CROSS-LINGUISTIC PROSODIC PERCEPTION: HOW CHINESE LEARNERS WEIGH SUPRASEGMENTAL CUES AND VOWEL QUALITY CUES ACROSS LANGUAGES

Abstract

This dissertation investigates how Chinese learners’ perception of word-level prosodic features is shaped by their first language (L1) and the specific target languages they are learning. The study examines how Mandarin Chinese speakers learning English as a second language (L2) and Russian or Japanese as a third language (L3) utilize suprasegmental and vowel quality cues when perceiving lexical stress in English and Russian and pitch accent in Japanese. Mandarin Chinese employs a tonal system where pitch encodes lexical meaning at the syllable level. This system lacks lexical stress and uses pitch for different purposes than the Japanese pitch accent system or English and Russian stress systems. Thus, Chinese speakers learning word-stress-based languages (English and Russian) and pitch accent systems (Japanese) need to establish new cue weighting procedures for prosodic prominence perception. The exploration of these procedures provides a unique opportunity to examine cross-linguistic influence and cue recalibration in multilingual prosodic acquisition.The languages were selected for their typological diversity in prosodic implementation: Russian uses lexical stress with minimal pitch cues and some other suprasegmental cues while heavily relying on a three-step vowel reduction system; Japanese employs a pitch accent system with no vowel quality variation; and English uses both suprasegmental cues and vowel quality in its stress system, serving as an intermediate case between the two extremes. This design allows for an examination of how the same learner population adapts perceptual strategies when confronted with different prosodic architectures and enables investigation of cue transfer patterns across L2 and L3 acquisition. Three experiments targeted different levels of prosodic processing. Experiment 1 employed AXB discrimination tasks to measure basic acoustic sensitivity to pitch, duration, and intensity contrasts. Experiment 2 used sequence recall tasks with nonwords containing varying prosodic and segmental patterns to assess sublexical processing abilities independent of lexical knowledge. Experiment 3 implemented lexical decision tasks with nonwords created by manipulating stress positions or pitch accent patterns and vowel quality in real words to examine how prosodic information is integrated during word recognition. A total of 127 participants were recruited across five groups: 20 English, 20 Russian, and 23 Japanese L1 speakers, and Chinese-L1 learners of Russian or Japanese as L3 (32 each). The results revealed several key findings. Across all groups, no significant differences in basic auditory discrimination ability were observed, confirming that group differences in subsequent tasks reflect linguistic experience rather than domain-general perceptual capacity. At the sublexical level, Chinese learners performed at or above L1 speaker levels in the sequence recall task across both suprasegmental and vowel quality conditions, with Russian learners outperforming L1 speakers in prosodic conditions. This is consistent with advantageous transfer from cumulative prosodic experience, though ceiling effects in English limit stronger conclusions. At the lexical level, both learner groups showed significantly lower nonword rejection accuracy than their respective L1 speaker groups, while the pattern differed by language: Russian learners demonstrated lower accuracy across salience conditions, while Japanese learners showed lower accuracy that was modulated by vowel change salience, pointing to language-specific rather than general prosodic processing challenges. Importantly, L3 proficiency modulated how vowel quality cues could be used at the lexical level in both languages: in Russian, the High-salience advantage emerged only among higher-proficiency learners, while in Japanese, the advantage was present at all proficiency levels but grew stronger as proficiency increased. An exploratory cross-linguistic prediction analysis at the lexical level revealed a cue-specific dissociation between the two learner groups. English Low-salience lexical decision performance predicted Russian L3 accuracy reflecting the high-level vowel quality demands of Russian vowel reduction, while English No-Change lexical decision performance predicted Japanese L3 accuracy, reflecting the suprasegmental demands of Japanese pitch accent in the absence of any vowel quality variation. This dissociation suggests that L3 prosodic accuracy is associated not with general English proficiency or overall prosodic ability, but with the particular type of vowel quality sensitivity a learner has developed in their L2, which is a sensitivity that maps selectively onto the specific cue demands of each L3 system. These findings advance theoretical understanding of how prosodic systems interact in multilingual minds. They extend the additive model of multilingual prosodic acquisition by demonstrating that transfer is selective and contrastive rather than uniformly cumulative, conditioned on the match between the perceptual skills cultivated in the L2 and the cue demands of the L3, and further modulated by L3 proficiency at the lexical level. The results also carry practical implications for language pedagogy, suggesting that L2 prosodic experience is associated with L3 learning outcomes in ways that are predictable from the phonetic properties of the cues involved and the learners’ current proficiency levels.