Gradient effects of tonal scaling in the segmentation of Korean speech: An artificial-language segmentation study

Abstract

French and Korean have similar intonational systems but differ in the alignment of the phrase-final High (H) tone and scaling of the following phrase-initial Low (L) tone. Tremblay et al. [1] found that Korean listeners have difficulty using tonal cues to segment French speech, raising the question of whether Korean listeners’ segmentation of French was inhibited by the different alignments of the phrase-final H tone or scaling of the phraseinitial L tone in the two languages. This study investigates this issue, thereby shedding light on the importance of fine-grained language-specific tonal cues in speech segmentation. Native Korean listeners completed three artificial-language (AL) segmentation tasks over three sessions. In Experiment 1, one AL contained no tonal cues to word-final boundaries (control), one contained French alignment cues, and one contained Korean alignment cues. Experiments 2 and 3 were identical to Experiment 1, except the phrase-initial L tone in the ALs containing prosodic cues was lowered by 20 Hz and by 40 Hz, respectively. The results showed that Korean listeners’ segmentation of the ALs improved as the phrase-initial L tone was lowered, highlighting the gradient effects of tonal scaling in Korean listeners’ speech segmentation, consistent with the intonational grammar of the language. © 2018, International Speech Communications Association. All Rights Reserved.