Wallis Lab
3. Neural computations underpinning the representation of cognitive maps
Lead Investigator: Zuzanna Balewski
Potential treatments: Depression, addiction, schizophrenia, Parkinson's
As we navigate through the world, we are often faced with decisions that involve comparing vastly different options. The neural circuits that allow us to make these decisions must be sufficiently flexible to adapt to novel scenarios. Reinforcement learning (RL) is a computational framework that formalizes how we learn to select actions to acquire reward and avoid punishment.
Orbitofrontal cortex (OFC) is an integral piece of this learning process. One hypothesis is that OFC constructs and updates a model of the environment, or a “cognitive map” of the structure of behavioral tasks. How this is implemented on the single-neuron level is not known.
OFC works in concert with other frontolimbic regions (e.g. dorsolateral prefrontal cortex, anterior cingulate cortex, caudate nucleus). We aim to better understand how computations across these areas influence each other on the single trial level.
We have developed experimental methods to record from a large number of neurons at the same time to overcome methodological difficulties in addressing these questions. Model-based RL is studied using re-evaluation manipulations, where subjects rapidly respond to changes in the task structure. We need a large sample size of neurons to study this fast learning on the single-trial level, especially since neural responses are heterogenous in frontal regions. These experiments would not be feasible with traditional neurophysiology methods, which record from only a handful of neurons at a time.
Lead Investigator: Zuzanna Balewski
Potential treatments: Depression, addiction, schizophrenia, Parkinson's
As we navigate through the world, we are often faced with decisions that involve comparing vastly different options. The neural circuits that allow us to make these decisions must be sufficiently flexible to adapt to novel scenarios. Reinforcement learning (RL) is a computational framework that formalizes how we learn to select actions to acquire reward and avoid punishment.
Orbitofrontal cortex (OFC) is an integral piece of this learning process. One hypothesis is that OFC constructs and updates a model of the environment, or a “cognitive map” of the structure of behavioral tasks. How this is implemented on the single-neuron level is not known.
OFC works in concert with other frontolimbic regions (e.g. dorsolateral prefrontal cortex, anterior cingulate cortex, caudate nucleus). We aim to better understand how computations across these areas influence each other on the single trial level.
We have developed experimental methods to record from a large number of neurons at the same time to overcome methodological difficulties in addressing these questions. Model-based RL is studied using re-evaluation manipulations, where subjects rapidly respond to changes in the task structure. We need a large sample size of neurons to study this fast learning on the single-trial level, especially since neural responses are heterogenous in frontal regions. These experiments would not be feasible with traditional neurophysiology methods, which record from only a handful of neurons at a time.