The effects of relative clause length on word order

Abstract

This paper mainly investigates how the length of relative clauses influences Cantonese speakers’ choice of word order. We manipulated both length (control/short/long) and target (subject/object) of relativisation of relative clauses modifying direct object. Following Yamashita and Chang (2001), participants’ task was to construct a sentence using sentence components and recall it. Hawkins (2004) suggests that long relative clauses increase the linear distance between the verb and its complements, making processing difficult. This predicts that speakers should produce word orders which can minimise the distance between heads. For control condition, we predict speakers to produce Zoeng1>Bei2>HNPS>Canonical constructions; for RC conditions we expect Zoeng1>HNPS>Bei2>Canonical constructions. After taking Hawkins’ Maximum On-line Processing principle into account, it predicts all conditions to have similar word order frequencies as RC conditions but with Bei2 being the most frequent. However, the accessibility account (Arnold et al., 2000) suggests that speakers should produce simple constituents first and long constituents later. This predicts that speakers should produce canonical construction for control condition and long RCs should increase HNPS constructions. We found a positive correlation between long subject RC and production of HNPS (control: 15%, subject long: 23%), and control condition significantly increases speakers’ preference of Canonical construction. However, we did not find any significant difference in the use of Bei2 and Zoeng1 constructions across conditions. The results follow predictions by the accessibility account, and partly support Hawkins’ theory, suggesting that both theories may influence Cantonese speaker’s word order choice.