omesticated horses mainly live in individual boxes, a housing system reported as compromising animal welfare. A common practice in riding schools involves offering a temporary period on pasture with conspecifics to alleviate the impact of long-term deprivation triggered by boxes. The aim of this study was to assess the influence of this practice using four behavioural indicators reflecting a compromised welfare state: stereotypies; aggressive behaviours towards humans; the "withdrawn posture" reflecting unresponsiveness to the environment; and the "alert posture" indicating hypervigilance. A group of 31 horses was monitored before, during and after a period of 1.5 months on pasture (intra-group comparisons) and their behaviours were compared to those of 29 horses kept in individual boxes during the study (inter-group comparisons). On pasture, no stereotypies and aggressive behaviours towards humans were observed, and the occurrence of the "alert posture" decreased, although the results were not significant. An increase in the expression of natural behaviours such as locomotion, exploration and social behaviours was observed. However, the expression of the "withdrawn posture" increased during the first five days on pasture (Friedman: P<0.001; Wilcoxon signed-rank: P<0.001) before a decrease was observed after 20 days, returning to the level previously observed when horses were in boxes (Wilcoxon signed-rank: P<0.01). These results suggest that going out to pasture can positively influence the welfare state of horses, but also that several days of adaptation are needed, probably due to the novelty of the environmental and social conditions. The most noticeable result occurred when horses returned to individual boxes. A sharp increase in the occurrence of stereotypies (Cochran test: P<0.001; Chi(2) of homogeneity: P = 0.05), of the "withdrawn" (Friedman: P<0.001; Wilcoxon rank-sum: P<0.05) and the "alert" postures (Friedman: P<0.01; Wilcoxon rank-sum: P<0.001) was observed during the first five days of returning to confinement. The expression of the majority of natural behaviours immediately returned to the level observed during the pre-pasture period. After three months, the expression of the four welfare indicators was not different from that in the pre-pasture period. These results demonstrate that the beneficial effects likely to be induced by the pasture do not last when horses return to individual boxes and that the environmental change causes deleterious shortterm effects on the animals' welfare state. It would thus be recommended to keep domestic horses permanently on pasture when possible.