Revealing the neural and computational bases of visual cognition

尽管我们可以很容易地理解外部视觉世界，但大脑执行的计算任务远非微不足道。设想一个自然的视觉场景，其中有许多对象，其中一些清晰可见，但另一些对象则部分遮挡在视线之外。为了解释这一场景，大脑不能只依赖输入的视觉信号，一方面包含的信息太少(遮挡)，另一方面包含太多(杂乱)。这引发了神经科学中的一个基本问题：大脑如何将传入的视觉信号转化为有意义的感知？在这项拟议的研究中，受训者和我将测试两种关于大脑如何将视觉信号转化为有意义的感知的有影响力的理论：预测编码和有偏见的竞争。这两种理论都认为，大脑产生的内部状态会主动与传入的感觉信号相互作用，从而引导视觉认知。这项拟议的研究通过在人类观察者身上实验操作这些内部状态，并使用一种新的集成功能磁共振成像(FMRI)和计算建模的方法来测试这些操作如何影响大脑的视觉处理。这种方法使我们能够揭示和整合视觉认知的神经和计算基础，从而为弥合认知和计算神经科学之间的鸿沟迈出了重要的一步。

Despite the ease with which we make sense of the outside visual world, the computational task performed by the brain is far from trivial. Consider a natural visual scene cluttered with many objects, some clearly visible but others partially occluded from view. To interpret this scene, the brain cannot only rely on the incoming visual signals, which on the one hand contain too little information (occlusion) and on the other hand too much (clutter). This raises a fundamental question in neuroscience: how does the brain transform the incoming visual signals into a meaningful percept? In the proposed research, trainees and I will test two influential theories about how the brain may transform visual signals into a meaningful percept: predictive coding and biased competition. Both theories propose that the brain generates internal states that actively interact with incoming sensory signals to guide visual cognition. The proposed research experimentally manipulates these internal states in human observers and uses a novel integrated approach of functional magnetic resonance imaging (fMRI) and computational modeling to test how these manipulations affect visual processing in the brain. This approach allows us to reveal and integrate the neural and computational bases of visual cognition, and therefore provides an important step toward bridging the gap between cognitive and computational neuroscience.