Figure 2.

Average decision bias. During the decision-making phase, subjects chose S2+ over S2− (mean S2+ bias 73.89% ± 3.7 SEM, t₍₂₉₎ = 6.4572, p < 0.001, g = 1.1482), which confirmed that conditioning during the preceding reward learning phase was successful. Subjects also showed memory guidance of decision making by choosing S1+ over S1− significantly more often than expected by chance (mean S1+ bias 64.96% ± 5.92 SEM, t₍₂₉₎ = 2.5282, p = 0.0172, g = 0.4616). Memory guidance generalized to new views of the same facial identities, as evidenced by subjects choosing S1*+ over S1*− (mean S1* bias 62.55% ± 5.88 SEM, t₍₂₉₎ = 2.1362, p = 0.0412, g = 0.3900) at about the same rate they chose S1+ over S1− (mean difference − 0.0240, t₍₂₉₎ = 0.8032, p = 0.4284, g = 0.0734). S1 and S1* decision biases are normalized by the corresponding S2 decision bias¹⁴. Error bars show the standard error of the mean. Dots indicate individual subject’s decision bias (n = 30).