Unit on Neural Computations in Learning

National Institute Of Mental Health

Investigators

Linked publications & trials

Abstract

In the last fiscal year, we have established the Unit on the Neural Computations in Learning and focused on building the computational theories that underpin our future empirical work on reward learning in rats and humans. We have been recruiting to open slots in the lab and have begun establishing new collaborations with our colleagues in the Intramural Research Program at NIMH. We have welcomed a postbaccalaureate fellow and a clinical biostatistication who will focus on online studies of human reward learning and have recruited a postdoctoral fellow who will work on neural recordings from animal models during associative learning tasks. In our computational work, we have made progress on understanding how complex information about the specific identity of outcomes is represented in reward predictions, finding that error signals in dopamine neurons contains information about the timing and flavor of rewards and not just their amount (as predicted by classic models of error-based learning in these brain circuits). This recent publication on multi-threaded reward predictions expands our understanding of how brain circuits build rich internal models of the environment as a part of reward learning. We have designed and are piloting a new learning task in rats to further test how the brain builds internal models to support temporally-precise reward predictions, and are acquiring new equipment to record neural activity across dopamine circuits, the striatum and prefrontal cortex in this task. In parallel, we have begun designing an analogous learning task for human participants in order to translate our neural findings to related human behaviors. We have begun setting up online systems for acquiring participant behavior in these online learning games and are refining experiment design in order to maximize the reliability and specificity of our behavioral measures and their relationship to self-reported measures of clinically-relevant symptoms in these large online cohorts.