We tested the hypothesis that a prior 30 s sprint exercise bout would significantly reduce the curvature constant (W') but not the power-asymptote (critical power, CP) of the power-duration relationship as assessed using a novel 3 min all-out cycling test. Seven physically active male subjects completed the 3 min all-out test on three occasions in random order: following no prior sprint exercise (control, C); following a 30 s sprint and a 2 min recovery (S2); and following a 30 s sprint and a 15 min recovery period (S15). The CP was estimated from the mean power output sustained over the final 30 s of the test and the W' was estimated as the power-time integral above the end-test power. There were no significant differences in the estimated CP between the control 3 min all-out trial and the two prior sprint conditions (C, 235 +/- 44 W; S2, 223 +/- 46 W; and S15, 232 +/- 50 W; P > 0.05; coefficients of variation 2, 3 and 6% for C-S2, C-S15 and S2-S15, respectively). However, the W' in S2 (16.5 +/- 3.3 kJ) was significantly lower than in C (20.8 +/- 3.9 kJ) and S15 (21.2 +/- 4.5 kJ; P < 0.05). The total work done was lower in S2 than in the other conditions (S2, 56.4 +/- 7.2 kJ; C, 63.5 +/- 6.6 kJ; and S15, 63.0 +/- 6.0 kJ; P < 0.05). The W', but not the CP, is sensitive to a bout of prior sprint exercise which would be expected to result in significant muscle phosphocreatine depletion. These findings support the fundamental principles of the power-duration relationship as applied to all-out exercise.