This study investigated whether higher-level object representations exert top-down effect in the grouping of visual elements. Stimuli with two uniformly connected regions were used. In one condition, they were meaningless while in another, they were meaningful Japanese kanji. In two experiments, participants judged whether two targets embedded within the stimulus figures were the same or different. The critical consideration, however, was whether the targets shared the same or different regions. Results revealed that when the two-region figure was meaningless, there was a significant advantage in responding to targets in the same region compared to those in different regions. However, when they were meaningful, this same region advantage vanished. These suggest that, in the meaningless figure, uniform connectedness influenced the parsing of the two-region figures into two distinct objects whereas, in the meaningful kanji, meaning representations influenced the grouping of the two-regions into a single object.