Metrical Segmentation in Dutch: Vowel Quality or Stress?

Abstract

Previous experiments using a word-spotting task suggest that English listeners use metrically strong syllables to segment continuous speech into discrete words (Cutler & Norris, 1988). The present study is concerned with this metrical segmentation strategy in Dutch. Although Dutch and English share general metrical properties, they differ in ways that may affect segmentation. First, the acoustic cues for metrically strong syllables are less salient in Dutch than in English; hence a metrical segmentation strategy is less likely to be applied by Dutch listeners. Second, vowel quality depends less on metrical structure in Dutch than in English; hence segmentation in Dutch is presumably triggered by other acoustic cues, namely, those related to stress.

Experiment 1 shows that stress strongly affects Dutch listeners' ability and speed in spotting Dutch monosyllabic words in disyllabic nonwords. Experiment 2, however, finds the same stress effect when only the target words are presented, without a subsequent syllable triggering segmentation. A third experiment shows a small effect of vowel quality on error scores, but not on latencies. These results suggest that Dutch listeners do not apply a metrical segmentation strategy. The discrepancy between the two languages suggests that segmentation strategies may depend on language-specific regularities in the phonology and in the lexicon.