Soutenance de thèse : Yiming Liang

Jeudi 14 Décembre 2023, 09:00
Yiming Liang (LLF)

LLF – Bât. ODG – 5e étage – Salle 531

Yiming Liang (LLF)
Quantitative Syntax, Formal Syntax and Information Theory: bridging gaps by studying French variation

This PhD thesis investigates morpho-syntactic variation in Spoken French through a dual lens of formal syntax and Information Theory. Employing quantitative methods, the study addresses two fundamental research questions: 1) To what extent can variationist studies shed light on formal syntactic models; 2) How can Information Theory account for speakers' preferences in variant choices?

The first part of the thesis bridges the gap between linguistic variation and formal syntax by examining three phenomena of variation: Future Temporal Reference (je vais manger vs. je mangerai), subject doubling (Marie_i elle_i mange vs. Marie mange), and object doubling (le film_i je l_i'aime bien vs. j'aime bien le film). My corpus studies reveal that all three are strongly influenced by grammatical factors. The second part of the thesis endeavors to incorporate Information Theory into the study of linguistic variation. Specifically, Future Temporal Reference and subject doubling are examined as instances of syntactic redundancy, where surprisal is predicted to play a role, following the Uniform Information Density hypothesis (Jaeger 2010). A GPT-2 model is used to provide surprisal estimates for both phenomena. My experiments do not reveal an effect of surprisal on future temporal reference beyond what is already explained by grammatical factors. In contrast, subject doubling shows a robust effect of nominal subject surprisal, in addition to grammatical and other cognitive factors, providing more empirical support to the Uniform Information Density hypothesis. In conclusion, this work demonstrates that variationist studies provide crucial insights into enhancing our understanding and refinement of the syntactic framework, and highlights the collaborative potential of formal syntax and Information Theory in understanding the complexities of morpho-syntactic variation in Spoken French.


  • Heather Burnett, Directrice de recherche (CNRS), Université Paris Cité, directrice de thèse
  • Pascal Amsili, Professeur, Université Sorbonne Nouvelle, co-directeur de thèse
  • Julie Auger, Professeure, Université de Montréal, rapporteure
  • Benedikt Szmrecsanyi, Professeur, KU Leuven, rapporteur
  • Caterina Donati, Professeure, Université Paris Cité, examinatrice
  • Philippe Muller, MCF-HDR, Université Toulouse III - Paul Sabatier, examinateur
  • Luigi Rizzi, Professeur, Collège de France, examinateur