A Neural Signature Encoding Decisions under Perceptual Ambiguity

Experiments 2 and 3. A–D, Experiment 2: face judgment task with anger-disgust morphed emotions. E–H, Experiment 3: animal judgment task with cat-dog morphs. A, E, Task. A face (A) or an animal (E) was presented for 1 s followed by a question asking subjects to identify the facial emotion (angry or disgusted; A) or animal category (cat or dog; E). Faces and animals are not shown to scale. B, F, Group average of psychometric curves showing the proportion of trials judged as angry (B) or dog (F) as a function of morph levels. Shaded area denotes ±SEM across subjects. C, G, RT. Subjects judged unambiguous faces (C) or animals (G) faster than ambiguous faces or animals. Error bars denote one SEM across subjects. Paired t test between adjacent levels of ambiguity: *, p < 0.05; **, p < 0.01; n.s., not significant. D, H, ERP at the electrode Pz differentiated ambiguity levels. Both experiments showed a larger LPP for anchors and a smaller LPP for high ambiguity, consistent with the face judgment task with fear-happy morphed emotions. Gray shaded area denotes the LPP interval (400–700 ms after stimulus onset). The top magenta bars illustrate the points with significant difference across three ambiguity levels (one-way repeated-measure ANOVA, p < 0.05, corrected by false discovery rate for Q < 0.05). I, J, Comparison between experiments 2 and 3 on all LPP (average across all conditions). K, L, Comparison between experiments 2 and 3 on the difference in LPP (anchor minus high). I, K, ERP. Shaded areas denote ±SEM across subjects. Gray shaded area denotes the LPP interval. The top magenta bars illustrate the points with significant difference between the two tasks (paired t test, p < 0.05, corrected by false discovery rate for Q < 0.05). There was no significant difference in the LPP between the two tasks for both all LPP (I) and the difference in LPP (K), although the animal task had more negative ERP ∼200 ms for all LPP (I). J, L, Mean LPP amplitude. LPP amplitudes were averaged across the entire interval (400–700 ms after stimulus onset). Error bars denote one SEM across subjects. There was no significant difference between the two tasks for both all LPP (J; paired t test, p = 0.45) and the difference in LPP (L; p = 0.41).

Experiments 4–6. A–E, Experiment 4: gender judgment task. F–J, Experiment 5: emotion judgment task. K–O, Experiment 6: wealth judgment task. A, F, K, Task. A question prompt was presented for 500 ms, followed by the stimulus. Subjects were instructed to respond as quickly as possible, and the stimulus stayed on the screen until subjects responded. B, G, L, Confidence ratings (CR). Confidence ratings systematically varied as a function of stimulus ambiguity for the emotion judgment task but not for the gender judgment task nor wealth judgment task. C, H, M, RT. RT can be considered as an implicit measure of confidence, and had a similar pattern as confidence ratings. D, I, N, ERP at the electrode Pz. Gray shaded area denotes the LPP interval. E, J, O, Mean LPP amplitude. Error bars denote one SEM across subjects. Paired t test between adjacent levels of ambiguity: +, p < 0.1; *, p < 0.05; **, p < 0.01; ***, p < 0.001; n.s.: not significant. P, Mean confidence rating for each experiment. Q, Difference in confidence rating between anchor and high-ambiguity stimuli for each experiment. R, Mean RT for each experiment. S, Difference in RT between anchor and high-ambiguity stimuli for each experiment. T, Mean LPP averaged across all conditions for each experiment. U, Difference in LPP between anchor and high-ambiguity stimuli for each experiment. Error bars denote one SEM across subjects. Paired t test between subjects: +, p < 0.1; *, p < 0.05; **, p < 0.01; ***, p < 0.001; n.s., not significant.

Direct behavioral ratings of stimulus ambiguity and judgment difficulty. A–E, Ratings of stimulus ambiguity. Twenty-one raters rated the ambiguity of stimulus on a 1–10 scale. A, Gender judgment task. For each stimulus, raters were asked, “how ambiguous is it to judge the gender of this face model?” We first averaged across stimuli for each ambiguity level within each subject and then averaged across subjects. Error bars denote one SEM across raters. B, Emotion judgment task. Raters were asked, “how ambiguous is it to judge the emotion of this face model?” C, Wealth judgment task. Raters were asked, “how ambiguous is it to judge the wealth of this face model?” D, Mean ratings across tasks. E, Difference in ratings between anchor and high-ambiguity stimuli. F–J, Ratings of judgment difficulty. The same raters rated the judgment difficulty on a 1–10 scale. F, Gender judgment task. Raters were asked, “how difficult is it to judge the gender of this face model?” G, Emotion judgment task. Raters were asked, “how difficult is it to judge the emotion of this face model?” H, Wealth judgment task. Raters were asked, “how difficult is it to judge the wealth of this face model?” I, Mean ratings across tasks. J, Difference in ratings between anchor and high ambiguity stimuli. Paired t test between adjacent levels of ambiguity: +, p < 0.1; *, p < 0.05; **, p < 0.01; ***, p < 0.001; n.s., not significant.