High-intensity exercise tolerance is well described by a hyperbolic function of the external power output (P), such that only a fixed amount of work (W′) can be performed above the critical power asymptote (CP) [1]. During supra-CP exercise, therefore, tolerance appears to be governed by a W′-linked mechanism [2], the physiological determinants of which remain unclear. Providing insight into the mechanism(s) of W′ is hampered by its demanding measurement which requires multiple constant-P symptom-limited tests each on a different day. Recently a single 3-min all-out test (AOT) was investigated but failed to provide robust estimates of W′ compared to the multi-test ‘gold standard’ method (GSM) [3,4]. The hypothesis that CP is the highest achievable P following W′ depletion [5] provides a potential means of estimating both CP and W′ in a single-test. We therefore examined whether a novel test combining constant- and variable-P exercise (CVT) would provide CP and W′ estimates similar to GSM. Nine healthy men (20-30 yr) consented to cycle ergometry (Excalibur Sport, Lode, NL) with cadence independent (constant-P) and dependent (variable-P) control. Breath-by-breath oxygen uptake (VO2) was measured by a mass-spectrometer and turbine (MSX, Morgan Medical, UK). CP, W′ and VO2max were determined using: (1) a series of four constant-P tests to intolerance (GSM); immediately followed by (2) all-out variable-P exercise for 3-min to determine the highest sustainable P after W′ depletion (CVT); (3) two 3-min all-out variable-P tests (AOT); and (4) constant-P tests just below CP estimates during which arterialised-venous blood lactate:pyruvate ratio was determined (GM7, Analox, UK). VO2max was achieved in GSM (4.04 ± 0.64 L/min), CVT (4.03 ± 0.62 L/min) and AOT (4.00 ± 0.64 L/min) (mean ± SD). CP in GSM (241 ± 55 W) was not different to that estimated in CVT (251 ± 44 W; p=0.27) but was lower than that from AOT (264 ± 52 W; p=0.03). Similarly W′ was not different between GSM (18.2 ± 3.4 kJ) and CVT (14.9 ± 4.8 kJ; p=0.46) but GSM was higher than the AOT estimate (12.8 ± 2.5 kJ; p=0.01). Though CVT estimations did not differ from GSM, they were poorly predictive with individual CP and W′ differences ranging from -19 to +26 W and -11 to +10 kJ, respectively. Only 1/9 subjects achieved a steady state in VO2 and lactate:pyruvate during exercise at 2% below the CVT-estimated CP (compared to 6/6 subjects at 2% below the GSM CP). Contrary to previous reports [3,4] the AOT protocol poorly characterised supra-CP exercise tolerance. While CVT estimates were closer to GSM, they were poorly predictive. This suggests that variable-P and all-out exercise cannot be extrapolated to infer the limits of constant-P exercise, perhaps owing to differences in muscular fibre recruitment patterns that engender different profiles of muscular fatigue in contributing to exercise limitation.