Commercial gestation housing systems for sows generally fail to cater fully for their needs in terms of comfort or the ability to perform highly motivated behaviours, which can lead to chronic stress and an impairment to welfare. We compared a typical gestation system (CONTROL) with an IMPROVED one as regards oral stereotypies, aggressive behaviour, skin lesions, locomotion, and tear staining. Sows were mixed into 12 stable groups (six groups per treatment, 20 sows per group), 29 days post-service in pens with free-access, full-length individual feeding/lying stalls. CONTROL pens had fully slatted concrete floors, with two blocks of wood and two chains suspended in the group area. IMPROVED pens were the same but with rubber mats and a length of manila rope in each feeding stall, and straw provided in three racks in the group area. Direct observations of oral stereotypical (30 instantaneous scans per sow per day) and aggressive (all-occurrence sampling, 3 h per sow per day) behaviours were conducted 72 h post-mixing, in mid and late gestation. Skin lesions were counted 24 h and three weeks post-mixing, and in late gestation. Sows’ locomotion (locomotory ability) was scored using a visual analogue scale in mid and late gestation. Right and left eye tear staining was scored in late gestation. Indications of better welfare in IMPROVED sows included performance of fewer oral stereotypies in mid and late gestation, and lower tear stain scores. These sows performed more aggression in late gestation, which was associated with access to enrichment, but skin lesion counts were not affected. Thus, the changes made in the IMPROVED treatment benefitted aspects of sow welfare.