Eq. (18) predicts that the total sum of the regression coefficient
for reward history is equal to the inverse temperature (here, β =
3.0) independent of the learning rate if the cut-off effect can be
neglected. Fig. 1(B) shows that this is the case with the exception of
the small learning/forgetting rate case where a negative deviation
from the predicted value (=3.0) was observed, particularly for a
short history length (e.g., Mr = 10). This exception is because of
the cutoff effect of reward history: for the small learning/forgetting
rate case, greater than 10 previous trials have an influence on the
current choice (Fig. 1(A)).Because these trials were not included in