Several theorists have suggested that infants use prosodic cues such as pauses, final lengthening, and pitch changes to identify linguistic units in speech. One potential difficulty with this proposal, however, is that the acoustic shape of an utterance is affected by many factors other than its syntax, including its phonetic, lexical, and discourse structure. This has raised questions about how the infant could use timing and pitch as cues to any aspect of linguistic structure without simultaneously factoring out other effects. Acoustic analyses of connected samples of spontaneous speech addressed to 13.5-14-month-old infants by American English- and by Japanese-speaking mothers revealed that both utterance- and phrase-level acoustic regularities were large enough to be detected in spontaneous speech without correcting for other influences on the same acoustic features. (1) Utterance-final vowels were lengthened and underwent exaggerated pitch changes in both languages, and (2) local acoustic changes in duration (English) or pitch (Japanese) were reliably associated with some phrase boundaries within utterances. These findings suggest that a naive listener could estimate a rough prosodic template for each language based on robust acoustic patterns in observed sentences. We discuss ways in which the learner could combine acoustic and distributional analyses across utterances to acquire language-specific variations in prosodic bracketing cues and to obtain indirect perceptual evidence for the internal structure of utterances.
Stored only when both papers are in the corpus — 136 citations exist upstream.