Animals use diverse forms of communication, from sound signals to body postures. Recent ethological studies have reported a unique syntactic communication of a songbird, the Bengalese finch (Lonchura striata var. domestica). Male Bengalese finches sing complex courtship songs, which can be reconstructed by finite automata, and female Bengalese finches prefer complex songs, as opposed to monotonous or random ones. These facts suggest that the song syntaxes of male birds may have evolved as a result of sexual selection by female birds. Inspired by this hypothesis, we developed a communication model that is a system coupling different types of automaton, one for song production by males and another for song evaluation by females. We applied this model to study the evolution of syntactic animal communication in terms of the self-organization of coevolving automata. Three types of courting strategies as well as a relationship between the song syntax and female preference emerged. We argue that despite the simple communication involved, the complexity and diversity of song syntaxes can evolve via diverse female preferences.