Dabas, A. | Search bioRxiv

bioRxiv2024

Learning from imagined experiences via an endogenous prediction error

Experiences shape preferences. This is particularly the case when they deviate from our expectations and thus elicit prediction errors. Here we show that prediction errors do not only occur in response to actual events - they also arise endogenously in response to merely imagined events. Specifically, we show that people acquire a preference for acquaintances as they imagine interacting with them in unexpectedly pleasant situations. This learning can best be accounted for by a computational model that calculates prediction errors based on these rewarding experiences. Using functional MRI, we show that the prediction error is mediated via striatal activity. This activity, in turn, seems to update preferences about the individuals by updating their cortical representations. Our findings demonstrate that imaginings can violate our own expectations and thus drive endogenous learning by coopting a neural system that implements reinforcement learning. They reveal fundamental principles how we acquire knowledge devoid of actual experiences.

neuroscience↗

bioRxiv2023

Concurrent representations of reinstated and transformed memories and their modulation by reward

An integral part of episodic retrieval is the reinstatement of neural activity that was present in the medial temporal lobe during encoding. However, neural memory representations do not remain static. Consolidation promotes the transformation of representations that are specific to individual episodes towards more generalized representations that reflect commonalities across episodes. Moreover, reward has been shown to augment episodic memory by enhancing consolidation, and it may accelerate the transformation of neural memory representations. We investigated this account with n=40 human participants using fMRI and an associative memory task. They encoded pictures of objects, each with one of four recurring scenes. Two scenes led to high reward, two led to low reward. The next day, participants encountered the objects again and retrieved the scenes from memory. Using representational similarity analysis, we demonstrate that retrieval is concurrently accompanied by the reinstatement of original neural representations and the activation of transformed, more generalized memories. Specifically, the parahippocampal cortex reinstates scene-specific patterns from the encoding phase during successful retrieval. In contrast, activity patterns in the medial prefrontal cortex and anterior hippocampus reflect transformed memories: They become more similar to each other for memories sharing the same scene, independent of memory success. Importantly, high reward enhances memory transformation in the anterior hippocampus. The brain thus maintains complementary memory representations: An episodic representation that resembles the original encoding pattern, and a generalized representation that summarizes commonalities across memories - in part for particularly valuable information.

neuroscience↗