Several proposals have been made as to how human listeners segment continuous speech into discrete words during spoken word recognition. These include segmentation through lexical competition and segmentation based on metrical structure. An experiment is presented which shows that another way listeners solve the segmentation problem is through the use of phonotactic constraints: a word boundary is more likely where the sequence of segments requires there to be a syllable boundary. Monosyllabic Dutch words were easier to detect in bisyllabic nonsense strings when they were aligned with a syllable boundary determined by phonotactics than when they were misaligned with such a boundary. Listeners appear to use several sources of information, phonotactic, metrical and lexical, in their segmentation of continuous speech.
Bibliographic reference. McQueen, James M. / Cox, Ethan (1995): "The use of phonotactic constraints in the segmentation of dutch", In EUROSPEECH-1995, 1707-1710.