The objectives of this study were first to describe the pattern of the epidemic of bovine spongiform encephalopathy (BSE) in Great Britain in terms of the temporal change in the proportion of all cattle holdings that had experienced at least one confirmed case of BSE to June 30, 1997, and secondly to identify risk factors that influenced the date of onset of a holding's first confirmed BSE case. The analyses were based on the population of British cattle at risk, derived from agricultural census data collected between 1986 and 1996, and the BSE case data collected up to June 30, 1997. The unit of interest was the cattle holding and included all those recorded at least once on annual agricultural censuses conducted between June 30,1986, and June 30,1996. The outcome of interest was the date on which clinical signs were recorded in a holding's first confirmed case of BSE, termed the BSE onset date. Univariate and multivariate survival analysis techniques were used to describe the temporal pattern of the epidemic. The BSE epidemic in Great Britain started in November 1986, with the majority of affected holdings having their BSE onset date after February 1992. After adjusting for the effect of the size and type of holding, holdings in the south of England (specifically those in the Eastern, South east and South west regions) had 2.22 to 2.43 (95 per cent confidence interval [ci] 2.07 to 2.58) times as great a monthly hazard of having a BSE index case as holdings in Scotland. After adjusting for the effect of region and type of holding, holdings with more than 53 adult cattle had 5.91 (95 per cent ci 5.62 to 6.21) times as great a monthly hazard of having a BSE index case as holdings with seven to 21 adult cattle. Dairy holdings had 3.06 (95 per cent ci 2.96 to 3.16) times as great a monthly hazard of having a BSE index case as beef suckler holdings. These analyses show that there were different rates of onset in different regions and in holdings of different sizes and types, that the epidemic was propagated most strongly in the south of the country, and that the growth of the epidemic followed essentially the same pattern in each region of the country, with modest temporal lags between them. The control measures imposed in 1988 and 1990 brought the expansion of the epidemic under control, although the rate of progress was slowed by those regions where the effectiveness of the control methods took some time to take full effect.