Measurement of the intrinsic positive end-expiratory pressure (PEEP_i) is important in planning the management of ventilated patients. Here, a new recursive least squares method for on-line monitoring of PEEP_i is proposed for mechanically ventilated patients. The procedure is based on the first-order model of respiratory mechanics applied to experimental measurements obtained from eight ventilator-dependent patients ventilated with four different ventilatory modes. The model PEEP_i (PEEP_i,mod) was recursively constructed on an inspiration-by-inspiration basis. The results were compared with two well-established techniques to assess PEEP_i: end-expiratory occlusion to measure static PEEP_i (PEEP_i,st) and change in airway pressure preceding the onset of inspiratory airflow to measure dynamic PEEP_i (PEEP_i,dyn). PEEP_i,mod was significantly correlated with both PEEP_i,dyn (r = 0.77) and PEEP_i,st (r = 0.90). PEEP_i,mod (5.6 ± 3.4 cmH₂O) was systematically >PEEP_i,dyn and PEEP_i,st (2.7 ± 1.9 and 8.1 ± 5.5 cmH₂O, respectively), in all the models without external PEEP. Focusing on the five patients with chronic obstructive pulmonary disease, PEEP_i,mod was significantly correlated with PEEP_i,st (r = 0.71), whereas PEEP_i,dyn (r = 0.22) was not. When PEEP was set 5 cmH₂O above PEEP_i,st, all the methods correctly estimated total PEEP, i.e., 11.8 ± 5.3, 12.5 ± 5.0, and 12.0 ± 4.7 cmH₂O for PEEP_i,mod, PEEP_i,st, and PEEP_i,dyn, respectively, and were highly correlated (0.97-0.99). We interpreted PEEP_i,mod as the lower bound of PEEP_i,st and concluded that our method is suitable for on-line monitoring of PEEP_i in mechanically ventilated patients.