Smith Maverick E, Zacks Jeffrey M, Reagh Zachariah M
Washington University in St. Louis.
Truman State University.
Curr Opin Behav Sci. 2025 Oct;65. doi: 10.1016/j.cobeha.2025.101581. Epub 2025 Jul 24.
The human mind constructs and updates models of events during comprehension. Event models are multidimensional, multi-timescale, and structured. They enable prediction, shape memory formation, and facilitate action control. Event models may be updated incrementally by replacing feature information as it changes or globally by constructing an entirely new model; there is evidence for both mechanisms. Default mode network components, particularly medial prefrontal cortex, are thought to implement key event model functions, utilizing a temporally graded architecture in which regions with longer timescales perform more integration and abstraction. Two signatures of event model representations are phasic changes in overall activity at event boundaries and shifts in neural patterns at those boundaries. Current theories propose multiple control structures for event model updating, including monitoring the quality of event model-driven predictions. Event model updating during comprehension has important consequences not only for processing information in the moment, but also for forming long-term memories.
人类思维在理解过程中构建并更新事件模型。事件模型是多维、多时间尺度且结构化的。它们能够进行预测、塑造记忆形成并促进动作控制。事件模型可以通过在特征信息变化时替换它来进行增量更新，或者通过构建全新模型进行全局更新；两种机制都有证据支持。默认模式网络组件，尤其是内侧前额叶皮层，被认为执行关键的事件模型功能，利用一种时间分级架构，其中时间尺度较长的区域执行更多的整合和抽象。事件模型表征的两个特征是事件边界处整体活动的相位变化以及这些边界处神经模式的转变。当前理论提出了用于事件模型更新的多种控制结构，包括监测事件模型驱动预测的质量。理解过程中的事件模型更新不仅对当下处理信息有重要影响，而且对形成长期记忆也有重要影响。
