A quantitative study of information structure and right dislocation in Cantonese spoken discourse

Abstract

This study investigated information structure in spoken Cantonese and the use of right dislocation (RD) in genres with different degrees of shared knowledge and planning time pressure. Previous studies based on written data have suggested that Chinese word order is greatly determined by information structure. The pre-verbal referent in a sentence is reserved for background or topical information; while focus information is expressed in the post-verbal or right-most position in a sentence. The preferred argument structure (PAS) account has proposed a pattern of information flow in spoken discourse. New referents are free to appear pre-verbally in intransitive clauses, but rarely found in the subject of transitive clauses. However, it has also been suggested that RD is a focus marking device motivated by limited planning time in conversation. In RD structures, less important information is dislocated to the end of a sentence.

The data of the present study were extracted from a recently developed corpus of oral narratives in Cantonese Chinese based on approximately 8 hours of recordings. The corpus consisted of speech samples from 144 native Cantonese speakers evenly distributed in three age groups, two education levels and gender. Three genres representing different degrees of shared knowledge and planning time pressure were chosen for analysis: procedural description, story-telling and recount of personal event. A total of 2,227 simple active clauses were examined for information structure analysis, and 117 instances of right dislocations (RD) were identified.

The results of the analysis of information structure based on personal recounts conformed to the PAS account. The number of new referents was statistically significantly higher in intransitive than transitive clauses. Besides, higher information pressure has resulted in a higher frequency of pre-verbal new referents in intransitive clauses. Investigation of RDs revealed that 1) a lower rate of occurrence of RD was found in the monologue data of this study than previous studies based on conversations; 2) the highest proportion of RD occurred in personal event recounting, which is assumed to have the least shared knowledge and the highest information pressure; 3) the highest proportion of RD appeared near the end of a narrative and tended to co-occur with disfluency; 4) RD was often used for the expression of emotions. The results show that information distribution in spoken discourse and the use of RD were influenced by genre type, information pressure and speakers’ assumption of shared knowledge.