The outstanding property of human language is its diversity, and yet acquisition data is only available for three percent of the world's 6000+ spoken languages. Due to the rapid pace of language loss, it may not be possible to document how children acquire half of the world's indigenous languages in as little as two decades. This loss permanently diminishes the scope of acquisition theory by removing its empirical base. In the face of pervasive language loss, the question of how best to document the language of the last children to acquire indigenous languages assumes critical importance. A collaborative effort by researchers is required to identify the most efficient procedures for documenting children's language, and share them worldwide. This paper makes the case for documenting diversity and outlines steps needed to accomplish this goal.