The ability to recall the location of a predator and later avoid it was tested in nine populations of rainbowfish (Melanotaenia spp.), representing three species from a variety of environments. Following the introduction of a model predator into a particular microhabitat, the model was removed, the arena rotated and the distribution of the fish recorded again. In this manner it could be determined what cues the fish relied on in order to recall the previous location of the predator model. Fish from all populations but one (Dirran Creek) were capable of avoiding the predator by remembering either the location and/or the microhabitat in which the predator was recently observed. Reliance on different types of visual cues appears to vary between populations but the reason for this variation remains elusive. Of the ecological variables tested (flow variability, predator density and habitat complexity), only the level of predation appeared to be correlated with the orientation technique employed by each population. There was no effect of species identity, which suggests that the habitat that each population occupies plays a strong role in the development of both predator avoidance responses and the cues used to track predators in the wild.
Brown, C. (2003). Habitat–predator association and avoidance in rainbowfish (Melanotaenia spp.). Ecology of Freshwater Fish, 12(2), 118-126.