Meaning and grammar seminar
Speaker: Hans Wilke
Title: The Role of Information Structure in Relative Clause Processing
Abstract: A long-standing question in linguistics centres on how comprehenders establish syntactic structure and what factors influence this process. Relative Clauses (RCs) provide a useful testing ground as their structural complexity challenges comprehenders’ cognitive capacities for holding syntactic components in memory to build long-distance dependencies. The observed complexity of certain RC types is often attributed to their syntax (Gibson et al. 2013). However, other evidence suggests the complexity of RC constructions is modulated by information structure (IS), which can facilitate processing of syntactically complex RCs (Gibson et al., 2005; Roland et al. 2012). Extending the range of information-structural factors and avoiding the potential confound found in Gibson et al. (2005), whose IS manipulation coincides with a restrictive vs. non-restrictive RC-distinction, we test how syntactic and information-structural factors influence the processing of RC-constructions.
In a web-based self-paced reading experiment, monolingual English participants (N=200) read experimental items containing a few introductory sentences, a target sentence, and a wrap-up sentence. Target sentences had the format of (1a-b); all RCs were non-restrictive. The given/new status of the RC and main clause were manipulated via the content of the introductory sentences, crossing position of givenness (given-new vs. new-given) and RC-modification (subject-modifying vs. object-modifying).
- a. Eric, [who was practising the piano]RC, [was teasing my sister]MAIN.subject-modifying RC
- b. Eric [was teasing my sister]MAIN, [who was practising the piano]RC.object-modifying RC
Syntax-driven predictions reflect two assumptions (Gibson et al. 2013): perspective shifts are harder to process (object-modifying RCs) and longer distance dependencies require more storage (subject-modifying RCs). We also consider two IS-related hypotheses: the Information Flow Hypothesis (Gibson et al., 2005), which postulates that new information is more easily processed later in a sentence, and an IS-clause type mapping from corpus-based evidence suggesting new information appears more in main clauses, and old information in subordinate clauses (e.g., Diessel, 2001). The first IS-related hypothesis predicts object-modifying RC-constructions with given content in the main clause, and subject-modifying RC-constructions with given content in the RC to be processed faster. The second IS-related hypothesis predicts that both subject- and object-modifying RC-constructions with given content in the RC and new content in the main clause would be processed faster than their counterparts.
Results show that RCs containing given information were read faster than those containing new information (p=<.001), and subject-modifying RCs were read faster than object modifying RCs (p=<.001). In absence of an interaction effect, this would support our second IS-related hypothesis. However, an analysis of main clause reading times contradicts this, as these are also read significantly faster when containing given information (p=<.001). These results cannot be generalised towards either IS-related hypothesis, as given information is always processed faster independent of clause type or position. While we found evidence that subject-modifying RCs are processed faster, we found that sentence-early main clauses were also processed faster (p=<.001); mitigating the slower reading times of the object-modifying RCs they precede. This could imply that RC-complexity/ease is balanced out by the main clause at the full sentence level. Further research is needed to confirm this.
If you'd like more information on dates and venues this semester, or if you'd like to present, please email Alex Lorson, or subscribe to our mailing list.
Meaning and grammar seminar
Room 1.16, Lister Learning and Teaching Centre, 5 Roxburgh Place, Edinburgh, EH8 9SU