Recent evidence suggests that environmental conditions may affect whether fishes do or do not respond to the presence of chemical alarm cues in water. We present a simple model which suggests that the combination of risk of predation and information from other sources will determine when fishes should react to these chemical cues. We tested this model with a laboratory experiment which manipulated the risk of predation by altering the animals (hungry or well fed), or their environment (presence or absence of cover). We also altered the availability of visual information by manipulating the water clarity. Consistent with our model, fishes were most likely to react to chemical alarm cues in the absence of visual information and when the perceived risk of predation was high. The manipulation of either parameter was able to extinguish this response.