Item

Abstract

In this paper, I present results from experimental data on switch-reference in the Oceanic language Whitesands (ISO: TNP), providing an evidence-based description of the system. Whitesands, like its sister languages of the southern Vanuatu sub-group, uses a switch-reference system across clauses primarily marking coreference (Lynch 1983, Lynch 2001:177, Crowley 2002:201, Hammond submitted). I present a description of the system in its most canonical form for Whitesands. I then tackle the properties of antecedents for the same subject clauses using two different experiments. I address the following questions: Does the Whitesands' system support claims that switch-reference systems are potentially sensitive to discourse topicality (Reesink 1983) or other extra-syntactic devices (Roberts 1988)? Furthermore, what is the preferred antecedent for Whitesands speakers in extended discourse? The m- ‘ER’ inflection is typically used when two adjacent predicates share the same subject. The m- replaces the person agreement and tense operators in the second clause. In (1) the m- indicates coreference of the subject of the predicate with the subject of the preceding predicate. The clause m-l-eru is underspecified for person and tense when taken out of context and is thus ungrammatical as in (1’). (1) k-l-eni ama [m-l-eru] 3.NPST-TRIAL-say just ER-TRIAL-see They (TRIAL) just talked and saw. (1’) * m-l-eru ER-TRIAL-see However, in the Whitesands corpus it is clear that a simple “antecedent equals subject” rule does not always hold for same subject clauses and that a notion of discourse topic might be a potential antecedent alternative. For example, there are topic chains that use the Echo Referent for continual reference whilst skipping immediately adjacent non-topical subjects. Further, there are forms where the prefix m- combines previously distinct arguments into a single argument slot. The first experiment presented here is a production experiment where speakers had to extend natural discourse using video and audio stimuli. The items were of four types and these were controlled for alignment of grammatical relations and topicality. The results suggest that a highly topical entity can indeed trigger a coreference pattern. The second experiment presented is a forced choice comprehension task. I investigate the relationship between same subject clauses (coreference) and antecedent types. I test what alternative constructions are considered grammatical. I conclude that the switch-reference clauses are much more likely to be aligned with a topical referent that is also the subject of the preceding clause than in other configurations tested.