Causal evidence of the role of ventral-tegmental area dopamine neurons in reinforcement learning
Learning based on rewarding outcomes of one’s behaviors is one of the most fundamental, and at the same time one of the most evolutionarily oldest, types of learning. Previous research has shed light on the potential underlying neural mechanisms, with non-human primate studies showing how dopaminergic neurons in ventral tegmental area and substantia nigra pars compacta increase their firing during unexpected rewards and decrease their firing during unexpected reward omission, with little change in activity upon occurrence of rewards that are well anticipated. These findings have strongly suggested that dopaminergic activity serves the purpose of providing reward prediction error signals that serve the purpose of teaching associations between rewards and preceding behaviors/events. Direct causal evidence for this mechanism has, however, been lacking due to methodological limitations.
In their recent study, Elizabeth Steinberg et al. (2013) optogenetically activated rat ventral tegmental area dopaminergic neurons to causally test the hypothesis that these neurons provide the prediction-error signal that guides associative learning. Specifically, the authors activated ventral-tegmental area dopamine neurons during both blocking (i.e., when there already is a fully reward-predicting stimulus present in the environment another temporally coinciding stimulus is not learned as a reward predictor under ordinary circumstances) and extinction (i.e., absence of a reward following reward-cued stimulus results in gradual un-learning). In both conditions, optogenetic activation of dopamine neurons that produced an artificial reward prediction error signal modulated learning in a manner consistent with the hypothesis that ventral-tegmental area dopamine neurons underlie associative learning.
These highly important findings provide yet another demonstration of the power offered by causal neuroscience methods in yielding evidence that confirms theoretical assumptions based on previously observed correlations between brain responses and behavioral effects. Indeed, the hypothesis that ventral-tegmental area dopaminergic neurons drive learning of reward-cue associations is not a new one, yet the definitive causal test of this hypothesis has been wanting. Therefore, the findings of Steinberg and her colleagues constitute a very important step forward in our understanding of the neural mechanisms that underlie reinforcement learning.
Reference: Steinberg EE, Keiflin R, Boivin JR, Witten IB, Deisseroth K, Janak PH. A causal link between prediction errors, dopamine neurons and learning. Nature Neuroscience (2013) e-publication ahead of print. http://dx.doi.org/10.1038/nn.3413