What happens in the brain when the floor starts to move?

Balance problems appear late in Parkinson’s disease, but the brain changes behind them are thought to begin much earlier. Researchers at Reykjavik University used a postural challenge — a platform that rocks like the sea, plus a virtual-reality headset — to provoke those changes and record them millisecond by millisecond. The brain patterns of people with Parkinson’s looked different throughout, even before the platform started to move.

Why balance is so hard to measure

Parkinson’s disease is usually diagnosed by watching for motor symptoms, which appear only after a great deal of damage has already been done to the brain. Difficulties with balance and posture come later still, even though the brain networks that govern them are thought to change long before. Researchers have therefore been searching for objective signals that could support earlier diagnosis and track how the disease progresses.

Electroencephalography, or EEG, is a promising candidate. It is cheap, painless and widely available, and it records brain activity thousands of times a second. The difficulty is that almost all EEG studies of Parkinson’s have been done with people sitting still. Balance problems, by their nature, do not show up when you are sitting still.

A laboratory that rocks like the sea

RU’s Institute of Biomedical and Neural Engineering has spent years building an answer to that problem: BioVRSea, a platform that moves in time with a virtual-reality seascape. Participants stand on it wearing VR goggles and a 64-channel EEG cap, with the amplifiers in a light backpack so they can move freely.

The protocol lasts five minutes and has four parts. First, a calm baseline on a still platform. Then 40 seconds of sea visuals with the platform still, so eyes and body disagree. Then two minutes of real movement, with waves building from 0.5 Hz to 3 Hz and up to two metres in amplitude, while participants hold a safety bar. Finally, the platform stops, but the visual waves continue.

Thirty people with early-stage Parkinson’s disease took part, alongside 26 healthy volunteers of similar age. The study was a collaboration between RU, Landspítali, the University of Iceland, Akureyri Hospital, Giorgio Di Lorenzo’s group at the University of Rome Tor Vergata, and MINDIG in Rennes.

What they found

Two brain patterns looked entirely different in Parkinson’s

Of the five patterns the team identified, three (A, B and C) looked much the same in both groups. Two did not. In the Parkinson’s group, patterns D and E bore almost no resemblance to those of the control group, or to the standard maps reported in the wider literature. The spatial similarity between the groups was 6.5% for D and less than 1% for E.

This is a striking difference, and the researchers treat it carefully. Because the two groups’ maps were not the same shape, comparing how long they lasted would mean comparing two different things. They therefore reported the difference in shape as a finding and excluded D and E from the rest of the analysis.

Two other patterns lingered

Among the patterns that were comparable, participants with Parkinson’s held patterns A and B for longer, and those patterns took up more of the recording, in every phase of the task. They did not appear more often; they simply lasted longer once they arrived.

The difference was there from the start

Perhaps the most interesting result is one the team did not expect. The differences barely changed between the calm baseline and the roughest waves. They were present throughout.

One likely explanation is that BioVRSea is demanding from the first second: simply standing still on the platform in a virtual seascape is already a balance task, so there is no genuinely easy phase to compare against. The authors are careful to say that their design cannot distinguish between a difference that is always there and one that appears the moment the task begins.

One sequence behaved differently

The brain’s patterns follow each other in sequences, and one sequence did shift with the task. Healthy participants became more likely to move from pattern A to pattern C as the protocol went on. Participants with Parkinson’s kept the same probability throughout, and the gap between the groups was widest once the platform had stopped, but the visual waves continued.

These findings open a new window on Parkinson’s research, paving the way for future studies combining postural tasks and EEG to assess early disease-related brain changes.

Paolo Gargiulo, Director of the Institute of Biomedical and Neural Engineering and Professor at Reykjavik University

What this does, and does not, mean

The authors are unusually explicit about the limits of their work, and it is worth repeating them. This is a pilot study with a modest sample. Everyone with Parkinson’s was tested while taking their medication, so the differences cannot be separated from the effects of the drugs. The study collected no measures of balance ability, symptom severity or cognition, so the brain patterns cannot be linked to anything a patient or clinician would notice.

The researchers therefore describe their results as candidate markers that require validation, not as a diagnostic test. What the study does establish is that this kind of measurement is possible at all: that meaningful, fine-grained brain dynamics can be recorded while a person is actively fighting to stay upright.

Why it matters

If brain activity during a balance task turns out to carry a consistent signature, it could one day give clinicians something objective to measure, both earlier in the disease and over time as it progresses. Reaching that point will need larger groups, clinical measurements alongside the EEG, patients tested both on and off medication, and repeat visits over the years.

It also matters as a demonstration of method. Most of what we know about the brain comes from people lying still in scanners or sitting quietly in chairs. Parkinson’s disease, however, makes itself felt when people stand, turn and walk. Bringing the laboratory to the wobble, rather than the other way round, is a step towards studying the brain as it actually works.

The research in brief

For researchers

EEG (64 channels, 4096 Hz, reduced to 52 channels after removal of muscle-contaminated sites) was downsampled to 512 Hz and band-pass filtered 2–20 Hz, cleaned with ICA (Picard/ICLabel) and re-referenced to the common average. Microstates were derived in MICROSTATELAB using modified k-means on GFP peaks, with a five-map solution selected by spatial similarity to meta-analytic templates and a nine-metric meta-criterion. Temporal parameters and adjusted transition probabilities were modelled with linear mixed-effects models (Group × Phase × Microstate, age as covariate, subject as random intercept), with FDR correction. Group main effects were significant for duration and coverage, but not occurrence, with Group × Microstate interactions across all three parameters; phase effects and higher-order interactions involving phase were not significant. Between-group temporal and transition analyses were restricted to maps A–C, since maps D and E were not topographically equivalent across groups.

Further reading

• Gelormini, C., Guerrini, L., Pescaglia, F., Maruotto, I., Aubonnet, R., Jónsson, H., Petersen, H., Þormar, G. Ö., Di Lorenzo, G., Hassan, M. & Gargiulo, P. (2026). Temporal dynamics of EEG microstates during postural control in Parkinson’s disease. Parkinsonism and Related Disorders, 152, 108968. https://doi.org/10.1016/j.parkreldis.2026.108968
• Gelormini, C. et al. (2025). Assessing brain network dynamics during a postural control task using EEG microstates. Brain Topography, 38(4), 47. https://doi.org/10.1007/s10548-025-01119-w
• Pescaglia, F. et al. (2025). Exploring dynamic alpha band connectivity in Parkinson’s disease: a novel approach to postural control assessment using the BioVRSea paradigm. Brain Topography, 39. https://doi.org/10.1007/s10548-025-01170-7
• Recenti, M. et al. (2021). Toward predicting motion sickness using virtual reality and a moving platform, assessing brain, muscles, and heart signals. Frontiers in Bioengineering and Biotechnology, 9, 635661. https://doi.org/10.3389/fbioe.2021.635661