This article comes from one of my Monday night livestreams, where I run neurofeedback on myself, demystify the process, and take questions about brain training. Audience questions have been anonymized. Tonight's topic: downtraining, a specialized neurofeedback technique, and a 2025 study in Communications Biology that tested what happens to movement when you push beta power up versus down.
What Is Downtraining in Neurofeedback?
Most amplitude neurofeedback rewards a brainwave for going in a direction you want. Downtraining inhibits a band instead. You gently tamp a frequency down, and that does more than move power in a linear direction. You exercise the waves in a way that breaks up stuck patterns, hypercoherence, and the coupling between a region you are training and the regions feeding into it.
Downtraining earns its place in a few situations. Some brainwaves resist being raised. Push alpha up directly and it can push back. Manipulate delta with a reward and you can get into risky territory. Inhibits give you a safer way to shift those bands. You also reach for downtraining when the brain is over-aroused and you do not want to add strength before you remove noise. Strong anxiety, autonomic over-arousal in some autistic clients, recent trauma, and developmental trauma all show persistent excess beta that runs all the time. In those cases I bring overactivation down before I build tissue strength up.
If you want the bigger picture on inhibition and the brakes-and-idling logic of the cortex, I cover the regulatory frequencies in Decoding Alpha Waves: Your Brain's Idle and Its Brakes and the sensorimotor band in SMR Neurofeedback: Train Sleep, Focus, and Self-Control.
The Windowed Squash: Leaving a Hole for Alpha
Here is a move I use constantly. When you inhibit 4 to 7 Hz theta and inhibit 12 to 20 Hz at the same time, you leave the 7 to 12 Hz window open. That window is alpha. Slow alpha (7 to 10 Hz) and fast alpha (10 to 13 Hz) rise on their own into the gap as the surrounding beta and theta drop away. Theta in that area is partly disinhibition, so as the brain gains self-control the theta recedes and the medium frequencies come up without you rewarding them directly.
This is the windowed squash. You target the biggest obstacles and leave a hole where a band needs room to surface. One audience question got at this directly: if you struggle to train alpha up, downtraining the cingulate can indirectly bring more alpha forward. That is exactly the mechanism.
You can also use difference training to move coherence. Train an FZ minus PZ montage and inhibit the difference between the sites, and you make them more similar, a kind of poor man's coherence work. Inhibit specific frequencies that are coupled across sites and you break up hypercoherent regions. A QEEG brain map shows you which frequencies and sites are driving which, so you can plan the squash instead of guessing.
Why Three Inhibit Bands Do Not Feel the Same
In tonight's session I set up an FZ minus PZ cingulate protocol: inhibit 4 to 7 Hz theta, downtrain 12 to 20 Hz, inhibit 20 to 32 Hz. Three inhibits, but my brain reacts to each one differently, for two structural reasons.
First, the one-over-f rule. Lower frequencies carry more power. A single hertz of delta contributes far more to a band than a single hertz of high beta. A high-beta burst has to be large to register much change against a 20 to 32 Hz band, while 4 to 7 Hz captures changes readily. Specific beta generators sit on the cingulate and the precentral gyri at C3 and C4, so a 12 to 20 Hz band can capture those local low-beta contributions.
Second, the inhibit percentages. Setting bands at 25, 65, and 15 percent biases the brain toward the band with the stronger pull. Band width, frequency range, and threshold strength together decide how each inhibit feels.
A few minutes into the theta downtraining I felt more even-keeled and still. I run a little high-energy and enthusiastic, and the theta-down component at FZ steadied that. To cap the cingulate work, which can be sensitive, I switched to a C4 minus A2 SMR reward at 11.5 Hz with a high inhibit. That adds weight and prevents the wired feeling some people get from cingulate manipulation.
The Beta and Motor Control Study
The study that prompted tonight's deep dive is Pierrieau and colleagues (2025), Changes in cortical beta power predict motor control flexibility, not vigor, published in Communications Biology (8:1041). It tested whether motor-cortex beta is about raw force and speed, or about flexible adaptation to changing demands.
The team trained 60 people total with EEG neurofeedback at C3, using a Laplacian montage for a narrow signal. They either rewarded or inhibited beta centered at 20 Hz plus or minus 5 (so 15 to 25 Hz). I would not reward 15 to 25 Hz without expecting some over-arousal, but they kept the doses tiny and ran up and down conditions within the same subjects, so the design controls for it.
Experiment One: Grip Strength
Thirty people did a grip-strength task. The prediction was that more beta meant more force. The result ran the other way. As beta trained down, grip strength went up. Training beta up produced less variability, more motor rigidity, behavior locked into the same pattern across trials. That fits what I see in the maps: pushing C3 beta too fast makes movement rigid and pushy.
Experiment Two: Speed and Slowness
The second 30 came in for two sessions weeks apart, one uptraining and one downtraining, each with sham trials inside the same session. The dependent task was movement speed, sometimes instructed to go as fast as possible, sometimes as slow as possible.
The striking finding was context-dependent. People who lowered their beta got better at following the instruction in either direction, faster when told to speed up, slower when told to slow down, without losing control. They gained fine-grained flexibility. People who raised their beta lost that variability. People who could not bring beta down failed to hit the slow target as well. Same band, same direction of manipulation, opposite functional effects depending on the task.
This challenges the old idea that beta power maps onto motor force or speed, and the related idea that exercising a band produces equivalent effects regardless of direction. Up and down do different things. My own PhD work showed frequency-specific event-related perturbations in the exact band you reward, so frequency specificity was already on the table. This study adds that the same band manipulated in the same person produces differential effects by task.
Which Frequencies Actually Carried the Effect
The authors did a follow-up analysis across 8 to 35 Hz in one-hertz increments. The training happened at 15 to 25 Hz, but the frequencies that predicted success sat lower, roughly 8 to 20 Hz, with the biggest single impact at 13 to 18 Hz. Anyone who has worked the motor strip knows why 13 to 18 Hz at C3 moves the motor system. Downtraining 15 to 25 Hz almost certainly produced changes in 12 to 15 Hz SMR and trimmed the 18 to 23 Hz range where anxiety and beta spindles live, plus body tension up through the filter roll-off at 28 to 30 Hz.
The model explains the result even where the methodology diverges from common practice. Rewarding 13 to 18 Hz properly and downtraining where beta is individually excessive would likely produce cleaner effects. There is a separate point worth flagging: if you do not inhibit fast beta, you teach the body to recruit muscle fibers, tightening up to manufacture artifactual reward. That is why a high inhibit always belongs in the protocol, and why the bottom edge of that inhibit should encompass the individual's own beta excesses.
I would have liked a sleep follow-up. Mucking with fast beta at C3 throws off sleep maintenance. Of those 60 people trained up and down, I would bet a handful had a night or two of disrupted sleep after each session. It was a one-or-two-session study, so it stayed safe; train that beta up for weeks and you would hear about it.
Notes from the Q&A
Substance use and beta. For long-term alcohol use you usually do not chase elevated broad beta directly. Train the regulatory features of low beta, get sleep shifting, then run alpha-theta to re-educate the brain's ability to drop into a GABAergic mode. Beta comes down on its own. The exception is hot cingulate beta driving intrusive, obsessive thoughts, which is closer to the protocol I ran tonight. For the obsessive-loop circuitry, see Biohacking OCD: Targeting the Cortico-Striatal Circuit.
Post-TBI theta. Eighteen months after a traumatic brain injury, elevated posterior theta means different things depending on the company it keeps. Theta high with delta normal and crashing alpha speed points to a corticothalamic pattern you can address with SMR. Delta and theta both elevated and both running fast looks inflammatory, as if the hit is still lingering, and that brain is hair-trigger for overtraining. Read it like an active event and stay conservative. For sensitive visual tissue and migraine, hemoencephalography (blood-flow training) can accelerate recovery. Reading the slow, medium, and fast bands together tells you where in the metabolic picture the person sits.
SMR and posture. SMR is the idling beta of the sensorimotor strip, the beta-band counterpart to the mu rhythm's idling alpha. Ascending proprioceptive information into the strip is enormous compared to descending motor commands. Train SMR and you become more aware of your body, more coordinated, less clumsy, a better athletic performer, which helps with ataxia and fine-motor difficulty.
Watching amplitudes in session. People stare at power and ratio changes during a session and adjust on the fly. I was trained that way and I am here to tell you it does not matter. In my PhD work I put a 64-channel cap over individual training electrodes, ran real versus sham, and tracked the signal of the training process itself (Hill, 2013). All 24 real-training participants showed clean EEG evolution across five days. Look across the first, second, and third ten minutes of a session and you see one evolution; look across sessions one, three, and five and you see a different one. The waveform does not move linearly in a way you can reliably eyeball. In a sizable share of people you cannot see a within-session change in the band you are training, yet the change shows up hours later, the next day, or a week on.
What you can see is the brain reacting to the applause. Within about five minutes for most people, an event-related spectral perturbation appears, a brief energy increase in the rewarded band time-locked to the beep, then settling back. That is the learning event. You can auto-threshold the whole thing and it works fine.
How to Use This
If you train yourself or work with a practitioner, downtraining is most useful when a brain is over-aroused, when a band resists being raised, or when you want a frequency to surface into a windowed gap. Always include a high inhibit on fast beta to capture muscle artifact. Plan the bands against a QEEG rather than training to the middle of a database. And do not chase visible in-session amplitude changes; the effect often arrives after you leave the chair. The clearest read on whether a protocol fits is the person's own report of how they feel, which always outranks any model you built of them beforehand.
If you want to map your own arousal pattern and beta load before training, a brain map is the starting point, and you can read more about cost and coverage in How Much Does Neurofeedback Cost in 2026?.