Mechanisms of Attentional Selection in Human Visual Cortex

在国家科学基金会的支持下，萨默斯博士将进行一项为期三年的研究，利用功能磁共振结合心理物理学和计算建模来研究人类视觉皮质中注意力选择的机制。研究内容包括：1)不同选择方式下注意的空间分布；2)抑制在选择中的作用及其与易化的关系；3)注意促进或抑制视觉皮层刺激表征的计算机制。人类的大脑经常会遇到复杂的环境，其中包含的信息超过了它在一瞬间所能处理的信息。注意力通过选择要处理的刺激，起到感知和认知的关键作用。聚光灯的比喻是选择空间的一个连续区域，长期以来一直被用来描述视觉注意(James，1890)。除了空间选择，还从行为上推断了基于对象和基于特征的选择。这一建议试图研究这些不同形式的选择对人类视觉皮质施加的调制的空间模式。使用功能磁共振成像，我们已经观察到，在适当的任务要求下，注意聚光灯可能被分离，以同时关注两个不同的空间区域，而不选择中间区域(见初步结果)。这项提案将研究分散注意力聚光灯的能力和限制。本研究还提出了基于特征和基于对象的注意如何在视网膜视觉皮质中跨空间运作的研究。进一步的研究将探索注意分辨率的空间限制与不同视觉皮质区域的处理能力之间的关系。注意选择是一个竞争的过程，胜利者是被选择的刺激，失败者是未被选择的刺激。有人认为，选择不仅包括促进过程，也包括抑制过程。事实上，已经观察到对视觉皮质中无人注意的刺激的注意抑制(例如，Somers等人，1999)。在这里，有人建议调查抑制是如何跨越空间起作用的，以及它与促进的关系。当注意力被转移到其他地方时，压抑纯粹是一个被动的过程(例如，撤退)吗？或者是否存在要抑制的激活零部件？是否需要主动抑制才能分割聚光灯？在无人看管的空间中，抑制是统一的，还是健壮的干扰物受到有针对性的抑制？抑制在多大程度上随任务形式和/或任务要求的不同而变化？即使在初级视觉皮质中，注意选择也会调节活动。这意味着注意力直接与低水平的刺激表征相互作用。在不破坏视觉皮层刺激编码的情况下，注意力选择刺激的计算机制是什么？注意力主要调节刺激驱动反应的增益(乘法效应)吗？或者，注意力主要偏向(加性效应)刺激反应，从而使他们在获得认知过程方面具有竞争优势？为了研究这些问题，提出了一系列参数fMRI/心理物理研究，其中当注意力指向或远离刺激时，物理刺激的性质将发生变化。这些数据将被用来构建有注意力和无注意力的fMRI刺激反应函数，并将允许我们推断注意力对神经元群体的主要计算效果。在进行这些实验的同时，还将进行计算建模研究。该模型将提供一个框架，不仅可以集成目前的实验结果，还可以集成关于视觉皮层单个单位反应的灵长类数据库。预计建模将完善对选择机制的理解，并有助于产生未来的实验问题。拟议研究的更广泛影响包括增加对正常人认知大脑机制的理解。这种关于思维和大脑之间关系的知识将为未来的临床研究提供信息，特别是对患有注意力障碍的人群。对大脑如何实现注意力的计算理解也可能有助于未来计算机体系结构的设计。这项研究还为本科生和研究生提供了培训机会，包括那些来自代表性不足群体的学生。

Mechanisms of Attentional Selection in Human Visual CortexAbstract With National Science Foundation support, Dr. Somers will conduct a three-year study to investigate the mechanisms of attentional selection in human visual cortex using functional MRI combined with psychophysics and computational modeling. Studies are proposed that will investigate 1) the spatial distribution of attention under different forms of selection, 2) the role of suppression in selection and its relationship to facilitation, and 3) the computational mechanisms by which attention facilitates or suppresses a stimulus representation in visual cortex. The human mind regularly encounters complex environments that contain more information than it can process at one instant. Attention acts as the critical gateway to perception and cognition by selecting the stimuli that will be processed. The metaphor of the spotlight, selecting a single contiguous region of space, has long been employed to describe visual attention (James, 1890). In addition to spatial selection, object-based and feature-based selection have also been inferred behaviorally. This proposal seeks to investigate the spatial patterns of modulations these differing forms of selection can exert on human visual cortex. Using fMRI, we have observed that under appropriate task demands, the attentional spotlight may be split to attend to two distinct regions of space simultaneously while leaving the intermediate region unselected (see preliminary results). The capabilities and limitations in splitting the attentional spotlight will be investigated in this proposal. Investigations of how feature-based and object-based attention operate across space in retinotopic visual cortex are also proposed. Further studies will investigate the relationship between the spatial limits of attentional resolution and the processing capabilities of different visual cortical areas. Attentional selection is a competitive process with winners, the selected stimuli, and losers, unselected stimuli. It has been argued that selection is comprised not only of facilitative processes, but also of suppressive processes. Indeed, attentional suppression of unattended stimuli in visual cortex has been observed (e.g., Somers et al, 1999). Here, it is proposed to investigate how suppression acts across space and its relationship to facilitation. Is suppression purely a passive process (e.g., withdrawal) that results when attention is directed elsewhere? Or is there an active component to suppression? Is active suppression required to split the spotlight? Is suppression uniform across non-attended space or do robust distractors receive targeted suppression? To what degree does suppression vary with task form and/or task demands? Attentional selection modulates activity even in primary visual cortex. This implies that attention directly interacts with low-level, stimulus representations. What are the computational mechanisms by which attention selects stimuli without disrupting stimulus encoding in visual cortex? Does attention primarily modulate the gain (multiplicative effect) of stimulus-driven responses? Or does attention primarily bias (additive effect) stimulus responses to give them a competitive advantage in gaining access to cognitive processes? To investigate these issues a series of parametric fMRI/psychophysical studies is proposed in which physical stimulus properties will be varied while attention is directed toward or away from the stimulus. These data will be used to construct fMRI stimulus response functions both with and without attention and will allow us to infer the primary computational effect of attention on a neuronal population. In concert with these experiments, computational modeling studies will be performed. The modeling will provide a framework to integrate not only the present experimental results, but also the primate database on single unit responses in visual cortex. It is expected that the modeling will refine understanding of selection mechanisms and help to generate future experimental questions. The broader impact of the proposed studies includes increased understanding of cognitive brain mechanisms in normal humans. This knowledge about the relationship between mind and brain will inform future clinical studies, particularly for populations with attentional dysfunction. Computational understanding of how the brain implements attention may also aid design of future computer architectures. This research also offers training opportunities for undergraduate and graduate students, including those from underrepresented groups.