During floods, river invertebrates may be swept downstream in large numbers. This so-called "catastrophic drift" leads to a major redistribution of animals, as well as reduced fitness and increased mortality among drifters. We present the first field evidence of the role of sediment movement in triggering catastrophic drift. Experiments indicate that the loss of invertebrates from the bed becomes exponential when shear stress reaches the threshold that entrains bedload. However, we found that low rates of bedload are sufficient to rapidly denude patches of riverbed of their invertebrates and so trigger mass drift. Such low bedload rates occur during small floods. As small floods occur frequently, our results suggest that episodes of catastrophic drift are frequent. This conclusion is counterintuitive, as the persistence of invertebrate communities on riverbeds suggests that such events cannot be truly catastrophic. Moreover, the drift losses that we observed occurred in the absence of significant geomorphic disturbance; this is inconsistent with the notion of catastrophic drift being a response to hydrological disturbance events. We argue that a new definition of catastrophic drift is needed, a definition based not on drift magnitude or the triggering role of sediment movement, but on the population consequences of downstream displacement.