“…We modeled participants' trial-by-trial behavior in the stress and control conditions using a reinforcement-learning framework (Sutton & Barto, 1998) that has been extensively used to ACUTE STRESS AND REWARD LEARNING 13 investigate the behavioral and neural impact of pharmacological manipulations and genetic variations in the dopaminergic system in humans (Diederen et al, 2017;Doll et al, 2011;Frank & Fossella, 2011;Frank et al, 2007;Grogan et al, 2017;Rutledge et al, 2009). Importantly, the fitted model included separate learning rates for positive (α + ) and negative (α -) prediction errors, to account both for the differential firing of dopaminergic neurons for positive and negative prediction errors (Daw & Tobler, 2014;Maia & Frank, 2011;Maia & Conceição, 2017) and the differential effects of dopamine onto the plasticity of the corticostriatal synapses implicated in action-value learning (Frank & O'Reilly, 2006;Maia & Frank, 2017;Maia & Conceição, 2017;Möller & Bogacz, 2019). This model also included the inverse temperature parameter, β, which controls the stochasticity of choice selection, or the exploration/exploitation trade-off (Daw, 2011;Sutton & Barto, 1998), as detailed below.…”