EUROSPEECH 2003 - INTERSPEECH 2003
Although word stress is usually seen as a powerful speech-segmentation cue, the results of several cross-modal fragment priming experiments reveal strong limitations to stress-based segmentation. When stress was pitted against phonotactic and coarticulatory cues, substantial effects of the latter two cues were found, but there was no evidence for stress-based segmentation. However, when the stimuli were presented in a background of noise, the pattern of results reversed: Strong syllables generated more priming than weak ones, regardless of coarticulation and phonotactics. Furthermore, a similar dependency was found between stress and lexicality. Priming was stronger when the prime was preceded by a real than a nonsense word, regardless of the stress pattern of the prime. Yet, again, a reversal in cue dominance was observed when the stimuli were played in noise. These results underscore the secondary role of stress-based segmentation in clear speech, and its efficiency in impoverished listening conditions. More generally, they call for an integrated, hierarchical, and signal-contingent approach to speech segmentation.
Bibliographic reference. Mattys, Sven L. (2003): "Stress-based speech segmentation revisited", In EUROSPEECH-2003, 121-124.