Reward prediction errors create event boundaries in memory.

Area of Science:

• Cognitive Neuroscience
• Memory Research
• Computational Neuroscience

Background:

• Reward prediction errors (RPEs) are crucial signals indicating discrepancies between expected and actual rewards.
• The impact of RPEs on memory formation and organization remains an active area of investigation.
• Existing models suggest RPEs might enhance memory encoding or signal environmental shifts for memory separation.

Purpose of the Study:

• To investigate how high-magnitude RPEs influence the encoding and organization of sequential memories.
• To determine if RPEs act as event boundaries, segmenting experiences into distinct latent contexts.
• To explore the effects of RPEs on recognition priming and temporal memory accuracy.

Main Methods:

• Four experiments involved participants learning reward associations with unique images.
• Recognition priming was assessed for sequential image pairs, with and without high-RPE events.
• Temporal memory was tested by evaluating participants' ability to order and estimate distances between events separated by RPEs.

Main Results:

• High-magnitude RPE events were preferentially recognized and maintained associative links with their predecessors.
• Significant evidence for RPE-modulated event boundaries was found, with reduced recognition priming across high-RPE events.
• Sequence memory accuracy was impaired for events separated by high-RPE events, unlike those separated by low-RPE events.

Conclusions:

• High-magnitude RPEs are strongly encoded, act as memory event boundaries, and disrupt the sequential integration of experiences.
• RPEs contribute to the formation of distinct latent contexts, influencing how subsequent information is processed and remembered.
• Findings support a modified Context Maintenance and Retrieval (CMR) model incorporating RPEs into memory encoding processes.