Constructing Filler-Gap Dependencies in Chinese Possessor Relative Clauses

Proceedings of PACLIC 19, the 19th Asia-Pacific Conference on Language, Information and Computation. Constructing Filler-Gap Dependencies in Chinese Possessor Relative Clauses Chien-Jer Charles Lin Department of Linguistics, The University of Arizona Tucson, AZ 85721-0028, USA clin@u.arizona.edu Sandiway Fong Department of Linguistics, The University of Arizona Tucson, AZ 85721-0028, USA sandiway@u.arizona.edu Thomas G. Bever Department of Linguistics, The University of Arizona Tucson, AZ 85721-0028, USA tgb@u.arizona.edu Abstract This article explores the construction of filler-gap dependencies in Chinese possessor relative clauses (PRCs), which are different from typical relative clauses (RCs) considered in the literature because Chinese PRCs contain no overt missing arguments (i.e. gaps). As Chinese RCs are prenominal, the gaps precede the head noun fillers. It has been suggested that when the gaps are close to the filler, the dependency is easier to construct; there is, thus, a processing advantage for object RCs over subject RCs (Hsiao & Gibson, 2003). The PRC data presented show that even in Chinese, a language with RCs that are head-final, it is possible to have a subject gap preference (over object) despite the fact that the subject is further away from the filler. Three experiments confirmed this subject preference with respect to naturalness and grammaticality ratings (Experiment 1), paraphrasing tasks (Experiment 2), and self-paced reading tasks (Experiment 3). The results support a theory of gap-searching which operates top down. Issues regarding locality and canonicity will also be discussed. 1. Introduction As human languages are replete with dependency relations within and across sentences, one crucial task of the language parser concerns the efficient recovery of such relations and their correct interpretation. This process involves various factors, including the required on-line processing load (often discussed in terms of working memory), the complexity of the processed materials, and human syntactic knowledge, etc.1 Most previous research focused on the first two factors. When a sentence consumes more processing resources, it is assumed to be more difficult and thus takes longer to understand. Similarly, when a sentence is more complicated or less usual, it requires longer processing time. However, the third factor, i.e. the structural properties of different syntactic positions in strategic on-line processing, has been less studied. Studies of focus positions and their salience for processing are both examples of this vein of inquiry (see for example Birch et al., 2000, and Frazier et al., 2005). In this paper, we present data that supports structural knowledge as a prominent factor in sentence processing. By structural or syntactic knowledge, we mean knowledge about the function of specific syntactic positions. This knowledge allows the human parser to recover structure in efficient fashions. For example, in a probe-goal model such as Chomsky (2001), this might mean specific and direct access to certain syntactic positions (e.g. a probe to spec-TP), followed by more general top-down search for matching goals. As the experimental data with possessor relative clauses (PRCs) will suggest, surface subject positions are the most likely candidates for direct gap-probing. In what follows, section 2 reviews findings on processing Chinese relative clauses …

Content truncated to 500 words