Figure 4

The role of stimulus-embedded valence and learning context in driving the generalization of explicit threat responses. (A) Shortly after the threat conditioning (CS+, 370 Hz; CS−, 784 Hz), participants (n = 12) were repeatedly administered with solely USs whose intensity was lower than in the conditioning phase. This procedure was aimed at devaluating the threatening value embedded into the CS+. Twenty-four hours later, they performed the 2AFC recognition task (NS+, 466 Hz; NS−, 1046 Hz). (B,C) In the explicit recognition task subjects successfully identified both the CS− and the CS+, thus indicating that devaluating the fearful outcome predicted by the CS+ limits the generalization of explicit responses. (D) Confidence judgments were higher for correct than incorrect responses for the CS− but not for the CS+. (E) In the control condition, participants (n = 12) were repeatedly exposed to USs whose intensity was kept constant from the conditioning phase. (F,G) Controls explicitly recognized the CS− but generalized the identification of the CS+ to the NS+, thus showing the threat-specific generalization pattern. (H) Confidence levels did not differ for both CSs recognition choices. (I) Twenty-four hours after the threat conditioning (CS+, 370 Hz; CS−, 784 Hz), subjects (n = 12) performed the explicit 2AFC test (NS+, 466 Hz; NS−, 1046 Hz) within a new physical context which was located inside of another building. (J,K) Participants exhibited a successful recognition of both the CS− and the CS+, thus indicating that encountering the CS+ in a different environment from the conditioning phase shifted the explicit threat-recognition pattern from generalization (old context) to discrimination (new context). (L) Participants were more confident when identifying than when misidentifying the CS− but not the CS+. *P < 0.05, **P < 0.01, ***P < 0.001. All data are mean and SEM. Wilcoxon signed-rank tests [(B,F,J)]; Mann–Whitney U tests [(D,H,L)].