The present study tested whether the D-linked object moves from its thematic position over the subject or it originates where it appears in non-canonical sentences in Japanese. To this aim, we conducted acceptability judgment experiments that employed island effects as a diagnosis of movement and assessed whether the D-linking status of an extracted object of non-canonical OSV sentences escaped island effects. The results revealed that D-linking did not improve an acceptability of island violations, and therefore, a D-linked object of OSV does have a status of a moved constituent.
The present result contributes to an understanding of a relationship between syntactic representation and processing of filler-gap dependencies. According to recent event-related brain potential (ERP) studies, non-canonical sentences with a filler-gap dependency elicits a P600 effect when there is no felicitous context, but they do not reveal any effect when the filler is discourse-old information. The present result is inconsistent with the interpretation that the D-linked filler does not have a status of a moved constituent, thereby resulting in no filler-gap dependency formation in Japanese sentence comprehension. Instead, the present result is consistent with the view that the P600 effect is not a neural cost of the reconstruction but is elicited by other cognitive processes, such as the resolution of the unsatisfied presupposition encoded by scrambling.