Communication is ubiquitous in biology, and agreement on terms essential for scientific progress. Yet there is no agreed definition of biological communication. Definitions couched in terms of adaptation are often used, but there is significant variability in exactly which criteria are invoked. An alternative is to define communication in terms of information transfer. This article reviews the merits of these approaches, and argues that the former is to be preferred, so long as we demand that both the signal and the response be adaptive, rather than just one or the other, as is common. Specific concerns with the definition are addressed, and it is then explained why an account of communication predicated on information transfer is necessarily derivative upon such an approach. Other alternatives and some variants of the adaptationist definition are also briefly discussed.