This is a write-up from one of my weekly live streams, where I run a neurofeedback session on myself, narrate what the software is doing, and take questions from whoever shows up. I have stripped out names and kept the substance. If you are new to brain training, this is a decent tour of how a session actually runs and what the signal means while it is happening.
What does a neurofeedback session actually look like?
I set up a one-channel referential montage at C4 minus A1. C4 sits on the right sensorimotor strip. A1 is the left ear. The amplifier subtracts one site from the other to build the signal, which is why most setups using a referential amplifier need a third wire for ground. I put the ground on the right ear.
The protocol was basic. I inhibited theta (4 to 7 Hz), rewarded a low beta band centered around 11.6 Hz, and inhibited a high beta band (22 to 34 Hz). I trained the reward band a little lower than usual because I had a mild headache and did not want to overtrain. That is a real consideration. You do not push hard through fatigue with brain training any more than you would max out a lift when your form is already breaking down.
Here is what the screen shows. The top trace is the raw signal coming off my head. Below it are the filtered bands. The software pulls each frequency range out of the total signal and tracks its amplitude in microvolts. When a band crosses its threshold in the wrong direction, the game pauses. When my brain holds all three bands on the right side of their thresholds for about half a second, I get a beep.
My theta started around 23 to 24 microvolts, which is high. End of the day, I was tired. As I concentrated, the theta dropped to 17, 16. Theta is something like an idle or automatic mode. When a region makes a lot of it, that region is running less reliably and less voluntarily. You can pull theta down a little with active concentration, which I demonstrated live. Beta is harder to push up cleanly, because most of what people add when they try is muscle tension, not brain signal. That shows up as inflation across all the bands at once, including the high beta inhibit. If only your reward band rises and your high beta stays flat, the change is real.
The mechanism underneath all of this is operant conditioning running below conscious awareness. The feedback rewards the brain when it drifts toward the target pattern and goes quiet when it does not. You are not steering it with your conscious mind. The brain learns the contingency on its own. If you want the longer version, the SMR neurofeedback guide covers how this band trains sleep, focus, and self-control.
Why does training the cingulate help some kinds of pain?
The cingulates hold attention on things. As you move toward the front of the brain you move toward the internal self; toward the back, you move toward the outside world. The cingulates reconcile the approach and avoid systems, the gas and the brakes, and they select among the competing thoughts and moments of internal awareness you are having.
If you calm an overactive cingulate, you can get the brain to hold attention on pain less rigidly. Some pain has a strong central component, where the brain echoes or resonates the pain signal long after the peripheral input. Loosen that resonance and the brain reacts less to what the body is sending.
The cingulate is not the only pain target, and often not the most effective one. I get strong results in self-training with an FZ minus C4 protocol where I train down the relevant bands, and with a C4 minus A2 alpha reward. The right side seems more involved. My working guess is the right ventral tegmental area and its connections to the cingulate, plus the right precentral gyrus and possibly the insula, which handles body awareness. The right hemisphere samples for what is negative, which may be why right-sided work seems to produce reliable pain relief. This is observation and inference, not settled anatomy.
In chronic pain you often see the vertex lit up in beta and fast beta. Some of that is muscle tension creeping in, but a lot of it is resonant beta. The person cannot let go. They are braced, buzzy, and on edge. That signature looks almost identical to chronic generalized anxiety, and it can be the same person. The time to run a pain protocol is when pain is acute and uncomfortable, because the brain shifts more, and you notice the shift more, when the signal is loud. For more on the anxiety side of that picture, see Biohacking Anxiety.
Where do the language centers live?
People vary, and language organization can be genuinely odd from one brain to the next. The general map: front left, Broca's area, is productive language, where you assemble speech. Back left, Wernicke's area, is receptive language. The arcuate fasciculus connects them, and timing problems along that path produce aphasias, stuttering, and some dyslexias.
On the right side, near the analog of Wernicke's, you find number and symbol processing, plus nonverbal communication like tone of voice, prosody, and the lilt that carries sarcasm and humor. There is something in the front right and something behind the right-side Wernicke's analog, neither of which is well understood.
A small minority of people have atypical language laterality. It is not a lefty-versus-righty story; most left-handers have the same left-dominant language as everyone else. A larger share of left-handers show atypical laterality than right-handers, but it is a few percent in both groups (Knecht et al., 2000). That can mean flipped brains, duplicated language areas, or imperfect timing between regions.
Receptive language for tone and meaning locks down around age eight or nine in girls and nine or ten in boys. Before that pruning, a child can hear and reproduce speech sounds, phonemes, they have never encountered. After it, the brain assumes any novel speech sound is a variant of something it already knows. That is the basis of accents, and why losing an accent after age ten or twelve is so hard. Those receptive and phonological processing areas have gotten less plastic. Semantic memory runs the other direction. It is one of the few memory systems that keeps improving across the lifespan, and it often stays intact even with significant impairment elsewhere.
What is the difference between near-infrared and passive-infrared HEG?
HEG, hemoencephalography, trains blood flow rather than electrical activity. There are two flavors. Near-infrared (NIR) is close to the commercial fNIRS devices like the Mendi. It emits red light and measures how much comes back, using the fact that oxygenated and deoxygenated blood absorb red light differently. That gives you a proxy for local perfusion, and you can move the sensor to target specific tissue.
Passive-infrared (PIR) is just an infrared camera strapped to the forehead midline. It reads a broad field of heat rolling off the brain with no fine spatial resolution. I think of it like sitting on the beach watching the waves to estimate how busy the ocean is. You are measuring waves of metabolic heat.
I have done a lot of work with both, and I lean on PIR. I find NIR weaker, and I have seen it produce headaches without much benefit. PIR seems impactful for moving blood flow and toning vascular dynamics, which makes it interesting for concussion, migraine, some developmental features in autism, strong concentration, and brightening an aging brain. Blood flow is semi-voluntary. You can bring it up by concentrating, somewhat like classic biofeedback, and it also runs involuntarily, so it sits in an odd hybrid space between biofeedback and neurofeedback.
One caution came up live. If your cingulate has a history of running hot, or you have a history of excess frontal beta, the midline PIR sensor will tend to activate the front midline anyway, because the field is broad. You probably do not need more frontal perfusion in that case. Map first, move slowly, or skip it, unless a metabolic problem like post-COVID fog or a concussion gives you a reason to push.
Can red light therapy and other metabolic tools support brain training?
I think of photobiomodulation, hyperbaric oxygen, and sauna as fuel for the neurofeedback process, especially when someone is fighting fog or aging. Red and near-infrared light in roughly the 1000-nanometer range and above passes through the skull and feeds the mitochondria fairly directly. For soft tissue you want the lower range, around 650 to 800 nanometers. There is plenty of consumer hardware below 800 nanometers and far less in the higher transcranial range.
Red light is a mild stressor as well as a fuel, which is part of why it works for some people. Do not push through fatigue. If the helmet makes you feel tired, congested, stuffy, or gives you a headache, take it off. That reaction is rare, but it is not something to train through. The photobiomodulation overview goes deeper on dosing and mechanism.
For peripheral neuropathy, like diabetic neuropathy, I would not start with neurofeedback. The research on metabolic support points toward benfotiamine, the fat-soluble B1 analog that reaches distal tissue, support for sugar metabolism, and tools like exogenous ketones, sauna, hyperbaric, and tissue-level red light. For central or generalized pain that the brain has been echoing for a long time, neurofeedback can shift things, often without eliminating the pain entirely.
Where do you train alpha and theta, and why?
You train a frequency where the brain actually makes it. Alpha is produced in the cingulates and in sensory tissue with the eyes closed. The classic alpha-theta sites sit along the back midline: PZ, POz, Oz. PZ sits over the posterior cingulate, Oz between the primary visual cortices. The large majority of the alpha-theta literature used PZ minus A1, and that is most of what has been done in practice as well.
Some practitioners historically trained theta up on right frontal sites, but that is raw and unpredictable, and several of the people who did it have moved toward the back of the head. Frontal areas are production sites for other functions; taking them briefly offline or into automatic mode is less predictable than rewarding the posterior tissue that makes these rhythms with the eyes closed.
Training low alpha up appears to raise GABA, which fits the calming effect. The alcohol-treatment work, the Peniston protocol, is alpha-theta, and alpha trains up strongly in those participants (Peniston & Kulkosky, 1989). There is also an immune angle: research by Gary Schummer reported that rewarding eyes-closed PZ alpha was associated with raised CD4-positive T-cell counts. For the cognitive-aging side of alpha, individual alpha frequency, the peak within the 8 to 12 Hz band, slows with age and tracks processing speed (Klimesch, 1999); the alpha waves explainer covers that in detail.
If you want to improve processing speed, I look at the left frontal approach system, at PZ as a major cortical alpha generator, and at C3-CZ for cleaning up sleep, since poor sleep drags speed down. Always find the bottlenecks and the foundational issues in your own map first. Those are usually the sites that move the most.
What about binaural beats, tones, and photic stimulation?
I am not convinced by audio entrainment or binaural beats in humans. For the twenty or thirty years that the various tone packs have circulated, the human research has mostly come back with non-findings and no reliable frequency-following response. There is some newer work I will keep reading, but the older evidence is weak. When tone-based approaches help inside a practice, I suspect the benefit comes from the practitioner, the individualized work, and the permission to focus and notice, which is another flavor of meditation, rather than from the sounds reshaping the brain.
Photic stimulation, flashing light, has a real use in QEEG when you are trying to provoke seizure activity to localize a focus. For routine SMR-style training it does not help in my experience, and the brain generally treats it as an annoying stimulus. If you want to see seizure-like activity, sleep deprivation works far better than photic stimulation. A midnight map, or a map at 6 a.m. after being up all night, will surface that activity much more reliably.
A few practical points from the Q&A
On hardware: with a clean signal from a good amplifier, the microvolt readings should be roughly the same across different amplifiers using the same montage. Paste thickness and setup shift things slightly, but a clean referential pair gives you strong common-mode rejection. Bad impedance shows up as a fuzzy, fat trace; 60 Hz line noise shows up as a clear artifact. Heavily filtered consumer forehead bands can look clean on screen while the smoothing has broken the timing relationship between the signal and the brain. That is part of why a device like the Muse may be better for grading a meditation session afterward than for real-time training.
On ground placement: it does not matter where the ground sits. Nose tip, forehead, ear, mastoid, even a forearm all work. The old sleep studies often taped it to the nose.
On reading session effects: a single surge of sadness, anger, or fatigue during a session is usually just the brain stretching, the way your arms shake holding a curl bar. The real measure of an abreaction is the two hours after the session and how sleep flexes that night. If a wave of emotion or fatigue clears within an hour or two and sleep is good, that is normal. If the negative effect lingers or magnifies, the protocol needs adjusting. I ease in with safer central-strip protocols, index how each person responds, and only move to stronger work once I understand their pattern.
That is the shape of a session and the science around it. If you want to see your own bottlenecks before training, start with a QEEG brain map, and if you are weighing whether this is worth doing at all, the research overview on neurofeedback lays out where the evidence is strong and where it is still emerging.