Are tones aligned with articulatory events? Evidence from Italian and French - Archive ouverte HAL Accéder directement au contenu
Chapitre D'ouvrage Année : 2007

Are tones aligned with articulatory events? Evidence from Italian and French

Résumé

Tonal alignment work has suggested that the temporal location of tonal targets relative to segmental "anchors" might be governed by principles of synchrony and stability (Arvaniti et al 1998, Ladd et al. 1999, inter alia). However, a number of discrepancies have emerged in the cross-linguistic study of alignment. For instance, despite some regularities in the alignment of L targets (Caspers and van Heuven 1993; Prieto et al. 1995), the alignment of H targets appears to be quite controversial. In fact, it is sometimes difficult to find definite segmental landmarks to which such targets might be aligned. Also, most of the alignment proposals so far inherently assume that if some anchors for tonal alignment do exist they must be acoustic in nature. A plausible alternative would be to assume that such anchors are primarily articulatory, which would explain why in some cases the underlying regularities would be masked. Hence, we adopt a new experimental paradigm for alignment research in which articulatory measures are performed simultaneously with acoustic measures. In order to test the constant alignment hypothesis, a preliminary study (D'Imperio et al. 2003) was conducted in which various latency measures, both acoustically and articulatorily based, were analyzed. Specifically, the kinematics of OPTOTRAK markers attached to the speaker's upper and lower lip was tracked over time during the production of the corpus sentences. The melodic target considered is the H tone of LH nuclear rises in Neapolitan Italian. In this variety, yes/no question LH rises are systematically later than (narrow focus) statement LH rises (D'Imperio 2000, 2001, 2002; D'Imperio and House 1997). In order to test the hypothesis of constant anchoring of H targets, the materials were produced with two different rates of speech, i.e. normal and fast. Summarizing the results, H targets of nuclear rises in Neapolitan statements and questions appear to be more closely phased with the articulatory dimension of between-lip distance than with two of the most commonly employed acoustic segmental landmarks for tonal alignment (i.e., onset and offset of stressed vowel). Statement H tones are phased with maximum between-lip distance within the stressed syllable. Note that this location does not correspond to any identifiable segmental boundary, acoustic event or phonological unit, and does not overlap with RMS peak amplitude. In fact, RMS peaks were generally much earlier than articulatory peaks, hence further away from H peaks. This calls for the collection and analysis of more articulatory data (especially jaw and tongue movements) to shed light on tonal alignment issues.In a second study, a French corpus was collected on the basis of the alignment contrast found by Welby (2003, in press). Welby's results show that listeners use the alignment of the initial rise (LHi) in French Accentual Phrases as a cue to speech segmentation. Specifically, listeners exploit the presence of an early rise to demarcate the beginning of a content word. In the present study, a corpus was built with a set of utterances displaying this specific alignment contrast. The kinematics of 10 pellets (8 on the face and tongue, 2 references) was tracked over time using an electromagnetometer (EMA, Carstens). The phasing of several articulatory events relative to the L and H part of the early rise were examined. The preliminary results seem to point to some kind of fine alignment specification for the L and H target. Specifically, we hypothesize that tonal target commands of Neapolitan as well as French rises are phased with commands of the supralaryngeal articulator involved to produce the segments to which the tone is associated. Regarding the word segmentation issue for French, it is important to study alignment in a diachronic perspective since we know of case of speech segmentation errors that can lead to lexical reinterpretation and change (l'abondance "abundance" > la bondance, from Welby 2003). We also take these results to suggest that not all rises align in the same way with the associated syllable. Though the role of articulatory constraints is important, the exact phasing properties of prosodic events are language-specific. Since prosody has recently become the realm of investigation of the Task Dynamics program (Byrd and Saltzman 2003), our alignment work will be cast under such a perspective.

Domaines

Linguistique
Fichier principal
Vignette du fichier
2197.pdf (757.04 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-00244489 , version 1 (07-02-2008)

Identifiants

  • HAL Id : hal-00244489 , version 1

Citer

Mariapaola d'Imperio, Robert Espesser, Hélène Loevenbruck, Caroline Menezes, Noël Nguyen, et al.. Are tones aligned with articulatory events? Evidence from Italian and French. Cole, Jennifer. Papers in Laboratory Phonology 9, Mouton de Gruyter, pp.577-608, 2007. ⟨hal-00244489⟩
460 Consultations
250 Téléchargements

Partager

Gmail Facebook X LinkedIn More