This piece comes from one of my weekly livestreams, where I take questions on neurofeedback, brain mapping, and brain training. I've anonymized the people who asked and organized their questions into the answers most of you are also asking. If you want to catch these live, I run them Mondays at 6pm Pacific.
What is neurofeedback actually doing?
Neurofeedback is exercise for brain waves. You put a couple of wires on your scalp, the computer measures your EEG in real time, and when your brain briefly flexes in the target direction, the computer applauds with sound and visuals. That feedback loop is the whole mechanism.
You parametrize the loop by choosing what you measure (a frequency band at a specific location) and then you move the goalpost. The reward fires only when, say, beta climbs at the site you're training. The software stays adaptive and waits for those small bursts, so the brain starts paying preferential attention to whatever the outside world consistently reacts to.
The learning runs below conscious awareness. This is operant conditioning, the same associative process a baby uses doing a clumsy push-up and then recruiting those same neurons again later. What fires together wires together, and repeated over time it builds. I cover the underlying mechanism in more detail in my piece on biohacking plasticity.
What do sessions feel like, and how long do they take?
You train for roughly half an hour. The work is mostly involuntary. You cannot feel your brain waves, so it does not feel like much. You are not trying, not focusing, not pushing. Your only job is to stay out of the way of clean signal, which means not moving around or piling on muscle tension.
After three or four sessions, most people notice something, either during the session, a couple hours later, or as a slow emergent shift over a day. On a bell curve, I look for about one standard deviation of change every couple of months, roughly 25 to 30 sessions. The conventional figure is 20 to 40 sessions; I'd say 40 is reasonable to reach a stable place for attention, anxiety, and sleep. Those are deep regulatory features practiced all day, so once you tune them, your life keeps the training going.
Can neurofeedback stimulate the thalamus?
Direct thalamic stimulation from the scalp is out of reach. A lot of what we do still taps into the thalamus through one of the core frequencies: sensorimotor rhythm (SMR), trained at C3, C4, and Cz along the central strip.
Picture the thalamus as a switchboard with dozens of nuclei handling input and output for vision, hearing, and the rest. The cortical strip running ear to ear sends information down into the thalamus and receives it back up through corticothalamic and thalamocortical connections. Neurons are mostly one-directional, so you need separate tracks in each direction.
Wrapping the entire thalamus is a structure called the reticular nucleus of the thalamus (NRT). Every thalamocortical and corticothalamic fiber that passes through gets synapsed on by the NRT, which carries inhibitory tone only. That makes it a gating mechanism. When you train SMR at the central sites, you reduce impulsivity, improve sleep, and cut seizure incidence. One leading theory is that you're strengthening the NRT's ability to regulate thalamic cross-talk. This is a GABAergic circuit, which is also why GABA-supporting interventions tend to help here. You can read more in my breakdown of SMR neurofeedback.
Does QEEG agree with SPECT and fMRI?
Yes, QEEG correlates well with both, though each tool has different strengths and shows different things.
SPECT is coarse spatially, so you see big broad features. fMRI is condition-based; you usually run subtraction conditions rather than a resting baseline, so there's no normative fMRI database the way there is for QEEG. With a 64- or 70-channel EEG plus source analysis through methods like LORETTA, you can approach fMRI's spatial precision while keeping EEG's timing precision.
QEEG's advantage is the age-matched population comparison. A brain map is close to raw data, cleaned of blinks and movement and compared against the average person your age. People are weird, though, so resting differences from typical do not mean much by themselves. The map shows data, not truth, and you have to interpret what patterns sit beneath it and what they mean functionally. I combine the map with an attention test for exactly this reason, so unusual features get a performance context. There's a full walkthrough in my QEEG brain mapping guide.
Are mini-QEEGs or dry electrodes good enough?
A two-, three-, or four-channel sequential QEEG produces data you cannot norm against databases. The way QEEG works is by reading static resting features that are stable across time and comparing them against population references. Running a few channels, especially sequentially, breaks most of the time-frequency relationships. The 19-channel montage rests on more than a century of EEG work, and we understand what those references and montages show. Fewer channels rob you of that context.
Dry electrodes also disappoint me. Most dry headsets use a Faraday-caged high-impedance contact at each electrode, which breaks the inter-electrode spread of EEG. A lot of what we read comes from how the signal spreads and mixes across sites, so dry systems create different norms. Their filtering below about 4.5 to 5 Hz is also unusable. Theta is one of the most important features to regulate, often stuck or elevated in the way, and losing it reliably is a dealbreaker. I've spent tens of thousands of dollars on dry headsets across the years, and nobody has solved the engineering. Stick wires to your head; you'll get over it.
Saline electrodes are a fairer middle ground. They work for some people, and I have clients who love them, including a couple of pregnant women avoiding the parabens in standard 10-20 paste. Others couldn't get clean signal and gave up. They're no less work to set up than paste.
Can neurofeedback get someone with ADHD off medication?
People do that themselves; I don't take anyone off meds. For relatively clean ADHD cases, even severe ones, without major trauma, depression, or concussion, the vast majority make a couple of standard deviations of change on executive function testing.
A practical signal worth watching: a few weeks into beta and SMR training, ADHD meds often start working two or three times stronger than expected. That tells you the dose can likely be backed off, in collaboration with your prescribing doctor. My goal is to teach you enough that you become your own expert at reading your data, the way you'd learn to read a lipid panel. For the bigger picture, see my neurofeedback for ADHD guide and the parent-focused piece on why ADHD kids trigger yelling.
As for long-term stimulant use, I haven't seen clear negative brain changes attributable to the meds themselves, though I do see secondary burnout and sleep disruption. You can map a brain with and without stimulants and see the overmedicated signature: a brain that looks woken up and fatigued at the same time. Caffeine produces a similar overdone pattern where performance starts to degrade.
Do high-functioning brains change more slowly?
Change tends to slow after the first several rounds for everyone, not specifically for peak performers. I look for about one standard deviation every 25 to 30 sessions, and two rounds covers a lot of goals: ADHD, stress, anxiety, even substance use. Nonverbal autism, major brain injury, and seizures take longer, usually six months minimum, but the change holds.
My long-term clients fall into two groups: people with severe needs and people with lofty goals, sometimes the same person. Peak performers keep training because they keep finding new ceilings. In long-term training the curve shifts from linear progress to a breakthrough pattern, where you train for a while, hit a breakthrough, train again, hit another. One client doing a lot of alpha and alpha-theta work for meditation and emotional access kept leveling up every couple of months that way. I cover that practice in biohacking meditation and the role of alpha in decoding alpha waves.
Why does a brain map say "dyslexia" when you expected ADHD?
Brain mapping does not diagnose, and the phenotype patterns cross diagnostic categories. ADHD patterns can look a lot like sleep issues, which is why I add fatigue-sensitive attention testing to tease them apart. When a database reports something like "83% in common with dyslexia" or "40% in common with autism," it's running a discriminant analysis against a reference population, not handing you a diagnosis.
The patterns these discriminants pick up are real but not specific. Tripping an "autism" discriminant might mean some quirky sensory or social processing behind the right ear, where sounds feel intense or eye contact is taxing, without any autism present. Most validated databases only detect dyslexia patterns reliably in children, partly because language features look similar across typical and atypical brains.
I showed a NeuroGuide map on the stream with a TBI discriminant to make this concrete. That tool compares a brain to a concussion population, and what it's sensitive to is brain fog: slowed waves, connectivity changes. Fog shows up with sleep problems, mold, Lyme, chemo, COVID, chronic stress, and apnea. A positive reading means patterns consistent with fog, not proof of a brain injury. The dotted line on the readout is the signal-detection threshold; sitting just right of center leaves real room for a false positive. For more on reading those features, see biohacking brain fog and biohacking with EEG phenotypes.
What can you do during a session, and what should you avoid?
Two principles guide this. First, avoid competing visual and audio sources. A dramatic movie obscures your ability to notice the variability in the reward stream, the dimming or the audio change that carries the learning. Reading is usually fine; boring lectures and still images are fine.
Second, neurofeedback relies on implicit learning, the brain's ability to extract rules and patterns without being told. Social engagement breaks that. The circuits that pick up social information run louder and earlier in processing than implicit learning, so they flood the channel. A boring task leaves the background learning intact.
A note on the visual attention system
Visual attention has several components running in each hemisphere: alerting to change, orienting in space, using spatial and environmental cues, and managing the act-versus-inhibit conflict on a subsecond scale. The supervisory and selective functions lean heavily right-hemisphere and involve the cingulate, especially the posterior cingulate, plus parietal association cortices that decide what you're looking at.
In ADHD you usually see C3 and C4 involved. Occasionally someone makes excess theta only at O1 and O2, often just at O2 behind the right occipital pole, which produces a visual-specific attention pattern without the broad ADHD signature. Training to bring down that theta can help, though right-hemisphere activation can add irritability, so short sessions are wise.
Where to start
If you want a look at your own brain, Peak Brain Institute runs QEEG mapping at multiple locations, including a low-cost option that gets you unlimited maps in the office for a year. Learning to read your own data over time and connecting it to how you feel is how you become your own expert. Book a consult or a discount QEEG through the links, and bring your questions to the next Monday livestream.