PSU Sternberg EEG-ECG

Dataset Description

This dataset contains physiological recordings and working memory performance data from participants who completed a Sternberg Working Memory task under different levels of cognitive load. The dataset includes Electroencephalogram (EEG), Electrocardiogram (ECG), Heart Rate Variability (HRV) measures, and behavioral performance data. The study initially enrolled 52 healthy participants. Following preprocessing and data-quality assessment, a final cohort of 47 participants was retained for subsequent analyses and is included in this dataset.

Participants

The study was approved by the Human Research Ethics Committee of Prince of Songkla University, Thailand (PSU-HREC: 2024-074-1-1; approval date: 31 October 2024). A total of 52 healthy participants were initially enrolled, including 19 men and 33 women. All participants provided informed consent before data collection and completed a demographic questionnaire for preliminary screening.

To control factors that could influence physiological signals, experimental sessions were conducted between 1:00 p.m. and 5:00 p.m., approximately 1–2 hours after a meal. Participants were instructed to refrain from caffeine-containing beverages and alcohol for at least 6 hours before the experiment, avoid intense physical exercise for at least 24 hours, and obtain at least 6 hours of sleep on the night before the experiment. Following preprocessing and quality assessment, 47 participants were included in the final dataset.

Experimental Procedure

The experiment began with a resting-state recording. Participants sat in a relaxed position with their eyes closed for 5 minutes during the Eyes Closed (EC) condition, followed by 5 minutes with their eyes open while fixating on a point on a screen during the Eyes Open (EO) condition. After an approximately 1-minute break, participants performed the Sternberg Working Memory Task under Low Workload (LW) and High Workload (HW) conditions, with 60 trials in each condition. The Working Memory task lasted approximately 15 minutes and was followed by a 5-minute recovery period.

In each trial of the Sternberg Working Memory Task, uppercase English letters were presented sequentially. Each letter was displayed for 300 ms, followed by a blank screen for 200 ms.

In the LW condition, participants were required to memorize three letters, whereas in the HW condition they memorized five letters. After the final letter, an asterisk was displayed for 1,200 ms, followed by a probe letter. Participants were instructed to determine as quickly and accurately as possible whether the probe letter had appeared in the preceding memory set. The task consisted of 120 trials in total, comprising 60 LW trials and 60 HW trials. Within each workload condition, 50% of the trials were match trials and 50% were non-match trials.

Signal Acquisition

EEG signals were recorded using a Brain Products actiCHamp Plus system at a sampling rate of 2,500 Hz. ECG was recorded using electrodes placed on the left and right sides of the chest in a Lead I configuration at a sampling rate of 1,000 Hz. During acquisition, the ECG signal was band-pass filtered between 0.3 and 200 Hz. EEG was recorded according to the international 10–10 electrode placement system and referenced online to Fz. EOG signals were also recorded for subsequent ocular artifact correction.

EEG Preprocessing

After acquisition, EEG signals were downsampled from 2,500 Hz to 1,024 Hz, notch filtered at 50 Hz, and band-pass filtered between 1 and 100 Hz. Ocular correction was performed using the vertical and horizontal electrooculography signals (VEOG and HEOG), followed by inspection and artifact rejection. For the EO condition, artifact-free EEG data were segmented into 2-second epochs. For the Working Memory conditions, EEG recordings were separated into LW and HW trials. Epochs were initially extracted around probe onset, and only correctly answered trials that passed artifact rejection were retained and exported as 5-second epochs. Therefore, the EEG data in the LW and HW files do not represent continuous recordings of the entire task period. Instead, they consist of selected 5-second epochs concatenated sequentially for each participant. Similarly, the EO EEG data consist of artifact-free 2-second epochs concatenated in sequence.

EEG Epoch Selection and Data Availability

The analysis protocol targeted up to:

120 EO epochs × 2 seconds 48 LW epochs × 5 seconds 48 HW epochs × 5 seconds

These target values correspond to a nominal maximum of 240 seconds of EEG data per condition. At a sampling rate of 1,024 Hz, this corresponds to 245,760 samples per EEG channel per condition.

The released dataset preserves the available signal values rather than artificially filling unavailable samples. Consequently, some participant/channel combinations may contain fewer usable samples than the nominal target. Blank or unavailable samples indicate that no valid data are present at those positions. They should not be interpreted as zero and should not automatically be replaced with zero. No artificial padding was applied to make participant recordings equal in length.

Dataset Structure

The dataset consists of four main files:

EEG_ECG_EO.xlsx — EEG and ECG data during the Eyes Open (EO) condition EEG_ECG_LW.xlsx — EEG and ECG data during the Low Workload (LW) condition EEG_ECG_HW.xlsx — EEG and ECG data during the High Workload (HW) condition Subject_Data.xlsx — participant-level information, HRV measures, and Working Memory performance

EEG_ECG_EO.xlsx contains data from the Eyes Open condition, which was used as the resting baseline.

EEG_ECG_LW.xlsx and EEG_ECG_HW.xlsx contain data obtained during the Sternberg Working Memory Task under the Low Workload and High Workload conditions, respectively.

EEG and ECG Worksheets

All three EEG/ECG files contain the same 20 worksheets:

Fp1, F3, F7, C3, T7, Pz, P3, P7, O1, Oz, O2, P4, P8, Cz, C4, T8, F4, F8, Fp2, and ECG.

Each worksheet represents one recording channel.

Within each worksheet:

one column represents one participant; rows contain sequential signal samples; EEG epochs retained for each participant are concatenated sequentially.

Participants are identified using anonymized IDs ranging from 001 to 047. The same ID is used consistently across all files. For example, column 015 in the EO, LW, and HW signal files corresponds to the same participant as ID = 015 in Subject_Data.xlsx.

EEG data length

EEG was sampled at 1,024 Hz after preprocessing. The nominal target for each condition is 240 seconds, corresponding to 245,760 samples. However, the actual number of usable samples may be lower for some participant/channel combinations. Blank cells occurring after or within the available signal indicate unavailable data and are not equivalent to zero-valued EEG samples.

ECG data

The ECG worksheets contain ECG recordings for the same participants and experimental conditions. The fixed EEG epoch targets described above should not be interpreted as requiring the ECG worksheets to contain exactly 245,760 samples, because EEG and ECG were acquired and processed at different sampling rates and for different analytical purposes.

Participant-Level Data

Subject_Data.xlsx contains participant-level information, with one row corresponding to one participant. Participant IDs are identical to those used in the EEG/ECG files and range from 001 to 047. The file contains participant sex, HRV measures, and Working Memory performance measures for the LW and HW conditions.

Heart Rate Variability

HRV measures were derived from ECG recorded during the EO condition, which served as the resting baseline immediately before the Working Memory task. Artifact-free RR intervals over a 5-minute period were analyzed using Kubios HRV Scientific 4.2.0, with automatic beat correction and medium-level noise detection. The participant-level file contains standard HRV variables obtained from this analysis.

Working Memory Performance

Working Memory performance is provided separately for the LW and HW conditions.

Variable

Description

ID

Participant identifier, ranging from 001 to 047

Gender

Participant sex

LW_Correct

Number of correct responses in the Low Workload condition

LW_Incorrect

Number of incorrect responses in the Low Workload condition

LW_Percent_Correct

Percentage of correct responses in the Low Workload condition

HW_Correct

Number of correct responses in the High Workload condition

HW_Incorrect

Number of incorrect responses in the High Workload condition

HW_Percent_Correct

Percentage of correct responses in the High Workload condition

Working Memory accuracy was calculated as the number of correct responses divided by the total number of trials in the corresponding workload condition. In addition to these behavioral variables, Subject_Data.xlsx contains standard HRV variables derived from ECG recorded during the EO condition.

Linking Data Across Files

Participant ID serves as the common identifier for linking information across files. The same ID always refers to the same participant throughout the dataset. For example, EEG and ECG data from column 025 in:

EEG_ECG_EO.xlsx EEG_ECG_LW.xlsx EEG_ECG_HW.xlsx

can be directly linked to the HRV measures and Working Memory performance associated with ID = 025 in Subject_Data.xlsx.

Verbatim fromZenodo· Licence: cc-by-4.0