This piece comes from one of my Monday night live streams, where I hook myself up for a neurofeedback session and then teach a topic with live audience questions. This session covered EEG phenotypes. I have anonymized the audience questions and pulled the teaching into one place. If you want the long-form companion piece, I keep an extended guide on biohacking with EEG phenotypes.
What is an EEG phenotype, and why doesn't it match a diagnosis?
A brain map shows you real patterns in brain activity. Those patterns are stable or semi-stable across time, and they cross diagnostic boundaries. That last part trips people up. You cannot line up QEEG brain mapping patterns perfectly against the DSM. The patterns are real. They are not deterministic. A given pattern does not lock you into a single diagnosis.
Most brain mapping works by comparing you to an average and looking for outliers. The mistake is treating an outlier as bad. Look at the brain tissue, the region, and the function involved, then make your judgment. People are weird, and that is fine. The map helps you build hypotheses about what is plausible. You then test whether those hypotheses fit your experience.
I learned to think this way from a few mentors in the field. Jay Gunkelman, Joy Lunde, and the late Jack Johnstone ran a QEEG processing house decades ago, when none of us could afford to process our own data. They saw the same patterns repeat across thousands of recordings every weekend, and in a 2005 paper they laid out the first framework of phenotype categories. At the time, several of those patterns were candidates without a known mechanism. Every one of that first set now has a genetic or medication-response link.
Features, biomarkers, and phenotypes
I use a three-level framework. A feature is something you see in the data that is real, not noise or artifact. You make this much theta, or your alpha runs at this speed. It is measured and interesting, but you may not know what it means.
A biomarker is a feature that shows up again and again, conserved across many people. Front midline theta. Slowed alpha. These repeat reliably.
A phenotype is a biomarker where we have some idea about the mechanism, often a genetic or pharmacological anchor, and a tighter link to specific patterns of experience or suffering.
How does neurofeedback actually change the brain?
During the session I ran a basic protocol at C3, on the left sensorimotor strip. Two things drive most band-training work. The first is SMR, low beta on the motor strip, which touches a wide range of regulatory resources. The second is excess theta, which is the single most common thing getting in the way. Theta in the four-to-seven Hertz range usually signals disinhibition. Too much of it in the front midline shows up as a song stuck in your head or a tic. On the right side it shows up as impulsivity. On the left, where I was training, it interferes with sustaining attention and initiating voluntary behavior.
The training itself runs below conscious awareness. I set a threshold so the feedback game runs smoothly when I produce more beta and less theta, and it dims or stops when my brain drifts the wrong way. I am not steering it with my conscious mind. You cannot feel your brain waves, and the squiggles on the screen bear no resemblance to your thoughts. The brain notices the pattern of the feedback starting and stopping, and gradually shifts toward the rewarded state through operant conditioning. This was first demonstrated on cats about sixty years ago, which is a useful detail, because cats are terrible instruction-followers. The effect does not require following instructions. Most people start feeling something after three or four sessions, subtle at first.
What do the common EEG phenotypes look like?
Low voltage fast (LVF)
LVF shows small, fast, spiky beta patterns without big slow waves. It can be typical and harmless, or it can be a problem. It involves a change in GABA-A metabolism or receptor function, and you see it in chronic drinkers, in anxiety, and sometimes during healing after concussion or TBI. The point is to notice the pattern, then test whether it matters for the person in front of you. (Low voltage slow is a different story and is always worth concern.)
Frontal slow
Here you see wide, slow waves over the frontal locations, frontal delta and theta. This is one of the more common clinically meaningful patterns, and it tends to bring difficulties with attention, mood, and executive function.
Slowed alpha peak frequency
This is the speed of the brain, not the amount. It tracks a change in the COMT gene and correlates with processing speed and IQ. If your primary pattern is slowed alpha, stimulants will mask the slowness without helping your attention. On a QEEG z-score plot you see the alpha numbers sliding below average, and the slowed alpha often impinges down into the theta range, so the alpha is worse than it first looks. For more on this rhythm, see my breakdown of alpha waves.
Spindling beta
Beta spindles are little bursts that climb and drop away. This is also a GABAergic pattern, and some medications make it climb. When you see it, you are usually looking at anxiety or fear. Where it shows up tells you how it presents. Frontal beta leans toward mood and anxiety. Front midline beta leans toward OCD. Behind the right ear it leans toward sensory and social overload.
Frontal alpha asymmetry
Based on Richard Davidson's work, too much alpha over the left frontal region pulls you out of approach mode, and too much beta over the right frontal region keeps you in avoid mode. I describe these as the happy kid on the left who wants to greet the world and the grumpy old man on the right who wants everyone off his lawn. The combination produces a negativity bias.
Theta-beta ratio
Joel Lubar and later Vincent Monastra showed in the 1990s that a high theta-beta ratio at the vertex sorts people into ADHD and non-ADHD buckets reasonably well. The statistic eroded over the following decade as replications got weaker. The likely reason is that the adolescent populations being tested got more and more sleep-deprived. A high theta-beta ratio predicts ADHD, and it also predicts a sleep problem. They look the same on the vertex.
How does the right temporo-parietal junction explain autism and social anxiety?
This came up around a question about distinguishing autism from developmental neglect. The right temporo-parietal junction (TPJ), the tissue behind the right ear, drinks in sensory and social information. I call it the princess and the pea, because it runs hot and resists relaxing. It runs hot in people on the autism spectrum, and it runs hot in people with strong social anxiety. Those experiences feel very different from the inside and can look very similar on a map. You might see the failure mode as a cramp up in strong beta in one person, a reduction of alpha in another, or excess theta in a third.
Developmental trauma usually has no clean cortical signature. The brain grows into an unsafe, unpredictable environment and cramps up against it. You see secondary phenomena instead. The amygdala has no EEG, but you can sometimes read a right frontal beta pattern suggesting avoid mode, because the cortex around the amygdala is activated. The periaqueductal gray is too deep to read, but the posterior cingulate may show threat sensitivity. To reach subcortical structures with neurofeedback, you work on the whole brain as a regulatory system, or you train cortical tissue connected to those deeper regions. Cynthia Kerson and Sebern Fisher's work is a good place to read more on the trauma side.
Autism is also heterogeneous on the map. You see short- and long-range connectivity changes, coherence and phase differences, plus front midline theta or beta, and weak left-side beta tied to inattention or unstable sleep. The diagnostic label does not map cleanly onto the phenotype set.
What about the mu rhythm?
A question came in about whether the mu rhythm at C3 and C4 is an autism phenotype. Mu is a normal rhythm we all make, a wicket-shaped idling of the mirror-neuron and motor system. It suppresses the moment you move or even imagine movement. The autism finding is that social observation does not suppress mu the same way it should. Mu still shows up in a quarter to a third of neurotypical recordings, so be cautious about pathologizing a normal variant.
Why neurotransmitter levels are the wrong target
A question about whether more beta raises norepinephrine gave me a chance to clear up the chemical-imbalance idea. Receptor type, density, and distribution govern neurotransmitter function far more than absolute concentration. A synapse is a sealed space. Neurotransmitters released there are not in general circulation, so the amount in one synapse is not tied to the amount in another.
ADHD tracks a specific distribution of D2 and D4 receptors and the dopaminergic signaling that follows from that distribution. Parkinson's makes the point sharply. You can lose 75 to 80 percent of your dopamine-generating neurons before symptoms appear, so for most of that loss the brain tunes around the absolute level. SSRIs raise serotonin in the synapse, which prompts the sending neuron to downregulate production, so weeks later your serotonin is lower than before, right when mood lifts. The lift tracks BDNF and plasticity factors in the hippocampus, not serotonin abundance.
Neurofeedback reliably improves selective attention, sustained attention, memory encoding, IQ, and processing speed. From those changes you can infer with confidence that signaling involving norepinephrine, acetylcholine, and dopamine shifts. The target is the information signaling, not a chemical level. One client of mine had built up enough plasticity through training that a microdose she barely used to feel hit her hard, because her brain had become more flexible and sensitive. Psychedelics, cannabis, and stimulants can all run two to three times stronger after a stretch of neurofeedback, which is worth knowing before you combine them.
What does change actually look like in the data?
Two cases from the session stand out. The first was a man who drank a bottle and a half of wine daily for 25 years, on Ativan and Ambien for maybe two hours of sleep. He came to my office straight from a 45-day detox, sober for over six weeks, and still shaking and hyperaroused. Years of alcohol kept GABA elevated, so his glutamate climbed to match it. With the alcohol gone, the brain could not pull the glutamate back down reliably, leaving a strong overactivation pattern, heavy beta and beta coherence. This is the glutamate rebound that keeps people from sleeping in early sobriety. After 30 sessions he called the office one morning, came in, and fell asleep on the couch to prove he could now choose to sleep. His maps showed the overarousal and beta coherence dropping out. Seeing the change gave him the agency to keep going.
The second was a brilliant, miserable man with full-blown ADHD, OCD, depression, and social anxiety on a single map: theta-beta ADHD on the vertex, front midline beta for the perseveration, left frontal alpha for the low mood and motivation, and left-side theta for sleep maintenance. After 30 sessions across a ten-week program he had gone back to grad school, moved back in with his girlfriend, and held the same job for a year.
I also showed an ADHD executive-function test that normalized across about 25 sessions, and a medication comparison where a woman's Concerta produced a paradoxical calming effect. Her map explained it. She is one of the few people who drop beta on stimulants, so she got a focus lift, a mood lift, and an anxiolytic effect at once. The map validated what she was already feeling. If you want the broader picture on whether this approach holds up, I cover the evidence in my guide on whether neurofeedback works for ADHD.
What about montaging and reference choices?
A few questions touched on why the choices of recording site and reference matter. You are never measuring at a single location. You measure the difference between locations. CZ referenced to the left ear is an inter-hemispheric mix dominated by the left, not a zero point. The closer two wires sit on the head, the smaller the resulting signal, because you subtract out the common information shared between them. That common-mode rejection is exactly why bipolar pairs tolerate noisy environments. C4 minus A2 references across the right side and gives a smaller, more spatially specific signal; C4 minus A1 references across the whole head and carries more information but less precision. Montaging is genuinely complex, and you will not find easy agreement on it.
How do you use phenotypes without over-reading them?
Brain mapping is a tool for learning, not for crisp diagnosis. When I look at a map with someone, it is rare that I find something they had no clue about. Most people already know the resources they want to work on. The map lets them see it and gives them agency to do something about it.
So the question stack runs like this. How are you weird? Does it matter? Would you like to take control of it? Walk down those questions and you move away from good and bad and toward the resources you are training. Hold your conclusions loosely. Nothing in the brain is ever one thing.
A closing note on a question about post-exertional malaise and exercise intolerance. In every case I have seen, there has been a strong neuroinflammatory signature alongside dysautonomia, POTS, long COVID, mold, or Lyme. Neurofeedback can help, but you go slowly, because these brains overtrain easily. The bigger levers in that metabolic territory are the hormetic stressors: sauna, cold exposure, photobiomodulation, light or deep ketosis, and a lot of walking, which is a strong plasticity booster. EEG work alone is not where the answer lives when the problem is metabolic insufficiency and stuck inflammation.
If you want to look at your own patterns, a brain map is the place to start. Bring the data back and we can build hypotheses about your resources together.