To determine the best procedure for reading maximal expiratory flow-volume curves, 2 sets of 5 curves were obtained 1 h apart in 89 subjects and processed digitally according to 8 different methods. Four indices were considered: the forced expiratory flows at 25, 50, and 75% of the forced vital capacity, and the maximal mid-expiratory flow. When selecting the curve yielding the largest forced vital capacity or the largest sum of forced vital capacity and forced expiratory volume in 1 s, flow values were significantly lower (P < 0.001) and were often less reproducible than those obtained with most of the other methods. Computing the mean of the indices among the curves with the 2 largest forced vital capacities also provided comparatively low values, but with better reproducibility. In contrast, maximal flows were probably overestimated by using the highest values among the curves having forced vital capacity or a surface area within 5% of the largest, or when reading the indices on a composite curve obtained by superimposing individual breaths at residual volume. More reproducible and, probably, unbiased data may be drawn from the composite curves obtained by superimposing the breaths either at total lung capacity or on the descending limb.