A neural computation model for decision-making times

We introduce two new models for decision-making times for a two-choice decision task with no a priori bias. One of the models is the mean-field Curie–Weiss model of neural computation, and the other is based on dynamics near an unstable equilibrium under a small noise perturbation. As in the existing literature, we interpret exit times as reaction times and show that our models lead to a specific shape of the exit time distributions in the vanishing noise limit. We test the distribution shape against experimental data and show that for almost 90% of the participants, reaction times are described well by the model. Among the features of our model are: the dependence of the exit distribution only on two parameters, the elegance of rigorous mathematical analysis, and the microscopic nature of the noise.

► A mean-field model of neuronal computation used to model decision-making times.
► Concise formula for exit distribution. Only 2 parameters: shift and scaling.
► A new diffusion model for decision-making times with no a priori bias, same results.
► Features: unstable equilibrium as initial indecisive state, smallness of noise.
► Agreement with experiment: the theoretical distribution is valid for 90% of participants.