MUST RELIABLE SIGNALS ALWAYS BE COSTLY

It has previously been shown that, in some circumstances, animal signals can be a reliable guide to the state of the signaller only if the signal is more costly than is needed to convey the information unambiguously. However, an analysis of a simple model, the Philip Sidney game, showed that there are also circumstances in which cost-free signals can be both reliable and evolutionarily stable. The essential requirement for a cost-free signal to be stable is that the two participants should put the possible outcomes of the interaction in the same rank order, where preference is determined by the effect on fitness: It is not necessary that they should have the same degree of preference. That conclusion, however, arose from a discrete model, in which the various parameters (pay-offs, signal costs) were taken as constant. This paper asks how far the conclusion holds in a more realistic case, in which the parameters vary in the population. Two models are analysed: The original Philip Sidney game, and an extension, in which both participants signal. It is shown that cost-free signals can be stable even if the parameters vary, but that there must be restrictions: If costs and benefits vary uniformly over the whole range, reliable signals must be costly. There is no logical reason why many animal signals should not be free of cost, except that needed to carry the information. © 1994 Academic Press, Inc.