Table 1.

Decision model comparison results

ModelParameters per subjectMean r2Median r2Model BICBIC- BICa-a
Prospect theory70.510.5013620883
Gain-loss learning110.520.51141601423
  • BIC measures are summed across the 30 subjects. Parameters per subject are across both placebo and l-DOPA sessions. All models include separate parameters for placebo and l-DOPA sessions that capture the weighting of losses relative to equivalent gains (loss aversion, λ), risk aversion in the gain domain (αgain), and risk aversion in the loss domain (αloss). All models except the dual-inverse-temperature model included a shared parameter across sessions for stochasticity in choice (inverse temperature, μ). The final column is the difference between the model BIC and BICa-a, the BIC for the approach–avoidance model. The approach–avoidance model was preferred (lower BIC) to the prospect theory model, dual-inverse-temperature model, and gain-loss learning model. The more complex approach-avoidance-mixed model included additional approach–avoidance parameters for the mixed trials and had the lowest BIC of the models tested.