Modelling the functions of tail–head linkage: a bayesian case study of Muyu (Papuan)

Abstract This study introduces a Bayesian hierarchical model for investigating the discourse functions of tail–head linkage (THL). THL is a cross-linguistically widespread discourse pattern that connects units of text through verbatim repetition or anaphoric reference. Previous research has largely concentrated on the structural properties of THL, while its proposed functions – such as cohesion and coherence, processing facilitation, and discourse structuring – have typically been inferred from qualitative analyses of individual examples. To address this gap, the present study applies a quantitative statistical model to a corpus of ten narrative texts from five speakers of the Papuan language Muyu. The results indicate that THL in Muyu is not used for referential coherence but rather contributes to maintaining coherence across narrative scenes set in different locations. No evidence was found for systematic stylistic variation among speakers. Beyond the case study, the paper demonstrates the methodological potential of Bayesian hierarchical modelling for discourse analysis. The approach allows for flexible inclusion of additional predictors, hierarchical grouping factors, and cross-linguistic data. It thus provides a framework for quantitatively exploring how discourse strategies such as THL contribute to textual organisation and coherence across languages.