A lvEolAr collapse is a negative consequence of monotonous mechanical ventilation (fixed rate and fixed low tidal volume) in patients with acute respiratory distress syndrome (ArDS). By varying the rate at which the mechanical ventilator is driven and the volume delivered, Mutch and colleagues [1][2][3] demonstrated that arterial oxygenation in an animal model of ArDS is improved over that attained with monotonous mechanical ventilation. It has been suggested 4 that the variability of peak airway pressure recruits collapsed alveoli, thereby leading to an increase in Pao 2 . In this issue of the Journal, Froehlich et al. 5 have gone one step further by asking whether the temporal order of the fluctuations in breathing rate and volume, as well as the fluctuations themselves, account for the beneficial effects of variable mechanical ventilation. The question arises from the fact that fluctuations in respiratory period and tidal volume, characterizing spontaneous breathing in healthy humans and other species, are time-scale invariant 6-8 rather than random in occurrence. Time-scale invariant (i.e., fractal) behaviour means that the fluctuations in the value of the measured parameter occurring on one time scale are proportional to some power to those occurring on yet longer time scales.9,10 As such, fractal behaviour reflects the linkage via nonlinear interactions of processes occurring on different time scales, thereby leading to long-range correlations among events. As a result, the current value of the measured parameter is influenced not only by recent events, but also by those occurring in the "distant" past. Such longrange correlations are viewed as a type of memory that may be useful in predicting the future behaviour of the respiratory system, and valuable in providing clues to the underlying dynamics governing the respiratory generator and its feedback controllers.In the current study, Froehlich et al. 5 used two types of computer-generated files to control the mechanical ventilation of pigs subjected to acute lung injury by the intravenous infusion of oleic acid. one of the files was physiologically-derived from the spontaneous breathing pattern of a healthy human at rest. This file contained long-range fractal correlations of the fluctuations in ventilatory rate. The other file was a white noise file with the same mean value and standard deviation, but with no correlations among events (current value of ventilatory rate is unrelated to past values). Thus, the objective was to determine whether the fractal ordering of events in the physiologically-derived file plays a role in reversing the effects of oleic acid-induced lung injury on Pao 2 and other respiratory parameters. For both types of files, the fluctuations in interbreath interval were linked to reciprocal changes in tidal volume, so that minute ventilation was fixed.The authors used relative dispersion (rD) analysis 11 and second moment diffusion analysis 12 to clearly distinguish between the fractal properties of the physiologically-derived file an...