Fluctuations in arousal reflect latent state transitions that facilitate behavioral optimization: EEG and eye tracking dataset
Data files
May 16, 2026 version files 31.44 GB
-
DryadUpload.zip
31.44 GB
-
README.md
3.27 KB
Abstract
People are often faced with surprising events that defy expectations. Such events elicit transient activity in the locus coeruleus/norepinephrine system and elevation of peripheral arousal markers, including pupil dilation and the EEG P300. Still, the function of these signals remains unclear. We propose that they reflect latent state transitions that dynamically control the mental context governing learning and perception. We tested and confirmed five preregistered predictions of this theory using EEG and pupil measurements collected from people performing a color prediction and reproduction task. The dataset uploaded here includes EEG and pupillometric data collected while participants performed this task.
Dataset DOI: 10.5061/dryad.wm37pvn36
Description of the data and file structure
We ran an eye-tracking and EEG study where participants were asked to do a task involving visual working memory and latent states.
Files and variables
DryadUpload.zip
compressed folder containing EEG and pupillometric data
Files formatted as "XXXX_ALP_FILT_STIM.mat."
Description: Matlab file containing cleaned EEG data from the time period 2 seconds before stimulus presentation to 2 seconds after stimulus presentation for each non-rejected trial.
Variables
- epochNumbers: list of trials kept in the cleaned dataset
- EEG: structure containing all EEG data with the following key fields
- data: EEG time series data represented in 3 dimensions as channels x timepoints x trials in microvolts
- chanlocs: channel labels and coordinates
- srate: sampling rate
- times: timestamps of each timepoint relative to stimulus onset in milliseconds
- nbchan: number of channels
- trials: number of epochs kept in the cleaned dataset
- icaweights: weights from ICA, components containing eye-related events have been removed
Files formatted as "XXXX.mat."
Description: Matlab file containing raw pupillometric data for each whole session.
Variables
- colheader: column headers for the data file, informs user what each column of "data" represents
- data: raw pupillometric data for whole session, columns for session where both eyes were tracked represent the following: sample time, left eye x position (in screen pixels), left eye y position (in screen pixels), left eye pupil area (in pixels), right eye x position (in screen pixels), right eye y position (in screen pixels), right eye pupil area (in pixels), and trigger number. Trigger numbers indicate the following events:
- 1: block start
- 2: instructions start
- 3: instructions end
- 4: stimulus on
- 5: stimulus off
- 6: reproduction prompt 1 on
- 7: reproduction 1 made
- 8: reproduction prompt 2 on
- 9: reproduction 2 made
- 10: prediction prompt 1 on
- 11: prediction 1 made
- 12: prediction prompt 2 on
- 13: prediction 2 made
- 14: session end
- 15: session start
- messages: contains both trigger and eye event (blink/saccade) times
- event: time of each trigger alongside trigger number labels
- eyeevent: information on blinks and saccades.
- othermessages: session information such as pupil thresholds, camera focal length, and screen coordinates
Code/software
Our code is available on our GitHub repository here. The software we used was MATLAB version 2023b with the EEGLAB package and helper functions included on our GitHub:
https://github.com/learning-memory-and-decision-lab/Li-Marble-2025
Human subjects data
All subjects gave explicit and informed consent to the publishing of de-identified data. Subject data present in the dataset is solely EEG and eye-tracking data with no PII attached. All data files are identified solely by an anonymous subject ID.