Abstract： One grouping principle of visual perception is to integrate spatially separated and partially occluded local parts into whole object representations. For example, in the processing of Kanizsa-type illusory contours (ICs), a subjective shape contour is perceived despite the inducers being isolated. The shape discrimination of the Kanizsa illusory contours depends on feedback signals in the brain. Thus, investigations of the adaptation to Kanizsa illusory contours can help us understand the temporal properties of top-down signals in the visual system. In the present study, four experiments were conducted to explore the adaptation to Kanizsa illusory contours. Participants adapted to +5° and −5° illusory contours in experiments 1a and 1b. In experiments 2a and 2b, in order to eliminate the possibility that the adaptation effect could be explained by the lines on the inducers, the inducer lines (experiment 2a) or the whole shape (experiment 2b) was used as the adapting images. For all of the experiments, illusory contours of varied degrees were used as test images. Subjects were instructed to respond as accurately and quickly as possible to the test images to judge whether they were fat or thin. When adapting to fat illusory contours, the participants tended to perceive the following tests as thin; on the contrary, when adapting to thin contours, they tended to perceive the test images as fat. These results showed significant adaption effect of Kanizsa contours. Further control experiments (2a and 2b) indicated that the adaptation effect was mainly caused by the illusory contour itself, rather than by the lines on the Pac-Man. These results revealed that adaptation effect existed in voluntary construction processes, indicating that the strength of feedback signals from higher-level visual cortex could become weak over time.