A documentation of phonetic detail in onset-embedded words
Katherine M. Crosswhite, Mikhail Masharov, Joyce M. McDonough, Michael K. Tanenhaus · The Journal of the Acoustical Society of America · 2001
As part of a study of online processing of onset-embedded words, we documented phonetic-level properties of speech production, such as duration differences, that may cue for word identification, segmentation and/or lexical access. Forty nouns (20 pairs like cap/captain, doll/dolphin) were placed in sentences in three frames (nuclear, prenuclear, and utterance-final position). Each sentence was read, then pronounced from memory by eight speakers randomly ordered. In keeping with previous research, instrumental analyses revealed strong effects of syllable count on vowel duration [Klatt, J. Phonetics 7, 279–312 (1979)] and initial consonant VOT [Lisker and Abramson, Lang. Speech 10, 1–28 (1967)]. A similar effect on initial consonant duration was also found. These effects were stable across speakers, words and contexts, suggesting they may be available as cues to overall word length in online segmentation. To test this, the noun pairs of one speaker were cross spliced to produce congruous and incongruous tokens, and presented to listeners in a ‘‘visual world’’ eyetracking paradigm [Allopenna et al., J. Mem. Lang. 38, 419–439 (1998)]. We will report on whether beneficial durational cues of congruous tokens and/or potentially misleading incongruous cues had more influence on online word segmentation of our subjects. [Work supported by NIH and NSF.]