Table 1.

Bayesian model variants and their parameters

Stage	Model
	Perceptual insight “μlikelihood”	“prior”	“μlikelihood + prior”	Late compensation “kchoice”	“μlikelihood + kchoice”	“kchoice confidence”
Early	μencoding × 2 (fixed) σencoding × 4	μencoding × 2 (fixed) σencoding × 4	μencoding × 2 (fixed) σencoding × 4	μencoding × 2 (fixed) σencoding × 4	μencoding × 2 (fixed) σencoding × 4	μencoding × 2 (fixed) σencoding × 4
Intermediate	μlikelihood × 4	prior × 4	μlikelihood × 4 prior × 4		μlikelihood × 4
Late	kconfidence × 4	kconfidence × 4	kconfidence × 4	kconfidence × 4 kchoice × 4	kconfidence × 4 kchoice × 4	kconfidence × 4 kchoice × 4

The cells show the model names and the parameters of each of the model variants, as well as their associated stages. × 4 means one parameter for each condition: No-Adapt-See, No-Adapt-Believe, Adapt-See and Adapt-Believe. × 2 (fixed) means that the μ_encoding parameters are fixed to the values fitted from the psychometric curves for No-Adapt-See and Adapt-See. In the last model, k_{choice confidence}, the k_choice factor depicted in Fig. 1 is allowed to shift simultaneously the confidence curves (as if apart from the arrow depicted from k_choice to C there was another arrow from k_choice to q).