The Hippocampus Rapidly Integrates Sequence Representations During Novel Multistep Predictions.

Area of Science:

• Neuroscience
• Cognitive Psychology
• Computational Neuroscience

Background:

• The hippocampus is crucial for forming memories of temporally extended sequences and predicting future events.
• Effective prediction requires updating these memories when environmental contexts change.

Purpose of the Study:

• To investigate how new learning modifies hippocampal representations of extended sequences.
• To examine the relationship between this updating process and flexible future event prediction.

Main Methods:

• Human participants learned sequences of environments in virtual reality.
• Novel transitions were introduced to connect previously separate sequences.
• Functional magnetic resonance imaging (fMRI) was used to measure brain activity during prediction tasks.

Main Results:

• The hippocampus integrated representations of the newly connected sequence, showing increased similarity in activity patterns across trials.
• These integrated representations formed rapidly after learning and included elements from original sequences plus novel activity patterns.
• Hippocampal activity patterns predicted participants' ability to update their predictions behaviorally.

Conclusions:

• The hippocampus dynamically integrates new information into existing sequence representations.
• This integration facilitates adaptive behavioral responses by enabling flexible predictions in changing environments.
• Structured knowledge emerges through continuous learning and hippocampal representation updating.