What does it take to learn a language?: Lessons from the field
Recent advances in large language models (LLMs) have revived longstanding debates about what it takes to learn a language. Much of the focus of these debates revolves around the minimal amount and kind of exposure that can support human language learning, making relevant all existing findings on cross-cultural variation in early linguistic input. Such empirically informed benchmarks are crucial to the continued development and discussion of language-using artificial agents. However, in this talk I will give a different answer to the question of "what does it take to learn a language?". Covering research on linguistic input and early phonological and lexical development in two small-scale communities (one Mayan and one Papuan), I will argue that discussions of learning that are solely cognitively, linguistically, and quantitatively driven can lead us to miss the social aspects of language learning that are essential to characterizing what it is children actually learn. I will close by attempting to sketch a picture of how we can study language acquisition in context.
