Horses’ behaviour in riding schools is of a wide importance regarding caretakers’ and users’ security. Previous studies showed that numerous intrinsic or extrinsic factors, such as breed, housing conditions, sire, and work could modulate horses’ emotional and cognitive abilities. Here, we compared the behaviour of 184 horses from 22 riding schools, practicing the same type of work, but differing in particular in terms of housing conditions proposed. Three emotionality tests (the arena and the novel object tests, with the horse released alone in a known arena and faced or not with a novel object, the bridge test with the horse led by the experimenter across a mattress) and one instrumental learning test (the chest test: in the box, the horse tries to open a wooden chest) were used to characterize the schools and determine how general management could explain the potential differences between sites. For the arena and novel object tests, emotionality indices were calculated, as well as frequencies of occurrence of active locomotion patterns (trot, canter and passage). For the bridge and the chest tests, times spent to cross or open were recorded. The ANOVAs and FCA showed that riding schools could be classified amongst four categories according to the behaviour of their horses. In accordance to previous studies, two factors appeared to be particularly relevant: breed, that impacts on the time to cross the bridge (Kruskall-Wallis, H(14, N = 184)=27.08, P<0.05) and housing conditions with box housed horses showing higher levels of emotionality (MW: P<0.05). These results underline the importance for riding schools owners to take into account the individual characteristics of the horse, and to have a questioning about the impact of general management on horses’ reactivity, and the consequences in terms of security. Other factors (feeding practices, work conditions...) remain to be tested.