Pattern Separation, Pattern Completion, and Encoding Dynamics in the MTL

学习和检索信息的能力对于我们有效应对日常生活的努力的成功至关重要，我们个人收集的长期陈述性记忆通过形成我们个人身份的核心来引导我们度过时间。此外，随着我们年龄的增长，长期记忆问题是最常见的认知问题，也是神经疾病的关键问题。因此，了解构成陈述性记忆的神经机制是神经科学的核心问题。在美国国家科学基金会的支持下，约翰·霍普金斯大学的克雷格·斯塔克和他的同事们在澄清记忆在人脑中存储和访问的方式方面取得了进展。内侧颞叶(MTL)的结构(包括海马体及其邻近的内嗅皮层、周围皮质和海马旁皮质)在学习和记忆事实和事件的能力中发挥着关键作用。然而，目前还不清楚每种结构扮演的角色是什么，以及这些结构是如何相互作用的，从而使我们具有快速学习新信息的非凡能力。在这个项目中，总共将进行十个实验，以加深我们对这个问题的理解。拟议中的实验利用了一系列广泛的技术来提供融合和互补的证据：健康志愿者大脑活动的功能磁共振成像(FMRI)，海马体受损患者的神经心理学研究，以及健康志愿者的认知行为测量。通过这样做，它将开发整合这些信息来源的新技术，并探索新的范例和新的理论基础。该项目分为三个目标。在目标1和2中，研究了模式完成和模式分离的神经解剖学基础。模式分离指的是将两个或更多相似的活动模式(要记住的事情)分离成不同的、不重叠的表示。当一个人需要分离可能相互干扰的相似信息片段时，就需要模式分离(例如，记住，今天你把车停在了B地段，尽管你通常把车停在A地段)。模式补全是模式分离的补充。它指的是根据你的知识将不完整、扭曲或退化的活动模式转换为完整的活动模式(例如，虽然没有对你今天的停车位置进行完美的记忆，在部分记忆中填入A批的细节……通常，但并不总是正确的)。两者都是成功记忆的关键，因为当差异很重要时，我们都必须隔离相似的记忆，当差异不重要时，我们必须利用相似记忆中存在的冗余来克服噪音和容量限制。计算模型和来自动物研究的证据强调了这些相互竞争的计算特性，并使用神经解剖学约束来分离海马体(甚至是海马体的一个子区域)在模式分离中的特定角色。在第三个目标中，使用后向掩蔽技术探索了单项和联想记忆编码的动力学。这项技术揭示了单项记忆和联想记忆在性能上的一种新的分离。这个项目将通过测试这种影响如何映射到现有的记忆理论，如陈述性/非陈述性和记忆/知道二分法，以及它如何受到海马体和MTL其余部分的损害来确定这种分离是否依赖于MTL。

The ability to learn and retrieve information is essential to the success of our efforts to effectively cope with daily life, and our individual collections of long-term declarative memories serve to guide us through time by forming the core of our personal identities. Further, as we age, long-term memory concerns are the most common cognitive complaint and are key concerns in neurological disorders. Understanding the neural mechanism that underlie declarative memory is thus a core problem in neuroscience. With support from the NSF Craig Stark and colleagues at Johns Hopkins University are making inroads into clarifying the way in which memories are stored and accessed in the human brain. Structures in the medial temporal lobe (MTL) (including the hippocampus and the adjacent entorhinal, perirhinal, and parahippocampal cortices) are known to play a critical role in the ability to learn and remember facts and events. However it is not yet clear what the role is that each structure plays and how the structures interoperate to give us this remarkable ability to rapidly learn new information. In this project a total of ten experiments will be conducted that will further our understanding of this problem. The proposed experiments utilize a wide array of techniques in order to provide converging and complementary evidence: functional magnetic resonance imaging (fMRI) of brain activity in healthy volunteers, neuropsychological study of patients with damage to the hippocampus, and cognitive behavioral measures on healthy volunteers. In so doing, it will develop novel techniques for integrating these sources of information and it will explore new paradigms and new theoretical grounds.The project is divided into three Aims. In the Aims 1 and 2, the neuroanatomical basis for pattern completion and pattern separation are examined. Pattern separation refers to the isolation of two or more similar patterns of activity (things to be remembered) into distinct, non-overlapping representations. Pattern separation is required whenever one needs to separate similar pieces of information that might otherwise interfere with each other (e.g., remembering that today, you parked your car in Lot B, although you usually park your car in Lot A). Pattern completion is pattern separation's complement. It refers to transforming an incomplete, distorted, or otherwise degraded pattern of activity into a complete pattern of activity based on your knowledge (e.g., while not having stored a perfect memory of where you parked today specifically, filling in the partial memory with details of Lot A...usually, but not always, correctly). Both are critical to successful memory, as we must both isolate similar memories when the differences are important and we must utilize the redundancy present in similar memories to overcome noise and capacity limitations when the differences do not matter. Computational models and evidence from animal studies have stressed these competing computational properties and have used neuroanatomical constraints to isolate a specific role for the hippocampus (or even a subregion of the hippocampus) in pattern separation. In the third Aim, the dynamics of single-item and associative memory encoding are explored using a backward-masking technique. This technique has revealed a novel dissociation in performance between single-item and associative memory. This project will determine whether the dissociation is dependent on the MTL by testing how the effect maps onto existing theories of memory such as the declarative / nondeclarative and remember / know dichotomies and how it is influenced by damage to the hippocampus and the rest of the MTL.