Abstract: We investigated 71 horses at five stud farms. Their breeds were Thoroughbred (n = 48), Anglo-Arab (n = 7), Thoroughbred mixed breed (n = 5), Appaloosa (n = 3), Selle Francais (n = 2), Dutch Warmblood, Haflinger, Pinto, Quarter Horse and Westfalen (n = 1, respectively), and one horse's breed was not known, but was a heavy horse. Their genders were stallion (n = 5), gelding (n = 41) and female (n = 25). Their ages ranged from 4 to 24 and the average was 11.1 +/- 5.3 (S.D.). The horses' behaviour was twice observed for 2 h both before and after feeding, using a scan sampling technique at 2 min intervals. We investigated the relationship between management factors, age, sex, breed and behavioural patterns of stabled horses, and the occurrence of weaving by a least-squares analysis of variance, a correlation analysis, Chi-square test and an analysis of behavioural sequences. Thoroughbred horses displayed weaving more than other breeds (P < 0.01) and horses in box stalls which were face to face with each other displayed weaving more than those in box stalls which were formed in a line (P < 0.01). The time budget of weaving correlated negatively with the amount of hay-cube fed (kg/day) (P < 0.05) and time budgets of drinking, bedding investigation, looking and coprophagia (P < 0.01, respectively). From the analysis of behavioural sequences, weaving followed resting (P < 0.01), looking (P < 0.01) and pawing (P < 0.05), and was followed by them (P < 0.01, P < 0.01 and P < 0.05, respectively). The least-squares analysis of variance revealed that Thoroughbred horses investigated the bedding more than other breeds (P < 0.05), horses that are usually in contact with `mainly familiar people' investigated the bedding significantly less than those usually in contact with `people including strangers' (P < 0.05), and horses in box stalls which were face to face with each other tended to investigate the bedding more than those in box stalls which were formed in a line. The time spent feeding was related to food type (min/kg), and tended to be negatively correlated with the time budget of bedding investigation behaviour. These results indicate that weaving in horses is affected by breed, stable design, feed type, and other behavioural traits.