The neural basis of top-down biasing of bottom-up visual processing

虽然我们看到的大部分东西都是由视觉刺激驱动的，并且几乎与物理现实完全一致，但有时我们可以看到图像或自然世界，并看到并不存在的东西。这就是许多视觉错觉的起源。然而，在其他时候，我们的视觉体验取决于我们如何处理传入的视觉刺激。例如，我们甚至可以看到我们想要看到的东西，就像我们选择在天空的云层中看到动物或面孔的形状一样。因此，我们的意图可以在视觉感知中发挥作用。注意力也影响我们如何看待事物。例如，我们更有可能注意到我们正在关注的事情的细节。注意力并不存在于世界或“自下而上”的视觉刺激中，而是我们大脑中一种“自上而下”的聚焦机制，使我们能够有选择地关注最相关的输入刺激。 因此，感知既可以受到“自上而下”的注意力的影响，也可以受到“自上而下”的意图的影响。我们称之为“自上而下”的处理，因为我们所看到的不仅仅是由“自下而上”的感官输入驱动的。它还受到我们的注意力集中在哪里以及我们的期望和意图的性质的影响。由自上而下处理驱动的一种视觉输入的一个例子是所谓的“视运动”。一个例子是，当你在高速公路上开车时，看到两个闪烁的灯对角对齐，在一个假想的正方形的对角上。当这些灯关闭时，两个不同的灯在不可见的正方形的另外两个角上打开。当这两种灯光配置依次打开和关闭时，您很可能会体验到一种明显的运动错觉，并看到闪烁的灯光似乎来回跳跃，即使灯光在任何给定时间都是静止的。有趣的是，有些人会自发地看到水平方向的明显运动，而另一些人则会在灯光成对闪烁时看到明显的垂直运动，即使灯泡实际上根本没有移动。更令人着迷的是，人们通常可以有意识地决定他们随后看到的是错觉的垂直运动还是水平运动，因此他们的意图可以塑造他们对这种明显运动错觉的感知。推动这项研究的一个中心问题是，大脑如何实现“自上而下”的意图，即看到垂直与水平运动。这个项目中的实验旨在揭示可能导致这种感知自愿重塑的神经机制。受试者在进行这些知觉实验时，使用功能性磁共振成像（fMRI）对他们的大脑进行非侵入性成像，以便在受试者参与任务时可以监测大脑活动。视觉感知实验是在扫描仪中进行的，受试者被指示尝试看到垂直或水平运动，然后在提示时切换方向。在大脑皮层中有视觉运动处理区域，不同的子区域编码垂直和水平运动方向。在收集到功能磁共振成像数据后，可以分析大脑图像，以精确定位那些允许预测受试者意图的大脑区域：即受试者是否打算在刺激发生任何运动之前看到水平或垂直运动。 根据大脑活动，也有可能破译他们最终会感知到什么运动，因为他们的意图是看到水平或垂直运动。这项研究的最终目标是发现大脑的哪些区域允许我们解码这些视觉意图，并了解大脑中自上而下的处理如何处理所有形式的感官知觉。除了学术研究外，项目组还希望通过多方面的外展活动，扩大公众对科学的参与，提高公众对科学研究的理解和兴趣，包括举办科学讲座，为公众撰写科学文章，以及为科学博物馆建立一个视觉错觉大厅。该项目由认知神经科学计划共同资助，该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Although much of what we see is simply driven by incoming visual stimuli and corresponds almost perfectly with physical reality, sometimes we can look at an image or the natural world and see things that aren’t really there. This is the origin of many visual illusions. Other times, however, what we visually experience depends on how we process incoming visual stimuli. For example, we can even see what we intend to see, as when we choose to see the forms of animals or faces in clouds in the sky. So, our intentions can play a role in visual perception. Attention is also influential in how we see. For example, we are more likely to notice details about the things to which we are paying attention. Attention is not in the world or in the “bottom-up” visual stimulus, but is a “top-down” focusing mechanism in our brain that enables us to selectively attend to the most relevant incoming stimuli. Perception can therefore be influenced by both “top-down” attention and “top-down” intentions. We call this “top-down” processing because what we see is not solely driven by what is in the “bottom-up” sensory input. It is also affected by where our attention is focused, and by the nature of our expectations and intentions. One example of a kind of visual input that is driven by top-down processing is so-called “apparent motion.” One example occurs when you are driving on the highway and see two flashing lights aligned diagonally, on opposite corners of an imaginary square. When these turn off, two different lights turn on at the other two corners of the invisible square. When the two light configurations toggle on and off sequentially, you are very likely to experience an illusion of apparent motion and see the flashed lights appear to jump back and forth, even though the lights are stationary at any given time. Interestingly, some people spontaneously see apparent motion in the horizontal direction, while others see apparent vertical motion between the lights when they flash on and off in pairs, even when no lightbulbs are actually moving at all. What is even more fascinating, is that people can often consciously decide whether they subsequently see illusory vertical or horizontal motion, so their intention can shape their perception of this apparent motion illusion. A central question driving this research is how a “top-down” intention to see vertical vs horizontal motion is realized in the brain. The experiments in this project are designed to unravel the neural mechanisms that can lead to this voluntary reshaping of perception. Subjects perform these perceptual experiments while their brains are being imaged non-invasively, using functional magnetic resonance imaging (fMRI), so that brain activity can be monitored while subjects engage in the task. Visual perception experiments are conducted in the scanner with human subjects who are instructed to try to see either vertical or horizontal motion, and then switch directions when cued. There are visual motion processing areas in the cortex of the brain, with different sub-areas encoding vertical versus horizontal motion directions. After collecting fMRI data, it is possible to analyze brain images to precisely locate those brain areas that allow predictions of a subject’s intention: i.e. whether subjects were intending to see horizontal or vertical motion, before any motion took place in the stimulus. Based on brain activity, it is also possible to decipher what motion they would actually end up perceiving, given their intention to see horizontal or vertical motion. The ultimate goal of the research is to discover what areas of the brain allow us to decode these visual intentions, and to understand how top-down processing works in the brain for all forms of sensory perception. In addition to scholarly research, the project team also hopes to engage in multiple outreach efforts to broaden participation in science and improve public understanding of and interest in scientific research, including giving public talks on science, writing about science for the general public, and setting up a hall of visual illusions for a museum of science.This project is jointly funded by the Cognitive Neuroscience Program, and the Established Program to Stimulate Competitive Research (EPSCoR).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.