Purpose: Variable ventilation is superior to control mode ventilation in a number of circumstances. The nature of the breathing file used to deliver the variable rate and tidal volume has not been formally examined. Methods: We compared two different noise files in a randomized prospective trial of variable ventilation. Pigs were anesthetized, intubated, and mechanically ventilated. Oleic acid was infused to introduce lung injury. The animals were ventilated at a tidal volume of 7 mL·kg-1, in variable mode, with either physiologically-derived noise (variability file - 1,587 breath intervals-obtained from a spontaneously breathing volunteer; n = 10) or a variability file of identical length derived from computer-generated white noise (n = 10). Results: The physiologically-derived noise had a power law α-exponent of -0.27 and a Hölder exponent of -0.38, indicative of auto-correlated noise. The computer-generated noise had an α-exponent of -0.52 and a Hölder exponent of -0.49, indicative of white noise. Both files showed multifractal characteristics. There were no differences between groups, at any time period, for PaO2, PaCO2, and static or dynamic respiratory system compliance. No differences were observed between groups for wet:dry lung weight ratios or for interleukin-8 in bronchoalveolar lavage fluid. Conclusion: This study demonstrates that the nature of the variability files, chosen to drive the variable ventilator, had no effect on indices of gas exchange or respiratory mechanics in this model. A considerable overlap of the multifractal files existed. The potential to drive a variable ventilator using algorithm-derived files with multifractal characteristics, thereby eliminating the requirement to use physiologically-derived signals, is discussed.