Processing long-distance wh-dependency in the Kyengsang dialect of Korean: an ERP study

Kyengsang Dialect Korean (KDK) is a wh-in-situ language that morphologically distinguishes content (or wh-) and polar questions via sentence-final question particles (QPs). This study investigates how KDK comprehenders build dependencies between wh-indeterminates and QPs, and how they compute question–answer concord. Two experiments−such as acceptability judgments and event-related potentials (ERPs) − tested sensitivity to feature matching and locality constraints. In the acceptability task, speakers showed robust interactions between QP type and wh-licensing configuration, with additional degradation in island environments. These patterns indicate that dependency resolution is guided by morpho-syntactic agreement while remaining sensitive to structural constraints, even in the absence of overt wh-movement. ERP recordings revealed three dissociable signatures that map onto successive stages of dependency formation. First, a right anterior negativity (RAN) emerged for feature mismatch between a wh-indeterminate’s +WH feature and a polar QP, consistent with the rapid detection of illicit licensing. Second, a left anterior negativity (LAN) indexed increased working-memory costs when the +WH feature had to be maintained or retrieved across an island boundary. Third, an extended anterior negativity (EXAN) reflected ongoing feature-match monitoring under question–answer discord. Together, the behavioral and neural results suggest that KDK speakers actively maintain and retrieve the +WH feature of an in-situ wh-indeterminate to establish syntactically licensed dependencies with matrix QPs, including configurations that challenge locality. Comparisons with Japanese and with wh-fronting languages (e.g., English/German) indicate that KDK engages similar incremental, feature-driven mechanisms for dependency resolution. The findings support a feature-based model in which the parser predicts and matches +WH with the appropriate QP at the earliest opportunity, providing neurocognitive evidence that wh-in-situ processing parallels filler–gap computation in wh-movement languages.