Four farms (540 sows) that group-housed sows from 2 weeks of lactation until weaning (G-farms) and 3 farms (300 sows)--used as controls--that kept the sows individually penned throughout the 5- to 6-week-long lactation period (C-farms) were compared in terms of reproductive performance. All sows were crossbred Swedish Yorkshire x Swedish Landrace. Sows were kept in groups in the breeding section and were also grouped on deep litter in the dry-sow section on all farms. Batchwise farrowing routines were used on all farms, and batch size ranged between 8 and 22 sows. Fertility and culling data were collected in connection with monthly visits to the farms for 16 months. Among the older (> or = 5th parity) sows, litter sizes were significantly (p = 0.02) smaller in the G-farm group compared with the C-farm group. The percentage of sows mated within 10 days post weaning was lower (p < 0.001) in the G-farm group than in the C-farm group, with the difference being most evident among the older (> or = 5th parity) sows. The frequency of repeat breeders among primiparous sows was similar in the 2 housing systems, but among the multiparous sows repeat-breeder frequency was higher (p = 0.04) in the G-farm group than in the C-farm group. In both groups, repeat-breeder frequency seemed to be highest from July to September. Repeat breeding/failure to farrow was a common reason for culling in the G-farm group. These results indicate that reproductive performance was impaired in the group-housing system. This impairment could have been due partly to the occurrence of lactational oestrus, which makes it difficult to maintain adequate routines for oestrous detection and mating/insemination.