Organization of complex memories in the mammalian brain - testing basic assumptions of systems consolidation theories

对经验的记忆包括储存在大脑不同区域的各种感觉模式。例如，对一顿非凡的晚餐的记忆可能包含关于食物的味道的知识，这些知识存储在味觉系统中，或者关于桌子上的花瓶看起来是什么样子的知识，这得到了中介对象知识的系统的支持。尽管共同构成这段经历的点点滴滴是以分散的方式存储的，但我们对这一事件的记忆是一个完整的整体。大脑如何实现这种整合被称为绑定问题。目前的理论提出了一种分层记忆系统，在该系统中，分布的新皮质部位通过中央中枢间接地联系在一起，中央中枢本身近似于一个自动联想网络，其组成神经元之间具有密集的互连。基于这一思想的两个最新的模型--系统整合的标准模型(SMC，例如，Alvarez&Squire，1994)和多踪迹理论(mtt，例如，Nadel&Moscovitch，1997)--表明，由于其解剖学和连接性，海马体连接新皮质区域，而新皮质区域编码共同构成事件记忆的刺激。这些记忆模型尚未显式地被开发来解释绑定问题，而是解释逆行性健忘症，即观察到最近的但不是远程的记忆严重依赖于海马体(例如，Zola-Morgan&Squire，1990年)。这两种理论都表明，随着时间的推移，存储记忆各种元素的分布的新皮质区域之间的联系是在海马体的帮助下形成的。关于海马体在远程记忆中的作用，这两种理论存在差异。虽然标准模型提出远程记忆缺乏相应的海马区成分，但MTT将其归因于记忆的整个生命周期在提供空间背景方面的作用。关于记忆的分布性的假设，以及关于记忆随着时间的质变和作为海马区损伤的函数的预测，几乎没有得到经验上的关注。例如，到目前为止，还没有行为测试可靠地评估分布式功能网络是否调节记忆，记忆是否随着时间的推移发生质的变化，以及海马体对这些记忆的定性和时间依赖贡献。我们利用系统神经科学中的先进概念和新的行为范式，允许测试记忆的质量性质是否随着时间的变化而变化。此外，我们提出了一项创新的任务，以研究海马体是否发挥着结合作用，如果是的话，它的作用是否是暂时的，直到有助于记忆的新皮质区域之间的连接成熟为止。这一范式是基于对由两个通道组成的刺激的识别记忆。根据理论，对于最近的记忆，在保持测试中呈现一种模式应该通过海马体激活另一种模式。然而，对于远程记忆，呈现一种模式应该导致第二种模式以一种独立于海马体的方式激活。总之，这些实验将直接从行为上测试海马体作为瞬时结合剂的假定功能作用。

Memory for experiences encompasses various sensory modalities that are stored in distinct brain areas. For example, a memory of a remarkable dinner may contain knowledge about how the food tasted, which is stored in the gustatory system, or of how a vase on the table looked like, which is supported by the system mediating object knowledge. Although the bits and pieces that together make up the experience were stored in a distributed manner, our memory of the event is an integrated whole. How the brain achieves this integration is referred to as the binding problem. Current theories propose a hierarchical memory system in which distributed neocortical sites are indirectly associated by ways of a central hub , which itself approximates an auto associative network with dense interconnections among its constituting neurons. The two most recent models based on this idea - the standard model of systems consolidation (SMC, e.g., Alvarez & Squire, 1994) and multiple-trace theory (MTT, e.g., Nadel & Moscovitch, 1997) - suggest that, due to its anatomy and connectivity, the hippocampus links the neocortical areas, which encode the stimuli that together constitute memory for an event.These memory models have not been explicitly developed to account for the binding problem, but to explain retrograde amnesia, i.e., the observation that recent but not remote memories critically depend on the hippocampus {e.g., Zola-Morgan & Squire, 1990). Both theories suggest that over time associations between the distributed neocortical areas that store the various elements of a memory are formed with the help of the hippocampus. The theories differ in respect to the role of the hippocampus in remote memories. While the standard model proposes that remote memories are lacking a corresponding hippocampal component, MTT ascribes to it a role for the entire lifetime of a memory in providing the spatial context.What has received little empirical attention are the assumptions regarding the distributed nature of memory, and the predictions regarding qualitative changes in memory over time and as a function of hippocampal damage. For example, so far no behavioral test reliably assesses whether a distributed functional network mediates memories, whether memories qualitatively change over time, and the qualitative and time-dependent contribution of the hippocampus to these memories. We exploit advanced concepts in systems neuroscience with new behavioral paradigms that permit testing whether the qualitative nature of memories changes with time. In addition, we propose an innovative task to study whether the hippocampus plays a binding role, and if so, if its role is transient until the connections between neocortical areas contributing to memory have matured. This paradigm is based on recognition memory for stimuli that are composed of two modalities. According to theory, for recent memories presentation of one modality on a retention test should activate the other modality via the hippocampus. However, for remote memories, presentation of one modality should lead to activation of the second one in a manner independent of the hippocampus. Together, these experiments will directly test behaviorally the postulated functional role of the hippocampus as a transient binding agent.