Long-term memory gated encoding in working memory

人类拥有掌握信息的非凡能力--将信息保存在工作记忆(WM)中，为持续的思考和行动做好准备。尽管WM无可争辩地仅限于存储少量信息，但很明显，我们可以通过利用长期记忆(LTM)中的现有知识来扩展这一容量。但我们如何将LTM中的先验知识联系起来，使WM受益？想象你需要记住一个电话号码：你意识到它由一个熟悉的部分组成，比如你的出生年份，它立即变得更容易记住。不是在WM中对电话号码的每一位进行编码，而是在策略上只对WM中的新部分进行编码，并为已知部分招募LTM。尽管这是一个公认的行为学发现，但在LTM和WM之间的联系机制上，理论存在分歧，神经理论从未经过经验检验。我提出了一种新颖的、在神经上可信的机制，在该机制中，WM的编码由LTM引导，以主动地降低系统的认知负荷：“LTM门控编码”。目标1将在一个新的行为任务中阐明这一机制的界限，目标2将用脑磁图(MEG)描述这一机制背后的神经相互作用，目标3将用计算模型描述这一机制。我将接受最先进的神经生理学和机器学习方法的培训，这将为我在计算神经科学领域的职业生涯做好准备。该项目将把主管的技术和计算技能与我的经验领域知识相结合，以建立和测试这个新的量化框架，与以前在该领域的工作形成对比。

Humans have the remarkable ability to hold onto information - preserving it in working memory (WM) where it is ready for ongoing thought and action. Although WM is undisputedly limited to storing only small amounts of information, it is clear that we can extend this capacity by capitalizing on existing knowledge in long-term memory (LTM). But how do we connect prior knowledge in LTM to benefit WM? Imagine you need to hold a phone number in mind: you recognize that it consists of a familiar chunk such as your birth year, and it immediately becomes easier to remember. Instead of encoding each digit of the phone number in WM, one can strategically encode only the novel parts in WM and recruit LTM for the known parts. Although this is a well-established behavioral finding, theories diverge in the mechanism of connection between LTM and WM, and neural theories have never been tested empirically. I propose a novel, neurally plausible mechanism in which encoding in WM is guided by LTM to proactively decrease the cognitive load on the system: "LTM-gated encoding". Objective 1 will elucidate the bounds of this mechanism in a novel behavioral task, Objective 2 will delineate the neural interactions underlying this mechanism with magnetoencephalography (MEG), andObjective 3 will describe the mechanism with a computational model. I will receive training in state-of-the-art neurophysiological and machine-learning methods which will prepare me for a career in computational neuroscience. The project will integrate the supervisor's technical and computational skills with my empirical domain knowledge, to build and test this new quantitative framework which contrasts with previous work in the field.