DOI: 10.47777/cankujhss.1902350 ISSN: 1309-6761

Ghosts of L1 Evidentiality: Constructional Transfer in L2 English Hedging

Fatih Ünal Bozdağ
Show PDF Cite
This study tests the ghost construction hypothesis (that obligatory L1 evidential categories persist as functional patterns in L2 hedging) by analyzing 1,011 English argumentative essays (752,967 words) by Turkish, Bulgarian, and Spanish learners (ICLE) and native English speakers (LOCNESS). Hedging devices were detected through a hybrid rule-based and LLM-assisted pipeline mapped onto Aikhenvald’s (2004) typology, and cross-group differences were assessed at the categorical, constructional, and propositional layers within a hierarchical Bayesian framework with L1 and text as random effects. At the categorical layer, the four groups produced similar evidential distributions, and eight directional predictions derived from the L1 typology were either null or directionally reversed. At the constructional layer, however, reportative subtype distributions diverged sharply. Turkish and Bulgarian writers allocated approximately 4% of reportative hedges to named attribution, compared to approximately 22% for Spanish and native writers, while preferring different source-nonspecific alternatives that paralleled their respective L1 evidential structures. At the propositional layer, a consistent L2–native distancing gap survived covariate adjustment for lexical diversity, syntactic complexity, and L2 exposure, and proved robust under an alternative distributional model. The results suggest that L1 evidentiality influences L2 hedging at the constructional rather than the categorical level.

More from our Archive