Déjà Vu as Overlap Between Dream Simulations and Waking Experience

This study proposes a novel predictive-processing framework to explain déjà vu as arising from partial overlaps between dream-generated internal simulations and waking experiences. Building on neuroscientific evidence that dreams function as offline generative models during REM sleep, the hypothesis suggests that when waking scenes share sensory, emotional, or contextual features with overnight dream fragments, they activate familiarity circuits, producing the characteristic feeling of "having been here before." To test this, the project will reanalyze existing datasets—such as dream reports, neuroimaging data, and recognition memory studies—using advanced NLP techniques to quantify content similarity, coupled with computational modeling to simulate how these overlaps generate familiarity signals. This approach aims to move beyond traditional theories of déjà vu rooted in memory misattribution or neural misfiring, offering a mechanistic, empirically testable explanation that bridges sleep, dreaming, and perceptual familiarity within a unified predictive-processing framework.

本研究提出一种全新的预测加工框架，用以解释既视感（déjà vu）的产生机制：其源于梦境生成的内部模拟与清醒状态下的感知体验之间存在部分重叠。本研究基于神经科学证据——快速眼动（REM）睡眠期间，梦境可作为离线生成模型——提出该假说：当清醒场景与夜间梦境片段在感官、情绪或情境特征上存在共通之处时，便会激活熟悉感通路，进而产生“曾经历过此场景”的典型感受。为验证该假说，本项目将对现有数据集（如梦境报告、神经影像学数据以及再认记忆研究数据）进行重新分析：采用先进的自然语言处理（NLP）技术量化内容相似度，并结合计算建模手段模拟上述重叠特征如何生成熟悉感信号。该研究思路旨在突破传统既视感理论的局限——传统理论多基于记忆错归因或神经误触发假说——并提供一种可实证检验的机制性解释，将睡眠、梦境与感知熟悉感整合至统一的预测加工框架之中。