Table 5.

The influence of previous prediction errors on choice as a function of stimulation Polarity and Condition^a

	Baseline		tDCS
	PE explore	PE exploit	PE explore	PE exploit
Anodal	0.68 ± 0.42	0.60 ± 1.91	0.68 ± 0.64	1.97 ± 2.56
Cathodal	0.67 ± 0.55	1.60 ± 2.53	1.03 ± 0.46	0.49 ± 1.78
Sham	0.84 ± 0.54	0.89 ± 2.23	0.95 ± 0.70	1.19 ± 3.13

^aThe values in this table represent the mean (± SD) coefficients of interest in Regression 9. Because of recoding of prediction errors (multiplication with −1), positive beta values indicate that the participant was more likely to explore following negative prediction errors. During stimulation, the anodal group's choice became significantly more influenced by the past negative prediction errors (PE) on exploitative choices (β₂ in Regression 9) such that the more stronger the recent negative prediction error, the more likely they were to explore. In contrast, the influence of PEs for the exploratory options (β₁ in Regression 9) on current choice did not vary significantly based on stimulation Polarity or Condition. This table corresponds to information presented in Figure 5 and tests of Hypothesis 3.