JMS says many insightful things about Zahavi's "handicap principle"
in the book "Animal Signals", but there is also an interesting
tidbit in the Introduction of the Canto edition of "The Theory
of Evolution". Smith writes:
Zahavi has suggested that reliable signals are costly to
make, and that it is this cost that ensures that they are
honest. I was one of those who was slow to accept this
argument, partly because it was expressed in a verbal
rather than in a mathematical form - I find verbal
arguments hard to follow. There is, therefore, a certain
irony in the fact that the correctness of Zahavi's argument
has been confirmed by Enquist and Grafen using mathematical
methods (game theory) that I was partly responsible for
introducing.
One interesting thing is that Maynard Smith's claim that he
finds verbal arguments "hard to follow" is not necessarily
false modesty. It may be the reason that he works so hard to
make his own verbal arguments so crystal clear - otherwise
he can't understand them himself! Which hypothesis may be
yet another permutation of the "handicap principle".