Comparing Large‐Grained and Fine‐Grained Syntactic Complexity in L2 Speaking

Syntactic complexity is a multidimensional construct that can be measured using indices at different levels of granularity. Previous studies have compared large‐grained and fine‐grained indices mainly in relation to the prediction of L2 writing quality, whereas direct comparisons of their ability to predict L2 speaking performance remain limited. This study compared 14 large‐grained and 24 fine‐grained syntactic complexity indices using monologic performances from 120 Chinese EFL learners in the TEM‐4 Oral Test. These indices were examined for differences across four speaking achievement groups and were then entered separately into Elastic Net models with 10‐fold cross‐validation to predict speaking scores. Results showed that 13 of the 14 large‐grained indices and 8 of the 24 fine‐grained indices differed significantly across achievement groups. The large‐grained model showed stronger predictive performance, with a CV‐ R 2 of 0.447 and an RMSE of 8.37, compared with 0.180 and 10.20, respectively, for the fine‐grained model. These results indicate that large‐grained indices can better distinguish L2 speaking achievement groups and predict speaking scores than fine‐grained indices. The findings have implications for the conceptualization and measurement of syntactic complexity in L2 speaking research and assessment.