French Speech Segmentation in Liaison Contexts by L1 and L2 Listeners

Abstract

In this study, we consider how native status and signal degradation influence French listeners' segmentation of an incoming speech stream containing liaison, a phonological process that misaligns word and syllable boundaries. In particular, we investigate how both first language (L1) and second language (L2) French listeners compensate for the syllable-word misalignment associated with liaison while segmenting French speech, and whether compensation-for-liaison strategies differ with decreasing signal-to-noise ratios. We consider the degree to which listeners rely on lexical knowledge, acoustic-phonetic cues, and distributional information to accomplish this compensation. Listeners completed a word identification task in which they heard adjective-noun sequences with or without liaison and were presented with the word or nonword alternatives for each noun that would result depending on whether the listener did or did not compensate for liaison. Results showed that both L1-French and L2-French listeners generally preferred lexically acceptable parses over those that resulted in a stranded nonword, and both groups gave significantly fewer lexically acceptable parses under harder listening conditions. However, the L2-French listeners demonstrated a pattern of boundary placement that indicated over-compensation for liaison, suggesting that they had successfully acquired, but not fully constrained, rules about liaison.

The Relationship between Production and Cognitive Processing of Liaison

“…Moreover, it should be noted that the research reviewed thus far has mostly focused on the temporal parameter: the analyses have concerned the duration of the target consonant and, in some cases, of the preceding vowel (very rarely on the VOT: Dejean De La Bâtie 1993; Wauquier-Gravelines 1996). The relative duration of consonants and preceding vowels has also been taken into account (Tremblay & manipulated so as to increase the difference between the two contexts (Shoemaker 2014;Gustafson & Bradlow 2016). Manipulations produced stimuli whose durations were still within the range of variation which was attested in the naturally recorded stimuli, and the studies showed that, when lexical ambiguity is maximal (as in les ailes 'the wings' vs. les zèles 'the zeals'), speakers may rely on such exaggerated acoustic differences to correctly assign the word to either category of vowel-initial or consonant-initial words.…”

Using allophonic variation in L2 word recognition: French listeners’ processing of English vowel nasalization

“…As for L2 listeners, some of this research has focused on how they deal with phonological variation, with mixed results. Advanced learners can range from native-or near-nativelike performance to failure to recover the intended word (Brand and Ernestus, 2018;Darcy, Peperkamp and Dupoux, 2007;Gustafson and Bradlow, 2016;Tremblay, 2011;Tremblay and Spinelli, 2014;Tuinman, Mitterer and Cutler, 2012). Not all phonological variation, though, is harmful for word recognition.…”