Adaptive scepticism: tolerance of unreliable callers varies with the cost of failing to respond

We have previously shown that the reliability of low-cost vocal signals such as chicken food calls is maintained by a social constraint. Playback experiments in which the apparent honesty of individual males was manipulated reveal that hens are sensitive to this attribute, responding differentially after three days of experience. The costs of such 'scepticism' should vary with signal type. If hens fail to respond to a food call, they might lose a preferred food item. In contrast, failure to respond to an aerial alarm call potentially increased predation risk. This logic generates the prediction that receivers should tolerate lower levels of signalling reliability when the costs of Type II error are high. We report the first test of this model. Spectrogram cross-correlation reveals that male identity accounts for 85% of the variance in aerial alarm call structure. Individual males are thus acoustically distinctive. We next tested the effects of variation in reliability. Hens heard a series of alarm calls from two males, one reliable (each playback was followed by a hawk animation presented overhead and the other unreliable (the monitor displayed a blank background). Playbacks were repeated daily for 10 days, using a large set of call exemplars and predator stimuli to prevent habituation. In striking contrast to the effects of identical experience with food calls, hens continued to respond to unreliable alarm callers. These results suggest that receivers do not use a fixed rule for assessing reliability, but have flexible criteria that reflect the environmental event predicted by the signal.