A concurrent case-control study of 12 UK racecourses was made between March 1, 2000, and August 31, 2001, to identify and quantify the factors associated with the risk of horses falling in steeplechase races. Cases were defined as a jumping effort at a steeplechase fence that resulted in a fall and controls were defined as a successful jumping effort over any steeplechase fence at any of the 12 racecourses within 14 days before or after the case fall. Information on the horse, the jockey and the race were collected and all the fences on all the courses were surveyed. Conditional logistic regression was used to examine the relationships between the predictor variables and the risk of falling. There was one fall per 254 jumping efforts. The risk of a horse falling decreased the more times it had raced on a particular racecourse. The number of fences, the distance from the previous fence and the nature of the previous fence also affected the risk of falling. If the previous fence was a water jump the risk of falling increased; fences that were sited on flat or slight uphill gradients (up to approximately 1 in 25) were associated with a lower risk of horses falling than downhill fences, and higher takeoff boards were associated with a higher risk of falling.