Domestic horses are faced with social challenges throughout their lives due to limitations in social contact, space restrictions and frequent changes in social companionship. This is in contrast to natural conditions where horses live in relatively stable harem bands. Currently, little is known about how repeated regrouping affect horse behaviour and welfare, and it is unknown whether horses may adapt to regrouping. In this study, we aimed to investigate the effects of an unstable group structure, caused by weekly regroupings, on behaviour and frequency of injuries in young horses. Forty-five horses were included in the study and were randomly assigned to the treatments; Stable (S; seven groups of three horses) or Unstable (U; eight groups of three horses). The experimental period lasted 7 weeks, during which horses in Stable groups remained in the same group, whereas one horse was exchanged between Unstable groups every week. The groups were kept in 80m×80m grass-covered enclosures and were fed additional roughage on the ground daily. Social interactions were recorded in Unstable groups immediately after each regrouping (30min), and in both Stable and Unstable groups on day 1, 3 and 6 after each regrouping (2×20min/group/day). Injuries were scored by the end of the experimental period. The level of aggression shown by horses in Unstable groups immediately after regrouping was not affected by week (F5,35=0.42, P=0.83), indicating that horses neither habituated, nor sensitized, to repeated regrouping. Compared to horses in Stable groups, more agonistic behaviour was shown by horses in Unstable groups (i.e. non-contact agonistic; F1,65=5.60, P=0.02), whereas there was no treatment effect on other variables. The level of play behaviour appeared, however, to be more variable in Unstable groups. There was a significant effect of week on the level of contact agonistic interactions as well as greeting behaviour, due to a high occurrence in weeks 4–6. Non-contact agonistic interactions constituted the major part of agonistic interactions (66%). Possibly as consequence, no serious injuries were registered and there was no treatment effect (U=184; P=0.11). We conclude that the behaviour of young horses is affected by group management, and that horses appear not to adapt to weekly regroupings.