Table 3. Model specification.

Models 1–4 were defined using different combinations of parameters for reward sensitivity and the mapping of expected values to the drift rate. A ‘static’ reward sensitivity means that pain increase and pain decrease were defined as –1 and 1, respectively (see Equation 4). A ‘scaled’ outcome sensitivity means that pain decrease was defined as ${\displaystyle -\rho }$ and pain decrease as $\rho$ (see Equation 5). A ‘linear’ drift rate mapping means that the drift rate ${\nu }_t$ for each trial was defined as the difference of expected values multiplied by $\nu$ (see Equation 7). A sigmoid mapping of the drift rate means that ${\nu }_t$ was defined by a sigmoid function bounded at $\pm {\nu }_{\mathrm{m}\mathrm{a}\mathrm{x}}$ . (see Equation 8 und Equation 9). All models included two learning rates (${\eta }_{+}$ , ${\eta }_{-}$), the non-decision time $\tau$, the boundary separation $\alpha$, and the a priori bias $\beta$.

Model	Outcome sensitivity	Drift rate mapping
Model 1	static	linear
Model 2	scaled	linear
Model 3	static	sigmoid
Model 4	scaled	sigmoid