Context-dependent rule learning in humans
Embargo Date
2022-02-10
OA Version
Citation
Abstract
Learning is a key process for developing flexible and efficient human behavior. This flexibility is driven in part by our ability to recognize relationships between events and the contexts that they occur in. Studies of context-dependent memory in humans have largely focused on retrieval of learned associations, and there have been fewer studies on learning. The goal of this project was to investigate how humans develop flexible rules for behavior and differences in learning across individuals.
Experiment 1 investigated how healthy, young adults (n = 34) inferred context-dependent and context-independent rules from experience, and how these inferences influenced learning success. The analysis utilized a Bayesian structural learning approach to infer the latent rules naive participants formed across time. The results suggested that poor rule learning performance was driven by incorrect inferences about rule structure during early learning. In unsuccessful learners, interference between rule types resulted in obstructed learning.
Experiment 2 used fMRI to investigate changes in brain activity as healthy, young adults (n = 30) progressively learned a series of context-dependent rules in the scanner. The results demonstrated that higher levels of feedback-related activity in regions of the prefrontal cortex (PFC) during learning were associated with better individual performance. Across the transition from learning to learned behavior, activity within the lateral and medial PFC shifted from being higher during feedback integration to higher during rule response production. After learning, shifts of context engaged regions across PFC, medial temporal lobe, and basal ganglia.
Experiment 3 used functional connectivity methods to explore connections between brain regions that were predictive of successful rule learning in healthy, young adults (n = 30). The results demonstrated that functional connectivity between lateral prefrontal cortex, parietal cortex, and basal ganglia were related to more successful learning across time.
These three experiments demonstrate that rule learning in naive participants is supported by behavioral and neural computations. Behaviorally, inferences about the causality of events have strong effects on successful and unsuccessful learning. Neurally, rule learning is supported by regions in prefrontal cortex, as well as medial temporal lobe and basal ganglia.