Summary A case-control study was performed to investigate the epidemiology of grass sickness in the United Kingdom from 1992 to 1995. Data were collected by means of postal questionnaire when cases of grass sickness were identified. Sets of three questionnaires were posted to owners of, or veterinary surgeons attending, cases of grass sickness, with a request to provide information on the case, on one healthy animal on the same premises as the case and on another healthy animal on other premises. Controls were matched to cases by date of onset. After univariate analyses, the probability of grass sickness in horses was modelled using conditional logistic regression techniques. Young animals were found to be at increased nick of grass sickness and females were less likely to becomeaffected, as were animals that had a history of contact with previous cases of the disease. The probability of grass sickness was higher in animals that were on premises where grass sickness had previously occurred, particularly if this was recent. Animals were at particular risk of disease if they changed fields within the previous 2 weeks; the risk thereafter reduced with time. The disease had a seasonal pattern, with a peak from April to June. More than 95% of cases had access to grazing, and 66% occurred after 2 week periods of predominantly dry weather.