New research has uncovered how the brain prevents “catastrophic forgetting,” a phenomenon in which the creation of new memories disrupts or erases older ones. By studying mice, scientists have found that the brain processes new and old memories in distinct phases of non-REM sleep, as indicated by changes in pupil size. This separation allows for effective memory consolidation while preserving prior knowledge, a discovery with implications for both neuroscience and artificial intelligence.
During sleep, the brain replays recent experiences, helping them transition to long-term memory. In mice, researchers identified two distinct sub-phases of non-REM sleep. During the small-pupil phase, newly acquired memories are reactivated and consolidated, while the large-pupil phase is dedicated to processing older memories formed days earlier. This separation minimizes interference between new and old information.
The study involved equipping mice with brain electrodes and miniature cameras to track pupil dynamics during sleep. Mice were taught tasks such as navigating mazes for rewards. By interrupting specific sleep phases, scientists observed that memory consolidation was disrupted when neuronal activity in the small-pupil phase was suppressed, causing the mice to forget newly learned tasks. However, disrupting activity in the large-pupil phase affected older memories but left recent ones intact. This alternating system—new learning during one phase and old knowledge during another—provides a mechanism for maintaining memory integrity.
This dual-phase system is believed to be crucial for integrating new information while preserving older memories. The findings highlight an intermediate timescale in the brain that separates these processes, enabling the coexistence of new and established knowledge. This mechanism may have evolved to optimize memory storage over time and prevent cognitive overload.
Beyond neuroscience, the study offers valuable insights into artificial intelligence. Neural networks, which emulate brain function, often face challenges related to catastrophic forgetting. Understanding how the brain avoids this issue could inspire algorithms to improve the efficiency of AI systems.
The implications of this research extend beyond memory consolidation in mice. Scientists hypothesize that similar mechanisms exist in humans, as memory processes are evolutionarily conserved across species. The findings may lead to improved treatments for cognitive disorders or age-related memory decline by leveraging the brain’s natural memory management strategies.
The study also suggests that enhancing these sleep phases might improve memory and learning. This could involve therapies targeting specific sleep patterns to boost memory retention and prevent interference between old and new information.
Supported by prominent institutions, including the National Institutes of Health and the Sloan Foundation, this research provides a foundation for understanding the intricate role of sleep in cognitive function. It bridges gaps between basic neuroscience and potential clinical applications, offering a path toward improving memory retention in both humans and artificial intelligence systems.
