One of the most fascinating aspects of language acquisition is that within a range of "normal" exposure and "typical" development, all children acquire the language in their environment, on a similar timescale. At the same time, the specific input a child gets dictates what she is in principle able to learn: a child who has never seen or heard of kangaroos will not learn the sound or meaning of that word. In this talk I examine the environment for early language acquisition, asking two central questions: (1) how much variability (or redundancy!) is there in the words that young infants see and hear, at the group and individual level, and (2) how does infants' home environment predict their own productions, and their performance on word comprehension measures in the lab. I will examine these questions in part within a rich multimodal longitudinal dataset dubbed SEEDLingS, discussing recent results of several eyetracking studies and corpus analyses probing early word learning. I will conclude by laying out new and ongoing work in my lab expanding beyond our typical 'boutique' samples to infants experiencing a broader range of learning environments.