The litter performance of 48 Large White x Berkshire sows mated to half-sib Berkshire boars was measured over three reproductive cycles. At first mating, animals were allocated to four dietary treatments according to mating weight and ultrasonically determined backfat cover. The four treatments were factorial, involving two levels of feed intake of either an all-meal control ration or a combination of meal and pasture herbage.The higher level of feeding increased both live litter weights and numbers in Parity 1. At the higher feeding level heavier live litter weights were found in all parities only after litter weights were adjusted for differences in the number of pigs. Increases in average piglet birth weight due to the level of feeding were 0.09, 0.14, and 0.18 kg approximately, in Parities 1, 2, and 3 respectively. Despite differences in average piglet birth weight, no difficulty was found in rearing the pigs even at the lower feeding level.At 4 weeks of age heavier and larger litters resulted from the higher level of feeding in Parity 1. In Parity 3, adjusted 4-week litter weights showed that the heaviest litters were produced by sows on the higher level of feeding. In the absence of differences in creep feed and lactation feed intakes, this finding was ascribed to differences in milk and milk energy production due to the depleted fat reserves of sows on the lower level of feeding in pregnancy.The partial replacement of meal by grass did not exert a consistent difference in farrowing performance, although litter performance of sows with access to pasture was adversely affected in Parity 2.The results suggest the need for differential feeding of pregnant sows. A higher level of feeding at the start of pregnancy was proposed to offset sow fat losses after lactation.