Ken Hale argued forcefully that linguistic diversity provides a crucial resource for linguistics. If we take this lesson seriously, this diversity provides the opportunity to recast the language sciences in a Darwinian mold — for we are the only species whose communication system differs fundamentally across social groups in both form and meaning. Starting from work by myself and colleagues in Island Melanesia, I argue that the patterns of diversity can be understood almost entirely in terms of cultural evolution over deep time. Language universals — understood as constraints specific to the language capacity — do not seem to strongly constrain these patterns: Absolute universals are almost always confronted by counterevidence, and Greenbergian statistical universals also turn out to be less robust than presumed. The puzzle that then arises is what exactly endows humans and humans alone with the language capacity? I argue that humans share an innate infrastructure for language which is largely pragmatic and ethological in character: the capacity for vocal learning, multimodal signaling, turn-taking, and especially intention-recognition. These capacities allow infants to bootstrap themselves into the local language. I suggest that the future for the language sciences involves embracing the diversity, and exploring how on the one hand one species can support such a diverse range of communication systems, and how conversely, within a speech community, internal variation can be damped down to produce relative uniformity.