Natural language processing captures memory content associated with shared neural patterns at encoding and retrieval

Area of Science:

• Neuroscience
• Cognitive Psychology
• Computational Linguistics

Background:

• Individuals interpret shared experiences uniquely, leading to distinct memories.
• The Default Mode Network (DMN) shows correlated activity across individuals during shared events, implying a role in processing high-level event features.
• The link between shared neural patterns in the DMN and memory content similarity remains underexplored.

Purpose of the Study:

• To investigate the correlation between memory similarity and intersubject spatial patterns within the DMN.
• To determine if shared neural activity during encoding and retrieval predicts the similarity of recalled event details.

Main Methods:

• Utilized a pre-existing dataset of 24 individuals watching and recounting two cartoon movies during fMRI scanning.
• Employed topic modeling to transform verbal recall into topic vectors, quantifying memory similarity.
• Analyzed functional magnetic resonance imaging (fMRI) data to assess intersubject activation patterns in the DMN.

Main Results:

• Higher similarity in recalled content correlated with stronger shared activation patterns in the DMN.
• This association was particularly evident in the posterior medial, medial prefrontal, and anterior temporal cortices.
• Findings demonstrate a neural basis for memory similarity linked to DMN activity.

Conclusions:

• The Default Mode Network (DMN) is crucial for encoding, interpreting, and recalling complex event features.
• Natural language processing tools effectively link memory representations to brain activity.
• Shared neural patterns in the DMN reflect and predict the similarity of memories formed from shared experiences.