The online resolution of filler-gap dependency in Korean relative clauses: An eye movement study

Abstract

The present study was designed to explore how Korean relatives─which should be interpreted based on the filler-gap dependency─are processed online by the parser (i.e., when the gap is identified and parsed in which manners.) and to determine whether a critical driving force for the gap searching and filling in Korean relatives is the filler itself or not (i.e., filler-driven parsing vs. gap-driven parsing). In order to examine how Korean parsers process Korean subject relatives (SR) and dative relatives (DR) online, the eye-tracking methodology was used. Following ‘the visual world paradigm,’ two kinds of stimuli (auditory and visual stimuli) were presented to subjects. Subjects’ eye movements and response times were recorded and analyzed. The result of response times (RTs) showed that SR was more accurately and rapidly processed than DR, suggesting that the subject gap in SR seems to be postulated more easily and fast than DR. The overall eye movement patterns observed in SR and DR were not significantly different. However, in some specific critical regions, significant differences of eye movements were found. Firstly, at the real gap position, there were no significant eye movements to the target picture, suggesting that the parser did not recognize the presence of the gap. Despite the richness of the morphological system in Korean, at the gap region, the lexical content and morphological information cannot offer clues making the parser recognize the existence of the gap. However, interestingly the proportion of looks to the target object increased at the relative verb region and the peak of the proportion of looks to the target object was found at the filler region (or head NP). Additionally at this region, the mean fixation duration on the target object was significantly longer than those in other argument regions. This indicates that using a morphological cue (i.e., a relative marker) at the relative verb region, the parser presumed the existence of the gap, and finally the filler-gap dependency was resolved at the filler region. This finding can be interpreted as partially supporting the filler-driven parsing account. Another important finding is that the different eye movement patterns were found at the accusative NP region (before the relative verb) and consequently this affected the processing of the relative clauses. This seems to occur due to the different combination of case markers or the different word order, suggesting that the case information can be immediately and incrementally utilized by the parser and affect the online processing of the Korean relative clauses.