关键词: 记忆巩固, 海马, 图式, 再激活

Abstract:

Memory consolidation refers to the process of transforming newly acquired memories into a more stable and lasting form that can be retrieved over long periods. Traditionally, two forms of consolidation have been identified: synaptic consolidation and system consolidation. Synaptic consolidation involves the molecular processes that occur immediately after learning, stabilizing the synaptic and cellular changes produced by learning. System consolidation, which is the focus of this review, entails the reorganization of memory representations within brain networks over longer periods and most often refers to the transfer of memories from the hippocampus to the neocortex. Although system consolidation has traditionally been considered a slow process, taking years or even longer, recent research suggests that memories can consolidate very rapidly under certain circumstances.

This paper reviews studies that demonstrate rapid memory consolidation and identifies four factors that accelerate system consolidation. First, prior learning experiences significantly impact the consolidation of new information. Early studies on the expertise effect have shown that chess masters can remember the positions of pieces on a chessboard significantly better than novices. Recent research indicates that prior learning experiences not only affect memory encoding but also expedite the transfer of memories from the hippocampus to the cortex. Second, specific encoding strategies, such as fast mapping and unitization encoding, can promote rapid consolidation. Fast mapping involves rapidly assigning labels to new concepts based on existing categories, while unitization involves grouping information into a single coherent unit. Both strategies have been shown to benefit amnesia patients with severe hippocampal damage in memory tests, suggesting that these encoding methods may produce rapid cortical consolidation by integrating new material directly into memory networks without relying on the hippocampus. Third, sleep has been shown to lead to changes associated with consolidation after even a single period of sleep following new learning. These changes include alterations in the neural basis and the nature of the memory, making the memory less dependent on the hippocampus and more abstract. Fourth, memory retrieval can serve as a shortcut to memory consolidation. Similar to sleep, retrieval can also produce changes in the neural basis and the nature of the memory. Moreover, the memory advantage of retrieval compared to re-study is more likely to be observed after longer delays, suggesting that retrieval benefits the consolidation phase rather than the encoding phase.

Our analysis of these factors suggests that system consolidation can be accelerated through two pathways. The first pathway is schema-based learning. Prior learning experiences, particularly those forming well-established knowledge structures (schemas), significantly impact how we consolidate new information. Schemas act as an organizing scaffold, facilitating the integration of new information with existing knowledge. When new information aligns with existing schemas, the neural representations activated in the medial prefrontal region partially overlap with established representations. This overlap gives the new information an advantage in Hebbian learning (where neurons that fire together wire together), accelerating its integration and absorption. This could also explain why encoding methods like fast mapping and unitization, which rely on pre-existing knowledge, promote rapid consolidation. The second pathway is repeated reactivation. Reactivation has been considered a key factor driving consolidation. During sleep, memories are spontaneously replayed at a compressed timescale compared to waking experience (20-60 times faster), allowing memories to be reactivated multiple times in a short period. During retrieval, due to the imprecision of recall, a large amount of information semantically or episodically related to the target is activated, and these activations spread to the target. As a result, the target memory is repeatedly activated.

Future research could explore the following areas: First, the role of the hippocampus in cortical learning needs further investigation. In schema-based learning, researchers emphasize direct cortical learning, with more attention given to the ventromedial prefrontal cortex. However, it is unclear whether the hippocampus facilitates or guides this direct cortical learning process. Second, future research should investigate the significance of interference suppression during memory consolidation. The organizing scaffold provided by schemas can reduce interference between similar information, and there are mechanisms for suppressing interference during sleep and retrieval. Third, the potential downsides of rapid consolidation deserve exploration. While learning methods that promote rapid memory consolidation enhance memory storage, they may also have side effects that lead to false memories.

Key words: memory consolidation, hippocampus, schema, reactivation