Fig 1.
Probability decision-making task.
(A) Task timeline. Participants were asked to decide between 2 options (left or right option). Each option had an associated reward indicated by a number. After a decision was made with a variable waiting time, feedback was provided. A green circle indicates that the participant won, whereas a red circle signals that he/she did not. In the ambiguity condition (bottom panel), a gray mask partially hides the color bars extension in the division. During the TMS-EEG session, a double TMS pulse is delivered −300 and −200 ms before feedback presentation, as represented in the gray rectangle over the superior right corner. (B) Schematic representation of the objective probabilities (Pobj) and assigned probabilities (Pas), and the relationship among the ambiguity probability (Pa) and the model parameters, τi and τb (see the Results and Materials and methods sections for details).
Fig 2.
(A) Rate of choice where individuals preferred the highest probability per conditions. (B) Rate of choice where subjects preferred the highest probability per degree of ambiguity. Black dots represent the rate per individual. Color rectangles indicate the number of individual decisions; red represents the maximum, and light gray represents the minimum account. The blue line represents the linear regression, and the gray area is the standard error. (C) Model fitting comparison using DIC, red, and LOOIC, green. (D) Posterior distribution of the key parameters for each model. Black dots represent the mean of the distribution, and black lines the 95% HDIs. The colored areas represent the complete posterior distribution. (E) Correlation between decision shift (difference between the rate of choices that the subject prefers the highest probability between condition, ambiguity less no-ambiguity conditions), and τi parameters. Red dots represent each subject. The solid blue line represents the linear regression, the dotted blue line the LOESS regression, and the gray area represents the standard error; * indicates p < 0.05, ** p < 0.01,***p < 0.001. See also S1 Fig and S1 Table. The data underlying this figure can be found at https://osf.io/zd3g7/. DIC, deviance information criterion; HDI, high-density interval; LOOIC, leave-one-out cross-validation information criterion.
Fig 3.
Brain activity during decision-making and feedback.
(A) Brain activity during decision-making. The reward magnitude (RW, yellow) of the chosen option is related to the activity in the ventral striatum (CTD Z = 3.1, cluster corrected p-value < 3e-8, for visualization threshold Z = 4). The degree of ambiguity (Pa, blue) correlates with the IPS and the PCC (cluster corrected p-value < 1e-10), among other areas. The probability assigned during ambiguity correlated with the right IPS (light blue, Pall[A > nA]:contrast Pall during ambiguity > Pall during no-ambiguity condition, corrected p-value = 0.0002). Underestimating uncertainty (red, Pall(τi = 0) calculated with τi = 0) correlated bilaterally with the IPS (corrected p-value = 0.0003) and the PCC (corrected p-value < 1e-10). Objective uncertainty (green, Pall(τi = 1) calculated with τi = 1) correlated with the IPS, the PCC, the somatosensory area in the left hemisphere, and the SMA (all corrected p-value < 1e-10). (B) Brain activity during feedback. The fact of winning (Win, yellow) correlated with ventral striatum activity (CTD Z = 3.1, cluster corrected p-value < 1e-5, for visualization threshold Z = 4). The probability prediction during ambiguity (uPE-Pall[A], red) correlated with activity in the MCC (CTD Z = 3.1, cluster corrected p-value < 1e-5). Contextual brain connectivity (PPI) (IPS seed from Pall[A > nA] contrast) showed that the ambiguity condition generates an increase in the correlation between the IPS and several brain regions, including the MCC and the ventral striatum. See also S2 Table. The data underlying this figure can be found at https://osf.io/zd3g7/. CTD, cluster threshold detection; IPS, intraparietal sulcus; MCC, midcingulate cortex; PPI, psychophysiological interaction.
Fig 4.
Behavioral result of interleaved EEG-TMS experiments.
(A) Target areas for TMS stimulation (right PPC x = 14, y = −64, z = 56; right IPS x = 46, y = −44, z = 57, and Scalp Vertex). (B) Decision shift (difference between the rate of choices subjects prefer with the highest probability between conditions, no-ambiguity less ambiguity) comparison between Vertex and Parietal TMS stimulation. (C) Posterior distribution effect of TMS stimulation on key parameters for cognitive models. Black dots represent the mean of the distribution, and black lines represent the 95% HDIs. The colored areas represent the complete posterior distribution. The data underlying this figure can be found at https://osf.io/zd3g7/. HDI, high-density interval; IPS, intraparietal sulcus; PPC, posterior parietal cortex.
Fig 5.
Oscillatory brain activity in frontal electrodes associated with unsigned prediction error during feedback.
(A) Time-frequency chart in frontal electrodes for the correlation between oscillatory power and unsigned prediction error is given by τi = 0 model (uPE-Pall(τi = 0)), τi = 1 (uPE-Pall(τi = 1)) model, and de join effect of 2 models for both Vertex TMS stimulation and the difference between vertex TMS and parietal TMS stimulation (TMS effect). The highlighted areas indicate time-frequency epochs showing significant modulation (without time-frequency a priori, whole-scalp cluster-based permutation test, CTD: p < 0.05 Wilcoxon test). Scalp topographies show oscillatory activity in the delta range. (B) Source estimation for delta activity correlated with unsigned prediction error given by τi = 0 model (uPE-Pall(τi = 0)) for Vertex and TMS effect. Sources that survive multiple comparison corrections are shown (FDR q < 0.05). The highlighted areas (green and red lines in the inserts) represent the coincident areas for EEG source estimation and BOLD activity for the fMRI experiment. All source results are shown in a high-resolution mesh only for visualization purposes. (C) A separate analysis of delta activity for TMS stimulation in the IPS and PPC and the differences between them. The data underlying this figure can be found at https://osf.io/zd3g7/. CTD, cluster threshold detection; FDR, false discovery rate; IPS, intraparietal sulcus; PPC, posterior parietal cortex.
Fig 6.
Oscillatory brain activity in frontal electrodes associated with the probability of the chosen option during the decision period.
(A) The time-frequency chart illustrates differences in frontal electrodes between the Ambiguity and No-ambiguity conditions in the correlation between oscillatory power and the probability of the chosen option (Pall[A > nA]). (B) The joint effect of 2 models, one for Vertex TMS stimulation and the other for the difference between Vertex TMS and parietal TMS stimulation (TMS effect). (C) A separate analysis of theta activity for TMS stimulation in IPS and PPC is conducted. (A, B) The highlighted areas indicate time-frequency epochs with significant modulation, determined using a cluster-based permutation test (CTD: p < 0.05, Wilcoxon test). (A–C) Scalp topographies display oscillatory activity in the theta range. Source estimation for theta activity is provided for Vertex and the TMS effect. Magenta lines on the cortex represent areas where BOLD activity correlates with Pa in the fMRI experiment (see Fig 3). All source results are presented on a high-resolution mesh for visualization purposes only. The data underlying this figure can be found at https://osf.io/zd3g7/. CTD, cluster threshold detection; IPS, intraparietal sulcus; PPC, posterior parietal cortex.