Cross-linguistic Prosody: A Comparative Analysis of Intonational Phonology Across Languages

Thesis Abstract

This study investigates cross-linguistic prosody through a comparative analysis of intonational phonology across typologically diverse languages, addressing a gap in understanding how language-specific prosodic systems interact with universal phonological constraints to shape discourse meaning and speaker intent. The problem centralizes inconsistent cross-language mappings between intonational categories and pragmatic functions, which hampers generalizable theories of suprasegmental structure and limits the applicability of prosodic models in multilingual settings. The aim is to delineate both universal and language-specific prosodic patterns and to test the applicability of intonational models across languages with contrasting prosodic typologies (tonal, syllable-timed, and stress-timed systems). Specific objectives are (1) to characterize the inventory of high, mid, and low pitch accents, boundary tones, and referential vs. emphatic intonation across Mandarin Chinese, English, Turkish, and Yoruba; (2) to examine cross-language alignment of prosodic boundaries with syntactic and discourse-wide units; (3) to assess the predictive validity of the Autosegmental-Metrical (AM) Theory and the ToBI framework in capturing language-specific prosodic events; (4) to identify how prosodic cues modulate perception of speaker stance and discourse structure in multilingual listeners; and (5) to propose an integrated cross-linguistic prosody model that accommodates both universal constraints and language-specific realizations. The methodology adopts a mixed-methods design combining corpus-based acoustic analysis with perceptual experimentation. The population comprises native speakers of Mandarin Chinese (n=40), English (n=40), Turkish (n=40), and Yoruba (n=40), balanced for gender and age (22–40 years). A corpus of 1,200 utterances per language is compiled from contemporary narrative, argumentative, and conversational genres, annotated for syntactic structure and discourse markers. Data collection instruments include high-fidelity audio recordings, a standardized elicitation task to produce target prosodic structures (focus, contrastive emphasis, topic change), and perception tests employing forced-choice and Likert-scale ratings to gauge listener judgments of stance, emphasis, and discourse segmentation. Acoustic analysis utilizes multi-dimensional prosodic features f0 trajectory (mean, slope, contour types), energy, duration, and spectral tilt, extracted via Praat and followed by functional data analysis. The theoretical framework integrates Autosegmental-Metrical Theory (AM) and Versioned ToBI annotations, complemented by the Perceptual Salience Model to connect acoustic patterns with perceived discourse function. Statistical analyses include mixed-effects regression models to assess the influence of language and discourse type on prosodic realizations, multivariate ANOVA to compare cross-language prosodic inventories, and spline-based temporal analyses to investigate boundary alignment with syntactic units. Perceptual data are analyzed using mixed-model ANOVA and signal-detection metrics to evaluate listeners’ discrimination accuracy across language-specific prosodic cues. Additionally, cluster analysis will identify recurrent prosodic archetypes and their association with discourse functions across languages. Expected findings indicate both universal patterns—such as the consistent association of rising boundary tones with continuation and focus marking in multiple languages—and language-specific deviations, such as Mandarin tonal constraints shaping boundary realizations and Yoruba discourse-related pitch movements modulated by lexical tone interactions. It is anticipated that AM framework explanations will capture cross-language commonalities, while language-specific ToBI adaptations will be necessary to account for typological variation. The study expects that listeners rely on a combination of f0 dynamics and durational cues to infer stance and discourse unit boundaries, with cross-language transfer effects evident in bilingual perception tasks. The study contributes to knowledge by advancing a nuanced cross-linguistic prosody model that reconciles universal prosodic functions with language-specific realizations, validating and extending ToBI-based annotation schemes, and providing empirical data to refine theoretical accounts of intonational phonology within multilingual communication contexts. The main conclusion posits that while universal prosodic cues underpin global discourse signaling, language-specific phonological constraints critically shape the realization and perceptual interpretation of intonation. Recommendations include refining cross-language annotation frameworks to incorporate language-specific prosodic phenomena, developing training materials for multilingual speech technologies that reflect typological diversity, and extending the corpus to include additional tonal and non-tonal languages to broaden generalizability.