The Role of Sleep in Insight and Generalization

 描述(申请人提供)：睡眠现在被认为在促进记忆的长期存储方面起着关键作用。文献中一些最耐人寻味的发现表明，睡眠不仅支持个人记忆的存储，还有助于揭示清醒时并不明显的经历之间的关系--它可以为这些经历提供新的“洞察力”。这种现象对我们如何理解我们的环境有着深远的影响，但睡眠是如何做到这一点的，这是一个谜。这项拟议中的研究将帮助我们理解睡眠如何能够促进我们对清醒时所经历的体验的洞察。来自神经网络计算模型的模拟表明，泛化的关键是以许多不同的顺序多次体验相关的信息--交织信息。我们的环境通常不会为我们提供这种交错接触新信息的机会。我们认为，睡眠可以提供一个机会，以交错的方式重播一天中的记忆，从而发现它们之间的关系。我们的第一个目标是进行一项实验，以检验这一新的假设，在这个实验中，人类参与者将了解三类相关物体的特征。在本培训的最后，参与者将接触到两个新对象，这两个对象具有跨越两个类别的特征，在类别之间创建了语义桥梁。为了测试通用性，我们将向参与者展示新的刺激，其中一些将具有跨越桥接类别的新特征配置，并要求他们评估它们看起来有多现实。我们预计，睡眠中桥刺激与其他刺激的交错将允许在两个测试期间睡觉的参与者在这一概括中有所改善，这样他们就更有可能感觉到具有桥接类别特征的新刺激是现实的。我们将用高密度脑电图仪记录睡眠多导睡眠图，使我们能够确定哪些睡眠阶段和睡眠的特定特征与改进的概括性有关。我们的第二个目标是扩展我们现有的睡眠中学习的计算模型，以纳入睡眠生理学，并展示睡眠的特定特征如何有助于提高泛化性能的模式。这项研究将帮助我们了解大脑中学习和记忆巩固的机制，这反过来可能导致对如何治疗伴有学习或记忆障碍的神经疾病的洞察。

 DESCRIPTION (provided by applicant): Sleep is now known to play a critical role in facilitating the long-term storage of memories. Some of the most intriguing findings in the literature suggest that sleep does not simply support the storage of individual memories, but can help to uncover relationships between experiences that were not apparent while awake - it can provide new "insight" into these experiences. This phenomenon has far-reaching implications for how we make sense of our environment, but it is a mystery how sleep accomplishes this. The proposed research will help us understand how sleep is able to promote insight into the experiences we have while awake. Simulations from neural network computational models suggest that the key to generalization is to experience related pieces of information many times, in many different orders - to interleave the information. Our environment usually does not provide us with this kind of interleaved exposure to new information. We propose that sleep may provide an opportunity to replay memories from the day in an interleaved fashion, allowing for discovery of the relationships between them. Our first aim is to run an experiment that tests this novel hypothesis, in which human participants will learn about the features of three categories of related objects. At the very end of this training, participants will be exposed to two new objects that have features spanning two of the categories, creating a semantic bridge between the categories. To test generalization, we will then show participants novel stimuli, some of which will have new configurations of features spanning the bridged categories, and ask them to rate how realistic they seem. We expect that the interleaving of the bridge stimuli with the other stimuli during sleep will allow participants who sleep in between two test periods to improve in this generalization, such that they become more likely to feel that novel stimuli with features from the bridged categories are realistic. We will record sleep polysomnographically with high-density electroencephalography, allowing us to determine which sleep stages and specific features of sleep are associated with the improved generalization. Our second aim is to extend our existing computational model of learning during sleep to incorporate sleep physiology, and demonstrate how specific features of sleep might contribute to the pattern of improved generalization performance. This research will help us understand the mechanisms of learning and memory consolidation in the brain, which may in turn lead to insights into how to treat neurological disorders with learning or memory disturbances.