Confinement of breeding sows and gilts is one of the most controversial issues in livestock production, together with floor space allowance and lack of bedding (Barnett et al., 2001). However, consideration of the welfare implications of group versus stall housing during gestation is complex. Both housing systems have some clear advantages and disadvantages in relation to pig welfare. For example, individual housing in stalls restricts the sows’ opportunities to exercise, socially interact and interact with other features of the environment, but on the other hand, group housing raises welfare issues concerning space allowance and level of aggression. About 70% of the breeding sows are kept in stalls during gestation (Barnett et al., 2001). The ongoing welfare debate about stall housing has led to the examination of alternative housing systems for breeding sows. There is little information in the scientific literature on the welfare of gestating sows in groups of more than 40 animals and there is no scientific literature on housing breeding sows or gilts in large groups on deep litter. The present experiment measured aspects of the welfare of gestating sows housed in either large groups on deep litter (Hoops) or conventional stalls (Stalls). Six hundred and forty sows were studied, with 40 recently mated sows weekly entering each treatment over an 8-week period; groups of 85 were formed using 40 experimental and 45 non-experimental animals. Sows in Hoops had a higher number of scratches, a higher return rate to oestrus after mating (13.20% versus 7.35%) and there was a trend for higher salivary cortisol concentrations in week 1 of gestation (6.29 nM versus 4.03 nM). Sows in Stalls had a higher incidence of lameness at weeks 9 and 15 of gestation (13.8% versus 0.8% at week 15). There were changes in some leucocyte sub-populations in the Stalls treatment late in gestation: the percentage of neutrophils was higher (46% versus 41% of WBC), the number and percentage of lymphocytes was lower (41.6% versus 46.5% of WBC) and consequently there was a higher neutrophil:lymphocyte ratio (1.22 versus 0.94). There was a trend for a lower reproductive failure in the Stalls treatment (14.5% versus 27.3%); farrowing rate was higher (76.8% versus 66%), and while sows in Stalls weaned fewer piglets per litter (8.31 versus 8.97), the average weaning weight of these piglets was higher (8.69 kg versus 8.01 kg). The combination of these reproductive parameters resulted in sows in the Stall treatment weaning the equivalent of 39 more piglets per 100-mated sows. The results suggest that sows in large groups on deep litter faced greater welfare challenges in the early stages of gestation based on the findings of increased scratches, a higher rate of return to oestrous and a trend for higher cortisol concentrations early in gestation, all possibly a consequence of aggression. In contrast sows in stalls faced greater welfare challenges later in gestation based on a higher incidence of lameness and an increased neutrophil:lymphocyte ratio perhaps as a consequence of increased stress. In conclusion, these data suggest that in both housing systems the welfare advantages and disadvantages change overtime.