Reports on reproducibility of lactate markers usually considered only two trials. The authors assessed reproducibility of power output at seven markers in 11 fit subjects over at least six trials under tightly controlled conditions. Subjects undertook incremental exercises (50 W start, +50 W every 3 min to exhaustion) on a cycle ergometer. At each trial blood lactate concentration was determined at rest and within the final 30 s of each stage. The Rest+1, 2.0 and 4.0 mmol/l markers were determined by interpolation, the D-max and nadir using a quadratic model and the lactate slope index using an exponential plus constant model, and a visual turnpoint was determined empirically. Intraclass correlations and coefficients of variation assessed reproducibility. Power output at all markers differed significantly between subjects, but not between trials. Intraclass correlation coefficients were respectively 0.799, 0.794, 0.807, 0.903, 0.677, 0.769 and 0.648, and corresponding standard errors of measurement 11.9, 9.2, 9.1, 2.5, 9.2, 10.8 and 24.7 W. Statistical powers of detecting a 30 W increment at these markers were 0.30, 0.43, 0.42, 0.98, 0.58, 0.38 and 0.18 respectively. These results indicate that only the D-max marker has good reproducibility and that it alone can identify small but meaningful changes in training status with sufficient statistical power.