The ability of the protein nutrition scheme proposed by the Agricultural Research Council (ARC, 1980, 1984) to predict responses to protein supply was examined in two experiments. In experiment 1, groups of nine British Friesian steers implanted with oestradiol-17β and nine non-implanted steers were fed from 133 to 300 kg live weight on each of four all-concentrate diets (metabolizable energy (ME) = 12 MJ/kg dry matter (DM)) containing ratios of soya-bean meal and formaldehyde-treated soya-bean meal such that undegradable protein (UDP) concentration was 19, 25, 32 or 42 g/kg DM. Rumen degradable protein (RDP) concentration was relatively constant (111 to 116 g/kg DM). Implantation did not affect DM intake (DMI, g/kg M0·75). Both live-weight gain (LWG) and food conversion efficiency (FCE) (LWG/DMI) were greater (P < 0·05) for the implanted cattle (LWG = 1·41 (s.e. 0·04) kg/day; FCE = 0·24 (s.e. 0·02)) than for the non-implanted cattle (LWG = 1·23 (s.e. 0·05) kg/day; FCE = 0·22 (s.e. 0·02)). There were no dietary effects on either DMI or FCE. In the non-implanted steers, UDP did not affect LWG but, for the implanted steers, there were positive, linear responses in LWG to both UDP concentration (P = 0·048) and UDP intake (P = 0·026). In experiment 2, groups of eight implanted steers were fed from 132 to 300 kg live weight on each of six diets (ME = 12 MJ/kg DM, soya-bean meal and formaldehyde-treated soya-bean meal as chief protein sources) supplying 17, 26, 34, 39, 54 and 82 g UDP per kg DM and concentrations of RDP up to 1·5 times ARC recommendations. There was no effect of UDP on either DMI or FCE but positive, linear responses in LWG to both UDP concentration (P = 0·019) and UDP intake (P = 0·010). In both experiments, mean DMI exceeded that predicted by ARC (1980) (P < 0·05). LWG predicted by the ME system and the ARC protein scheme was not significantly different from observed LWG for non-implanted steers, but was an underestimate of observed LWG for implanted steers (P = 0·002 in experiment 1, P < 0·001 in experiment 2). Observed responses in LWG to UDP concentration were also poorly predicted (regressions of observed on predicted LWG: experiment 1, r2 = 0·12, residual s.d. = 0·126; experiment 2, r2 = 0·004, residual s.d. = 0·119). It was concluded that oestradiol-17p has significant effects on responses in LWG to UDP, which are not predicted by the current ARC (1980, 1984) protein nutrition scheme.