CAREER: Testing a unified theory of perception and memory in the medial temporal lobe

记忆对我们的日常生活至关重要，它使我们能够认出一个熟人的脸，记住我们把车停在哪里，或者生动地回忆起上周发生的一件事的细节。然而，它是脆弱的：随着年龄的增长，记忆力会明显下降，大脑内侧颞叶受损的人记忆力恶化得更厉害。直到最近，记忆理论都认为内侧颞叶只负责记忆功能。然而，最近的证据表明，这些大脑区域对其他认知任务也很重要，包括想象未来和视觉上区分两个高度相似的物体。因此，当前知识的一个关键缺口是记忆如何与高级视觉感知相互作用的理论，以及为什么这两种认知功能都依赖于内侧颞叶。在NSF CAREER奖的支持下，Rosie Cowell博士将发展这样一个理论，并通过实证研究对其进行检验。通过将该理论应用于脑损伤引起的健忘症和正常衰老引起的更中度的记忆丧失，该项目将研究这两种形式的记忆丧失是否可以用相同的机制来解释。此外，该项目将通过以下方式更广泛地造福社会：(1)提供独特的综合研究培训制度，使研究生获得理论构建和假设检验的直接经验；(2)通过为来自弱势群体的高中女学生提供实践科学研讨会，以及通过在马萨诸塞州当地社区的公共宣传，扩大STEM领域的参与。为了建立记忆和高级感知的统一理论，建议的研究将采用行为研究，功能磁共振成像（fMRI）和计算建模。这些研究将在两种认知受损人群中检验知觉干扰对记忆的影响：内侧颞叶（MTL）受损的个体和健康的老年人。这一提议的中心假设是，MTL容纳了对记忆和感知都至关重要的物体和事件的联合表征。连词对感知至关重要，因为它们允许区分两个相似的物体或事件，对记忆至关重要，因为它们屏蔽了持续不断的感知干扰，否则会导致对物体或事件的戏剧性遗忘。据推测，MTL损伤和正常衰老都会损害连接表征，迫使一个人依赖于保留在感觉皮层中的碎片性、基于特征的表征。基于特征的表征不足以区分相似的项目（导致知觉障碍），并且极易受到知觉干扰（导致记忆问题）。计划中的研究预计将揭示MTL结构有助于任何认知任务，无论是助记还是感知，只要任务解决方案需要复杂的、连接的刺激表征。通过开发一个形式化的、神经计算上合理的模型，将神经表征与认知表现以明确的机制联系起来，将充分利用这种理解认知的表征方法。

Memory is critical to our everyday existence, enabling us to recognize an acquaintance's face, remember where we have parked our car, or recall an event from last week in vivid detail. And yet, it is fragile: memory performance can decline noticeably with age and deteriorates even more dramatically in individuals with damage to the medial temporal lobes of the brain. Until recently, theories of memory assumed that the medial temporal lobes are dedicated solely to memory function. However, recent evidence suggests that these brain regions are important for other cognitive tasks, including imagining the future and visually discriminating between two highly similar objects. Therefore, a critical gap in current knowledge is a theory of how memory interacts with high-level visual perception and why both of these cognitive functions depend upon the medial temporal lobes. With the support of this NSF CAREER award, Dr. Rosie Cowell will develop such a theory and test it with empirical studies. By applying the theory to both amnesia caused by brain damage and the more moderate memory loss caused by normal aging, this project will investigate whether these two forms of memory loss can be explained by the same mechanisms. In addition, the project will benefit society more broadly by (1) providing a uniquely integrated research training regime in which graduate students receive direct experience with both theory-building and hypothesis-testing, and (2) broadening participation in STEM fields via hands-on science workshops for female students from high-schools that serve under-represented groups, and via public outreach in the local Massachusetts community.To develop a unifying theory of memory and high-level perception, the proposed research will employ behavioral studies, functional magnetic resonance imaging (fMRI) and computational modeling. The studies will examine the effect of perceptual interference on memory in two cognitively-impaired populations: individuals with medial temporal lobe (MTL) damage and heathy, older adults. The central hypothesis of this proposal is that the MTL houses conjunctive representations of objects and events that are critical for both memory and perception. The conjunctions are critical for perception because they allow differentiation between two similar objects or events, and critical for memory because they shield against a constant stream of perceptual interference that would otherwise cause dramatic forgetting of an object or event. It is assumed that both MTL damage and normal aging compromise the conjunctive representations, forcing a person to rely upon the fragmental, feature-based representations that remain in sensory cortex. Feature-based representations are inadequate for differentiating similar items (leading to perceptual impairments) and highly susceptible to perceptual interference (leading to memory problems). The planned studies are expected to reveal that MTL structures contribute to any cognitive task, mnemonic or perceptual, whenever the task solution requires complex, conjunctive representations of stimuli. This representational approach to understanding cognition will be fully exploited by developing a formal, neuro-computationally plausible model that links neural representations to cognitive performance with an explicit mechanism.