# Does Motion-Based Visual Stimulation Work for BCI Without a Flickering Screen?

**85.67% mean accuracy. 2.61 seconds average selection time. No flickering.** A new [brain-computer interface](https://bciintel.com/glossary/brain-computer-interface) paradigm from Hanneke Scheppink, Rainer Herpers, Jordy Thielen, and Ivan Volosyak — published on arXiv (2605.15801v2) — demonstrates that pseudo-random motion sequences can drive reliable [EEG](https://bciintel.com/glossary/eeg)-based BCI classification without the photosensitive flicker that has long been an accessibility barrier for SSVEP and c-VEP systems.

The approach, called code-modulated motion visual evoked potential (c-MVEP), benchmarks favorably against steady-state motion VEP (SSMVEP) — which it significantly outperforms — while falling short of conventional code-modulated VEP (c-VEP) and SSVEP on raw accuracy and speed. That performance gap is the central tension the field now needs to resolve.

This is a small feasibility study, not a controlled clinical trial. Results should be interpreted as proof-of-concept data requiring replication in larger, more diverse participant cohorts before c-MVEP can be considered a validated clinical tool.

---

## What c-MVEP Is and How It Differs From Existing Paradigms

Classical SSVEP and c-VEP [brain-computer interfaces](https://bciintel.com/glossary/brain-computer-interface) drive cortical responses by rapidly flickering visual targets — typically at frequencies between roughly 8 and 40 Hz. The approach is effective: SSVEP is among the highest-throughput non-invasive BCI paradigms available. The problem is physiological and practical. Sustained exposure to flickering stimuli causes visual discomfort and fatigue, and poses non-trivial photosensitive seizure risk for a subset of users, limiting the population who can safely use these systems.

SSMVEP attempted to solve this by replacing flicker with oscillatory motion — targets physically move rather than change luminance. The motion-only approach reduces discomfort but has historically suffered from poor signal quality and low classification accuracy.

The c-MVEP paradigm from Scheppink et al. takes a different route. Instead of sinusoidal motion oscillations, it drives target objects using **pseudo-random binary sequences** — the same code-modulation strategy that gives c-VEP its strong decoding performance. Objects move (rather than flicker) according to these sequences, generating motion visual evoked potentials that share temporal and broadband spectral characteristics with c-VEP responses while avoiding luminance changes.

---

## The Numbers: Offline Signal Characterization

The team first ran an offline experiment comparing EEG responses across all four conditions — c-MVEP, c-VEP, SSMVEP, and SSVEP — during sequential stimulation of a single target.

Key signal-level findings from the source text:

- **c-MVEP evoked similar temporal and broadband spectral responses as c-VEP**, with comparable signal-to-noise ratio (SNR)
- **c-MVEP responses were more concentrated in the lower frequency range** compared to c-VEP
- **SSVEP yielded higher SNRs** than SSMVEP, consistent with the known performance hierarchy
- **Spatially**, both motion-based paradigms (c-MVEP and SSMVEP) peaked at electrode Oz but spread across multiple electrodes; flicker-based paradigms (c-VEP and SSVEP) were more spatially localized at Oz

The broader spatial distribution of motion-driven responses is worth flagging. It suggests the cortical generators of motion VEPs are less discretely confined to primary visual cortex than flicker-driven responses — which could complicate decoding in multi-target configurations but might also offer useful signal diversity for advanced classifiers.

---

## The Numbers: Online Four-Target BCI Performance

The online experiment evaluated a **four-target BCI** under all four conditions. This is the most clinically relevant benchmark because it reflects real-world operation with simultaneous competing targets.

Accuracy and mean selection time, as reported:

| Paradigm | Mean Accuracy | Avg Selection Time |
|----------|--------------|-------------------|
| c-VEP | 97.81% | 1.15s |
| SSVEP | 93.42% | 1.94s |
| **c-MVEP** | **85.67%** | **2.61s** |
| SSMVEP | 64.91% | 4.18s |

The statistical picture is clear from the source: c-MVEP accuracy was **significantly lower than c-VEP and SSVEP**, but **significantly higher than SSMVEP**. On selection time, c-MVEP was significantly slower than c-VEP and SSVEP, but significantly faster than SSMVEP.

For context: 85.67% accuracy in a four-class EEG BCI is operationally meaningful — it's above the chance floor of 25% by a wide margin, and for a first-generation non-flickering paradigm, it establishes a viable baseline. The gap from c-VEP (97.81%) represents roughly one error in every seven selections versus roughly one in forty-five — a difference that matters for communication BCI users.

---

## User Comfort: Surprisingly Flat Across Conditions

Subjective ratings revealed **no clear preference between motion- and flicker-based paradigms**, indicating comparable user comfort. This finding cuts against a central assumption in the field: that motion stimulation would be strongly preferred by users because it avoids flicker discomfort. The absence of a clear preference signal either means the discomfort differences are smaller than expected in short experimental sessions, or that the subjective comfort instruments used were insufficiently sensitive. Longer-duration real-world use may tell a different story — but that data doesn't exist yet for c-MVEP.

---

## What the Performance Gap Means for BCI Translation

The 12-percentage-point accuracy deficit versus c-VEP is not a dealbreaker, but it does define where the engineering work needs to go. Several paths are available:

**Decoder optimization.** The pseudo-random sequence structure that makes c-VEP decodable is preserved in c-MVEP. Template-matching and canonical correlation analysis approaches should transfer; the question is whether the broader spatial distribution and lower-frequency concentration of c-MVEP responses requires adapted filter banks or spatial filters.

**Sequence design.** The choice of pseudo-random sequence length, motion amplitude, and motion direction are all free parameters. The current study establishes a baseline; systematic optimization of these parameters could meaningfully close the accuracy gap.

**Hybrid approaches.** Combining motion-based primary stimulation with subtle luminance cues below photosensitivity thresholds is a plausible middle path that no group has fully characterized.

For patient populations where flicker-induced fatigue or seizure risk is the binding constraint — including some users with [ALS](https://bciintel.com/glossary/als) or photosensitive epilepsy — an 85.67%-accurate non-flickering system may already be preferable to a higher-accuracy system they cannot safely use. Clinical prioritization is population-specific.

---

## Industry and Translation Implications

Non-invasive visual BCI remains a critical access tier. Intracortical systems from Neuralink, [Synchron](https://bciintel.com/companies/synchron), and [Precision Neuroscience](https://bciintel.com/companies/precision-neuroscience) offer higher bandwidth but require surgical intervention — ruling them out for the majority of potential BCI users. EEG-based visual paradigms are the realistic near-term path for users who need communication or control assistance but are not surgical candidates.

SSVEP and c-VEP dominate current commercial and research EEG-BCI deployments, in part because their performance metrics are well-characterized and their decoders are mature. The c-MVEP paradigm, if it can close the accuracy gap through decoder and stimulus optimization, would expand the usable population for visual BCI by removing flicker as an exclusionary factor.

The broader spatial EEG distribution of c-MVEP responses also has potential implications for passive neural monitoring systems — a space that companies like [Neurable](https://bciintel.com/companies/neurable) and [EMOTIV](https://bciintel.com/companies/emotiv) operate in, where sustained comfortable stimulation matters more than peak decoding speed.

---

## Key Takeaways

- **c-MVEP achieves 85.67% mean accuracy** in a four-target online EEG BCI using motion stimulation alone — no luminance flickering
- **Average selection time of 2.61 seconds** is significantly faster than SSMVEP (4.18s) but slower than c-VEP (1.15s) and SSVEP (1.94s)
- **Signal characteristics differ from c-VEP**: more distributed spatially, more concentrated in lower frequencies, but comparable broadband SNR
- **User comfort ratings showed no clear preference** between motion and flicker paradigms in this small study — the assumed comfort advantage of motion stimulation was not confirmed subjectively
- **This is small-scale feasibility research**, not a clinical trial; replication in larger cohorts is required
- The paradigm's primary near-term value is expanding BCI access to users for whom flicker stimulation is contraindicated or intolerable

---

## Frequently Asked Questions

**What is c-MVEP and how does it differ from SSVEP?**
c-MVEP (code-modulated motion visual evoked potential) stimulates visual targets using pseudo-random motion sequences rather than luminance flickering. SSVEP uses rhythmic on-off or contrast-reversal flicker at fixed frequencies. c-MVEP avoids flicker entirely; SSVEP generally achieves higher SNR and accuracy in current implementations.

**How accurate is the c-MVEP BCI system?**
In the online four-target experiment reported by Scheppink et al., c-MVEP reached a mean accuracy of 85.67% with an average selection time of 2.61 seconds. This is significantly lower than c-VEP (97.81%) and SSVEP (93.42%), but significantly higher than SSMVEP (64.91%).

**Why does flickering matter for BCI users?**
Sustained flickering visual stimuli cause eye strain and fatigue, and can trigger photosensitive seizures in susceptible individuals. For users who need to interact with a BCI over extended daily sessions — such as people with ALS or high-level spinal cord injury — flicker-induced fatigue is a real barrier to practical use.

**Is c-MVEP ready for clinical use?**
No. This is a feasibility study establishing proof-of-concept. It has not been evaluated in clinical populations, over long sessions, or in individuals with motor or sensory impairments. Significant further development and validation work is required before clinical deployment.

**What is the key technical challenge c-MVEP needs to solve?**
Closing the accuracy and speed gap versus c-VEP — approximately 12 percentage points and 1.46 seconds per selection — through optimized motion sequence design, spatial filtering adapted to c-MVEP's broader cortical distribution, and decoder refinement tuned to its lower-frequency signal concentration.