Frequency of occurrence is one of the main drivers of children’s early vocabularies. However, multiple authors have concluded that it fails to account for one of the most important phenomena in word learning— the order in which children acquire nouns and verbs. Across many languages, there is an early noun bias despite nouns occurring far less frequently than verbs in caregiver speech. In this study we use large-scale caregiver–child corpora and a computational model to reexamine noun and verb learning in English, German, Spanish, and Mandarin. Our results show clear evidence for variations in the strength of the noun bias across languages that can be directly explained by a combination of the number of different nouns and verbs used by caregivers and their suffi cient frequency: how often most nouns and verbs need to occur in caregiver speech to be learned. Our fi ndings provide further evidence for frequency shaping early children’s language devel- opment, in this case mapping onto the relative distributions of nouns and verbs in their mental lexicons.