Twelve sows in good quality unstrawed stalls, three groups of five sows in strawed pens with individual feeding stalls and sows in a 38-sow group in a strawed yard with an electronic sow feeder were compared during the first four parities. They originated from the same source, were about 9 months of age and in the 7th week of their first pregnancy at the start of the experiment and were kept in adjacent rooms in a building, cared for by the same staff and given the same diets at a rate of 2·2 kg/day per animal. No new animals were added to the groups or stalls during the study and animals returned to the same condition after periods in farrowing and service accommodation. Using a wide range of welfare indicators, it was clear that stall-housed sows had more problems than group-housed sows and that tliese problems were worse in the fourth than in the first pregnancy. By the fourth pregnancy, stall-housed sows spent proportionately 0·14 of time showing activities which were clearly stereotypies and much time on activities which were sometimes stereotyped, i.e. ‘drinking’ and rooting or chewing at pen fittings making a total of proportionately 0·50 of time. Comparable figures for group-housed sows were much lower (0·037 and 0·081 in total). Stall-housed sows were also more aggressive than group-housed by the fourth pregnancy and their body weights were lower. There were no differences using physiological or immunological tests or measures of reproductive output. When the two group-housing systems were compared, sows in the electronic feeder system showed more fighting, especially soon after initial mixing, but fewer total agonistic interactions than sows in groups of five during the first pregnancy. Oral stereotypies were slightly higher in small groups, perhaps because of smaller pen space, than in larger groups but much lower than in stalls. By the fourth pregnancy there were few differences between sows in small and large groups and all seemed to have adapted well to the conditions. Evaluation of welfare in different housing systems requires use of a wide range of measures and of long-term studies.
