Elsevier

Neurocomputing

Volumes 32–33, June 2000, Pages 979-986
Neurocomputing

Perceptual segmentation and neural grouping in tilt illusion

https://doi.org/10.1016/S0925-2312(00)00269-1Get rights and content

Abstract

We investigate which neural grouping is relevant to the perception of the global orientations of bars. We examine psychophysically whether perceptual segmentation affects the apparent tilt in X-shaped junctions constructed by solid or illusory bars. The results show that perceptual segmentation is not relevant to the global orientation, except for illusory stimuli. Early level processing, probably in the primary visual cortex (V1), seems to be involved in the determination of global orientation. We propose a network model consisting of V1 neurons and a grouping mechanism which synchronizes the cellular activities based on colinear coupling. The computed bar orientations agree quantitatively with the apparent tilt measured in comparable psychophysical experiments reported elsewhere.

Introduction

Humans are incapable of determining correctly the orientation of a bar in the presence of another intersecting bar, as is observed in the tilt illusions shown in Fig. 1. Tilt illusion seems to result from the inherent properties of the cortical mechanisms relevant to the perception of the global orientation of bars. Investigation of the mechanisms underlying tilt illusion is expected to lead to an understanding of the fundamental processes of the perception of global orientation.

We have previously investigated the retinotopic distribution of the responses of the cells in the primary visual cortex (V1) to bar junctions by simulating a large-scale V1 network model [5]. The simulation results showed that cells tuned exactly to the actual bar orientation did not respond strongly around the junction. Cells tuned to a few to 30° away from the bar orientation responded most strongly, depending on the retinotopic position of the cell and the geometric configuration of the junction. It is most likely that the grouping of these V1 cells is responsible for the misjudgement of bar orientation.

Section snippets

Perceptual segmentation

We investigate psychophysically the phenomenological mechanisms for the neural grouping in the perception of bar orientation, and the level of cortical processing involved in the grouping. To determine the global orientation of a bar in the presence of another intersecting bar, we need to determine the neural grouping, i.e. which neurons contribute to the perception of one bar, and which to the perception of the other. First, we examine whether the amount of apparent tilt depends on the

Neural grouping

The results of psychophysical experiments suggest that early level processing is involved in the neural grouping for the determination of bar orientation. We investigate what neural mechanisms underlie the neural grouping. It has been reported that colinear connections in the superficial layers of V1 are critical in contour perception [1], [8]. We propose that colinear coupling provides the basis for neural grouping in the determination of global orientation. In order to investigate the role of

Conclusions

The psychophysical study shows that the amount of apparent tilt is independent of the perceptual segmentation for the stimuli consisting of solid bars, while the stimuli consisting of subjective contours show dependence on the perceptual segmentation. These results indicate that the intersection of bars is not taken into account for the determination of apparent tilt, and that early level processing, probably in V1, is involved in the grouping process. We propose a network model consisting of

Ko Sakai received his BEng and MEng from the University of Electro-communications, Tokyo, and a Ph.D. in Bioengineering from the University of Pennsylvania. He is currently a researcher at RIKEN Brain Science Institute, working on modeling visual perception.

References (8)

There are more references available in the full text version of this article.

Cited by (1)

Ko Sakai received his BEng and MEng from the University of Electro-communications, Tokyo, and a Ph.D. in Bioengineering from the University of Pennsylvania. He is currently a researcher at RIKEN Brain Science Institute, working on modeling visual perception.

View full text