When English-speaking two-year-olds begin producing polysyllabic words, they often omit unstressed syllables that precede syllables with primary stress (Allen & Hawkins, 1980; Klein, 1981; Gerken, 1994a). One proposed mechanism for these omissions is that children omit syllables at a phonological level, due to prosodic constraints that act on outputs. Under such accounts, it has been largely assumed that these syllables are simply missing, or deleted, from children's outputs. The present research consists of a pair of experiments that tested this assumption by investigating the acoustic properties of utterances manifesting or lacking weak initial syllable omissions. In the two experiments, 33 two-year-old children were asked to imitate sentences like ‘He kissed Lucinda’ (often reduced as expected to a disyllabic trochaic form, e.g. ‘He kissed _cinda’) and ‘He kissed Cindy’. Durations of each child's imitations were measured from the onset of the verb to the onset of the name, for each pair of sentences containing the reduced or unreduced disyllabic forms, for example, ‘kissed _cinda’ vs. ‘kissed Cindy’. Our results yielded a significantly longer duration for the verb-onset to name-onset portion of sentences containing reduced ‘_cinda’-type names than for sentences with ‘Cindy’-type names. This finding provides evidence that children do not completely delete weak syllables. Rather, the data from the phonetic analysis indicate that some prosodic trace exists of the omitted syllable.