Housing soivs in large groups is increasing in popularity but some group-housing systems may compromise subsequent reproductive performance. An experiment was designed to test this hypothesis using multiparous sows weaned at 21 ±2 days. Two treatments were studied: a stable group of 12 sows and a dynamic group of 18 sows. Both treatments had similar space allowances in the lying area (1·24 m2 per sow). Sows were allocated to treatment within 1 week of service and remained on treatment until after day 40 of pregnancy. Each stable treatment group remained unchanged, whereas sows were replaced weekly in the dynamic treatment. Sows were fed using a single electronic feeder for each pen. All sows were treated similarly for the rest of pregnancy. Eighty and 120 sows from the stable and dynamic treatments respectively completed pregnancy. Five second parity sows from the dynamic treatment were removed having suffered from aggression. The farrowing rate was 0·78 and 0·85 for stable and dynamic treatment sows respectively. The stable treatment sows had larger litters and a higher proportion of litters having 10 or more total piglets born (P < 0·01). The stable treatment showed a significantly heavier (P < 0·05) litter weight for total pigs born (16·9 v. 14.6 (s.e.d. 0·57) kg per litter). The parities were uneven so the data must be treated with caution. However, the results indicate that the present advice, that soivs should be housed in stable groups for the first 4 weeks of pregnancy, should remain unchanged.