The energy cost of internal work and its relationships with lower limb mass and pedalling frequency were studied in four male subjects [age 22.2 (SD 1.5) years, body mass 81.0 (SD 5.1) kg, maximal O-2 uptake (VO2max) above resting 3.06 (SD 0.4) l . min(-1)]. The subjects cycled at 40, 60, 80 and 100 rpm and at five different exercise intensities for every pedalling frequency (unloaded condition, UL); the same exercises were repeated after having increased the lower limbs' masses by 40% (loaded condition, L). The exercise intensities were chosen so that the oxygen consumption (VO2) did not exceed 75% of VO2max. For all the subjects and all the conditions, the rate of VO2 above resting increased linearly with the mechanical power (W). The y-intercepts of the linear regressions of VO2 on W, normalised per kilogram of overall lower limbs' mass were the same in both UL and L and increased with the 4.165 power of pedalling frequency (f(p)). These intercepts were taken to represent the metabolic counterpart of the internal power dissipation in cycling; they amounted to 0.78, 0.34, 3.29 and 10.30 W . kg(-1) for pedalling frequencies of 40, 60, 80 and 100 rpm respectively. The slope of the regression lines (Delta W/Delta VO2) represents the Delta efficiency of cycle ergometer exercise; this was also affected by f(p), ranging, on average, from 22.9% to 32.0%. These data allowed us to obtain a comprehensive description of the effects of f(p) (per minute), exercise intensity (W, watts) and lower limbs' mass with or without added loads (m(L), kg), on VO2 (ml . min(-1)) during cycling: VO2 = [m(L) .(4.3 . 10(-8). f(p)(4.165)/0.35)] + {1/[(3.594 . 10(-5). f(p)(2) - 0.003 . f(p) + 0.326). 0.35]}. W. The mean percentage error between the VO2 predicted from this equation and the actual value was 12.6%. This equation showed that the fraction of the overall VO2 due to internal work, for a normal 70-kg subject pedalling at 60 rpm and 100 W was of the order of 0.2.