This piece comes from one of my weekly Neurofeedback & Chill livestreams, where I run a session on my own head and break down the technique while answering questions live. I have anonymized the audience questions. What follows is the substance: how I think about QEEG for peak performers, how I built a two-channel SMR/beta protocol on myself, what I see with tinnitus, the Sebern attachment protocols, and why I use LORETA z-score training sparingly.
What does a QEEG actually tell you about peak performance?
A common assumption is that the goal of neurofeedback guided by a QEEG brain map is to normalize the map, to squash hotspots and pull everything toward the middle of the bell curve. That is not how a skilled provider works.
The middle of the bell curve is statistically average. A yardstick is a reference instrument, not a performance target. A meter stick is equally arbitrary. When you train someone's brain only to push outliers toward zero, you can change the QEEG without meeting the person's goals. Some automated systems do exactly this, and so do courses run by a supervisor who is not adjusting protocols carefully (ten sessions of one thing, ten of another).
Use the QEEG as a lens. The data are valid and stable, and the map gives you agency to make change when you read it as a picture of how a particular brain is configured right now. For more on what the scan shows and what it does not, see my QEEG brain mapping guide.
When you go after high performance rather than fixing a clear problem, the bottlenecks are often the same across people. Brains cramp up in a handful of recognizable ways: rumination, slow processing speed, weak short-term memory, impulsivity. The difference between a high performer who is leaving capacity on the table and someone who is suffering is mostly how far off those features are and how reliably the person can access their resource.
A high performer is often fine until 5 or 6 PM, then falls apart once they get home. The resource is there; the stability of the resource is not. So you clear the bottlenecks first. Sleep, stress, and attention are the best places to start. Then you work on the features that feel like peak performance for that person: processing speed for verbal fluency and that crisp, intelligent feeling, and alpha-theta protocols for flow state and access to nonlinear states. You push up beta power, speed up alpha, and run alpha-theta training, all while protecting sleep, stress, and attention as you go.
How do I build a two-channel SMR and beta protocol?
For my session I ran a two-channel protocol, training both hemispheres at once with linked criteria.
On the left side at C3 (left pre/postcentral gyrus, the mind-body connection that receives ascending sensory information and sends descending motor information), I rewarded 14.75 Hz beta. Left-side beta is about vigilance, sustaining attention, and maintaining sleep through the night. On the right side at C4, I rewarded 11.5 Hz, which sits in the SMR (sensorimotor rhythm) range. Right SMR is about self-control, sitting still, poised and supervised attention. On both channels I inhibited 4 to 7 Hz theta and a fast beta band above the reward.
Here is what makes a dual protocol different from running one side and then the other. In dual mode you have to meet all six criteria at once: theta down on both sides, fast beta down on both sides, beta up on the left, SMR up on the right. Left beta and right SMR work together. Training the combination ties the two hemispheres into a stabilizing relationship while emphasizing the individual resource in each. You often see less visible change within a session this way and a stronger effect later.
A note on the equipment, because it matters for clean data. I use electrolyte paste and clip references in each ear. I put same-colored wires on the same ear deliberately, so I can glance down and verify placements by color rather than tracking each one visually. That habit prevents the most common setup error: plugging things into the wrong sockets. When I got animated and talked too much during the stream, you could watch my right-channel signal drop and the reward beeps fall off. Two-channel training demands you stay still. Movement at the electrode is the fastest way to destroy signal quality.
Why do people have two alpha peaks?
One question was about seeing two alpha peaks at FZ. The short answer: a single electrode location samples multiple sets of tissue and connections to other regions, and those circuits can run at different speeds.
Most of us carry roughly two alphas. There is usually an alpha-1 peak somewhere in the 8 to 10 Hz range and an alpha-2 peak somewhere in 10 to 12 Hz, with an overall peak commonly near 10 Hz. The same logic applies to theta. A paper out of Scott Makeig's lab at UCSD showed that in a Parkinsonian population, frontal midline theta in the 4 to 7 Hz range contained two distinct components that overlapped but behaved differently in the math. Different circuits, different speeds, recorded at one spot. If you want the deeper background on these rhythms, I wrote about alpha waves and their function.
Do brain phenotypes exist?
Humans come in flavors. Some people are short, some tall, some have blue eyes. Brains are the same. You see the big cortical hubs (default mode, salience, executive, sensory) lit up with characteristic styles of function, and yes, those styles cluster into phenotypes. I work with these patterns directly, and you can read more in my piece on EEG phenotypes.
The area behind the right ear handles sensory and social processing. For some people that region runs hot, and that strong activity produces sensory sensitivity and social anxiety because the person is picking up an enormous amount of information.
Phenotypes do not track diagnostic labels. An outlier in the map tells you a complaint is likely, not certain. You have to hold your model loosely. The data can be real while your interpretation of what it means for this person is still a hypothesis. People are too weird to be predicted with confidence from a map alone. Use the ways a brain sticks out as interesting and informative, and let the person confirm or correct you.
Does neurofeedback work for tinnitus?
Tinnitus is brutal, persistent ringing, and people suffer with it for years. Neurofeedback helps about half the people who try it, in my clinical experience. When it works, it tends to work very well, reducing or eliminating the ringing. When it does not, it simply does not. Most things in neurofeedback have high efficacy; tinnitus is one of the harder phenomena.
Confidence rises when you can see the tinnitus in the EEG. You will sometimes find a blob of delta over the auditory tissue on one side and not the other, or a blob of theta on the front midline over the anterior cingulate. The literature describes tinnitus as carrying both an auditory flavor and an OCD flavor, and you see both in the brain maps. The cases that look purely auditory seem to respond better.
When the pattern is visible, look at the one-Hz-wide bins for that brain and design a range to support what is in the way. You generally do not want to inhibit delta, which carries some risk. If you find weak beta or weak alpha alongside excess theta or excess alpha, those can respond to a more linear approach. Adding a stability protocol (SMR, or FCPZ alpha to support the cingulate) helps people who are going to respond.
Tinnitus is one of the few situations where you train the temporal lobes directly: T3 minus T4, sometimes T5 minus T6, rather than bracketing them as my C3-A1 and C4-A2 montage does. I have also seen cervical spine adjustment help, especially when the tinnitus follows a concussion from a car accident, where a stuck cervical vertebra appears to free up when released. With persistent tinnitus or vertigo, I would send someone for upper cervical imaging.
When tinnitus follows a mold exposure or COVID, you often see a bilateral inflammatory picture: excess delta, weak slow alpha, low overall power. Jack Johnstone described this "low and slow" phenotype in his 2005 paper. These cases resist focal work and call for broader, more systemic training to help the whole brain sleep more deeply and wake more fully.
What about the Sebern Fisher attachment protocols and FPO2?
A question came up about training theta upward at FPO2, the orbital site just medial to the midpoint of the brow arch, on the underside of the brow. This is the right-frontal attachment work associated with Sebern Fisher and Ruth Lanius, developed for clients with C-PTSD and developmental trauma.
FPO2 is a hard location to train. You train it eyes-closed with a neutral wire, and it leaves people feeling raw, crunchy, even angry, like they speed-ran some difficult processing in a day. The theory points at connections between the periaqueductal gray and the right-frontal insular cortex.
When I developed my version of this protocol about ten years ago, I followed a similar path and then backed off to FP2, as Sebern did. FP2 is easier to place, sits further from the eye, and produces effects that are nearly as strong without being quite as intense. I will reward theta there at times, but not in everyone. The contraindications are the same ones you screen for before alpha-theta training: no excess theta at the site, no seizure disorder, no very high delta, no major instability or dissociation. As a technician early in my career I watched an FPO2 delta reward pull out a seizure in a seven-year-old with severe reactive attachment and right-frontal seizure activity, because we rewarded too close to the delta range.
Sebern has moved toward PZ training because it seems impactful, even though the mechanism is not fully understood. There is a paper showing that a single session of alpha-down training toward this posterior region changed the fMRI coupling between the periaqueductal gray and the amygdala in around 85% of people with dissociative PTSD. I developed my own hybrid alpha-down protocol at PZ, which we normally would not do, and it hits more things at once in a high-regulatory way. Some of this work sits on solid physiological ground and some of it is field lore that simply works. This sits in between.
What is my opinion on LORETA z-score training?
Z-score training runs a QEEG in real time. You place a full cap (19 channels plus references and ground) and train toward a set of database parameters live. LORETA adds a source analysis, estimating where in the brain the scalp signal originates. The idea is interesting. In practice I use it sparingly, for several reasons.
The first issue is signal quality during training. Source analysis assumes a clean resting signal. During training the person is moving and clenching their jaw, and any noise gets pulled into LORETA and solved into fake sources. It is a noisy data environment.
The second issue is database approximation. To make the math fast enough for near-real-time feedback, the database becomes an approximation. Many of these tools scale the bell curve, so what would be two standard deviations out of range on a resting QEEG reads as one standard deviation on the LORETA z-score system.
The third issue is clinical understanding. Clinicians often feed in a list of complaints, get 75 parameters out, and let it run. That can work. When it does not, the provider does not know what to do next because they do not understand the circuits being trained. A smart brain will sometimes earn the rewards without producing the effect you wanted.
Add the setup time and the complexity that pulls the clinician out of the loop, and one-, two-, three-, and four-channel training, iterated through protocols, usually yields bigger, more durable, more reliable results across people. The rule of thumb is that LORETA takes 10 to 20 twenty-minute sessions while traditional work takes 20 to 40 thirty-minute sessions, which lands at about the same chair time once you account for the slower cap setup. LORETA can accelerate the response, and it can accelerate adverse effects too, triggering anxiety or sleep disruption quickly when the protocol is not well suited.
I bought into the full-head, dry, LORETA z-source approach. I spent $24,000 on a headset and several thousand more on NeuroGuide modules. I do not use it now. Dry caps wrap each electrode in a little Faraday pin cage, which breaks the transmission between electrodes and costs you coherence, phase, and the other classic connectivity measures. Dry systems are also poor below about 4.5 to 5 Hz, and a great deal of what is clinically in the way for people sits in the theta range from 3 to 5 Hz. A century of EEG has been built on pasted electrodes measuring sleep and seizure. Dry, high-impedance, contactless acquisition is not yet apples to apples with that baseline, so the judgments do not transfer directly.
If you are working with one confusing case at a time, very skilled, with the attention to manage it finely, LORETA can get specific. You can usually get there anyway by taking a resting EEG, running a LORETA analysis to learn about the brain, then doing surface electrode training. You are always training scalp sources regardless of the label.
How should you actually steer a course of neurofeedback?
The person's experience is the arbiter of truth, not the provider's model. Providers read brain maps and build protocols. Only the person knows how the training is actually working.
The workflow is straightforward. Start from phenotypes. Build protocols around the person's goals, not toward zero on the bell curve. Make a change, ideally toward those goals. Then iterate, eliciting subjective experience to validate your read of the brain before plowing ahead.
This is why a narrow protocol on a big regulatory feature teaches you so much. After tonight's session, if I have trouble falling asleep, my frequencies ran too fast. If I keep waking through the night, the C3 frequency ran too slow. If I am irritable, the right hemisphere was too sensitive. If I cannot filter the sensory world or feel over-focused, the left was too sensitive. The response tells you about the person, often more accurately than your starting assumptions did. Counterintuitive moves matter here too: heavy beta under C3 with sleep-maintenance problems can be compensatory, and you may need to train that beta up rather than down. For the self-regulation mechanics underneath all of this, the principle is operant conditioning of specific circuits, aimed at that person's goals.
Most people get a brain-evoked event within five or ten minutes of starting real neurofeedback. About a third get measurable in-session change. The rest may show no visible change during training and a clear change in the resting QEEG later. The process has to be tailored to each brain to get the effect you are after.
The second brain map is where prediction sharpens. Once you see large changes in performance, self-report, and the resting QEEG over a couple of months, you can make finer-grained reads, and the person usually says, "Yeah, that is what I have been telling you." By their second or third map, many people read their own brains better than I can, because they know how those pattern shifts feel and I do not.
If you want to track your own brain, Peak Brain runs QEEG mapping in Los Angeles, New York City, St. Louis, and Orange County, with remote mapping available across the US and one-off mapping in Stockholm and London. The goal of these streams, and of the work, is to make you your own expert so you can advocate for your brain the next time you need support. Read a map. Learn one-, two-, three-, and four-channel arousal and phenotype training. The skilled provider knows the brain a little better than the software does, and that gap is where the real gains live.