The welfare of sows kept in two different group housing systems, namely voluntary cubicle pens housing small static groups and a split-yard housing system housing a large dynamic group, were examined. Each week, four newly weaned sows were introduced into either a voluntary cubicle pen or the split-yard system, with a total of eight groups of four sows introduced into each system. Behavioural and skin lesion data were recorded during the sows’ first 5 weeks in either system. Agonistic behaviours were performed more frequently during week 1 than in subsequent weeks in both housing systems (P < 0001). Sows in the voluntary cubicle pens spent more time standing inactive (P < 0001) and spent less time in exploratory (P < 005) and locomotory (P < 001) behaviours. However, sows in the split-yard system had higher skin lesion scores (P < 001) and engaged in more social (P < 005) and agonistic interactions (P < 005). These agonistic interactions included attacks (P < 005) and fleeing (P < 0001) throughout the 5-week observation period and fighting (P < 0001) during week 1 only. The split-yard system, therefore, appeared to offer sows a more stimulating social and physical environment than the voluntary cubicle pens, but also led to higher levels of aggression and skin damage. Both housing systems, therefore, appear to compromise sow welfare in different ways.