Many language acquisition theories underline the importance of high-quality, child-centered interactions, in consonance with some correlational and intervention evidence. However, anthropological reports suggest that, in some cultures, such interactions are vanishingly rare. Should we then predict strong cultural variation in the timeline of early language acquisition? In this talk, I first present the results of a meta-analysis that quantitatively confirms the extent of variation in input quantity. I then provide an overview of modeling studies on the learnability properties of child-directed speech, compared to other registers, which suggests small and unstable advantages for the former. Finally, I lay out a framework for acquisition that accommodates both previous findings and theories, showing that cultural differences in acquisition are only predicted in a small range of theories and language domains.