People often perceive configurations rather than the elements they comprise, a bias that may emerge because configurations often predict outcomes. But how does the brain learn to associate configurations with outcomes and how does this learning differ from learning about individual elements? We combined behavior, reinforcement learning models, and functional imaging to understand how people learn to associate configurations of cues with outcomes. We found that configural learning depended on the relative predictive strength of elements versus configurations and was related to both the strength of BOLD activity and patterns of BOLD activity in the hippocampus. Configural learning was further related to functional connectivity between the hippocampus and nucleus accumbens. Moreover, configural learning was associated with flexible knowledge about associations and differential eye movements during choice. Together, this suggests that configural learning is associated with a distinct computational, cognitive, and neural profile that is well suited to support flexible and adaptive behavior. Duncan et al. investigate how people learn to predict outcomes using cue configurations. They show that configural learning is characterized by unique computational, behavioral, and neural signatures, including hippocampal activity, interactions between the hippocampus and striatum, and enhanced flexible knowledge.
Duncan, K., Doll, B. B., Daw, N. D., & Shohamy, D. (2018). More Than the Sum of Its Parts: A Role for the Hippocampus in Configural Reinforcement Learning. Neuron, 98(3), 645-657.e6. https://doi.org/10.1016/j.neuron.2018.03.042