| The study of decision making poses new methodological challenges for systems neuroscience. Whereas the traditional approach linked neural activity to external variables that the experimenter directly observed and manipulated, many of the key elements that contribute to decisions are internal to the decider. Variables such as subjective value or subjective probability may be influenced by experimental conditions and manipulations, but can neither be directly measured nor precisely controlled. Pioneering work on the neural basis of decision circumvented this difficulty by studying behavior at steady state, where knowledge of the average state of these quantities was sufficient. More recently, a methodology has developed to confront the conundrum of internal decision variables more directly by leveraging quantitative behavioral models as intermediaries. We motivate and introduce this approach, wherein a model's dynamic internal variables are used as proxies for the unobservable decision variables that actually drive the decider's behavior.; We present a specific case study where this methodology was fruitfully applied to the study of value-based decision in a simple foraging task. We first describe a simple mechanistic behavioral model, based on Herrnstein's Matching Law, that captures the essential nature of the behavior, and show that neurons in parietal cortex covary with relative value as captured by the internal variables of this model. We then develop more advanced techniques for reconstructing the best behavioral model for these data, within the class of Linear-Nonlinear-Poisson models. Through rigorous predictive and generative testing, we quantify the success of this model in capturing behavior and show that it compares favorably to alternative models. Ultimately, we show that this refined model provides not only a better description of behavior, but better proxy variables for extracting valuerelated signals from neural firing rates in the parietal lobe.; Finally, we outline ongoing experiments to localize neural representations of value outside the parietal lobe using functional magnetic resonance imaging. The analysis and interpretation of these data are facilitated by this same methodology of constructing proxy variables based on intermediary behavioral models. We discuss preliminary findings, and future directions for this line of research. |
