The Context Layer for Real-World EEG

What is Eventag?

The short version: Eventag is a wearable device that automatically recognises events in a participant's environment and writes them, as precisely timed markers, straight into an EEG recording. The context that makes real-world brain data interpretable is captured while the data is collected, not reconstructed by hand weeks later.

If you run EEG outside the lab, the rest of this explains where that fits in work you're already doing, and why the timing detail matters more than it first appears.

The problem mobile EEG inherited

EEG earns its interpretive power from a simple dependency: a neural response only means something if you know what the brain was responding to. In a conventional lab study that dependency is satisfied by design. The stimulus computer controls every image and sound and injects a marker into the data stream at the moment each is presented, historically through a parallel-port trigger, today often through Lab Streaming Layer with sub-millisecond accuracy. The experimenter always knows the ground truth, because the experimenter created it.

Mobile and ambulatory EEG removed the chair and the screen, which is the entire point: a brain studied while someone moves through an office, a ward, or a street yields findings with far more ecological validity than one studied staring at a monitor. But it also removed the scaffolding that supplied that ground truth. The amplifier still records the response to a car horn, a greeting, a transition from quiet into noise, yet nothing tells the recording that any of those things happened, or precisely when. The signal is intact, the cause is missing.

This is the annotation problem, or the context gap, and it is one of the reasons "real-world" studies so often quietly revert to scripted, semi-controlled protocols, which forfeits much of the reason for leaving the lab in the first place.

Why the existing workarounds don't fully close it

The field has reasonable workarounds. Each trades away something EEG cannot easily spare.

Participant button presses create a marker on demand, but they carry human reaction time and drift, and the simple reality that people miss events while mid-task or mid-conversation, and the dual-task burden of remembering to press can itself alter the very signal under study.

Post-hoc video coding, recording the surroundings and later hand-annotating the footage against the EEG timeline in software such as ELAN, is thorough but slow, easily consuming weeks of a PhD student's time, and the eye-matching of footage to a data timeline introduces its own alignment error.

Wearable cameras carry the same labour cost and add a substantial privacy liability. Inertial and motion sensors reliably tell you that the participant moved, but not why, a movement spike does not distinguish dodging a cyclist from reaching for a phone.

None of these is wrong. They are simply not built for the two things naturalistic EEG most needs at once: temporal precision and scale.

What Eventag does

Eventag is a compact device worn alongside the EEG system during a recording. On-device models recognise meaningful events in the participant's surroundings, a notable sound, a vocal interaction, a shift from one acoustic environment into another and emit a labelled, timestamped marker into the recording in real time, aligned to the moment the event occurred. It is designed to slot into the synchronisation pipeline a lab already runs.

There is no button to press and no footage to review afterwards. The participant behaves normally, the labelling happens continuously in the background. By the time the data reaches the analyst, the environmental context already sits inside it, ready to model.

Many of the most informative EEG responses unfold within a few hundred milliseconds of their trigger. A marker displaced by even tens of milliseconds can wash out the effect entirely.

Why timing is the whole game

This is the line between a convenience and a change in what is analysable. Many of the most informative EEG responses, event-related potentials, unfold within a few hundred milliseconds of their trigger. Across trials, temporal jitter in the event markers smears the averaged response; a marker displaced by even tens of milliseconds can place an analysis on the wrong segment of activity, or wash out the effect entirely.

This is exactly where reaction-time-bound button presses and eye-aligned video coding struggle. Eventag is built to inject markers with millisecond-level synchronisation, so that a detected event lines up tightly enough with the neural response to be treated as an event-related effect rather than an approximate one.

Privacy, by design

A device that monitors sound invites a fair question, and the architecture is meant to answer it deliberately rather than after the fact. Eventag is designed to process sound on the device, classify the event, and retain only the resulting label and timestamp, not a recording of what was said or heard. The researcher ends up with a structured, timestamped account of what category of event occurred and when, rather than an audio archive of a participant's day. For studies run in public or semi-public space, that is both easier to analyse and cleaner to defend before an ethics board or under GDPR.

It works with the EEG you already run

Eventag is amplifier-agnostic. It is built to operate alongside common research-grade EEG platforms and to integrate into an existing recording pipeline, rather than to replace hardware a lab has already invested in and validated. Adoption is meant to be additive, not a migration.

Where Eventag is today

To be straight with the people we most want to work with: Eventag is a functional prototype at an early technology-readiness stage. The core hardware and detection pipeline are working, it has been exercised across multiple EEG systems, and it has been validated in controlled setups, with naturalistic validation ongoing. We are actively looking to develop it together with research groups running mobile and ambulatory studies, partly because the people absorbing the manual annotation today are the people who should shape what replaces it.

Who this is for

Most directly, anyone running EEG beyond the traditional lab: principal investigators designing naturalistic studies, the PhD students, postdocs, and research assistants who currently carry the manual annotation, and BCI researchers who need neural data tied to real-world triggers rather than scripted stimuli.

The underlying problem generalises past EEG. Anyone working with field biosignals such as EMG, heart-rate variability, electrodermal activity, or mobile eye-tracking meets the same wall: a clean physiological signal with no record of what produced it. The same reasoning applies directly.