L2 speech learning of European Portuguese /l/ and /ɾ/ by L1-Mandarin learners: Experimental evidence and theoretical modelling

This doctoral dissertation explores what constrains L2 phonological acquisition of European Portuguese /l/ and /ɾ/ across prosodic contexts and how different speech modalities interact during this process. The poor distinction between /l/ and /ɾ/ has been considered one of the most perceptible characteristics in Chinese-accented Portuguese. Recent empirical research has revealed that this notorious L2 speech learning difficulty goes beyond the confusion of the two L2 categories—different modalities, representational levels, and prosody all seem to shape this learning process (Zhou 2017). This dissertation aims to deepen our current understanding of how L2 phonological categories /l/ and /ɾ/ are created and developed by L1-Mandarin learners across syllable positions (onset and coda) and of how different modalities (speech perception, production, and visual) interact during their acquisition. To achieve these goals, both experimental tasks and theoretical modelling are employed. The first study of this dissertation examines the role of cross-linguistic influence and orthography on L2 category creation. Following the Full Transfer Hypothesis adopted to L2 speech learning (Escudero & Boersma 2004), we assume that learners initially rely on their L1 phonology to construct representations for the L2 sounds. A delayed-imitation task is administrated to 19 Mandarin native speakers without any knowledge of Portuguese. Moreover, whether orthography conditions L2 category creation was tested by manipulating the input types given in the experiment (auditory input alone vs. auditory + written input). Results show that the naïve imitators’ responses matched with those of L1-Mandarin learners, suggesting that L2 category creation is subject to cross-linguistic influence. Additionally, the Mandarin [ɻ] (a repair strategy for /ɾ/) occurred almost exclusively when the written form was given, providing evidence for the crosslinguistic interaction between phonological categorization and orthography during the construction of L2 phonological categories. In the second study, we first investigate the relationship between L2 speech perception and production by examining whether the deviant L2 productions for the target /l/ and /ɾ/ stem from misperception and whether the acquisition order in L2 speech perception mirrors that in production. Secondly, we assess whether these L2 phonological categories become more target-like with increasing L2 experience. Two perceptual experiments (a forced-identification and an AXB discrimination task) are conducted with 61 L1-Mandarin learners. Results indicate that L2 speech perception and production are related: (i) the acquisition order was consistent in the two speech modalities; (ii) most deviant forms observed in L2 production indeed have a perceptual motivation ([w] for the velarised lateral; [l] and [ɾə] for the tap). However, such correlation does not always hold true, evidenced by the fact that a repair strategy can be production-specific (deletion of syllable-final tap) and L2 production can precede perception: The distinction between /l/ and /ɾ/ is somehow maintained in production (/ɾ/ → [l], */l/ → [ɾ]), but not in perception (/ɾ/ → [l], /l/ → [ɾ]). To assess the plasticity of the L2 categories /l/ and /ɾ/, two groups of L1-Mandarin learners who differed substantially with respect to L2 experience (quantified as years of received formal instruction and length of immersion) were recruited. No LANGUAGE ACQUISITION 2022, VOL. 29, NO. 1, 105–106 https://doi.org/10.1080/10489223.2021.1952591