Hippocampal representations switch from errors to predictions during acquisition of predictive associations

Area of Science:

• Neuroscience
• Cognitive Science
• Psychology

Background:

• The hippocampus is crucial for learning and utilizing environmental statistical regularities to guide perception and generate predictions.
• A key question is how the hippocampus manages the balance between encoding novel predictive associations and retrieving established ones.

Purpose of the Study:

• To investigate the role of the hippocampus in acquiring and exploiting predictive associations.
• To determine how the hippocampus represents prediction errors versus predicted outcomes during learning.

Main Methods:

• Two high-resolution human functional Magnetic Resonance Imaging (fMRI) studies were conducted with 24 participants each.
• Participants underwent auditory cueing predicting visual shape identity with 75% validity.
• Multivoxel decoding analysis was employed to examine hippocampal activity patterns.

Main Results:

• Initially, the hippocampus preferentially represented unexpected visual shapes that violated learned cue regularities.
• As learning progressed, hippocampal activity shifted to represent the cue-predicted shape, irrespective of the actual visual stimulus presented.
• This demonstrates a dynamic shift in hippocampal representation from prediction errors to predictions.

Conclusions:

• The hippocampus is actively involved in both the acquisition and exploitation of predictive associations in the environment.
• Hippocampal function is modulated by learning state, prioritizing prediction errors during initial learning and shifting to represent predictions once learning is established.