Sociolinguistic Variation in Multilingual Urban Speech Communities: A Corpus Study

Thesis Abstract

This study investigates sociolinguistic variation in multilingual urban speech communities through a corpus-based approach, addressing the gap between micro-level speaker alternations and macro-level urban language dynamics. The central aim is to quantify how language contact, social identities, and urban affordances shape variable linguistic features across multilingual speakers in metropolitan settings. Specific objectives are (1) to identify patterns of lexical, phonological, and syntactic variation associated with language contact zones (e.g., streets, marketplaces, transit hubs); (2) to examine the relationship between speaker social variables (age, gender, education, ethnicity) and variation in selected features; (3) to test the applicability of variationist models and sociolinguistic theories to corpus-derived data; (4) to map variation across districts with differing levels of linguistic diversity and language prestige; and (5) to assess the reliability of automated annotation pipelines for multimodal urban speech data. A mixed-methods design is employed, combining corpus linguistics with traditional variationist analysis. The population comprises adult multilingual residents within a major global city known for high linguistic diversity. A stratified random sample of 200 speakers (balanced across age cohorts 18–30, 31–45, 46–65; and gender) produces approximately 180 hours of conversational speech collected via portable audio recorders and smartphone apps over a six-month period, complemented by 40 hours of ethnographic observation in purposively selected locales representing high and low linguistic density. Data collection instruments include a multimodal urban speech corpus, sociolinguistic questionnaires capturing language use, language proficiency, and community affiliation, and ethnographic field notes. Audio data will be transcribed with time-aligned orthography and phonetic annotation using the ELAN framework, while automatic speech recognition outputs will be manually corrected for reliability. Statistical analysis will integrate regression-based models and multivariate techniques. Variation in selected features—phonetic realizations of high-frequency phonemes, code-switching indices, lexical collocations, and syntactic alternations—will be analyzed with hierarchical linear modeling to account for speaker-level and locale-level effects. ANOVA and post hoc tests will examine district-based differences, and logistic regression will assess predictors of code-switching behavior. Corpus-based methods, including n-gram analysis, collocation networks, and topic modeling (LDA) on accompanying discourse, will contextualize quantitative results within discourse domains. A conceptual framework drawing on classic sociolinguistic theory (Labovian variationism) and contemporary models of multilingual repertoires (Gumperz’s conversational codeswitching, Active Emergence of Varieties, and Network Sociolinguistics) will guide interpretation. The study also engages sociolinguistic constructs of language prestige, ethnicity indices, and social identity performance as articulated in the theoretical sections. Expected findings anticipate systematic variation linked to contact intensity and social positioning. For instance, higher rates of code-switching and lexical borrowing are expected in commercially dense districts with greater language prestige contrasts, while morpho-syntactic alternations may correlate with age and education, reflecting ongoing acquisition and identity negotiation. Phonological variables are anticipated to show gradient shifts driven by inter-speaker interaction frequency and acoustic environment. The study further expects that automated annotation will achieve high reliability when supplemented with targeted human correction, validating scalable methods for large urban corpora. Contributions to knowledge include an empirically grounded model of urban multilingual variation that integrates corpus-derived evidence with sociolinguistic theory, offering fine-grained mappings of linguistic variation across spatial, social, and discursive dimensions. Methodologically, the research demonstrates the feasibility of combining standard variationist techniques with corpus linguistics and spatial stratification to study multilingual urban speech, contributing robust data and replicable procedures for future cross-city comparisons. The main conclusion posits that sociolinguistic variation in multilingual urban speech is best understood as the outcome of dynamic interactions between language contact networks, participant identities, and urban affordances, rather than solely as static codeswitching phenomena. Recommendations include adopting scalable corpus-analytic pipelines for urban linguistic surveillance, informing language planning and educational strategies in multilingual cities, and extending the framework to include social media and digital communication channels to capture trans-urban linguistic evolution.