Since the pioneering work by Durand (1936), several studies of the French "enchaînement" phenomenon have isolated subtle phonetic contrasts that contribute to the segmentation of lexically ambiguous strings. Word-strings with enchaînement like "tante roublarde" have been shown to differ from "temps troublant" both in the perceptual duration attributed to the pre-boundary vowel (as a function of pitch factors), and in the articulatory properties of the following consonants (Delattre 1940; Malmberg 1971). Using lexically ambiguous sequences (e.g. t (#)t(#)ru) extracted from similar pairs of two-word strings (cf. "temps troublant" vs. "tante roublarde"), we recently showed that both the production and the perception of differences in such pairs depended upon the cluster class (Dumay et al. 1999). In obstruent-liquid but not in /s/-obstruent clusters, the pre-boundary vowel and the cluster second consonant were found to be longer when the word-boundary was located within the cluster. Moreover, word-spotting experiments demonstrated consistent initial (V#CCVC) as well as final (CV#CCV) misalignment effects between the CVC target and the intended word-boundary, in obstruent-liquid clusters only. The present research explored the relation between the effect of final misalignment and the activation of overlapping lexical competitors that begin at the second syllable onset. Two experiments were performed in which respectively CVC (t t) or CV (t ) initially-embedded words had to be spotted in aligned or misaligned CV(#)C(#)CV sequences. In one condition (t #tru vs. t t#ru), the CCV final portion (tru) of the carrier was a lexical competitor of the CVC target due to obstruent overlap. In another condition, the CCV (trø) and CV (rø) final portions of the carrier were nonwords (t #trø from "temps treublant" vs. t t#rø from "tante reublarde"). Spotting the CVC words showed an alignment effect only when a competitor was present at the following onset (t #tru). No effect was obtained in the absence of competitor (t #trø). Spotting the CV words showed the reverse pattern of results: the alignment effect was found when the CCV portion was a nonword (t t#rø) but not when it was a word (t t#ru). The re-assignment of the obstruent to a misaligned CVC target seems thus harder when the obstruent already supports another lexical hypothesis. Similarly, the rejection of the obstruent to overcome the misalignment of a CV target seems easier when the obstruent can form a word with the rest of the sequence. We conclude that the location of word-boundaries in ambiguous sequences is tied to both the nature of competing lexical hypotheses and to word-boundary acoustic-phonetic cues which modulate the activation level of lexical candidates.
Delattre, P. (1940). Le mot est-il une entité phonétique en français? Le Français Moderne, 8 (1), 47-56.
Dumay, N., Content, A., & Frauenfelder, U.H. (1999). Acoustic-phonetic cues to word boundary location: evidence from word-spotting. In Proceedings of the International Congress of Phonetic Sciences, (pp. 281-284), San Francisco, CA, USA.
Durand, M. (1936). Le genre grammatical en français parlé Paris et en région parisienne. Bibliothèque du Français Moderne, Paris.
Malmberg, B. (1971). Phonétique générale et romane: Etudes en allemand, anglais, espagnol et français. The Hague/Paris, Mouton.