Abstract: REASONS FOR PERFORMING STUDY: Studies on the prevalence of behavioural disorders in horses and on associated risk factors have revealed inconsistent results. There are many studies on the neuropharmacological, surgical or mechanical therapy of stereotypies, but little is known about their causation. OBJECTIVES: To explore risk factors associated with the occurrence of behavioural disorders in horses. METHODS: A sample of horse owners, selected randomly and representative for Switzerland, was contacted in a postal survey. Answers were provided for 622 stables (response rate 35.2%). Individual data of 2,341 horses were examined with path analysis (multivariable linear and logistic regression), and adjustment made for possible confounding effects due to age and breed. RESULTS: Out of 60 possible risk factors, 11 were associated with the outcome at the univariable level (null-hypothesis path model) and 3 factors remained after the backward logistic regression procedure. Mature Warmbloods and Thoroughbreds, assessed by the owners to be reactive, fed 4 times a day and without daily pasture, had increased odds of displaying crib-biting, weaving and box-walking. Furthermore, indirect associations of 5 factors with the outcome were identified. CONCLUSIONS: The final logistic regression model of risk factors leads to the hypotheses that causal prevention of stereotypic behaviours should be based upon housing and management conditions which allow tactile contact with other horses (e.g. mutual grooming), daily free movement (paddock or pasture), as well as the provision of high amounts of roughage but of little or no concentrates. POTENTIAL CLINICAL RELEVANCE: It is one of the aims of population medicine to prevent the development of behavioural disorders. Further research is needed to test the concluding hypotheses in experimental studies or to verify them in the context of similar observational studies.