Table 2.

Model parameters used for simulating learning. See Table 1 for abbreviations

Varying values (default value)
Learner parameters			Task parameters
${\sigma }_m^2$	${\sigma }_{\eta \ast }^2$†	$\alpha$‡	Target amplitude (units of ${\sigma }_m$)	Reward criterion (R(t) = 1 if: …)
1	1	0	0	Random:
				50% of trials
4	4	0.1	2	Adaptive (median):
				If EP < target: $median(E{P}_{t-1:t-10})$ ≤ EP ≤ target + 1
				If EP within fixed reward zone: target – 1 ≤ EP ≤ target + 1
				If EP > target: target – 1 ≤ EP ≤ $median(E{P}_{t-1:t-10})$
9	16	0.15	4	Adaptive (mean):
				If EP < target: ${\overline{\mathit{EP}}}_{t-1:t-10}$ ≤ EP ≤ target + 1
				If EP within fixed reward zone: target – 1 ≤ EP ≤ target + 1
				If EP > target: target – 1 ≤ EP ≤ ${\overline{\mathit{EP}}}_{t-1:t-10}$
16	36	0.2	6	Fixed:
				If EP within fixed reward zone: target – 1 ≤ EP ≤ target + 1
25	64	1	8	Fixed with lower target fraction (target fraction = 2):
				If EP within fixed reward zone: target – 1 ≤ EP ≤ target + 1

† The input ${\sigma }_{\eta \ast }^2$ is equal to the exploratory variance used following non-successful trials in the model of Cashaback19 and Therrien16. Using a variability control function (see Eq. 10, Fig. 3), it defines the two exploratory variances in the Therrien18 model, and a whole range of variances in the Dhawale19 model

‡ The values 0.1, 0.15 and 0.2 are not used in the Therrien16 and Therrien18 models, as their learning parameter is fixed at 1 (Eq. 4)