Limits to reliability in avian aggressive signals

Early game theory models of aggressive signaling predicted that aggressive signals would be unreliable because all signalers would be selected to exaggerate their signals regardless of their aggressive intentions. Recently some signals have been shown to convey reliable information about aggressive intentions, but in all cases reliability is limited. Here we test whether limits to reliability are due to exaggeration of aggressive intentions, or 'bluffing', as originally envisioned. Our earlier work on two related songbird species, song sparrows (Melospiza melodia) and swamp sparrows (M. georgiana), has shown that low intensity 'soft song' is the signal that best predicts attack in both cases. Here we test two predictions of the bluffing hypothesis: (1) that the distribution of the number of soft songs given per signaler should be skewed towards high values because of widespread exaggeration and (2) that unreliability should arise from over-signaling (giving many displays and not attacking) rather than from under-signaling (giving few or no displays and attacking). Neither prediction is upheld in either species. We propose that the surprising prevalence of under-signaling can be explained by the opportunity costs of aggressive signaling.