Neurophysiology of Visual Perception

我们认为我们的感知是理所当然的，很少质疑它的准确性。然而，患有大脑紊乱的患者却没有这种奢侈。例如，精神分裂症患者听到真实的声音或经历视觉幻觉，无法将真实与想象的刺激和事件区分开来。在另一个例子中，视觉偏斜，这是一种神经系统疾病，通常伴随着右脑顶叶的损伤，患者可能完全不知道世界的左半部分。这种感知错觉对我们大多数人来说甚至很难想象：在半忽视中，问题不在于病人看不到左边的东西，就好像视力被遮挡在那里一样。奇怪的是，对那个病人来说，左半世界本身完全不存在了，甚至病人自己都没有意识到问题的存在。这些和其他临床综合症表明，我们的感知世界，在其核心，是大脑的产物或结构，有时与外部世界不匹配。只要想想夜间的梦就行了。

We take our perception for granted, and seldom question its accuracy. Yet patients afflicted with disturbances of the brain do not have this luxury. A schizophrenic patient, for example, who hears real sounding voices or experiences a visual hallucination is not able to disentangle real from imaginary stimuli and events. In another example, visual hemineglect, which a neurological affliction that commonly follows damage to the parietal lobe of the right cerebral hemisphere, a patient may become entirely unaware of the left half of the world. Such perceptual delusions are difficult for most of us to even imagine: in hemineglect, the problem is not that the patient does not see what is on the left, as if vision were occluded there. Strangely, for that patient, the left half of the world itself entirely ceases to exist, often without even the patient realizing that a problem exists. These and other clinical syndromes illustrate that our perceptual world is, at its very core, a product or construction of the brain which sometimes does not match the external world. One need only think of nighttime dreaming. Human and nonhuman primates rely primarily on vision to guide us through the environment, manipulate objects, and interact with one another. As visual animals, our representation of shape, color, and space is fundamental to how we see the world. While scientists, owing to decades of research in animals, understand many of the steps in processing the images striking the retina, there still does not presently exist a good conceptual framework for understanding the brain circuitry that determines the contents of our perception. One mission of the Unit on Cognitive Neurophysiology and Imaging is to use novel techniques and paradigms in an attempt to reframe, and gain a deeper understanding of, this challenging problem. For the brain, all visual information is derived from the retinal images, which is covered with a regular 2-D array of light- and color-sensitive receptors. However, the world is three dimensional. In order for the brain to interpret, and ultimately construct, an internal 3-D representation of its surroundings, it must apply an inferential and interpretive approach to the retinal images. Research has shown that the brain is adept at interpreting depth cues from shadows, texture gradients, relative motion, occlusion, and stereoscopic disparity. These processes are done so immediately and effortlessly that few would even consider the problem at all. Yet, for a primate, human or monkey, needing to react quickly to catch a ball or grasp a branch, accurate visual computation of the third dimension is critical for survival. But how does one investigate such subjective processes that require so much prior knowledge and assumptions on the part of the brain? One approach is to employ stimuli for which a single stimulus offers more than one perceptual solution. Stimuli in this category, such as the famous Necker cube, or Rubins face-vs-vase stimulus, are commonly found in introductory psychology textbooks. Such stimuli are typically called perceptually multistable, owing to their tendency to lead to a sequence of perceptual reversals over time, and offer a window into studying the neurophysiological basis of the interpretive processes in vision. We have approached this problem using two variants of multistable patterns: (1) bistable stimuli, with exactly two solutions, and (2) perceptual suppression, where a salient visual pattern, which is normal visible and stable, can be induced to abruptly disappear. In the latter case we ask, "when a bright visual stimulus is perceptually suppressed, does it continue to activate the visually responsive regions of the brain?" Finding an answer to this question is surprisingly difficult. We have recently published findings related to activity of neurons in the thalamus and visual cortex during periods of perceptual suppression in alert monkeys, trained to report the visibility of a stimulus. We found that under conditions in which a salient, bright pattern is physically present on a screen, but unperceived because of a visual illusion, neurons in these structures are divided in their responses. Some neurons, such as those in the lateral geniculate nucleus (LGN) of the thalamus, and primary visual area of the cortex, reliably respond to the bright stimulus, whether it is perceived or not. In other words, these areas can be seen as sensory channels, transmitting information into the brain with little regard to what the perceptual outcome may be. However, in sharp contrast, neighboring areas of the brain, such as the pulvinar nuclei of the thalamus, and visual association areas of the cortex, show an entirely different pattern. Many neurons in these regions respond according to whether the animal subjectively perceived the stimulus on a given trial. On trials in which the stimulus was perceived, neurons would fire. On those in which it was not perceived, neurons would become quiet. Importantly, the physical stimulus was the same in each case. Thus, in the thalamocortical circuitry involved in the processing of visual information, there exist both pre- and post-interpretive elements, with the latter more closely allied than the former to the monkeys perceptual state. In a related study, we have further explored the role of the thalamus in perception using local pharmacological inactivation of specific subnuclei. Given our neurophysiological results, we were particularly interested whether inactivating the LGN would bring about a different result from inactivating the pulvinar. We carried this out by injecting microliter quantities of muscimol (a GABA-agonist that shuts down local neural activity) into each area. Upon temporary inactivation of the LGN, the animal experienced a scotoma, in which a portion of the visual field was essentially blind. For a few hours, the animal was unable to detect small targets presented to that region of visual space, a finding consistent with the LGN playing a critical role in transmitting sensory information to the cortex. The deficit was temporary, and after a few hours, the animal was back to normal. In sharp contrast, inactivating the pulvinar (its dorsal aspect) led to an entirely different perceptual outcome, albeit also temporary. Specifically, following such inactivation, monkeys displayed a perceptual syndrome that, when tested, was highly similar in its nature to that experienced with human hemineglect, with the apparent loss of subjective representation of the side of visual space contralateral to the injection. As in the human patients, all evidence suggested that the subjective representation of space itself had disappeared. These findings taken together raise multiple questions about how to think about the site of visual perception, particularly in light of the different tools used to study it. The conventional metaphor for visual processing in the brain, as a sequence of filters acting upon an image, does not adequately capture the interpretive elements that account for phenomena such as multistability and hemineglect. Our research program therefore aims to break new ground in how to conceptualize how the brain constructs its perceptual world based on its sensory input.