The Emory Multimodal Learning Test, ported to virtual reality

Cognitive tasks in the lab tend to look nothing like the situations they are meant to stand in for. A memory task is usually a sequence of images on a monitor, with a chin rest and a button box. That buys experimental control, and it costs something real: the thing being measured is a long way from remembering where you left your keys.

The Emory Multimodal Learning Test — a gamified memory paradigm built to assess aspects of memory that a standard epilepsy surgery evaluation rarely reaches, and which won Best Poster at the CTSA Program Meeting in 2025 — is being ported from a screen into a virtual environment, where a participant moves through a space and learns it the way people actually learn spaces. The point of moving it is not the novelty of the headset. It is that a headset already measures where someone is looking, how they are moving, and how their body is responding, all on the same clock as the task itself.

That makes the behavioural side of the work richer without adding equipment to a hospital room, and it makes the task portable — the same environment can run in a lab, at a bedside, or on a laptop, which matters when your participants are inpatients on a monitoring unit.

›

Methods and technical detail

What is being built

A virtual-reality port of an existing multimodal learning paradigm, with the task logic, the environment, and the data pipeline written to run either in a headset or in a desktop fallback mode, so the protocol does not depend on a patient tolerating a headset.

Signals

Behavioural responses and trial timing, gaze and fixation, head and controller pose, and physiological channels from integrated biosensing. Every stream is written with a shared timebase so that a fixation, a heart-rate change, and a response can be placed against the same trial without post-hoc alignment guesswork.

Engineering constraints

Three constraints shaped the build. Sessions have to survive an interruption, because a clinical unit interrupts. Data has to land in a format the existing Python analysis stack already reads, rather than a bespoke export nobody else can open. And the whole thing has to run without a network, on hardware that can be wheeled into a room.

Why this one is here

Design plan §5.2 asks that at least one project be framed around what was built rather than what was found. This is that project: it is an instrument, and its output is a platform other studies can run on.