Schools of herring exposed to progressive hypoxia show a peak in velocity during severe hypoxia, at 15–34% oxygen saturation, followed by a decrease in swimming speed until school disruption occurred. The observed increase in swimming speed during severe hypoxia reveals a graded response, since the lower the fish's swimming speed prior to severe hypoxia ( U 95−50 , the speed at oxygen saturations between 95 and 50%), the greater the relative increase in swimming speed. The oxygen saturations at which both peak velocity and school disruption occurred were lower for fish with lowest U 95−50 , suggesting that the fish with the slowest speed U 95−50 reach their critical P O2 (at which there is respiratory distress) last, i.e. at lower oxygen saturation. At a functional level, it is suggested that herring encountering hypoxia increase their speed in order to find more favourable conditions, and the magnitude of this increase is modulated by their respiratory distress. It is also hypothesised that the observed increase in speed may be related to an increase in the rate of position shifting within the school. Since the oxygen saturation at which the response to hypoxia occurs and the magnitude of the response are related to the fish's preferred speed prior to severe hypoxia, it is suggested that such a preferred speed should be measured in experiments testing the effect of hypoxia on fish behaviour.