Why trace and delay conditioning are sometimes (but not always) hippocampal dependent: a computational model

Area of Science:

• Neuroscience
• Computational Neuroscience
• Learning and Memory

Background:

• Classical conditioning research traditionally separates delay and trace conditioning based on hippocampal involvement.
• Delay conditioning (overlapping CS and US) is often considered hippocampal-independent, while trace conditioning (CS precedes US) is hippocampal-dependent.
• Recent findings challenge this dichotomy, showing exceptions for both conditioning types.

Purpose of the Study:

• To present a unified recurrent-network model explaining hippocampal function in temporal information processing during classical conditioning.
• To account for recent contradictory findings in delay and trace conditioning paradigms.
• To propose the hippocampus as a general predictor of stimulus states rather than being specialized for timing.

Main Methods:

• Extension of prior trial-level models of hippocampal function.
• Incorporation of adaptive recurrent collateral connections to represent intra-trial temporal information.
• Simulation of empirical data from various classical conditioning paradigms (delay, trace) with varied timing parameters.

Main Results:

• The model successfully explains why both delay and trace conditioning can be hippocampal-dependent or spared.
• It predicts critical hippocampal involvement in conditioning with long CS-US delays, not just trace intervals.
• Simulations accurately replicate data across diverse classical conditioning variants.

Conclusions:

• The hippocampus acts as a general-purpose system for learning and predicting stimulus states based on temporal information.
• Recurrent collateral connections are crucial for representing intra-trial temporal dynamics.
• The model offers novel empirical predictions for future research on hippocampal function and learning.