This study examined the processing of two types of Japanese causative cleft constructions (subject-gap vs. object-gap) by conducting an event-related brain potential experiment to clarify the processing mechanism of long-distance dependencies. The results demonstrated that the subject-gap constructions elicited larger P600 effects than the object-gap constructions. Based on these findings, we argue that the linear distance rather than the structural distance between the extracted argument (filler) and its original gap position is a crucial factor for determining processing costs of gap-filler dependency in Japanese causative cleft constructions. This argument indicates that (at least) some types of long-distance dependencies are sensitive to linear distance.