Table 1.

Base models. Median [25–75 percentile] of subject-level parameter estimates in model space. See Appendix 5 for subject-level / top-level parameters in sampling space (i.e. untransformed). Absolute WAIC is reported at the top as the estimate of model evidence, where a smaller WAIC indicates higher evidence.

DOI: http://dx.doi.org/10.7554/eLife.22169.012

	Base models
	M1	M2	M3a	M3b	M4
WAIC	71014	69038	67678	67602	66987
ρ	42.7 [19.3 79.8]	41.6 [18.7 72.4]	35.2 [15.8 66.4]	33.4 [13.9 59.8]	32.5 [14.9 56.4]
ε0	0.013 [0.008 0.059]	0.015 [0.008 0.054]	0.017 [0.009 0.064]	0.022 [0.010 0.070]	0.021 [0.010 0.063]
b		−0.25 [−0.45 0.04]	−0.25 [−0.46 0.04]	.01 [−0.33 0.27]	−0.03 [−0.29 0.19]
π			0.47 [0.02 1.00]		0.12 [−0.29 0.70]
ε rewarded Go(ε0+κ)				0.037 [0.016 0.122]	0.034 [0.016 0.109]
ε punished NoGo(ε0-κ)				0.006 [0.002 0.014]	0.008 [0.003 0.022]