There is very little information about the effect of inspiratory muscle training on inspiratory flow (V'I) and thus on power output (PO) in patients with chronic obstructive pulmonary disease (COPD). In this study we aimed to evaluate the changes induced by training on the determinants of PO. Thirty one patients with severe COPD were randomly divided into: Group 1, trained with 30% maximal inspiratory pressure (PI,max); Group 2, with 10% PI,max; and Group 3 also trained with 30% PI,max, but the breathing pattern was evaluated while performing the training manoeuvres along inspiratory muscle training (IMT). All groups used a threshold device for 10 weeks. The PO for each of the loads during an incremental threshold test was evaluated prior to and after training. Maximal PO (POmax) increased in all groups, but the increment was higher in groups trained with 30% PI,max (p<0.005), mainly due to an increase in V'I. Group 3 showed a progressive increase in V'I (p<0.001) during the training manoeuvres in spite of an increase in load along IMT. In addition, the load after IMT was overcome with a shorter inspiratory time (tI) (p<0.02), a smaller tI/total duration of the respiratory cycle (t(tot)), (p<0.001) with no change in tidal volume or t(tot). The increment in POmax in this group correlated with the V'I generated while training (r=0.85; p<0.0001). We conclude that in patients with chronic obstructive pulmonary disease, the use of an intermediate threshold load for training improves power output mainly by increasing inspiratory flow, an effect consistent with an increase in shortening velocity of inspiratory muscles.