Back

Representation of Task Structure in Human Hippocampus

Mizrak, E.; Bouffard, N.; Libby, L. A.; Boorman, E.; Ranganath, C.

2019-10-07 neuroscience
10.1101/794305 bioRxiv
Show abstract

The hippocampus is thought to support episodic memory, or memory for specific events, but recent work also suggests that it may be involved in extracting structure from the world to guide future decisions and predictions. Recent evidence in rodents suggests that the hippocampus supports decision-making in cooperation with orbitofrontal cortex (OFC), possibly based on representation of task structure. Here, we used functional magnetic resonance imaging (fMRI) to test how the human hippocampus and OFC represents decision-relevant information extracted from previous experiences. Participants performed a task in which they learned values of different foods in grocery store contexts. The task was structured such that we could examine the degree to which neural representations could reflect generalized information about different task structures. Specifically, we manipulated whether a foods desirability varied with store context or not. Some foods were desirable in some store contexts and not in others; some foods were always desirable or undesirable. Participants needed to extract these two task sub-structures (i.e., context-determined vs. context-invariant) from the task structure. We examined hippocampal and OFC activity patterns during a decision-making task after participants were trained with the task structure. Our results showed that both hippocampus and OFC carried task structure information that was relevant to the decision outcomes. Hippocampal and lateral OFC representations differentiated between context-determined (deterministic) and context-invariant (probabilistic) task structures. The degree of this differentiation, an index of task structure representation, was highly correlated between hippocampus and lateral OFC. These results add to a mounting evidence suggesting that the hippocampus and OFC support decision-making by representing task relevant information to the decision outcomes after the task structure is learned.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.