Milk production and composition were measured for 63 days in 153 Holstein-Friesian cows offered either 7, 10 or 13 kg/day (as fed) of a rolled, 74% barley/26% lupin grain mixture together with either no feed additive, 300 mg/day of virginiamycin (VM) or 300 mg/day of virginiamycin plus 200 g/day of sodium bicarbonate (VM + NaHCO3). All cows were fed 1 kg/day of a pelleted mineral supplement containing the additives. The cows were grazed as a single herd on perennial ryegrass-subterranean clover pasture with pasture silage available during periods of pasture shortage. Rumen fluid was analysed for pH and volatile fatty acids, blood for plasma glucose, beta-hydroxy-butyrate, urea and D- and L-lactic acid, faeces for dry matter and pH, and both urine and milk for urea and sodium. The incidence of grain bloat in all treatments was low. There was no main effect of dietary additive treatment on grain intake, but at the highest grain level, cows offered VM or VM + NaHCO3 ate more grain than those offered no dietary additive. There were no significant differences between the dietary additive treatments in milk production, milk composition, cell count, liveweight and condition score. Mean daily covariance-corrected yields of milk and protein (kg), and milk protein content (g/kg) respectively, were significantly (P<0,01) greater for cows fed 11 (28.0, 0.86, 30.6) and 14 kg/day (28.7, 0.88, 31.0) of concentrate compared with those fed 8 kg/day (26.4, 0.78, 29.7). Mean milk fat content (g/kg) was significantly (P<0.01) lower in milk from cows fed 14 kg/day (32.0) of concentrate compared with those fed 8 (35.9) or 11 (34.7) kg/day. There were no significant differences between concentrate feeding levels in milk fat yield or milk somatic cell count. Covariance-corrected liveweight and condition score were significantly (P<0.01) higher for cows fed 14 kg/day of concentrate compared with cows fed at the lower concentrate levels. There were no significant interactions between concentrate level and dietary additive for all milk production parameters. No significant differences were recorded between the different levels of concentrate in the concentration of total rumen volatile fatty acids, or in blood plasma concentrations of glucose and L-lactate. The rumen molar proportions of acetate and butyrate were significantly (P<0.01) lower, and propionate and valerate significantly (P<0.01) higher at the higher levels of concentrate offered. The molar ratio of acetate plus butyrate to propionate was significantly (P<0.01) higher in cows fed 8 kg/day of concentrate compared with cows fed 11 and 14 kg/day. It was concluded that the inclusion of virginiamycin, or virginiamycin together with sodium bicarbonate, in high grain rations cannot be recommended for improving milk production in grazing dairy cows already adapted to high levels of barley and lupin grain.