Depth Perception Through Motion PDF Download

"When we move through the world, the motion of objects provides a sufficient cue for depth perception. For accurate depth measurements, the brain needs to resolve the depth-sign of objects (that is, whether the object is near or far relative to fixation). This is no easy task as depth-sign can be ambiguous based solely on visual motion. MT neurons are selective for depth-sign from motion parallax by combining retinal inputs and eye movement signals. We addressed three fundamental questions about how the brain uses motion parallax to code depth information. In the first experiment, we asked whether MT neurons are functionally linked to the perception of depth from motion parallax. Responses were recorded while macaque monkeys judged the depth-sign of visual stimuli containing motion parallax cues. We found that trial-by-trial variability of neural responses was correlated with the animal's perceptual decisions in the discrimination task. Greater responses predicted choices toward the depth preference of the recorded neurons. These results provide evidence that MT neurons may be involved in the perception of depth from motion parallax. In the second study, we investigated the nature of response modulation by eye movements. Direction-dependent modulation by eye movements yields the depth-sign selectivity of MT neurons. Responses of near-preferring neurons are suppressed when the eye moves toward the anti-preferred direction of neuron, whereas responses of far-preferring neurons are suppressed during eye movements toward the preferred direction. This response modulation exhibited both multiplicative and additive components, but the depth-sign selectivity of neurons was predicted only by the multiplicative gain change component. Using computer simulations, we show that a population of gain-modulated MT neurons can compute depth from motion parallax. Movement of an observer produces large background motion. In the third study, we hypothesized that neurons can use a visual consequence of self-motion (dynamic perspective cues) to compute depth-sign from motion parallax. We show that MT neurons can disambiguate depth-sign based on large-field background motion, in the absence of eye movements, and that these depth-sign preferences are correlated with those obtained when the animal is physically translated. MT neurons also contribute to depth perception from binocular disparity. It is likely that both eye movements and large field motion modulate MT responses to binocular images in a systematic way to encode the 3D spatial information of objects. These insights provide a deeper understanding of 3D information processing during navigation"--Pages v-vi.