Repetition and novelty in six million words of child-directed speech

Fausey, C. 1 & Willits, J. 2

1 University of Oregon
2 Indiana University

The number of opportunities to encounter a word matters for how we learn and use language. In early life, the frequency landscape changes dramatically as infants encounter words for the very first time followed by opportunities to build knowledge via repetitions of these words. We present three new discoveries about frequency landscapes by analyzing all of the child-directed speech available in CHILDES. Lexical encounters are (1) Early: In their first year, infants encounter 1 new word type per 100 tokens, which drops to 1 new type per 300 tokens in their third year; (2) Bursty: Within most noun categories, children encounter many new word types within the same month; (3) Repeated: After the first year, over two-thirds of the words that children encounter are words they?ve heard before. The combination of an early sweep through the range followed by massive practice may facilitate the precocious language learning of human infants.