This project seeks to describe the development of phonological structure in child and child-directed speech using metrics of word similarity (e.g. neighborhood density, Levenshtein distance) and phonological probability (aggregated syllable probabilities, positional bigram frequencies, etc.). A related goal is to determine which levels of analysis (e.g. the phoneme, syllable, articulatory gesture) are most appropriate for child speech. We are working towards developing measures of phonological complexity that can be used to predict later language development such as decoding skills in reading. At present, there are two major efforts in this project. One seeks to use survival modeling to estimate how a word’s phonological and other properties impact the age at which it enters children’s productive lexicons. The other uses network theory to build graphs of the child and child-directed speech lexicons, and to compare them on global graph-theoretic properties such as the degree distribution, mean shortest path, size of the giant component, clustering coefficient, transitivity, and assortative mixing by degree.