Abstract
Topic chain (TC) is a prevalent grammatical device that can be readily identified in Chinese discourse. Tsao (1979) regards it as a semantic domain under which a topic extends over more than one sentence with or without connectors. According to his model, a topic chain is led by a definite, sentence-initial chain topic, which controls all the subsequent coreferential deletions within the chain. Following Tsao’s TC framework, this study is primarily intended to investigate whether the use of TC is likely to display any difference between spoken and written narrative in terms of the domain and frequency of occurrences of topic-controlled deletions (TCD). The narrative data for my analysis were elicited through the picture-based technique on 20 native subjects producing 40 stories with 20 spoken and 20 written. Based on these data, through this study I hope to answer the following questions. First, is written TC longer than spoken TC? Second, in which modality, spoken or written, does TCD occur more frequently? Third, if a topic could undergo coreferential deletion but turns out not to, what alternative devices would instead appear? Which modality employs these devices more? And why?The quantitative investigation of TCD in the two modalities reveals that written narrative exhibits a higher percentage of occurrences of TCD and long chains than spoken narrative. Moreover, written-prior narrative may potentially raise the percentage of TCD occurrences for spoken narrative carried out later. On the other hand, the qualitative examination of anaphora indicates that in some cases where written narrative makes use of zero anaphora (coreferentially deleted topic), spoken narrative instead favors the use of pronominalization or NP repetition. All these findings tend to demonstrate a gap between spoken and written narrative with respect to the production of TCD, and I argue that such a gap is related to mental planning. In unplanned discourse, since speakers undergo processing pressure caused by speech spontaneity, they are often unable to employ devices such as long-chain zero anaphora that requires forethought and organizational preparation (Ochs, 1979; Lakoff, 1982). Furthermore, speakers’ preference of pronoun and NP suggests that due to memory limitations, they need, more often than not, to confirm the status of the current topic in discourse.