Humans have an innate capacity to learn language. This is an undisputed fact. However, what this capacity actually consists of has yet to be worked out in full detail. The main reason for this is that language is an abstract capacity which manifests in thousands of languages that vary widely in every possible domain, and which change continuously over time. Furthermore, empirical research shows that linguistic features of individual languages can influence the learning process and so bias any generalizations about the human capacity for language learning. Thus, any understanding of language or its development must be considered in a cross-linguistic perspective. However, current research in the field has focused on only a small subset of the world’s languages; we have only scratched the surface of what children must learn and how they do it. In this chapter, I offer a solution to this sampling bias. The solution is a maximal diversity approach, which samples from languages that are as structurally diverse as possible. This allows us to simulate the linguistic variability that children must be able deal with in being able to learn any language. Maximum diversity sampling promises insights into the general mechanisms which underlie language development and how distributions of linguistic features in the input work hand in hand with these mechanisms.
