The long-term goal of research in our laboratory is to understand how the processes of choosing appropriate behaviors and evaluating the outcomes of chosen actions are implemented by the neural networks in the cerebral cortex and basal ganglia of the primate brain. 

Neuroeconomics & Decision Making

During decision making, many different types of information about reward and penalty, such as their quality, magnitude, probability and delay, must be estimated and combined appropriately.  For example, we investigate the process of determining the animal's preference for temporally delayed rewards.  Economic theories posit that the value of delayed reward is discounted as a function of reward delay according to a discount function.  We found that neurons in the prefrontal cortex often modulate their activity as a function of both reward magnitude and delay, and thus encode temporally discounted values.  We also study the role of the anterior cingulate cortex and other prefrontal cortical areas during decision making that involves gains and losses of token reinforcers. 

Sequence Learning

In our daily lives, reward is obtained only intermittently, and it is often necessary to produce a series of movements in an orderly and timely manner to maximize the rate of reward.  However, how the information necessary to choose the optimal sequence of movements is learned and selected is largely unknown.  In our experiments, we train non-human primates to choose between multiple sequences of movements, and seek to determine how the parameters related to the number of movements and reward delay are encoded in the medial and lateral prefrontal cortex. 

Spike Statistics

Single-neuron and local field potential (LFP) activity recorded in the brain often display complex temporal structures.  However, how much information is encoded uniquely in such temporal structures separately from the spike rates and average evoked potentials is still poorly understood.  Our goal is to understand whether specific types of information processed during decision making can be read out from the high-order statistics of spike trains and local field potential activity.