Processing resyllabified words in French

Abstract

In French, the final [ ] of dernier is not pronounced in dernier train (last train), but is pronounced, in the following syllable, in a liaison environment like dernier oignon (last onion). Due to liaison, dernier oignon becomes homophonous with dernier rognon (last kidney). In four pairs of cross-modal priming experiments, French participants made visual lexical decisions to vowel-or consonant-initial targets (e.g., oignon, rognon) following both versions of spoken sentences like C'est le dernier oignon/rognon. Facilitation was found for both types of target when targets matched the speakerÕs intended segmentation, but was weaker when they mismatched the intended segmentation. In unambiguous sentences there was facilitation only for targets matching the speakerÕs intentions. The consonants in the liaison environments were shorter than the word-initial consonants (e.g., [ ] in dernier oignon vs. rognon). Word recognition therefore appears to be influenced by subphonemic cues to the words that speakers intend. Ó 2002 Elsevier Science (USA). All rights reserved. The process of liaison in French speech might appear to create a problem for French listeners. When a French speaker says the word dernier (last), for example, the final [ ] will not be produced if the next word begins with a consonant (e.g., dernier train, last train), but will be produced if the next word begins with a vowel (e.g., dernier oignon, last onion). Furthermore, when the [ ] is produced, it appears in the initial position of the first syllable of the following word. This phenomenon might thus have little effect on the recognition of the first word, since this word is phonologically identical up to its last vowel whether the [ ] appears in the following syllable or not. But liaison would appear to make recognition of the second word more difficult, since it makes vowelinitial words seemingly consonant-initial. In the worst case, liaison can generate another word (such as rognon, kidney 2 ). How then do French listeners recognize vowel-initial words in liaison environments? We address this question by examining spoken word recognition in sentences which, according to phonological analysis, have the same sequence of phonemes and the same syllabification, and hence are lexically ambiguous (e.g., C'est le dernier oignon/C'est le dernier rognon; ItÕs the last onion/ kidney). Our findings have important implications for theories of continuous speech recognition. Since it is often assumed that word boundaries tend to coincide with syllable boundaries, syllable onsets have been proposed as locations where word boundaries are more likely to occur (Content, Kearns, & Frauenfelder, 2001a;Content, Meunier, Kearns, & Frauenfelder, 2001b;Cutler & Norris, 1988;Norris, McQueen, Cutler, & Butterfield, 1997;Vroomen & de Gelder, 1997 2 Although they are spelled differently, oignon and ognon in rognon sound the same.0749-596X/02/$ -see front matter Ó 2002 Elsevier Science (USA). All rights reserved. PII: S 0 7 4 9 -5 9 6 X ( 0 2 ) 0 0 5 1 3 -2 are differen...

The role of prosodic boundaries in the resolution of lexical embedding in speech comprehension

“…This finding adds to a growing body of research that suggests that fine-grained subphonemic information in the speech signal can modulate lexical activation, both in the recognition of individual words (Andruski, Blumstein, & Burton, 1994;Dahan, Magnuson, Tanenhaus, & Hogan, 2001;Marslen-Wilson & Warren, 1994;McQueen, Norris, & Cutler, 1999) and in the recognition of words in continuous speech (Gow, 2002;Gow & Gordon, 1995;Spinelli, McQueen, & Cutler, 2003;Tabossi et al, 2000). Our results are also consistent with Davis et al (2002), who showed that subphonemic cues can be used to resolve ambiguities caused by lexical embedding.…”

Spoken-word recognition in foreign-accented speech by L2 listeners

“…A number of studies, for instance, have found that assimilated word forms, like greem for English green, are recognized correctly only in assimilation-licensing contexts such as greem bench where the /m/ of greem and the /b/ of bench match in place of articulation (e.g., Coenen, Zwitserlood, & Boelte, 2001;Gaskell & Marslen-Wilson, 1996Gow, 2001Gow, , 2002Mitterer & Blomert, 2003). Similarly, reduced word forms like posman for postman (Ernestus, Baayen, & Schreuder, 2002;Mitterer & Ernestus, 2006;Sumner & Samuel, 2005) or resyllabified forms (e.g., Spinelli, McQueen, & Cutler, 2003;Vroomen & de Gelder, 1999) are recognizable in contexts that license the surface change. All these findings suggest that listeners can recover the underlying form from surface variation, but restoration only happens if the variation occurs in viable contexts.…”

La liaison : approches contemporaines

“…On peut cependant noter à cet égard que la faiblesse acoustique et perceptuelle des consonnes de liaison, observée notamment par Dejean-de-laBâtie (1993), , Spinelli et al (2002Spinelli et al ( , 2003, Coquillon & Astésano (2008) et , n'appuie pas de façon générale l'analyse morphémique des consonnes de liaison. Certaines études ont montré que les consonnes ayant un statut morphémique sont plus longues que les consonnes identiques non morphémiques, par exemple le [s] flexionnel de l'anglais wrecks par rapport au [s] lexical de Rex (Walsh & Parker, 1983 ;Losiewicz, 1992).…”

“…Les indices acoustico-phonétiques associés à la liaison peuvent influencer l'accès lexical également dans les énoncés ambigüs au niveau segmental. Spinelli, McQueen et Cutler (2003) [ʁ]oignon < dernier rognon (ambiguë), demi rognon (pas ambiguë), ancien nitrate (contrôle)) ; en outre, les mots cibles à consonne initiale (p.ex. ROGNON) étaient davantage activés lorsqu'ils étaient contenus dans l'amorce que lorsqu'ils n'étaient pas contenus dans l'amorce (p.ex.…”

“…Notre expérience s'inspire en partie de celle intermodale employée dans Spinelli et al (2002) Nos prédictions sont les suivantes : si les différences acousticophonétiques sont perçues et employées par les auditeurs pour désambi-guïser les situations de liaison, les syntagmes hybrides créés par la technique de splicing croisé, qui contiennent des indices de segmentation contradictoires, devraient moins favoriser les cibles et au contraire augmenter l'activation des compétiteurs. Par ailleurs, si, tout comme dans Spinelli et al (2003), les mots à consonne initiale font davantage compé-tition que ceux à voyelle initiale, il se peut que l'information acousticophonétique ne suffise pas à augmenter l'activation des cibles à voyelle initiale en contexte de liaison et diminuer l'activation des compétiteurs à consonne initiale.…”