The intermediate disturbance hypothesis has been influential in the development of ecological theory and has important practical implications for the maintenance of biodiversity but has received few rigorous tests. We tested the hypothesis that maximum taxon richness of macroinvertebrates will occur in communities subject to intermediate levels of disturbance at 54 stream sites that differed in the frequency and intensity of flood‐related episodes of bed movement. Our results support the intermediate disturbance hypothesis, with both highly mobile and relatively sedentary taxa conforming to the predicted bell‐shaped curve. Taxon richness was not related to habitat area (stream width), distance from the headwater, or the diversity of microhabitats (particle size categories) but was significantly and negatively related to the proportion of the substratum made up of small particles. Of all the factors measured, however, bed disturbance was by far the best at accounting for variation in taxonomic richness. We also quantified several kinds of potential refugia for invertebrates and found a positive relationship between richness and a refugia axis that combines amount of dead space with proportion of large substratum particles.