When Methods Fail: The Myth of the 'Uncontrollable' Alpha

I love discussing new findings — even the critical ones — because they push our field to refine its methods and sharpen its theories. But occasionally, I come across studies where the conclusions say more about poor experimental design than about neurofeedback itself. Too many researchers, often well-intentioned but inexperienced in neurofeedback practice, make sweeping claims that risk discrediting decades of careful clinical and experimental work. And this is one of those times.

A recent preprint by Maaz et al. (2025), titled “Alpha power increases no matter what during a neurofeedback session,”claims that participants cannot volitionally control their alpha activity, suggesting that alpha neurofeedback is ineffective. At first glance, this sounds alarming — if alpha power truly cannot be controlled, what does that say about the thousands of practitioners and clients who use alpha training for relaxation, anxiety reduction, and creativity enhancement? But on closer inspection, the flaws in this study are both conceptual and procedural. The findings, while intriguing in their consistency, cannot be generalized to real-world neurofeedback because the experimental design itself was destined to produce the null result it reports.

This post dissects why: from the inappropriate use of visual feedback in an eyes-open paradigm, to the absence of proper baselines, to violations of consensus methodological standards. In short, this paper doesn’t prove that alpha is uncontrollable — it proves that if you design your neurofeedback study incorrectly, you’ll get meaningless results.

Methods

Maaz et al. (2025) implemented a single-session EEG neurofeedback design in healthy adults. The participants were instructed to either increase (Alpha-Up), decrease (Alpha-Down), or perform a sham version of feedback modulation. The task involved keeping a dynamic circle large on the screen — an entirely visual feedback display reflecting parietal alpha (8–12 Hz) amplitude derived from a six-electrode OpenBCI setup centered on Pz.

Session structure: The study consisted of four blocks — one baseline, three training, and one transfer block — with feedback updated at 1, 5, or 10 Hz. The only instruction was to “make the circle as large as possible using your thoughts.” No specific strategy training, reward shaping, or coaching was provided. Participants were tested in an eyes-open condition, fixating on a screen for approximately 20 minutes.

Equipment and signal acquisition: EEG was recorded using a low-density OpenBCI Cyton system at 250 Hz, with gold-cup electrodes at Pz and neighboring sites, referenced to the earlobes. Data were band-pass filtered between 8 and 12 Hz for alpha analysis, and feedback was calculated in real time. The same visual feedback protocol was used for all conditions, including the sham group, which received pseudo-random feedback derived from prerecorded EEG.

Analytic approach: Bayesian hierarchical models compared alpha power across groups and blocks. The authors pooled their new Alpha-Down cohort with previously collected Alpha-Up and Sham datasets, reanalyzing them collectively to test whether feedback direction or authenticity affected alpha trajectories.

While the analytical rigor is sound, the design choices — particularly the use of eyes-open visual feedback and the pooling of distinct participant cohorts — severely compromise the interpretability of the results.

Results

The study reports that alpha amplitude increased significantly across training blocks, regardless of whether participants were instructed to raise or lower alpha, and regardless of whether feedback was genuine or sham. Theta and SMR bands also increased slightly over time, while beta remained stable. Participants’ self-reported sense of control did not differ between groups, and Bayesian analyses supported the null hypothesis of no feedback-specific effects.

On the surface, these results seem to indicate that alpha self-regulation is impossible — that alpha “increases no matter what.” But the data more plausibly reflect time-on-task effects (gradual relaxation, visual fatigue, reduced attentional engagement) rather than an inability to modulate alpha. Without adequate baseline comparisons, eyes-closed conditions, or control for visual entrainment, these results cannot be used to generalize about neurofeedback learning.

Discussion

Conceptual Misunderstanding of Alpha Neurofeedback

The study’s premise misinterprets the very foundations of alpha neurofeedback. Alpha rhythms are context-dependent: they rise with decreased visual input and lower arousal, and they drop during visual or cognitive engagement. Attempting to suppress alpha while forcing visual fixation and feedback processing is methodologically contradictory. Visual neurofeedback targeting alpha guarantees an upward drift due to visual fatigue and arousal fluctuations — a predictable artifact, not a revelation about neurofeedback inefficacy.

Inappropriate Feedback Modality

From Kamiya’s (1968) early experiments to modern alpha training protocols, auditory feedback has remained the gold standard precisely because it minimizes interference with visual cortex activity. By using a dynamic visual target in an eyes-open setting, the authors directly activated occipital-parietal pathways that generate alpha — producing the illusion of generalized alpha increases.

Missing Baselines and Calibration

No eyes-closed baseline, rest calibration, or individualized alpha frequency (IAF) assessment was conducted. Without such comparisons, the authors cannot determine whether alpha changes were due to feedback, relaxation, or simple sensory adaptation. In clinical practice, alpha training always begins with IAF determination, eyes-closed baselines, and multi-session reinforcement — none of which are present here.

Overinterpretation of Null Results

Bayesian evidence for null differences across conditions does not invalidate alpha neurofeedback as a whole; it only reflects the inadequacy of this particular paradigm. As with any learning process, neurofeedback requires repeated exposure, strategy refinement, and individualized protocols. Testing it in one uncoached session is akin to concluding that humans cannot learn piano because no one mastered a sonata after 15 minutes.

Absence of Strategy Coaching and Reinforcement

Instructional design matters. The study’s vague directive — “make the circle bigger” — omits critical learning supports like relaxation cues, attentional anchoring, and reinforcement schedules. Without them, participants are left guessing, reducing both motivation and learning efficiency. The result: noise, not neuroscience.

Visual Entrainment Artifacts

The feedback refresh rates (1, 5, and 10 Hz) overlap with intrinsic neural frequencies, introducing rhythmic visual flicker that can entrain brain activity independent of volitional control. This overlap may have driven the observed alpha increases — not participant effort — further undermining the study’s interpretability.

Ignoring Consensus Methodology

Pretty much every serious neurofeedback study I've ever seen stresses eyes-closed auditory feedback for alpha protocols to minimize confounds (there are a few notable exceptions using visual fractal feedback). Maaz et al. cite some of these standards but then disregard them, weakening their methodological credibility and rendering their results incomparable to established research.

Brendan’s Perspective

As I’ve said before — neurofeedback doesn’t fail, poor designs do. The Maaz et al. study highlights exactly why protocol fidelity and context matter. In real clinical and research practice, alpha neurofeedback is not a one-size-fits-all process; it’s an adaptive dialogue between brain, body, and context.

If we wanted to test alpha regulation meaningfully, here’s how we’d do it:

• Condition: Eyes-closed or dimly lit room to minimize occipital interference.

• Feedback: Auditory tones or gentle musical modulation, avoiding visual entrainment.

• Target: Individual alpha frequency (IAF) or upper-alpha range, rather than a fixed 8–12 Hz band.

• Protocol: Minimum 6-10 sessions, incorporating reinforcement shaping and transfer tasks. 3-5 sessions are usually required to demonstrate volition control, possible in about 85% of people.

• Instrumentation: High-SNR EEG (at least 19-channel qEEG) for accurate localization and artifact control.

• Outcome metrics: Both EEG modulation and behavioral correlates — stress reduction, cognitive flexibility, or mood stabilization.

In my clinical experience, properly implemented alpha training reliably enhances relaxation, reduces anxiety, and improves sleep quality. The patterns learned through repeated feedback consolidate through neuroplasticity and generalize into daily life — the hallmark of effective neurofeedback.

But studies like this, which strip the paradigm of its essential components, risk misleading policymakers and the public. When readers encounter headlines claiming “alpha cannot be controlled,” they rarely dig deep enough to uncover that what’s really uncontrollable is the experimental design itself.

This field needs more collaboration between experimentalists and certified practitioners — not fewer studies, but betterones. When research aligns with clinical wisdom, we get robust, reproducible science that reflects how neurofeedback actually works.

Conclusion

The Maaz et al. (2025) preprint demonstrates one thing clearly: if you design an alpha neurofeedback study in a way that violates every core principle of alpha physiology and training methodology, you’ll get inconclusive results. Their findings do not disprove alpha self-regulation — they expose the pitfalls of testing it under the wrong conditions.

Rather than dismissing neurofeedback, the takeaway should be this: design matters. Feedback modality, participant state, baseline conditions, and learning structure all determine whether neurofeedback works. When these foundations are ignored, what fails is not the brain’s ability to learn, but the experiment’s ability to teach.

If we continue to conflate flawed designs with failed paradigms, we risk turning good science into bad headlines. The real lesson from this paper isn’t that alpha can’t be controlled — it’s that good neurofeedback can’t be reduced to a single, poorly constructed session. In the end, alpha isn’t uncontrollable — poor methodology is.

References

Maaz, J., Dia, A., Waroquier, L., Paban, V., & Rey, A. (2025). Alpha power increases no matter what during a neurofeedback session [Preprint]. HAL. https://amu.hal.science/hal-05312109v1