The temporal dynamics of blood flow in the human lung have been largely unexplored due to the lack of appropriate technology. Using the magnetic resonance imaging method of arterial spin labeling (ASL) with subject-gated breathing, we produced a dynamic series of flow-weighted images in a single sagittal slice of the right lung with a spatial resolution of ~1 cm(3) and a temporal resolution of ~10 s. The mean flow pattern determined from a set of reference images was removed to produce a time series of blood flow fluctuations. The fluctuation dispersion (FD), defined as the spatial standard deviation of each flow fluctuation map, was used to quantify the changes in distribution of flow in six healthy subjects in response to 100 breaths of hypoxia (FI(O(2)) = 0.125) or hyperoxia (FI(O(2)) = 1.0). Two reference frames were used in calculation, one determined from the initial set of images (FD(global)), and one determined from the mean of each corresponding baseline or challenge period (FD(local)). FD(local) thus represented changes in temporal variability as a result of intervention, whereas FD(global) encompasses both FD(local) and any generalized redistribution of flow associated with switching between two steady-state patterns. Hypoxic challenge resulted in a significant increase (96%, P < 0.001) in FD(global) from the normoxic control period and in FD(local) (46%, P = 0.0048), but there was no corresponding increase in spatial relative dispersion (spatial standard deviation of the images divided by the mean; 8%, not significant). There was a smaller increase in FD(global) in response to hyperoxia (47%, P = 0.0015) for the single slice, suggestive of a more general response of the pulmonary circulation to a change from normoxia to hyperoxia. These results clearly demonstrate a temporal change in the sampled distribution of pulmonary blood flow in response to hypoxia, which is not observed when considering only the relative dispersion of the spatial distribution.