Perceptual segmentation and apparent tilt: psychophysical and computational analyses of neural grouping in tilt illusion

I propose a network model for the perception of the global orientation of a bar including the tilt illusion, the misjudgment of the bar orientation in the presence of another intersecting bar. It is most likely that the perception of global orientation, including tilt illusions, results from the grouping of V1 cell responses. I investigated what neural grouping is involved in the perception of global orientation. Psychophysical experiments were carried out in order to investigate phenomenological mechanisms for the neural grouping, and a level of cortical processing involved in the grouping. The specific aim of the experiments is to examine whether the amount of apparent tilt depends on the perceptual segmentation of the bars. The results show that apparent tilt is independent of that which bar the junction belongs to. This suggests that the visual system does not take into account the orientation information at the junction region for the determination of global orientation. In order to investigate the cortical mechanisms underlying the grouping, I developed a neural oscillator network based on colinear synchronization, and simulated the model with various bar junctions. The simulation results show that the intersection of the bars is not grouped together with either bar, which agrees with the phenomenological grouping observed in the psychophysical experiments. The model reproduces, with good quantitative agreement, the apparent tilt measured in comparable psychophysical experiments reported elsewhere.