The development of adaptive specificity in learning and memory

项目摘要 许多形式的精神病理学在青春期出现或达到顶峰，包括焦虑和 抑郁症与基于价值的学习和情景记忆中的过度概括有关。理解 基于价值的学习计算和记忆表征的特异性的发展变化可能 提供了对精神病理学发病高风险时期的深入了解。关键是，为了有效地引导 行为跨越不同的环境，学习和记忆系统应该灵活地适应环境， 当类似状态的奖励统计收敛时， 当需要更详细的表示来指导行为时的粒度。在成年期， 海马体和内侧前额叶皮层之间的相互作用支持了 个体范例以及在学习过程中引导概括的更抽象的表示。而 先前对基于价值的学习和情景记忆的研究表明，儿童在表征信息时， 特异性比成年人，没有研究已经研究了如何认知和神经机制，支持 灵活地适应价值学习计算和记忆表示的需求， 随着发展，情况发生变化。拟议的工作将使用一种新的强化学习任务， 用先进的计算模型和神经成像方法来解决这个问题。具体目标1将 检查与年龄相关的变化，在上下文不变的偏见，更具体或更普遍的价值协会 在学习过程中，以及在他们的灵活性，以适应不同的情况下的奖励统计。具体目标2 我将探讨价值-学习关联的特异性如何影响信息记忆的特异性 在学习过程中遇到的，无论是立即和随着时间的推移。最后，具体目标3将研究如何奖励 学习环境的统计影响海马内表征的粒度， 内侧前额叶皮层的发展。我们假设正常的学习和记忆发展 可以通过增加神经认知表征的特异性来标记， 背景，并在他们的灵活性，以适应不同的学习环境的奖励统计。拟议 工作将阐明具体的价值观为基础的学习和记忆之间的机械关系， 发展，增强我们对出现神经认知障碍的潜在神经认知风险因素的理解。 与过度概括有关的精神病理学症状

PROJECT SUMMARY Many forms of psychopathology that emerge or peak in prevalence in adolescence, including anxiety and depression, are associated with overgenerality in value-based learning and episodic memory. Understanding developmental change in the specificity of value-based learning computations and memory representations may provide insight into periods of heightened risk for the onset of psychopathology. Critically, to effectively guide behavior across diverse contexts, learning and memory systems should flexibly adapt to the environment, representing information more generally when the reward statistics of similar states converge and with a higher degree of granularity when more detailed representations are needed to guide behavior. In adulthood, interactions between the hippocampus and medial prefrontal cortex support both specific representations of individual exemplars as well as more abstract representations that guide generalization during learning. While prior studies of value-based learning and episodic memory suggest that children represent information with less specificity than adults, no research has examined how the cognitive and neural mechanisms that support the flexible adaptation of value-learning computations and memory representations to the demands of varied contexts change over development. The proposed work will use a novel reinforcement learning task, coupled with advanced computational modeling and neuroimaging methods, to address this question. Specific Aim 1 will examine age-related change in context-invariant biases toward more specific or more general value associations during learning, as well as in their flexibility to adapt to the reward statistics of different contexts. Specific Aim 2 will probe how the specificity of value-learning associations influences the specificity of memory for information encountered during learning, both immediately and over time. Finally, Specific Aim 3 will examine how the reward statistics of the learning environment influence the granularity of representations within the hippocampus and medial prefrontal cortex across development. We hypothesize that normative learning and memory development may be marked both by increases in the specificity of neurocognitive representations that emerge across contexts, and in their flexibility to adapt to the reward statistics of varied learning environments. The proposed work will elucidate the mechanistic relations between specificity in value-based learning and in memory across development, enhancing our understanding of potential neurocognitive risk factors for the emergence of psychopathological symptoms associated with overgeneralization.