This research examines phonological neighbourhoods in the lexicons of children acquiring English. Analyses of neighbourhood densities were done on children's earliest words and on a corpus of spontaneous speech, used to measure neighbours in the target language. Neighbourhood densities were analyzed for words created by changing segments in word-onset position (rhyme neighbours as in pin/bin), vowel position (consonant neighbours as in pin/pan/) and word-offset position (lead neighbours as in pin/pit). Results indicated that neighbours in children's early lexicons are significantly more often distinguished in word-onset position (rhyme neighbours) and significantly less often distinguished in word-offset position (lead neighbours). Moreover, patterns in child language are more extreme than in the target language. Findings are discussed within the PRIMIR framework (Processing Rich Information from Multidimensional Interaction Representations; Werker & Curtin, 2005). It is argued that early perceptual sensitivity aids lexical acquisition, supporting continuity across speech perception and lexical acquisition.