Many species produce specific alarm vocalizations when they encounter predators. There is considerable interest in the degree to which bird, ground-dwelling sciurid rodent, and primate alarm calls denote the species or type of predator that elicited the vocalization. When there is a tight association between the type or species of predator eliciting an alarm call, and when a played-back alarm call elicits antipredator responses qualitatively similar to those seen when individuals personally encounter a predator, the alarm calls are said to be functionally referential. In this essay I aim to make two simple points about the evolution of functionally referential alarm communication. Firstly, functionally referential communication is likely to be present only when a species produces acoustically distinct alarm vocalizations. Thus, to understand its evolution we must study factors that influence the evolution of alarm call repertoire size. Secondly, and potentially decoupled from the ability to produce acoustically distinctive alarm vocalizations, species must have the perceptual and motor abilities to respond differently to acoustically-distinct alarm vocalizations. Thus, to understand the evolution of functionally referential communication we also must study factors that influence the evolution of context-independent perception. While some factors may select for functionally referential alarm communication, constraints on production or perception may prevent its evolution.