Users in a real-time fMRI-scanner learn to control virtual avatar in under one hour via real-time fMRI and “intrinsic manifold alignment”.

Researchers at Yale University have developed a non-invasive brain-computer interface (BCI) that enables users to control a virtual avatar using only signals from their brain — and notably, be familiar with the system in less than an hour.

Detailed in a 9 June 2026 paper in Nature Neuroscience, the work highlights how aligning BCI design with the brain’s natural organizational structure can significantly speed up user adaptation.

The project, led by Erica Busch and Nicholas Turk-Browne from Yale’s Department of Psychology, relied on real-time functional MRI (fMRI) to monitor and interpret brain activity. Participants (working in a real-time fMRI scanner in a neuroscience lab) were asked to navigate a virtual environment by intentionally modulating activity in brain regions associated with spatial awareness.

Central to the study is the concept of the brain’s “intrinsic manifold”, which refers to the underlying geometric patterns governing neural activity. The researchers used a data diffusion technique to map this structure and then aligned the interface controls accordingly.

When the system’s mapping of neural signals to avatar movement stayed consistent with this intrinsic geometry, participants were able to quickly regain control even after disruptions. In contrast, when the mapping deviated from the brain’s natural structure, users struggled and often failed to learn control altogether.

Implications for future neurotechnology

The findings suggest that the brain’s intrinsic geometry may act as a limiting framework for how humans acquire new cognitive skills, particularly in complex, higher-order tasks. For the BCI field, this points to a design shift: systems that adapt to the brain’s existing patterns rather than forcing users to learn arbitrary control schemes.

The work reinforces earlier findings from invasive BCI experiments in animal models, extending the concept of neural manifolds to human subjects using noninvasive imaging. This could have meaningful implications for the development of more intuitive neurotechnology in both medical and consumer applications.