Foreign-Language Knowledge Enhances Artificial-Language Segmentation

Annie Tremblay, Mirjam Broersma

This study investigates whether and how foreign-language knowledge affects the use of non-native cues in speech segmentation. It does so by testing whether Dutch listeners’ French knowledge enhances their use of word-final fundamental-frequency (F0) rise — consistent with the typical French prosodic pattern — in artificial-language (AL) speech segmentation. More specifically, this study examines whether Dutch listeners with good French knowledge outperform Dutch listeners with limited French knowledge in the selection of AL words over (nonword or partword) foils, following exposure to an AL with word-final F0 rises. Dutch listeners with good French knowledge completed the AL-segmentation task from Kim et al.’s [2] word-final F0-rise condition. The results were compared to Kim et al.’s [2] Dutch listeners with limited French knowledge and Tremblay et al.’s [1] native French listeners in the same condition. Dutch listeners with good French knowledge performed more accurately than Dutch listeners with limited French knowledge but less accurately than native French listeners on trials with partword foils, with the three groups not differing on trials with nonword foils. Given these results, we propose that foreign-language knowledge can help listeners compute the conditional probability of co-occurrence of successive syllables in an AL and can thus enhance AL speech segmentation.

 DOI: 10.21437/Interspeech.2019-2446

Cite as: Tremblay, A., Broersma, M. (2019) Foreign-Language Knowledge Enhances Artificial-Language Segmentation. Proc. Interspeech 2019, 2658-2662, DOI: 10.21437/Interspeech.2019-2446.

  author={Annie Tremblay and Mirjam Broersma},
  title={{Foreign-Language Knowledge Enhances Artificial-Language Segmentation}},
  booktitle={Proc. Interspeech 2019},