The study of language has relied (to a varying degree) on the patterns of worldwide linguistic diversity. However, language structures and linguistic lineages are not ideally distributed in space and time: we have too many of a few and too few (or nothing) of many. Hence, it is not surprising that researchers try to either “fix” the complicated patterns of linguistic diversity (by e.g. choosing curated samples of languages) or directly abandon that route in favor of alternative venues of research, such as laboratory experiments with artificial languages or toy computational models of language. In this presentation I will argue that a reappraisal of linguistic diversity as a primary source for linguistic theories is in place if we are willing to get our hands dirty with the wider range of human sciences, including the study of behavior, culture, cognition and history.