Figure 1. The model.

(A) The decision making network consists of two populations of sensory neurons N_i, corresponding to the two targets, and two populations of premotor neurons M_i, corresponding to the two actions. Choice is determined by comparing the activities of the two populations of premotor neurons (see text). (B) The effect of the synaptic plasticity rule on synaptic efficacy. The decision making model was simulated in a concurrent VI reward schedule (see Materials and Methods) with equal baiting probabilities, and the efficacy of one of the synapses is plotted as a function of trial number. During the first 300 trials (blue), the synaptic efficacies evolved according to Eq. (2) with α = 0 and β = 1 (and thus γ = 0), resulting in small fluctuations of the efficacy around the initial conditions. A 10% mistuning of the mean subtraction after 300 trials (red arrow) to β = 0.9 (γ = 0.1) resulted in a linear divergence of the efficacy (red line). The addition of a linear decay term to the plasticity rule (Eq. (4) with ρ = 1) after 600 trials (black arrow) resulted in small fluctuations of the efficacy around 0.04 (black line).