While organisms are capable of detecting predators via chemical cues, how well these detection abilities are matched to different historical predation regimes remains poorly understood. The Least Killifish, Heterandria formosa, is a small live-bearing fish whose predominant predators differ among habitats. We performed two experiments to see how H. formosa from different populations respond to a chemical cue from either a familiar predator (abundant in that population's habitat) or novel predator (absent from its habitat but abundant in another habitat inhabited by a different population of H. formosa). Our first experiment compared fish from two populations exposed to a familiar predator, a novel predator, and water from each habitat. Fish from the population with lower historical predation risk were more active regardless of treatment but were especially active when cues from a novel predator were offered. Our second experiment compared fish from four populations exposed to either a familiar or novel predator at the same time. Large fish were more active than small fish and, as in the first experiment, fish from the population with the lowest historical predation risk were more active regardless of the identity of the predator. These results suggest that predator recognition is not specific to individual populations but that historical levels of predation risk have selected for different characteristic levels of activity in different populations.