A perception study of syntactic disambiguation using resynthesized speech
Mariko Sugahara, Yuki Hirose · The Journal of the Acoustical Society of America · 2006
It is already well known that when a preceding word (W1) is accented, the following word (W2) is likely to undergo a postaccent downstep. Such a downstep is especially evident in a left-branching structure (henceforth LB) as in [[W1 W2] W3], while it is mitigated in a right-branching structure (henceforth RB) as in [W1 [W2 W3]]. That is, ups and downs of the F0 peak of W2 disambiguate the RB and the LB structure when a postaccent downstep is available (Selkirk and Tateishi, 1991; Kubozono, 1993, Hirose, 2005). Two questions, however, emerge. One is whether or not the magnitude of W3’s downstep would ever contribute to the branching disambiguation. The other is what would happen when all words are unaccented, i.e., when no postaccent downstep is available. We carried out a perception study using resynthesized speech as stimuli, to examine those questions. Our results show that neither the W3 peak values of a postaccent environment nor the W2 peak values of an unaccented environment had significant effect on the LB vs. RB disambiguation. Such results will be compared with Hirose’s (2005) production and perception study on spontaneous speech.