Abstract: Removing a horse from its social group may be considered risky, both for the handler and the horse, because other horses can interfere in the catching process. The main aim of this study was to identify where and when these risk situations occur while removing a horse from its group. A potential risk situation was defined by the closeness of loose horses in the group or by any physical contact with them. Whether the number of horses following would be influenced by the social rank of the horse being led out, and whether more horses would follow to the gate when a larger proportion of the group was removed compared to when a single horse was taken out were also investigated. Thirty-two mares (1–2 years) were kept in groups of four. All horses were taken out of their home paddock twice alone (64 tests) and twice with a companion (32 tests). One handler (or two handlers when two horses were removed) was asked to approach (phase 1) and catch the target horse (phase 2), walk it to the centre of the paddock and remain stationary at a post for 30 s (phase 3), walk to the paddock entrance (phase 4) and through the gate (phase 5). The number of horses following, and the number of loose horses in proximity (<2 m, 2–5 m) to the target horse and handler was estimated, and horse–horse and horse–human interactions were recorded continuously for the five scoring phases. Significantly more loose horses were within 2 m of a single target horse during the phases approach (mean ± SD: 1.5 ± 0.8), catch (1.6 ± 0.9) and post (1.7 ± 0.7) than during walk (1.0 ± 0.5) and gate (1.1 ± 0.6). Rank did not influence the number of horses following to the gate (high rank: 2.4 ± 0.7; lower rank: 2.0 ± 1.0; P = 0.396) and interactions between horses were rare. A greater proportion of the loose horses followed when two horses (0.9 ± 0.2) were removed compared to when a single horse (0.7 ± 0.3) was taken out (P = 0.011). In conclusion, maintaining a distance to other horses in the group by reducing the time being relatively stationary, so giving loose horses fewer chances to approach, is likely to contribute to improved handler's safety. Removing a small proportion of the group may also decrease the probability of the other horses following.