Background & Summary

Electroencephalography (EEG) has long been a key tool for investigating brain activity related to somatosensory processing. By examining the multi-stage dynamics of Somatosensory Evoked Potentials (SEPs), researchers have explored the cortical sources and functional roles of various SEP components during tactile perception. Early SEP components, such as the N20 and P50, originate from the primary somatosensory cortex and are modulated by stimulus features such as intensity1 and duration2. In contrast, later components—such as the N140 and P3b—are more likely to reflect higher-order functions, including attention, awareness, reporting, and task relevance3,4,5,6,7,8,9. Within these late components, however, the possibility to distinguish between neural markers of sensory awareness and those reflecting post-perceptual processes has been challenged by confounds inherent in experimental designs, as the perception of near-threshold stimuli is typically assessed through explicit subject reports. Although true “no-report” conditions are difficult to implement10, recent intracortical studies have introduced experimental paradigms that minimize reporting biases by employing passive tactile stimulation above the individual sensory threshold11,12,13. These studies have proposed parieto-opercular sustained activity as a neural signature of somatosensory awareness—characterized by features such as late latency, long duration, and origin in higher-order somatosensory cortices—which are also shared by the Somatosensory Awareness Negativity (SAN, or N140) observed in EEG14. The P3b component, on the other hand, has been more consistently linked to task relevance and report generation, although its precise role remains to be fully clarified, and its relationship with sensory awareness is still unresolved8.

Our dataset is specifically designed to address this gap through an ad-hoc experimental procedure, with the aim of disentangling the role of report and task relevance from the potential correlates of somatosensory awareness. We collected high-density EEG (hd-EEG) data from a cohort of 30 healthy participants who received electrical somatosensory stimulation at their individual sensory and motor thresholds. Each participant completed two sequential experimental blocks. In the first “No-Report” block (NR), they passively received tactile stimulation with no further instructions. In the second “Report” block (R), they provided reports either about the tactile stimuli (Report Task-Relevant, R-TR) or about distractor stimuli (Report Task-Irrelevant, R-TI; see Methods).

This dataset makes a significant contribution to the ongoing debate on the neural correlates of conscious perception and the optimal procedures for their investigation. It enables an in-depth exploration of the role of the P3b component in report production and task relevance, while also shedding light on the interaction between task relevance and sensory awareness. In interpreting these dynamics, it is worth considering that differences between Report and No-Report conditions may also be accompanied by variations in arousal, which is not matched across our conditions. In addition, components such as the N140 and the P3b could reflect the contribution of multiple cortical sources15 during overlapping functional processes, thereby weakening a direct link to the activity of specific areas but, at the same time, broadening the opportunities for analysis and data exploitation.

Methods

Task

Individual motor and sensory threshold identification

Median nerve stimulation was delivered via a Digitimer DS7A to electrodes (pre-gelled disposable electrodes with a conductive fabric backing) on the subjects’ right wrist (both cathode and anode). The motor threshold (MT) was determined as the minimum amplitude eliciting a visible thumb twitch. The sensory threshold (ST) was identified using an adaptive staircase procedure: participants reported whether a stimulus occurred after one of two sounds. If they responded correctly three times, the amplitude decreased by 10% until errors occurred, then was fine-tuned with smaller adjustments (5%, 3%, 1%, and 0.5%). The final ST was calculated as the average of the last six adjustments16. Both MT and ST were both further increased by 10% (respectively MT + 10% and ST + 10%) to avoid habituation effects17.

Stimuli

The experiment included two blocks: No-Report (NR) and Report (R). In the NR condition, participants passively received two trains of 100 tactile stimuli (0.2 ms of duration, square waveform), spaced 1–2 seconds apart, at MT + 10% and ST + 10% separately to ensure detectability and reduce habituation. In the R block, 40 sequences of tactile stimuli were randomly administered at ST + 10% or MT + 10%, interwoven with visual stimuli, consisting of flashes (i.e. white screen image presented after a black one, duration ∼16 ms), and presented on a 24-inch screen. Two subjects (subject 2 and 5) completed 28 and 32 series of stimulation respectively. Participants were instructed to count either visual (report task-irrelevant, R-TI) or tactile (report task-relevant, R-TR) stimuli and select the correct response on a keyboard (with response options ranging from a to c). Participants were instructed to respond at their own pace using their left hand (i.e. the unstimulated hand), without any requirement to use specific fingers for the response. The experiments in Matlab using the Psychophysics Toolbox extensions18,19,20. The experimental procedure is consistent with21 and summarized in Fig. 1.

Fig. 1
Fig. 1
Full size image

Experimental Procedure. (A) Motor and sensory thresholds assessment: The motor threshold was determined as the minimum amplitude that elicited a noticeable thumb twitch. To identify the sensory threshold, participants underwent an adaptive staircase procedure. Initially, they listened to two sounds and were asked to indicate whether tactile stimulation occurred after the first or second sound. If they answered correctly three consecutive times, the stimulus amplitude was decreased by 10%. If an error occurred, or the participant could no longer perceive the stimulation, the amplitude was gradually adjusted by increasing it by 5% and then decreasing it by 3% after three correct responses. This process continued with finer adjustments (1% and 0.5%) for a total of six cycles. The sensory threshold was calculated by averaging the amplitude levels from the final steps of the procedure. (B) Scalp-EEG recording sessions: Participants underwent two separate sessions. In the first session, they passively perceived two trains of tactile stimuli delivered at sensory threshold + 10% and motor threshold + 10% (NR condition). In the second session, participants were presented with 40 unimodal stimulus trains (20 tactile and 20 visual), with 20 trains at ST + 10% and 20 at MT + 10%, presented randomly. Participants were instructed to count either visual stimuli (task-irrelevant, R-TI) or tactile stimuli (task-relevant, R-TR) based on on-screen instructions. After each train, three numbers with a letter appeared, and participants pressed the letter corresponding to the correct count.

Subjects

We recruited 30 right-handed participants (age: 24 ± 3, 11 males), all of whom were unaware of the study’s purpose. They had normal or corrected-to-normal vision and no history of psychiatric or neurological disorders. The study was approved by the local ethical committee (CNR Commission for Ethics and Integrity in Research, n. 0065527/2019) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from each participant before the experiment, including consent to the publication of anonymized data.

Data recordings

The EEG was continuously recorded at a sampling rate of 500 Hz (vertex reference) using the 128-channel Geodesic EEG System (Electrical Geodesics Inc.) and the HydroCel Geodesic Sensor Net, including 19 sensors (AgCl-coated electrodes) arranged in a geodesic pattern over the scalp at positions corresponding to the standard 10–20 system. Consistent positioning was achieved by aligning the Sensor Net with skull landmarks (nasion, vertex, and preauricular points). We used the Net Amps300 high-input impedance amplifier. Low-noise EEG data were obtained, guaranteeing sensor-skin impedances below 50 kΩ except for the reference one, which was kept below 10 kΩ. We excluded the outermost belt of electrodes of the sensor net, prone to show residual muscular artifacts, thus discarding 19 peripheral channels located on the cheeks and nape. The recordings were all performed inside a silent cabin which guarantees a noise reduction ≥25 dB for frequencies ≥125 Hz, as well as a magnetic shielding of ≥30 dB and electrical shielding ≥50 dB for frequencies ≥100 kHz.

Data preprocessing

EEG continuous data were preprocessed with EEGLAB22 v2021.1. Data were band-pass filtered between 0.1 and 100 Hz (i.e default zero-phase Hamming-windowed FIR filters) and bad channels were detected using an EEGLAB plug-in clean_rawdata23. In particular, channels meeting the following criteria were interpolated: a) flat channels for more than 5 seconds, b) channels with line noise exceeding four standard deviations relative to their own signal, based on the total channel signal and c) channels with a correlation with their neighbours lower than 0.8. After interpolation, data were epoched in an interval of [−400 ms, +800 ms] around the stimulus delivery.

Data Records

Both the raw (128-channel, without filtering or interpolation) and preprocessed EEG data are publicly available in the Open Science Framework (OSF) repository: (OSF, https://osf.io/hqkym)24. The preprocessed data are provided in a single zip file, whereas the raw data are split into two separate zip folders corresponding to subjects 1–15 and 16–30. Once decompressed, each dataset includes one folder for each subject, including NR and R conditions. The NR folder contains the epoched data in.mat format (channels × timepoints × trials, i.e. 110 × 600 × 100) for the stimuli delivered at ST + 10% and MT + 10%.The R folder contains 40 subfolders (expect for subjects 2 and 5 containing respectively 28 and 32 subfolders), whose name corresponds to the number of the sequence, the intensity of the stimulus (TACTSS + 10% and TACTSM + 10%), and the instruction given to the subjects (CV for the R-TI condition and CT for the R-TR condition, respectively; e.g., 1_ST_CV). Within each subfolder it is stored the 3-D matrix in.mat format containing the epoched data. Each folder contains the labels of the electrodes and their positions in.mat format (see Fig. 2).

Fig. 2
Fig. 2
Full size image

Data Organization and File Structure. For each released ZIP archive contains one folder per subject (n = 30). Within each subject folder are two condition-specific subfolders: No-Report (NR) and Report (R). The NR folder contains epoched data saved as.mat files (dimensions: channels × time points × trials, e.g., 110 × 600 × 100) for stimuli at ST + 10% and MT + 10%. The R folder: Contains 40 trial subfolders (except for subjects 2 and 5, which have 28 and 32 subfolders, respectively). Each subfolder is named using the trial sequence number, stimulus intensity (TACTST + 10% or TACTSM + 10%), and task instruction (CV for R-TI, CT for R-TR), for example 01_ST_CV. Inside each train folder you will find a 3D.mat matrix of epoched EEG data (channels × time points × trials) and a.mat file listing electrode labels and their 3D positions.

Technical Validation

Before validation, we prepared the data by removing line noise (48–52 Hz), applying band-pass filtering (2–100 Hz), re-referencing to a common average, and visually inspecting the data to remove corrupted trials. Following these steps, we extracted Independent Components (ICs) using the runICA algorithm (EEGLAB v2021.0), retaining components that explained 99% of the data variance. Artifactual ICs were identified using the ICLabel toolbox25 and subsequently verified through visual inspection. Subsequently, the raw EEG was band-pass filtered ([0.1, 30] Hz) and segmented around median nerve stimulation ([−100, 500] ms). Previously identified ICA weights were applied to these data, and bad components were removed. A final visual inspection was performed to reject possible remaining bad trials. Here, we show the SEP and the scalp topographic distribution, averaged across subjects for NR and R conditions, at 10% above the subjects’ individual motor threshold (Fig. 3).

Fig. 3
Fig. 3
Full size image

Technical Validation. Panel A shows scalp topographies averaged across subjects at different time intervals, grouped by stimulation threshold (ST + 10% or MT + 10%) and by condition (NR, R-TR, and R-TI). For the MT + 10% scalp topographies, a small inset displays the same maps using a different scale (−2 to +2 µV). Panel B shows SEPs for the six experimental conditions (violet: NR, green: R-TR, yellow: R-TI). SEPs (±SE) are averaged across the left parietal electrodes indicated in the inset. Panel C indicates the outermost belt of electrodes excluded from the analysis (in red) and the 19 electrodes positioned according to the 10–20 system (in green).

It is worth emphasizing that scalp activity at around 50 ms from the stimulus delivery is mainly expressed by contralateral centro-parietal electrodes, while N140 components are exhibited bilaterally, with a fronto-central topography. Finally, P300 has a central organization, in line with previous literature8,26,27,28,29.