The possible-word constraint in Cantonese speech segmentation

Introduction Speech segmentation is a central issue of spoken language comprehension research (Cutler, 2001). And, recently, one important solution comes from the discovery of a mechanism operated in our lexical system, the Possible-Word Constraint (PWC). In their word-spotting experiments, Norris, McQueen, Cutler and Butterfield (1997) observed that listeners usually found it more difficult to spot the real word apple in the nonsense word string fapple than in vuffapple. Both f and vuff are not words in English, but only the latter one could be a possible word in the sense that the word vuff generally satisfied all the phonological conditions for becoming a word. The consonant of f would never been a possible word under any circumstances. Hence, Norris et al. (1997) proposed that our lexical system would be sensitive to this kind of discrepancy and inhibit the activation of those words that stranded an “impossible word candidate” residue in the speech signal during speech processing. This procedure was called the Possible Word Constraint (PWC). However, realization of this mechanism has been done in English (Norris, McQueen, Cutler, Butterfield & Kearns, 2001); Japanese (McQueen, Otake, & Cutler, 2001) and Sesotho (Cutler, Demuth & McQueen, 2002) so far. Hence, the present study aims to further examine the efficiency of the PWC in Cantonese, a language that differs significantly from most Indo-European languages (English, Dutch, French) in its use of lexical tones, phonotactic structure and its special morphemic nature of words.