Grillo, N. 1 , Santi, A. 2, 3 , Grodzinsky, Y. 2 & Wagner, M. 2
1 Centro de Linguística da Universidade Nova de Lisboa
2 McGill University
3 Max Planck Institute for Human Cognitive and Brain Sciences
This paper investigates relative clause attachment comparing Self-Paced Reading (SPR) with Planned Production (PR), and provides evidence that they show comparable duration effects. The PR data suggests a new perspective on some previous SPR results.
In PR, participants read sentences silently until they are ready to produce them. We segmented the sentences into zones parallel to SPR-data and extracted duration and other acoustic measures. This method leads to richer data (various acoustic measures, distinction between lengthening and pause), and reduces some extraneous factors (ie, plausibility effects). We used PR to look at subject- and object modifying relative clauses (SRC and ORC), crossed with subject- vs. object extraction inside the RC. A mixed model regression showed a difference across extraction site and RC location for relative clause duration (ps<.002) that replicated previous results. Additionally, there was a significant effect at the matrix verb (p<.01): longer for the SRC than the ORC condition. Furthermore, the matrix clause excluding the RC was significantly longer in the SRC than in the ORC condition (p<.001), but no significant difference was found over the entire sentence. These results challenge a previous conclusion that ORC are more difficult than SRC. We propose that the ‘speed-up’ on the RC in the SRC structure may not be due to a lower complexity of SRC, as previously claimed, but rather a reflection of the parser trying to minimize the temporal distance in the dependency relation between subject and predicate. Thus, we claim that not all durational differences directly reflect complexity. The results reported here are based on durational measures, to mirror SPR results. In the presentation we will discuss differential effects of duration due to lengthening and pause duration, and other acoustic variables, exploring the richness of the PR data.