Main content
Segmentation of highly vocalic speech via statistical learning: Initial results from Danish, Norwegian, and English.
- Fabio Trecca
- Stewart M. McCauley
- Sofie Riis Andersen
- Dorthe Bleses
- Hans Basbøll
- Anders Højen
- Thomas O. Madsen
- Ingeborg Sophie Bjønness Ribu
- Morten H. Christiansen
Date created: | Last Updated:
: DOI | ARK
Creating DOI. Please wait...
Category: Project
Description: Research has shown that contoids (i.e., phonetically-defined consonants) may provide more robust and reliable cues to syllable and word boundaries than vocoids (i.e., phonetically-defined vowels). Recent studies of Danish — a language characterized by frequent long sequences of vocoids in speech — have suggested that the frequent lack reduced occurrence of consonantal sounds contoids may make speech intrinsically harder to segment than in closely related languages, such as Norwegian. In this study, we addressed this hypothesis empirically in an artificial language learning experiment with native speakers of Danish, Norwegian, and English. We tested whether artificial speech consisting of concatenated contoid-vocoid syllables is easier to segment than speech consisting of vocoid-vocoid syllables (where the first segment is a semivowel and the second a full vowel). Contrary to what was expected, we found no effect of the phonetic makeup of the syllables on speech segmentability. Possible interpretations and implications of this result are discussed.