Grammatically-guided resolution of filler-gap dependencies: An investigation of Chinese multiple dependencies.

摘要

Previous studies have provided evidence that the parser avoids positing gaps in grammatically unlicensed positions such as islands, suggesting that the grammar constrains the construction of filler-gap dependencies (e.g., Stowe, 1986). However, such evidence does not provide evidence against an alternative interpretation which posits that that islands are simply processing bottlenecks and thus the parser is unable to establish a filler-gap dependency within these domains (e.g., Kluender & Kutas, 1993; Hofmeister & Sag, 2010). To this end, Wagers and Phillips (2009) provide evidence that grammatical knowledge is not only used to avoid establishing wh-dependencies when prohibited, but is also in motivating the active search for gap positions when required by the grammar. Building on Wagers and Phillips (2009), the current study examines whether gap-filling is predictive and grammatically-guided in a new domain: Chinese topicalization. In a self-paced reading study with Chinese native speakers (n=40), we compared the processing of sentences with coordinated verb phrases to that of sentences with post-verbal adjunct phrases. Coordinated VPs in English and Chinese require across the board extraction (Who does John know _ and love _/*Mary?) while extraction from the second position of post-verbal adjunct phrases is optional but not required (What did John eat _ after washing _/the dishes?). It is found that the parser actively searches for a second gap after positing a gap in the first coordinated VP while no gap-filling is observed in the post-verbal adjunct phrases. These results demonstrate that grammatical knowledge actively guides the incremental resolution of filler-gap dependencies.