A comprehensive multimodal MRI and EEG-TMS dataset on the impact of parietal cortex inhibition on decision-making under ambiguity

Abstract

In daily life, we often face decisions where potential outcomes are unclear, creating uncertainty. The complete or partial lack of knowledge regarding outcome probabilities—referred to as ambiguity—poses significant challenges for individuals. While recent studies have linked ambiguity in decision-making to neural activity in the parietal cortex, the precise role of this region and its interactions with other brain areas remain poorly understood.

Here, we present a comprehensive dataset on human decision-making under conditions of risk and ambiguity. The dataset includes two experimental sessions. The first one corresponds to the MRI setting, which includes structural MRI (T1- and T2-weighted images, n = 52), diffusion-weighted imaging (n = 45), and task-based functional MRI (n = 38). The second session corresponds to the EEG setting combined with inhibitory transcranial magnetic stimulation (TMS), targeting two parietal regions and the vertex (n = 24). TMS targets were defined from group-level fMRI activations obtained in the first session and then transformed to individual anatomy. Ten participants completed both fMRI and EEG-TMS recordings.

This dataset, partially analyzed in previous work, now includes newly acquired and previously unexamined data—such as diffusion-weighted imaging, T2-weighted images—and is fully organized according to the Brain Imaging Data Structure (BIDS) standard. It provides valuable opportunities to investigate the neurobiological decision-making mechanisms under ambiguity, focusing on the parietal cortex.

Specifications Table

Subject	Biology
Specific subject area	Neuroscience
Type of data	Raw data: Behaviours, RI, and EEG
Data collection	Data were collected in two sessions: 1. First session: Behavioral responses were recorded during the functional MRI acquisitions. Additionally, MRI and diffusion-weighted imaging (DWI) were acquired. 2. Second session: Behavioral responses were recorded alongside EEG measurements, during which transcranial magnetic stimulation (TMS) was delivered over the parietal cortex or vertex.
Data source location	Universidad del Desarrollo, Santiago de Chile, for EEG-TMS session. Clínica Alemana de Santiago, Santiago de Chile, for MRI session.
Data accessibility	Repository name: Probability Decision-making Task with Ambiguity Data identification number: 10.18112/openneuro.ds004917.v1.0.1 Direct URL to data: https://openneuro.org/datasets/ds004917
Related research article	Valdebenito-Oyarzo, G. et al. The parietal cortex has a causal role in ambiguity computations in humans. PLOS Biol. 22, e3002452 (2024). https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002452

1. Value of the Data

• •Comprehensive dataset for multiple brain measures, including structural MRI, functional MRI, and EEG-TMS recordings.
• •The behavioural task examines understudied aspects of decision-making (i.e., selecting among alternative options under risk, uncertainty, or ambiguity), providing valuable data for developing and validating novel behavioural models.
• •TMS experiments can be used to examine different effects on behaviour and brain activity.
• •The dataset enables robust between-subject and modality-specific analyses. Within-subject analyses are also feasible, particularly for investigating structure–function relationships (e.g., MRI–fMRI, MRI–EEG–TMS). However, cross-functional modality analyses are limited due to the smaller number of participants (n = 10) with fMRI and EEG–TMS data.
• •The dataset can be combined with others to investigate further brain signatures associated with risk preferences and other key behavioural features of decision-making, including relationships between structural (e.g., cortical thickness, white matter integrity) and behavioural measures.
• •This dataset may serve as a reference for comparative studies in clinical populations with decision-making deficits and inform artificial intelligence models of human choice under uncertainty.
• •This dataset can serve as a reference for comparative studies in clinical populations with decision-making deficits and can inform artificial intelligence models of human choice behaviour under uncertainty.
• •The dataset includes experimental data from a Latin American sample, contributing to increased MRI and cognitive neuroscience research diversity.

The behavioural task examines understudied aspects of decision-making (i.e., the cognitive process of selecting alternative options under risk, uncertainty, or ambiguity), providing valuable data for developing and validating novel behavioural models.

TMS experiments can be used to examine different effects on behaviour and brain activity, particularly in the context of

2. Background

Understanding how individuals navigate decision-making in uncertain environments has emerged as a crucial area of investigation. Ambiguity—incomplete or missing information about outcome probabilities—significantly challenges optimal decision-making [1]. The prevalence of ambiguity in ecological situations underscores the importance of resolving uncertainty and understanding the neurobiological mechanisms underlying its perception and management during decision-making. Life often demands decisions based on incomplete information, prompting a need for a deeper understanding of how individuals confront and resolve ambiguity [2]^.

Neurobiological research has made significant strides in implicating specific brain regions, such as the ventromedial prefrontal cortex [3], orbitofrontal cortex [4], ventral striatum, intraparietal sulcus, and dorsal posterior parietal cortex [5], in perceiving, comparing, and evaluating uncertainty during decision-making [6]. Recent research has linked the intraparietal sulcus (IPS) and dorsal posterior parietal cortex (PPC) to the perception and comparison of uncertainty levels during decision-making [7]. Additionally, activity in the frontal and parietal areas correlates with the degree of uncertainty and the updating process to reduce it [8]. Despite this evidence, the precise causal role and computational processing implemented in the parietal cortex and its connectivity in human decision-making remain unclear.

We present a multimodal human dataset comprising 53 participants (ages 18–31), collected across two experimental sessions. The first session included structural MRI (52 participants) and functional MRI recordings (38 participants; see details in Fig. 1). The second session included EEG–TMS recordings from 24 participants. The overlap of participants across modalities and sessions is illustrated in Fig. 1. fMRI and EEG–TMS recordings were acquired during uncertain decision-making tasks under conditions of risk and ambiguity (Fig. 2).

Fig. 1.

Overview of the experimental workflow and data validation process. The diagram illustrates the progression from participant recruitment and screening through data collection (MRI and EEG/TMS sessions), to validation and final deposition in BIDS format on OpenNeuro. The figure also specifies the number of participants per modality and the sequence of experimental appointments. The blue matrix indicated the overlap of participants between modalities.

Fig. 2.

(A) Task timeline. Probability decision-making task (PDM). Participants performed a probabilistic decision-making task in which they chose between two options, each defined by a reward magnitude and a probability of obtaining it. Choices (left or right) were visually represented by colored bars (indicating probability) and numeric values (indicating reward). The task included a no-ambiguity condition with complete probability information and an ambiguity condition (bottom panel) in which a grey mask partially obscured the probability information, simulating uncertainty. After a decision was made with a variable waiting time, feedback was provided. A green circle indicates the participant won, whereas a red circle signals that they did not. During the TMS-EEG session, a double TMS pulse is delivered −300 and −200 ms before feedback presentation, as represented in the grey rectangle over the superior right corner. (B) Graphical representation of the trial descriptors and the relation between variables. The figure was adapted from [9] illustrates the same experimental paradigm applied in the present dataset.

The aim is to provide comprehensive data to understand the neurobiological mechanisms underlying decision-making processes, offering the opportunity to test causal relationships between biological activity and behaviour. Our original article [9] demonstrated that parietal cortex activity tracks both the objective degree of ambiguity and a tendency to underestimate uncertainty. Moreover, disrupting parietal activity with TMS increased perceived uncertainty and reduced oscillatory responses in mid-cingulate and lateral frontal areas. Despite these previously published results from part of this dataset [9], it has several opportunities for reuse and extension. This dataset extends existing resources by providing multimodal neurobiological signals (fMRI, EEG, TMS) collected during identical behavioural tasks, allowing cross-modal analyses. Using a novel task design, it focuses on the understudied domain of decision-making under unavoidable uncertainty and ambiguity. Thus, the novel dataset will be a valuable resource for researchers in the field, serving as a reference for new experiments, testing new computational models, or replicating findings of other discovery data sets related to decision-making under ambiguity. Since our data is in the brain imaging data structure (BIDS) format [10,11] and is openly available on OpenNeuro.org [12], it can be easily combined with other datasets.

Additionally, it is expected to streamline the work for researchers interested in studying decision-making through empirical neuroimaging studies and behavioural experiments. Second, including diverse neural sources in our dataset allows us to integrate empirical information, enhancing our understanding and addressing critical gaps in the neurobiological mechanisms of decision-making processes. The current dataset extends beyond our original research article by including newly acquired and previously unexamined data, such as diffusion-weighted imaging (DWI), T2-weighted (T2w) images, and a BIDS (Brain Imaging Data Structure)-compliant organization. This expanded dataset enables data-driven and hypothesis-driven approaches to investigate decision-making under ambiguity. It supports diverse new analyses, for example, (1) cross-modal validation (oscillatory EEG activity vs. gray matter thickness in IPS/PPC), (2) structural-behavioural links (white matter connectivity vs. ambiguity aversion), (3) structural-TMS sensitivity (white matter connectivity vs. TMS effects), and (4) clinical benchmarking via our Latin American sample (N = 53) for cross-cultural comparisons. The data also informs real-world applications like financial modelling (e.g., modelling investor behaviour under market uncertainty), clinical assessments (obsessive-compulsive disorder and addiction), and AI development. Hence, this experimental dataset derived from healthy participants can be utilized to examine and contrast clinical populations that have documented impacts on cognitive function [[13], [14], [15], [16]]. These examples underscore the value of our data and its potential utility.

3. Data Description

The dataset follows the BIDS directory naming conventions and is available on OpenNeuro (accession number: ds004917). To access specific data, users can filter by session (MRI or EEG-TMS) or task (Probability Decision-Making, PDM) using BIDS-compatible tools (e.g., PyBIDS, FMRIPREP). The repository includes a full dataset description and validation metrics in the linked publication for further details. Each participant is assigned a unique identifier in the `sub-<IDe>` format, where `<ID>` ranges from `01` to `53`.

3.1. Folder structure

Each participant directory contains up to five subfolders, depending on available data:

• •anat: Anatomical images (T1w and T2w)
• •fmap: Field map images
• •func: Functional MRI (fMRI) data
• •dwi: Diffusion-weighted imaging (DWI) data
• •eeg: EEG recordings

3.2. File Naming Conventions

For fMRI and EEG data, filenames include the task label (Probability Decision-Making with Ambiguity, PDM) and follow this format: `sub-<ID>_task-pdm_<modality>`, where `<modality>` is `bold` (fMRI) or `eeg` (EEG). Each data file has a corresponding `.tsv` event file (e.g., `sub-<ID>_task-pdm_events.tsv`) listing task events and their timing.

3.3. Additional metadata

Behavioural data: Stored in *_task-pdm_events.tsv files within the func and eeg folders (included as pdm_events.zip in supplementary files).

EEG specifications: Electrode details: *_electrodes.tsv; channel information: *_channels.tsv; coordinate systems: *_coordsystem.json

Root directory files: participants.tsv: Demographic information (file also added as supplementary material, see below). task-pdm_events.json: complete documentation of task events and their structure

Further details about the dataset organisation are provided in the following sections.

4. Experimental Design, Materials, and Methods

4.1. Methods

Initially, we conducted a comprehensive literature review to identify the key experimental and neurobiological frameworks that guided our data collection. Simultaneously, participants engaged in decision-making tasks involving uncertainty.

4.2. Participants

As described in our previous work [9], fifty-three healthy participants between the ages of 18 and 31 took part in the experimental protocol (see Fig. 1 and Table 2), which was approved by the Ethics Committee of the Universidad del Desarrollo, Chile (Folio 2020–67). Participants were recruited by social media (non-probabilistic sampling method), and demographic information is summarized in Table 1, with additional details available to the participants.tsv file in the OpenNeuro repository and in the Supplementary Material.

Table 2.

Demographic Characteristics of Participants.

Parameter	MRI Session (n=52)	EEG-TMS Session (n=24)
Age (years)	24.05 ± 3.76	23.95 ± 3.75
Median Age	24	24
Gender (F/M)	27/26	13/11

Table 1.

Summary of Data Organization and File Types in the BIDS-Formatted Dataset on OpenNeuro. This table describes the available imaging and behavioral data modalities, their corresponding BIDS folder structure, file formats, and contents. A total of 10 participants completed both fMRI and EEG sessions, 37 completed both fMRI and DWI sessions, and 17 completed both EEG and DWI sessions.

Modality	BIDS Folder	File Formats	Contents	n
Structural MRI	anat/	.nii.gz, .json	T1-weighted and T2-weighted anatomical images	52
Field Map	fmap/	.nii.gz, .json	Field mapping images for distortion correction	48
Diffusion MRI	dwi/	.nii.gz, .bval, .bvec, .json	Diffusion-weighted imaging data and gradient directions	45
Functional MRI	func/	.nii.gz, .json, .tsv	fMRI data during PDM task (_task-pdm_bold.nii.gz), event files (_events.tsv)	38
EEG	eeg/	.eeg, .vhdr, .vmrk, .tsv, .json	EEG recordings during the PDM task and TMS, including electrode positions and impedance data	24
Behavioral	func/, eeg/	.tsv	Task performance data for each session (_task-pdm_events.tsv)	53

Of the sample, 52 participants underwent anatomical MRI, 45 completed diffusion tensor imaging (DTI), 38 participated in the fMRI session, and 24 participated in the EEG–TMS session. All participants had normal or corrected-to-normal vision, no color vision impairment, no history of neurological disorders, and no current psychiatric diagnosis or use of psychotropic medications. Written informed consent was obtained from all participants.

To minimize fatigue effects: (1) fMRI tasks were completed within the first 20 min of scanning, and (2) EEG–TMS sessions included 5-minute rest breaks between blocks for hydration and recovery. No participants reported fatigue-related issues during either session.

Experiments were conducted at the Social Neuroscience and Neuromodulation Laboratory of the Centro de Investigación en Complejidad Social (neuroCICS) at Universidad del Desarrollo, and at the Unidad de Imágenes Cuantitativas Avanzadas (UNICA) at Clínica Alemana de Santiago.

Session 1 corresponds to the structural MRI session (n = 52), and Session 2 corresponds to the EEG–TMS session (n = 24). Values are reported as mean ± standard deviation for age. Gender is noted as the number of female (F) and male (M) participants.

4.3. Set up data collection

4.3.1. Task

All participants completed the probabilistic decision-making (PDM) task [17] illustrated in Fig. 2. Participants had to choose between two probabilistic options with rewards in the PDM task. Each option was represented by a coloured bar (one on each side of the screen) associated with a probability of being selected, indicated by the length of the bar placed in the centre of the screen, and a reward, represented by a number placed above each coloured bar. These numbers represented real monetary incentives (see below). The options had random, complementary probabilities and rewards, with the option having the highest visible bar (highest probability) having the lowest reward and vice versa. After the participant made a selection (within approximately 3 to 6 s), the rewarded option was indicated with a green circle if they chose the rewarded option or with a red circle otherwise. Feedback (red or green circle) was presented for 3 s. If the participant chose the rewarded option, they received the associated reward; otherwise, they received no money.

Participants completed this task under two conditions: no ambiguity and ambiguity. In the no-ambiguity condition, participants saw the full extension of the colour bar, providing complete information related to the probability distribution of possible outcomes (i.e., risk or first-order uncertainty). In the ambiguity condition, a grey mask partially obscured the extension of both bars. The size of this mask varied from 40 % to 80 %, resulting in incomplete information about the probability distribution of possible outcomes (i.e., ambiguity or second-order uncertainty). The task was programmed and presented using Presentation Software (Neurobehavioral Systems™). All variables that describe the properties of a trial are shown in Fig. 2B and are stored for each experimental session in the *_pdm_task_events.tsv files when event 10 occurs (the beginning of each trial; see Fig. 2A, B, Table 3, and pdm_events.zip supplementary file).

Table 3.

Stimulus and Response Coding Schema. A summary of key variables extracted from the task-pdm_events.json metadata file provides detailed technical specifications for the experimental task. Variable: Field name (matching the JSON keys). Description: Purpose/context. Units/Levels: Numeric: Units (ms, s) or range (0–1). Categorical: Code → Meaning (e.g., 31=Win). Binary: 0/1 flags (e.g., TMS sites). Key Notes: Time vars: Relative to experiment start (onset=0). See Fig. 2B and the pdm_events.zip supplementary file.

Variable	Description	Units/Levels
onset	Start time of the stimulus	seconds
duration	Duration of the code	milliseconds
value	Stimulus type code	9: Payoff display (both options) 10: Probability display (both options) 20: Start of waiting time for the result 31: Positive feedback (win) 30: Negative feedback (no win)
response	Participant’s choice	1: Left option 2: Right option
rt	Reaction time	milliseconds
good	Trial correctly displayed	1: Correct 0: Incorrect
a	Ambiguity level (hidden area proportion)	proportion [0, 0.8]
bias	Bias in hidden option distribution	proportion (0: left, 0.5: middle, 1: right)
p	True probability of the left option	proportion [0, 1] (right = 1–p)
pv	Visible probability of the left option (excluding the hidden area)	proportion [0, 1] (right = 1 – pv)
ronda	Trial/round number	count
TMSips	TMS applied to the intraparietal sulcus	0: No 1: Yes
TMSppc	TMS applied to the posterior parietal cortex	0: No 1: Yes
TMSvertex	TMS applied to the vertex	0: No 1: Yes
delta	Timing discrepancy (stimulus vs. EEG recording)	milliseconds
fixed	Recovered lost stimulus code (EEG recording)	0: Original 1: Recovered

In the fMRI experimental session, participants completed 40 trials: 20 for the no-ambiguity condition and 20 for the ambiguity condition, organised in 5-trial blocks. In the TMS-EEG experimental session, participants completed 240 trials in 10-trial blocks per condition (no-ambiguity and ambiguity). Each participant completed six runs of TMS stimulations, consisting of 2 runs of 40 trials per TMS condition (TMS interference on the PPC, IPS, and vertex, see below for further details).

4.3.1.1. Imaging acquisition

The participants underwent (i) a sagittal 3D anatomical MPRAGE T1-weighted imaging (repetition time [TR]/ echo time [TE]=2530/2.19 ms, inversion time [TI]=1100 ms, flip angle=7°; 1 × 1 × 1 mm3 voxels), (ii) a sagittal 3D anatomical SPC T2-weighted (TR/TE=3200/412 ms, flip angle=120°; echo train length [ETL]=258; 1 × 1 × 1 mm3 voxels), (iii) an axial 3D echo-planar imaging (EPI) (TR/TE=8600/95 ms, 2 × 2 × 2 mm3 voxels, flip angle = 90°) with diffusion gradients applied in 30 non-collinear directions and two optimised b factors (b1 = 0 and b2 = 1000 s/mm2) with three repetitions, and (iv) a functional image (fMRI) weighted echo-planner T2* (TR/TE=2390/35 ms, flip angle=90°, 3 × 3 × 3 mm voxels). All imaging was acquired on a 3T Siemens Skyra (Siemens AG, Erlangen, Germany) MR scanner with a gradient of 45 mT/m and a maximum slew rate of 200 mT/m/s.

4.3.2. Functional MRI data analyses

Participant volumes were coregistered to 2-mm standard imaging using the nonlinear algorithm implemented in FSL. We investigated brain activity related to feedback processing, specifically looking for win activity in the ventral striatum. We compared win vs. no-win feedback in a simple model to validate the data (see below).

4.3.3. EEG recordings

We used TMS-compatible EEG equipment (BrainAmp 64 DC, BrainProducts, http://www.brainproducts.com/). EEG was continuously acquired from 64 channels, a reference (FCz), and a ground. TMS-compatible sintered Ag/AgCl-pin electrodes were used. The signal was band-pass filtered from DC to 1000 Hz and digitised at a sampling rate of 5000 Hz. Skin/electrode impedance was maintained below five kΩ, and electrode impedances were re-tested during pauses to ensure stable values throughout the experiment. The positions of the EEG electrodes were estimated using the neuronavigation system employed for the TMS. This information is detailed for each recording in the *_electrodes.tsv and *_channels.tsv files in the eeg directory of each participant's folder.

4.3.4. EEG-TMS protocol

TMS was applied during task performance during the EEG session. During the recording blocks, participants were instructed to maintain central fixation and minimise eye blinks and movements. Double TMS pulses were delivered over the right IPS (TMSips, MNI [46, −44, 57]), the right PPC (TMSppc, MNI [14, −64, 56]), and the Vertex (TMSvertex, MNI [0, −29, 77]) using a 70 mm figure-of-eight TMS coil connected to a Mag and More Stimulator. TMS targets were defined from group-level fMRI activations obtained in 38 participants (see [9]) and then mapped to each subject’s native anatomy using non-linear transformations. The peak coordinates were transformed from MNI space into each subject’s native anatomy using inverse normalization (non-linear inverse coregistration, FSL algorithm, default parameters). A neuronavigation system was used to identify individual stimulation targets by mapping the nearest grey matter locations on the gyral crowns of each participant's native-space structural MRI. TMS coil positioning and orientation about brain x, y, and z axes (yaw, pitch, and roll) were optimised so that the electric field impacted perpendicular to the target region, maximising the induced current strength. This approach results in all subjects with approximately an angulation in a horizontal plane (yaw) regarding the interhemispheric fissure of 45° for the IPS and 0° for the PPC and the vertex. For each trial and both tasks, two consecutive single TMS pulses were delivered before the feedback presentation (−300 and −200 ms pre-stimulus onset) with an interpulse interval of 100 ms to interfere with target activity, with a 100 to 200 ms window that has been demonstrated to inhibit motor potential.

TMS intensity was fixed at 120 % of the individual resting motor threshold (TMS intensity ranging from 54 % to 78 % of the maximum machine power and a mean of 63 %). Each TMS session included six runs. In each run, 40 two-pulse TMS bursts were delivered trial by trial, leading to 80 pulses per run over a block duration of about 11 min. Pauses for a minimum of 5 min separate each run. Each TMS-EEG experiment thus contained 480 active TMS pulses (including those delivered at the vertex). The rationale for this block design was to obtain a balance that maximises the number of trials per condition while maintaining a single TMS-EEG session for each subject. This approach increased our statistical power through within-subject analysis. Two 5-minute EEG resting-state recordings were conducted before and after the six blocks. The entire experiment took approximately 180 min: 1 hour for setting up the EEG electrodes to ensure stable and adequate impedances, 1.5 h for recordings, and 30 min for electrode MRI localisation and finalisation. The TMS protocol consistently adhered to past and current safety guidelines regarding stimulation parameters (intensity, number of pulses, and ethical requirements).

4.4. Data availability

All data records are organised according to the BIDS (Brain Imaging Data Structure) standard (version 1.9.0) [10,11] and are publicly archived within the OpenNeuro open-access repository ds004917 (https://openneuro.org/datasets/ds004917), a free and open platform for validating and sharing BIDS-compliant MRI, PET, MEG, EEG, and iEEG data [12]. The data can be accessed through multiple methods:

• 1.Browser download: Direct download via the OpenNeuro web interface
• 2.
  Command-line tools:

  • ○
    Using OpenNeuro CLI: download– snapshot 1.0.1 ds004917 ds004917-download/
  • ○
    Using DataLad: datalad install https://github.com/OpenNeuroDatasets/ds004917.git
  • ○
    AWS S3: aws s3 sync –no-sign-request s3://openneuro.org/ds004917 ds004917-download/

Complete download documentation is available at: https://openneuro.org/datasets/ds004917/versions/1.0.1/download

4.5. Code availability

The present dataset does not include preprocessed data; however, we provide the preprocessing scripts used for this study in the following repositories: https://osf.io/zd3g7; https://github.com/neurocics/Martinez-Molina2024; https://github.com/neurocics/LAN_current; https://github.com/neurocics/Figueroa-Vargas_Navarrete_2025_Scientific_Reports; see also [9,18,19]. We recommend using the TESA toolbox for EEG-TMS analyses, which includes comprehensive documentation (https://nigelrogasch.github.io/TESA/). Additionally, we suggest this tutorial to investigate behavioural and neurobiological mechanisms [20].

4.6. Technical validation

We implement several analyses to ensure the technical validation of electroencephalographic data, just as we do with structural and functional neuroimaging data. For each data type, we present results for all participants. The results of these validation analyses are shown below.

4.6.1. Behavioural data

All participants completed the task. To ensure that all participants understood the task and did not choose randomly, we analysed their responses when one factor was neutral (i.e., equal for both choices). Therefore, no conflict was generated under the condition of no ambiguity, and one of the options was the optimal choice. For both factors, participants chose the rational option in 84 % of cases, significantly greater than the 50 % expected for random choice (Wilcoxon test, p < 0.0001). We then tested whether experiential conditioning produced the expected behavioral changes, testing participants' preferences for choosing the option with the highest reward versus the highest probability. In the no-ambiguity condition, participants preferred the option with the higher probability of reward over the one with the higher reward magnitude. This preference was significantly reduced under ambiguity. Specifically, in the behavioral fMRI experiment, the change in choice rate was statistically significant (n = 38, Δ rate = 0.14, Wilcoxon test, p = 8 × 10⁻⁵). A similar pattern was observed in the behavioral TMS experiment under Vertex stimulation (n = 24, Δ rate = 0.11, Wilcoxon test, p = 0.02). Additionally, we evaluated this rate between TMS conditions to validate that the TMS perturbation produced an observable behavioural effect. Active TMS targeting the parietal region further increased this effect (n = 24, delta rate = 0.20; difference in delta rate compared to Vertex = 0.09, Wilcoxon test, p = 0.01). To assess potential learning effects, we analysed choice consistency across trial blocks (no significant change in probability-based preferences under ambiguity, p = 0.14). All these results were further corroborated using specific cognitive computational models [9]. The parameter τᵢ, which represents ambiguity perception, correlated significantly with the Δ rate (ρ = −0.76, p = 3 × 10⁻⁷, n = 38) and also differentiated TMS-induced changes in decision-making. Specifically, the interaction between τᵢ and TMS condition showed a mean effect of 0.66 (high-density interval = [0.07, 0.88], p = 0.008; for further details, see [9]).

4.6.1.1. Electroencephalographic data

We visually inspected the EEG data, and the acceptable quality of the obtained EEG signals was confirmed. The mean level of impedance obtained during each EEG study is included in the dataset, confirming high EEG acquisition standards. Additionally, we carried out two analyses to check the quality of the data. We calculated the Fourier transform, looking for the 1/F decay of the signal (Fig. 3. A-B). Then, we calculated the wavelet transform to obtain the time-frequency chart for all subjects, contrasting positive and negative feedback and looking for the known theta power increases (Fig. 3. C).

Fig. 3.

Validation of EEG and fMRI data. (A) Power spectrum of the FCz electrode. Each line represents a participant, while the thick purple line represents the sample mean. The X-axis represents frequencies (Hz), and the Y-axis represents power density. (B) Electrode positions, with the FCz electrode highlighted in red. (C) Time-frequency chart related to the feedback presentations. Colours indicate power in decibels (dB) to contrast positive and negative feedback. The X-axis represents time (seconds) and the Y-axis represents frequencies (Hz). (D) Head movement for fMRI experiments. Each line represents a participant, and the legend indicates the axis. (E) Significant BOLD activity for the contrast win > no win for the feedback periods (p 〈 0.001, cluster-corrected, cluster threshold detection z 〉 3.1; yellow areas represent z > 4.5 for visualisation purposes).

4.6.1.2. Structural neuroimaging data

All structural data were examined by a radiologist who checked the quality of the data and was informed of any possible clinically relevant findings. All data included in this dataset were catalogued as normal.

4.6.1.3. Functional neuroimaging data

We applied two approaches to validating functional data. Firstly, we extracted the intra-task movement of all subjects, looking for relative movements <1 mm (Fig. 3. D). Then, to ensure that the data and the experimental setting could generate activity, we studied the well-known striatal activity when the subjects received rewards. For this, we contrasted the trials where subjects won against those where subjects did not win in the feedback period. We found the expected activity in the striatum (Fig. 3. E).

Limitations

While this multimodal dataset provides comprehensive measures of brain activity during decision-making, several constraints should be noted. First, the moderate sample size (N = 53) limits the capacity to capture age- or demography-related variations, reducing generalizability. Second, only 10 participants completed all experimental sessions (MRI and EEG), which (1) restricts robust within-subject multimodal analyses between these modalities and (2) diminishes the power for repeated-measures designs across sessions. Researchers should, therefore, prioritise between-subject comparisons or group-level analyses when examining cross-modal relationships. Future work could address these limitations through larger-scale replications or meta-analytic integration with complementary datasets.

Ethics Statement

All protocols and experiments were approved by the Ethics Committee of the Universidad del Desarrollo, Chile (Folio 2020–67). The Declaration of Helsinki followed all ethical regulations relevant to human research participants.

CRediT authorstatement

AF-V: Methodology, Validation, Formal analysis, Data Curation, Writing - Original Draft, Writing - Review & Editing, Project administration. GV-O: Investigation, Methodology, Validation, Formal analysis, Writing - Review & Editing. MM-M: Methodology, Validation, Writing - Review & Editing. FZ: Resources. PB: Conceptualization, Software, Methodology, Validation, Formal analysis, Data Curation, Writing - Original Draft, Writing - Review & Editing, Supervision, Funding acquisition.

Data Availability

• openNeurorobability Decision-making Task with ambiguity (Original data)