In particular, the ability of the handicap theory to explain the design of signals has never been properly tested.
The handicap theory of biological signalling (Zahavi, 1975, 1977, 1987, 1991; Grafen, 1990b) is now widely applied to many areas of animal communication such as mate choice (Hamilton and Zuk, 1982; Pomiankowski, 1987; Grafen, 1990a), male-male competition (Clutton-Brock and Albon, 1979), offspring begging calls (Godfray, 1991), and pursuit deteffence (Fitzgibbon and Fanshaw, 1988).
A major obstacle to applying the handicap theory to warning signals is that the theory itself has been interpreted in several different ways (see Maynard Smith, 1985; Grafen, 1990b).
Grafen (1990b) put forward a less specific version of the handicap theory that does not make the assumption that conspicuously colored animals necessarily signal their unpalatability by making themselves more palatable.
We have already argued that if it is to provide a viable explanation of warning coloration even the handicap theory has to be combined with a learned component, but its primary emphasis is that prey are avoided because their signals are inherently costly.
If toxins or other defenses act immediately, the handicap theory explains conspicuousness because it demonstrates that a prey individual is prepared to increase the risk of attacks by predators because it can survive them.
In our combined model the handicap theory requires that predators learn to associate degree of conspicuousness with degree of unpalatability.
In contrast, the handicap theory only requires conspicuousness where cheating would otherwise be beneficial.
The most general prediction of such a theory is that, in contrast to the handicap theory, experienced receivers should not treat conspicuousness level as inherently signalling degree of unpalatability.
However, there is a complex way in which the handicap theory could also explain some of the demonstrated enhanced learning features of warning signals.
Under the handicap theory of warning coloration conspicuousness level alone indicates quality, so patterns do not have to be similar, except in conspicuousness, to have the same signal meaning.