Shingo Tanaka and Ichiro Fujita
Neural Computation of Object Size using Binocular Disparity as a Cue for Distance in Macaque Area V4
The perceived size of an object remains relatively stable despite changes
in the size of its retinal image that accompany changes in distance from the
observer. This perceptual phenomenon is known as size constancy. The brain
uses both distance and retinal image size to estimate the size of objects.
In spite of the importance of object size information for object recognition,
there has been few studies on the neural representation of object size in the
Brain lesion studies suggest that the prestriate cortex of monkeys plays an important role in size constancy1 and area V4, a prestriate cortex, is essential for size discrimination2. Because neurons in area V4 are tuned to both stimulus size and stereoscopic depth, we hypothesized that binocular disparity information and retinal image size information are integrated to represent object size in area V4.
Here, we searched for neurons that encode object size, not retinal image size, in area V4. Neurons that invariantly encode the size of an object irrespective of its distance should systematically change their preferred retinal image size depending on the distance to the object. While monkeys were engaged in a simple fixation task, a disk of binocularly correlated random dot stereograms (RDSs) was presented over the receptive field of V4 neurons. We examined responses of V4 neurons to various combinations of binocular disparities and “cyclopean” image sizes of the disk. The background surrounding the disk was filled with binocularly uncorrelated dots, therefore the changes in binocular disparity or size of the disk did not accompany any change in monocular images.
A majority of V4 neurons were selective for both binocular disparity and cyclopean image size. They scaled the size tuning curves and shifted the preferred image size according to the change in binocular disparity; the preferred image size became larger as a stimulus was presented nearer. This result was consistent with the psychophysical performance tested with human observers; they perceived the larger stimulus presented at nearer position as the same size as the smaller stimulus presented at farther position. The magnitude of shifts varied from cell to cell. It was not affected by vergence angle, a cue for fixation distance. The results indicate that each of the object-size coding cells is used to calibrate the retinal image size when fixating on their appropriate fixation distances. The response property of V4 neurons was explained by an extension of disparity energy model. From these results, we suggest that V4 neurons compute object size using binocular disparity as a cue for distance.
1. L. Ungerleider et al., (1977) Exp Brain Res, 27:251-269.
2. P. H. Schiller and K. Lee, (1991) Science, 251:1251-1253.