A probability distribution of dependencies in interlanguage

Abstract The diversity of syntactic units in second language has attracted much scholarly attention. Most existing studies on syntactic diversity have focused on only a small number of syntactic structures, and it is difficult to find studies that consider the full range of syntactic dependencies present in the dataset. Based on a syntactic annotated interlanguage corpus that we constructed, this paper is a quantitative study of dependencies in English-speaking Chinese learners’ interlanguage across proficiency levels. We fit the frequency distributions of dependency type, word class (both as dependent and governor), verb as a governor, and noun as a dependent with a modified right-truncated Zipf-Alekseev distribution and Zipf’s law. Our findings show that: (1) from the mathematical model, interlanguage followed distributional regularities like natural languages in terms of the syntactic structure distribution; (2) most of the determination coefficients’ R2 were high, indicating that the investigated distributions in interlanguage fit the distributional law finding in natural languages. This also demonstrated that both interlanguages and natural languages consistently conform to the law of linguistic diversity and uniformity; (3) the dependency relation distribution parameters a and b manifest the developmental trend of L2 learners’ proficiency levels, demonstrating that the parameters had universal applicability in reflecting interlanguage proficiency.