How does visual long-term memory implement control over visual attention?

在我们的日常生活中，我们经常应用我们的先验知识，同时与复杂的自然界互动。考虑一次去杂货店的简单旅行。首先，你可能不得不在一张凌乱的桌子上找到你的车钥匙。你的办公桌上可能有一些东西，比如假日照片、卷笔刀或电脑，这些东西自然会引起人们的注意，但在去商店时却毫无用处，所以它们必须被忽略。然后，在找到钥匙并开始开车后，一个人必须忘记钥匙，而是记住照顾正在玩耍的儿童、行人、骑自行车的人、其他司机、交通信号灯、街道标志和地标，以安全地沿着正确的路线到达商店。尽管在这项看似简单的任务中存在所有明显的挑战，但人们显然可以轻松地完成它，以及复杂得多的任务。这个项目试图更好地理解先前的知识如何指导我们目前的行动和对环境的看法。更好地处理这一过程可以帮助我们理解人类能力的一个基本要素，即使是我们最先进的人工智能系统也无法理解。这项研究本身对社会有很大的好处：了解并因此改善司机或空中交通管制员的注意力控制的潜力。除了对心理学和神经科学的更广泛影响外，科学家还将参与公共推广活动，如年度脑爆炸事件、大脑生物播客，以及K-12教师-神经科学家合作伙伴关系，以提高对神经科学的理解和欣赏。为了了解先前知识如何指导感知和当前行为，这项提议涉及到非侵入性记录人们在杂乱场景中寻找特定物体时的脑电活动(用脑电图仪或EEG)。这些场景是计算机屏幕上随机排列的对象的图片。研究人员将使用这些录音来调查是否需要刻意记住先前的知识(视觉长期记忆中的模板)，以指导视觉注意和视觉空间行为，如视觉目标检测。使用这种方法，研究人员可以测量大脑活动，跟踪你存储在记忆中的东西，以及你在场景中关注的东西。对这些数据的分析将揭示出，与物体记忆相关的大脑活动是否就是你关注该物体的全部。这些实验将解决的一个关键问题是，长期物体记忆和短期工作记忆对注意物体搜索的相对贡献。相互竞争的假说将通过脑电波和非侵入性经颅脑刺激相结合的方式进行检验。科学家们还将研究，如果你之前在场景中发现了一个物体，那么它在未来是否会变得特别难以忽视。这个假设是，以前搜索过对象会导致在未来的搜索中吸引更多的注意力到该对象上。实验将测试这种效应是否可以在大脑的反应中观察到。初步结果表明，记忆塑造了人们对特定对象的注意力，此外，人们可以在记忆中同时持有多个对象，并使用这些搜索模板同时寻找多个对象，这是目前人工智能系统不可能实现的壮举。这些发现将被用来通过测试人类注意力和记忆的模型来取得科学进展，并帮助我们设计更好的人工智能系统，这些系统可以匹配人类的能力，并可能更好地与人类合作。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

In our everyday lives we often apply our prior knowledge while simultaneously interacting with the natural world in all its complexity. Consider a simple trip to the grocery store. First, one might have to find your car keys on a messy desk. Your desk may have items like a holiday photo, pencil sharpener or a computer that naturally draw attention but are of no use in going to the store, so they must be ignored. Then, having found the keys and starting to drive, one must forget the keys and instead remember to attend to children at play, pedestrians, cyclists, other drivers, traffic signals, street signs and landmarks to safely follow the right route to the store. Despite all the challenges apparent in even this seemingly simple task, people accomplish it, and far more complex tasks, with apparent ease. This project seeks to gain a better understanding of how prior knowledge guides our current actions and perceptions of the environment. A better handle on this process can help us understand a fundamental element of human ability that still eludes even our most advanced artificial intelligence systems. A substantial benefit to society is inherent in the research: the potential to understand and thereby improve attentional control for drivers or air traffic controllers. In addition to the broader impact in psychology and neuroscience, scientists will engage in public outreach such as the Annual Brain Blast events, the Brain Bios podcast, and the K-12 Teacher-Neuroscientist partnership to improve understanding and appreciation of neuroscience.To understand how prior knowledge guides perception and current behavior, this proposal involves non-invasively recording electrical brain activity (with electroencephalography or EEG) from people searching for particular objects in cluttered scenes. These scenes are picture of objects arranged randomly on a computer screen. The researchers will use these recordings to investigate whether prior knowledge (templates in visual long-term memory) needs to be intentionally held in mind to guide visual attention and visuospatial behavior such as visual target detection. Using this approach, researchers can measure brain activity that tracks what you store in memory, as well as what you pay attention to in the scenes. Analysis of these data will reveal whether brain activity related to object memory is all you need to pay attention to that object. A key question these experiments will resolve is the relative contribution of long-term object memory and short-term working memory to attentional object search. Competing hypotheses will be tested using EEG combined with noninvasive transcranial brain stimulation. Scientists will also study whether if you have previously found an object in a scene whether it then becomes particularly difficult to ignore in the future. The hypothesis is that previously having searched for an object will lead to more attention being drawn to that object in a future search. Experiments will test whether this effect can be observed in the response of the brain. Preliminary results suggest that memory shapes attention for specific objects and moreover, that people can simultaneously hold multiple objects in memory, and use these search templates to look for multiple objects at the same time, a feat that is currently impossible for artificial intelligence systems. These findings will be used to make scientific progress by testing models of human attention and memory, as well as helping us design better artificial intelligence systems that can match human ability and potentially better partner with humans as well.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.