The addition of nocturnal, Hoplias malabaricus, and diurnal, Crenicichla alta, predatory fishes downstream of barrier waterfalls increases predation threat for a killifish, Rivulus hartii, in Trinidadian streams. We hypothesized that the diel patterning of predation risk would affect prey movement rates, and tested this hypothesis by comparing movement in river sites/zones containing both the nocturnal and diurnal predator with movement in river sites/zones containing only the nocturnal taxon. We evaluated this prediction in the framework of an intermediate threat hypothesis (ITH) that holds that movement will be highest at some intermediate level of threat. We marked prey fish in study sites in two watersheds of a river, each with waterfalls that divided the river into three zones: a predator absent zone (P0), a zone with one nocturnal predator (P1), and a zone with one nocturnal and one diurnal predator (P2), and tested the ITH prediction that movement will be ordered as P0 P2. The single predator promoted longitudinal movement by Rivulus (P0 P2) as predicted by the ITH. However, movement by larger, less vulnerable Rivulus remained elevated (P1=P2 or P2>P1). A displacement experiment in each zone found that threat tended to reduce the probability of a displaced fish reaching home, but the two predator zones did not differ from one another in their effect on this probability. Hence, the prediction that predator activity over the full 24 h diel cycle would retard movement, P2<P1, was not supported with respect to homing. Because habitat and predator communities change predictably from headwater streams to larger rivers in many lotic ecosystems, we present a conceptual model for predicting fish movement behavior along this continuum. The model posits an important role for predation threat, and the size and spacing of refuge patches, suggesting that human alterations of these factors will affect the natural movement of fish in streams.