Emotion and language are very common in young children's everyday lives. Hour by hour, they play, listen, vocalize, react, and emote. Despite the centrality of emotion and language to toddlers' local environments, the dynamic interplay of these communicative signals is practically unexplored. Here, we investigated how fluctuations in caregiver and child affect are linked to caregiver-child communication and children's emerging knowledge of words. Multiday household audio recordings and densely-sampled ratings of affect revealed that, in a US-based sample, children (24-30 months) were more likely to know words that they heard frequently in moments with more positive valence or higher arousal. These moments were also associated with denser communication, suggesting that moments of higher valence or arousal facilitate word knowledge in part by supporting mutually engaging communication. This investigation underscores the importance of natural affective states for understanding how children learn language.