Although food deprivation is a major ecological pressure in fishes, there is wide individual variation in tolerance of fasting, whose mechanistic bases are poorly understood. Two thousand individually tagged juvenile European sea bass were submitted to two 'fasting/feeding' cycles each comprising 3 weeks of food deprivation followed by 3 weeks of ad libitum feeding at 25 degrees C. Rates of mass loss during the two fasting periods were averaged for each individual to calculate a population mean. Extreme fasting tolerant (FT) and sensitive (FS) phenotypes were identified that were at least one and a half standard deviations, on opposing sides, from this mean. Respirometry was used to investigate two main hypotheses: (1) tolerance of food deprivation reflects lower mass-corrected routine metabolic rate (RMR) in FT phenotypes when fasting, and (2) tolerance reflects differences in substrate utilisation; FT phenotypes use relatively less proteins as metabolic fuels during fasting, measured as their ammonia quotient (AQ), the simultaneous ratio of ammonia excretion to RMR. There was no difference in mean RMR between FT and FS over 7 days fasting, being 6.70 +/- 0.24 mmol h(-1) fish(-1) (mean +/- s.e.m., N=18) versus 6.76 +/- 0.22 mmol h(-1) fish(-1) (N=17), respectively, when corrected to a body mass of 130 g. For any given RMR, however, the FT lost mass at a significantly lower rate than FS, overall 7-day average being 0.72 +/- 0.05 versus 0.90 +/- 0.05 g day(-1) fish(-1), respectively (P<0.01, t-test). At 20 h after receiving a ration equivalent to 2% body mass as food pellets, ammonia excretion and simultaneous RMR were elevated and similar in FT and FS, with AQs of 0.105 +/- 0.009 and 0.089 +/- 0.007, respectively. At the end of the period of fasting, ammonia excretion and RMR had fallen in both phenotypes, but AQ was significantly lower in FT than FS, being 0.038 +/- 0.004 versus 0.061 +/- 0.005, respectively (P<0.001, t-test). There was a direct linear relationship between individual fasted AQ and rate of mass loss, with FT and FS individuals distributed at opposing lower and upper extremities, respectively. Thus the difference between the phenotypes in their tolerance of food deprivation did not depend upon their routine energy use when fasting. Rather, it depended upon their relative use of tissue proteins as metabolic fuels when fasting, which was significantly lower in FT phenotypes.