The effect of progressive hypoxia on the structure and dynamics of herring (Clupea harengus) schools in laboratory conditions was investigated. The length, width and depth of schools of about 20 individuals were measured from video recordings to test the hypothesis that during hypoxia fish schools change their shape and volume. School shape (calculated as the ratios of length/depth, width/depth and length/width) did not change significantly during hypoxia. School length, width, depth, area and volume were all significantly increased at 20% oxygen saturation. Volume, area and width were more sensitive to hypoxia; volume and width were also increased at 25% and area at 30% oxygen saturation. The degree of position changing (shuffling) of individuals within the school was also analysed. Shuffling in normoxia was observed to occur largely through 'O-turn' manoeuvres, a 360( degrees )turn executed laterally to the school that allowed fishes in the front to move to the back. O-turn frequency during normoxia was 0.69 O-turns fish(-1) min(-1) but significantly decreased with hypoxia to 0.37 O-turns fish(-1) min(-1) at 30% oxygen saturation. Shuffling was also investigated by measuring the persistence time of individual herring in leading positions (i.e. the first half of the school). No significant changes occurred during hypoxia, indicating that the decrease in O-turn frequency does not affect shuffling rate during hypoxia, and that position shuffling in hypoxic conditions is mainly due to overtaking or falling back by individual fishes. School integrity and positional dynamics are the outcome of trade-offs among a number of biotic factors, such as food, predator defence, mating behaviour and various physical factors that may impose certain limits. Among these, our results indicate that oxygen level modulates schooling behaviour. Oxygen alters whole-school parameters at oxygen saturation values that can be encountered by herring in the field, indicating that oxygen availability is an important factor in the trade-offs that determine school volume. An increase in school volume in the wild may increase the oxygen available to each individual. However, shuffling rate is not affected by hypoxia, indicating that the internal dynamics of positioning is the result of the balance of other factors, for example related to the nutritional state of each individual fish as suggested by previous studies.