Table 1 Parameters for base and best-fitting reinforcement learning models in the food task

M0: base (Rescorla-Wagner)

0.44

3.04

baseline

M1: reward sensitivity

0.43

3.39

1

ρ = 0.96

M2: asymmetric learning rate

α + = 0.81

α- = 0.33

2.91

1

(see α)

M3: choice persistence

0.56

2.22

1

φ = 0.44

M4: initial values

0.56

2.36

1

Q₀ = −0.85

asymmetric learning rate (M2)

+ initial values (M4)

α + = 0.76

α- = 0.50

2.33

2

(see α)

Q₀ = −0.67

asymmetric learning rate (M2)

+ initial values (M4)

+ group differences (by block)

α + = 0.72

α- = 0.50

2.37

8

(see α)

Q₀ = −0.67

(see Fig. 3 for difference parameters)