coverage, and direct age-sex standardisation was used to take account of different population structures. Comparisons were made between the prevalence in Scotland and that in the rest of the United Kingdom. A total of 727 (definite and probable) and 880 cases (early, probable and possible) were identified using the criteria of Poser et al. and those of Allison and Millar in a population of 395,600 (1995 mid-year estimate). The prevalence of MS on 1 September 1996 was 184/100,000 (95% confidence interval 171-198) and 222/100,000 (95% confidence interval 210-240), respectively. The two-source capture-recapture model estimated that the survey was between 93% and 99% complete. Age-sex standardisation eliminated certain north-south differences in prevalence when comparisons were made with previous surveys. Diagnostic misclassification may also have influenced reported prevalence statistics. The prevalence is similar to that found in revised figures from the Grampian region in Scotland but significantly higher than recent estimates from England and Wales. Methodological differences may account for most of the reported differences between north and south, although there is still evidence to suggest that MS is more prevalent in northern Great Britain and Northern Ireland than in England and Wales.