# VR-COL: EEG during an academic auditory sustained attention task in red vs. blue virtual environments

This dataset accompanies the article:

> G. Ávila-Muñoz, M. Á. Lopez-Gordo and M. Rodríguez-Álvarez, "Exploratory Study of Color Influence During an Academic Auditory Sustained Attention Test and the Implications for Educational Environments", *International Journal of Neural Systems* **36**(11) (2026) 2650040. https://doi.org/10.1142/S0129065726500401

Please cite the article above when using these data. Further protocol details are given in the preceding study: G. Ávila-Muñoz *et al.*, in *Artificial Intelligence for Neuroscience and Emotional Systems* (Springer, 2024), pp. 528–538, https://doi.org/10.1007/978-3-031-61140-7_50.

## Overview

University students performed five academic listening tests inside a virtual reality (VR) environment whose colour was either red or blue (between-subjects factor). Continuous 16-channel EEG was recorded throughout the session, together with a log of experimental events and the responses to all questionnaires.

- **Participants:** 34 students of the University of Granada, with normal or corrected-to-normal vision (Ishihara test) and no auditory impairments or cognitive disorders. Data are anonymised.
- **Ethics:** approved by the Research Ethics Committee of the University of Granada (3702/CEIH/2023). All participants gave written informed consent.
- **EEG:** Bitbrain Versatile EEG, 16 channels (AF3, AF4, F3, Fz, F4, FCz, C3, Cz, C4, P3, P4, PO7, POz, PO8, O1, O2; 10–20 system). Ground at AFz, reference at the right earlobe, sampling rate 256 Hz. Signals are provided raw (unfiltered).
- **VR:** Meta Quest Pro, with an environment developed in Unity 2022.3.5f1. Red = RGBA (255, 86, 83, 255); blue = RGBA (0, 248, 255, 255).
- **Task:** five C2-level listening tests from ELE (Spanish as a Foreign Language). Each lasts about 5 min (4:46–5:03) and has six chronologically ordered two-option questions.
- **Acquisition:** subjects 1–14, November–December 2023; subjects 15–35, April–May 2025.

**Relation to the published analysis.** The article analysed the 29 participants whose sessions took place in the morning. Subjects 20, 29, 33, 34 and 35 (16:00 sessions) are included here but were excluded from the article because they fell outside the predefined time window. Subject ID 8 is not part of the dataset.

## Experimental protocol

The protocol was conducted in Spanish. Phase names below are those used in `events_log.csv`.

1. `Calibration`: EEG cap and VR headset set-up, with real-time impedance monitoring.
2. `SubjQuest` (Pre-SQ): five-item subjective questionnaire in a white environment.
3. `ColorChange`: the environment switches to the assigned colour (red or blue).
4. `Immersion` (RS1): 3 min eyes-open resting state. An alarm sounded if the participant kept their eyes closed.
5. `Listening` ×5 (L1–L5): instructions, about 1 min to read the questions, then the audio. Questions were answered in VR via hand tracking while listening. Test order was randomised per participant.
6. `Immersion` (RS2): 3 min eyes-open resting state.
7. `SubjQuest` (Post-SQ).

## File structure

```
VR-COL/
├── correct_answers.csv
├── SubjectsInfo.csv
└── <subject_id>/
    ├── EEG_audios.csv
    ├── events_log.csv
    └── performance.csv
```

### `SubjectsInfo.csv`

One row per participant.

| Column | Description |
|---|---|
| `Subject` | Subject ID (matches the folder name) |
| `Time` | Session start time (local time) |
| `Gender`, `Age` | Gender; age in years |
| `Field of Study`, `Level`, `Course` | Academic field, degree level and year of study |
| `Color` | Colour of the virtual environment (`Red`/`Blue`) |
| `Order` | Presentation order of the tests: the *k*-th value is the ID of the test presented in position *k* (e.g. `3,1,2,4,5` = `Examen3`, `Examen1`, `Examen2`, `Examen4`, `Examen5`) |
| `Test 1`–`Test 5` | Number of correct answers (0–6) in the test presented in position 1–5 (L1–L5). Use `Order` to map positions to test IDs |
| `Percentage` | Overall percentage of correct answers (sum of `Test 1`–`Test 5` / 30 × 100) |
| `SQ1 1`–`SQ1 5` | Pre-SQ, items 1–5 (1–5 scale) |
| `SQ2 1`–`SQ2 5` | Post-SQ, items 1–5 (1–5 scale) |

The subjective questionnaire items (originally in Spanish) are:

1. What has your level of attention been so far?
2. What is your stress level right now?
3. What is your level of interest so far?
4. What has your level of mental effort exerted so far?
5. What is your level of relaxation?

### `EEG_audios.csv`

Continuous EEG for the whole session, one row per sample.

| Column | Description |
|---|---|
| `timestamp` | Acquisition timestamp (ns) provided by the Bitbrain Versatile recording software, computed from the uptime of the Windows host running the recording application |
| `date` | `timestamp` converted to date and time |
| `ch1-AF3` … `ch16-O2` | EEG signal per electrode (µV) |
| `n-audio` | Segment label: `1`–`5` = audio of `Examen1`–`Examen5`; `-1` = RS1; `-2` = RS2; `0` = unlabelled (calibration, questionnaires, instructions, etc.) |

The `n-audio` label was aligned with the audio playback through a digital synchronisation signal and cross-correlation, so it does not depend on the clocks of either the recording host or the VR application. It allows direct segmentation. Each labelled segment has the same number of samples in every subject: tests take about 286–303 s, and RS1/RS2 take 48,923 samples (about 191 s) each.

### `events_log.csv`

Event log produced by the VR application.

| Column | Description |
|---|---|
| `date` | Date and time of the event |
| `startup` | Time since application start (s) |
| `action` | Event type |
| `context` | Event argument (phase name, test ID, etc.) |

The relevant `action` values are listed below.

- `Start Phase`, `Ends Phase` and `Instructions Phase` (context = phase name).
- `Audio Listening Start` / `Audio Listening Finish` (context = test ID).
- `Audio Relax Start` / `Audio Relax Finish` (RS1/RS2).
- `Cuestionario seleccionado` ("test selected"; context = next test ID).
- `Manual Interaction` (experimenter actions: `Force Start`, `Play check sound`).
- `END`.

### `performance.csv`

One row per questionnaire, in presentation order.

| Column | Description |
|---|---|
| `date` | Submission date and time |
| `startup` | Time since application start (s) |
| `type` | `SubjQuest` (first row = Pre-SQ, last row = Post-SQ) or `Examen1`–`Examen5` |
| `q1`–`q6` | Responses. For `SubjQuest` (5 items), scale answer from 0 to 4. For `Examen*` (6 two-option items), `0`/`1` = selected option and `3` = no answer |

## Known issues

- **Timestamps.** In subjects 1–14, `timestamp`/`date` are assigned per data packet, so consecutive samples share the same value, and they carry no time-zone offset. In subjects 15–35, they are per sample and include the UTC offset (`+02:00`). In subject 16, `timestamp` is stored in floating-point notation. For precise timing, use the sample index divided by 256 Hz.
- **Subject 11.** The recording host clock was wrong, so the EEG `timestamp`/`date` (2023-11-24 04:35) do not match the event log (2023-12-15 13:10). The `n-audio` labels are unaffected and correct.

## Funding

This work was supported by:

- Grant PID2021-128529OA-I00, funded by MCIN/AEI/10.13039/501100011033 and by ERDF "A way of making Europe".
- Grant P21_00084, Junta de Andalucía.
- The EU H2020 project IMOCOe4.0 (101007311) and Spanish national funding (PCI2021-121925).

## Contact

Miguel Ángel Lopez-Gordo (corresponding author), University of Granada, malg@ugr.es
