This experiment measured the consequences of manipulating body fat reserves at farrowing by nutritional means in three sow breed types. A 3 ×2× 2 factorial experiment involving 83 second parity sows (previously given food to achieve a standard backfat measurement of 25 mm at first farrowing) was used to investigate the influence of breed type (Meishan synthetic 50% (M) or purebred European Landrace (LR) and Large White (LW) on response to isoenergetic diets of differing protein levels (180 g crude protein (CP) per kg, 9 g lysine per kg; or 240 g CP per kg, 12 g lysine per kg) offered to appetite over a 28-day lactation. Body fat reserves were adjusted during the preceding pregnancy by supplementing a set quantity of basal diet (131 g CP per kg, 12·7 MJ digestible energy per kg) with an energy source (maize starch + soya oil), to attain a backfat at P2 of either 23 mm (T) or 28 mm (F) at farrowing. No lactation diet effects on performance were observed. There were no significant breed or diet differences in backfat at farrowing. M sows were lighter at farrowing than LW and LR (M = 19·4, LW = 22·3, LR = 21·4 kg, P < 0·001). M consumed more food throughout lactation than LR and LW (M = 7·02, LW = 5·02, LR = 5·86 kg/day, P < 0·01) but weight loss was higher for M than for LW and LR sows (M = 11·7, LW = 5·4, LR = 6·1 kg, P < 0·05). Fat loss also showed a breed effect (M = 3·6, LW = 2·1, LR = 2·7 mm P2, P < 0·05), as did litter growth rates (M = 2·33, LW = 1·74, LR = 2·07 kg/day, P < 0·01). However, the breed effect disappeared when litter size was added as a covariate. The proportion of sows that was served within 5 days of weaning was higher for M than for the White breed types (W) (M = 0·74 , W = 0·51 , P < 0·05). The Meishan synthetics still mobilized more body tissue and produced greater milk yields than the W sows when backfat levels were standardized between breeds. The use of a very high protein food did not offer any advantage for the sow in terms of milk production or conservation of maternal body reserves.