|Titre||Mandarin relative clause processing or the joy of replication.|
|Année de publication||2016|
|Authors||Pozniak, Celine, and Barbara Hemforth|
|Titre de la conférence||29th Annual CUNY Conference on Human Sentence Processing|
|Date de publication||03/2016|
Mandarin relative clause processing or the joy of replication.
Céline Pozniak, Barbara Hemforth (LLF, CNRS, Univ. Paris-Diderot, Labex EFL)
Subject relative clauses (SRs) are considered as being easier to process than object relative clauses (ORs) cross-linguistically. However, Mandarin has been claimed to behave differently, since Hsiao & Gibson (2003) found an OR advantage compatible with linear distance effects as predicted by the Dependency Locality Theory (DLT) (Gibson, 2000). Chen et al. (2011), however, found that SRs are easier to process in Mandarin, in line with frequency/structure based predictions. To explain these diverging results, Vasishth et al. (2013) suggested in a meta-analysis that there may actually be no significant difference at all or just a slight SR advantage possibly because the two effects (frequency/structure and linear distance effects) cancel each other out. Still, this meta-analysis included experiments with known methodological problems (such as contexts allowing for non-RC analysis). Thus, we ran a series of replication of experiments with different groups of participants to test the robustness of the preferences.
Our study We designed an experiment to test frequency/structure based versus linear distance based theories presenting RCs in contexts with very high anticipation of restrictive RCs. Experiments We ran a Visual World Eye-Tracking experiment (5 items per condition) with 41 native Mandarin Chinese speakers for Exp1, 23 native Mandarin Chinese speakers for Exp2 (both groups living in Paris) and 34 native Mandarin Chinese speakers in China (Nanjing) (Exp3). Exp1 and Exp2 were identical except for the speed of presentation of the auditory material. Presentation was slowed down in Exp1 in order to clearly capture fixation patterns. We sped up the presentation in Exp2 to enhance cognitive load. Exp2 and Exp3 were identical. We tested reversible SRs and ORs (1-2). Participants listened to a sentence while viewing a pair of pictures with the same three characters each performing different actions. The task was to find the correct picture corresponding to the sentence. One of the pictures was only compatible with an SR interpretation, the other one only with an OR interpretation.
Results Linear mixed models showed no significant difference between SRs and ORs in Exp1 (Figure 1), probably because of the reduced speed of the presentation. Increasing the speed in Exp2 (Figure 2) resulted in a significant SR advantage in Mandarin (ps<.05) but not in Exp3 (Figure 3) where numeric but no significant difference could be established. Conclusion In a study with constraining context replicated three times, we find only a slight SR advantage when combining the results from the three experiments (Figure 4). We follow Vasishth et al. ‘s proposition that RCs processing involves a competition between frequency/structure and linear distance effects, which may cancel each other out. Our results do not show any evidence for an OR preference, disconfirming DLT. The lack of stability of preferences in Mandarin relative clause processing pertains even when using the same paradigm. This shows why it is so important in psycholinguistics to replicate experiments in order to test the robustness of the factors playing a role in sentence processing.
Back to Friday Poster Schedule