With interest we have read the paper by Carvalho and colleagues, in which the authors report the results of their elegant study on the role of positive end-expiratory pressure (PEEP) in acute lung injury [1]. Carvalho and colleagues demonstrated in an acute respiratory distress syndrome (ARDS) model that alveolar hyperinflation and nonaerated areas may coincide during a stepwise reduction of PEEP. Their finding is in line with our findings in a mathematical ARDS model and with other studies [2,3], and emphasises that the PEEP at which the respiratory system compliance is maximal coincides with a compromise between hyperinflation and poor aeration [3]. The accompanying commentary by Rouby and colleagues, however, raises questions about the contribution by Carvalho and colleagues on which we would like to comment [4]. First, Rouby and colleagues state that PEEP may produce end-expiratory hyperinflation of aerated areas. This statement is clear, but the authors also state that PEEP produces end-expiratory reaeration of nonaerated parts of the lung (recruitment). This last statement is puzzling, since PEEP is an end-expiratory phenomenon and as such cannot in itself recruit collapsed lung units. PEEP can maintain volume that has been recruited by higher pressures. Second, we fully agree that a lung-protective ventilator strategy should not only include a reduced tidal volume, but also should combine a PEEP level that corresponds to an optimal compromise between alveolar recruitment and hyperinflation. Rouby and colleagues conclude that this compromise yields equal numbers of hyperinflated and non-aerated lung regions. Evidently, both regional consolidation and regional hyperinflation may lead to ventilator-induced lung injury [5]. To our knowledge, however, whether both are equally harmful has yet to be demonstrated Furthermore (at least during high-frequency oscillatory ventilation), airway pressures at which oxygenation is maximal do not equal pressures at which the respiratory system compliance is maximal [6]. We therefore propose that for the survival of the individual patient it is unclear whether to aim at optimal oxygenation or at optimal compliance.