Fig. 6: MAGNet provides a framework reconciling conflicting results on the motivational role for dopamine.

a In MAGNet simulations, peripheral DA release after the reinforcement of a central location affects the probability that the e-mouse directs towards the central location (i.e. the goal-encoding attractor), compared to a baseline DA level. MAGNet simulations predict that such an increase in goal-directed locomotion depends on the distance between the initial state and the attractor state. b Experimental data show that VTA DA photostimulation in the Reward context affects the probability that mice direct toward the central location when mice are initially away from the central location (paired t-test ON versus OFF : T₍₁₀₎=4.20, p = 0.0018) but not when VTA DA photostimulation occurs in proximal locations (T₍₈₎=1.13, p = 0.29). c Simulation of Hamid et al. data (port-choice task) with the MAGNet reduced model. The action latency after DA stimulation decreases only for animals not already engaged in the task. This differential increase in action pace is accounted for by the reduced model where increased action probability at high DA only occurs when considering a distant (versus proximal) initial state (here, animal position) from the goal-encoding attractor situated at the rewarded location. d Simulation of Fischbach-Weisset et al. data (leverpress task) with the MAGNet reduced model, in which action probability decreases after VTA DA inhibition when the initial state is distant from the goal-encoding attractor. e Simulation of Lee et al. and Coddington & Dudman (head-fixed licking task) with the MAGNet reduced model, in which the action probability does not decrease after DA inhibition when the initial state is near or upon the goal-encoding attractor. Data are presented as mean ± s.e.m. See also source data file.