When determining the neural correlates of decision making, a major difficulty is the non-stationarity of the brain's response. The evaluation of the same action should vary over time depending on the subject's choice and reward history. The speed at which such subjective evaluations changes should differ among subjects, or even in the same subject in different sessions. A recently emerging paradigm for coping with this difficulty is to estimate the time course of the internal variables of a mathematical model of decision making from each subject's sequence of stimuli, actions, and obtained...

When determining the neural correlates of decision making, a major difficulty is the non-stationarity of the brain's response. The evaluation of the same action should vary over time depending on the subject's choice and reward history. The speed at which such subjective evaluations changes should differ among subjects, or even in the same subject in different sessions. A recently emerging paradigm for coping with this difficulty is to estimate the time course of the internal variables of a mathematical model of decision making from each subject's sequence of stimuli, actions, and obtained rewards in each experimental session, and use that as a marker for detecting the neurons or brain regions implicated. This chapter reviews mathematical models describing the adaptive process of decision making, computational methods for estimating the model variables from observed data, and examples of applications of such model-based analysis to behavioural tests, neural recording, and functional brain imaging.