A model is proposed in which stochastic choice results from noise in cognitive processing rather than random variation in preferences. The mental process used to make a choice is nonetheless optimal, subject to a constraint on available information-processing capacity that is partially motivated by neurophysiological evidence. The optimal information-constrained model is found to offer a better fit to experimental data on choice frequencies and reaction times than either a purely mechanical process model of choice (the drift-diffusion model) or an optimizing model with fewer constraints on feasible choice processes (the rational inattention model).Experimental studies often find that choice has an apparently random element (e.g., Hey, 1995;Ballinger and Wilcox, 1997;Cheremukhin, Popova and Tutino, 2011). It is plausible, especially in light of the similarity between the stochasticity of choice behavior and the the stochastic responses observed in perceptual discrimination tasks (e.g., Green and Swets, 1966), to think of this as reflecting noise in the cognitive processing of the choice situation, rather than variation in the subject's genuine preferences. Such a theory, however, makes specific predictions, even of a probabilistic nature, only under a specific hypothesis about the processing errors.Here we offer such a theory, in which the mental process used to make a choice is optimal, in the sense of maximizing expected utility, subject to a constraint on the information-processing capacity of the neural pathways that supply information about the current choice situation. In proposing such an information constraint, the model is in the spirit of Sims' (2010) theory of "rational inattention." However, unlike Sims' theory, the model proposed here imposes additional constraints on the class of choice