π Join us LIVE for NeuroNoodleβs Q&A Panel! This is your chance to drive the conversationβyou ask, we answer! Bring your neurofeedback and neuropsychology questions to our world-class expert panel, featuring: π§ Jay Gunkelman β The legend with over 500,000 brain scans analyzed π Dr. Andrew Hill β Peak Brain Institute Founder & Cognitive Performance Expert π Dr. Mari Swingle β Author of i-Minds and a leader in neurodevelopment π‘ Dianne Kosto β Neurofeedback Pioneer and Advocate π€ Anthony Ramos β Neurofeedback Advocate & Practitioner π Santiago Brand β Global Neurofeedback Trainer (joining when available) ποΈ Pete Jansons β Host of the NeuroNoodle Podcast π What Weβll Cover: βοΈ Cutting-edge neurofeedback strategies βοΈ Advances in brain health and psychology βοΈ Practical tips for clinicians and practitioners βοΈ Your Questions β YOU drive the content with your curiosity! π When: January 29, 2025, at 6 PM CST π Hit the π notification bell to join us and participate in real-time!
Episode Summary
I joined Jay Gunkelman, Dr. Mari Swingle, and the crew on the NeuroNoodle podcast for a live Q&A panel. The conversation ranged from running neurofeedback across a dozen time zones to gamma and consciousness, the real limits of AI in EEG analysis, and how to pick an EEG device that is worth owning. This piece pulls together my own answers from that discussion. You can watch the original conversation on the NeuroNoodle channel.
What does running a remote neurofeedback program actually require?
I coach about a hundred or more remote clients at any given point. We have remote services out of Florida now and a partner in Nova Scotia, so we cover the East Coast from New York down to Florida and up into Canada. I also work with clients in India, Australia, Hong Kong, Dubai, Sweden, and across the Middle East.
The neurofeedback itself is the straightforward part. I train an arousal model, a regulatory model, and phenotype-driven protocols off a QEEG. Once you have done neurofeedback with a thousand or more brains, you get the hang of that style of training. The hard part of a remote program is logistics: staff seven days a week, twelve hours a day, clients in different time zones, and occasional language barriers.
Language matters less for neurofeedback than it would for talk-based work. I am asking about regulatory shifts in someone's experience, their stress, sleep, attention, and mood. Those are fluctuating resources that move day to day, so I can track and validate them even when I am working through a client's daughter or a friend who translates. If you want the full picture of how distance training works, I cover it in Remote Neurofeedback: How It Works and What to Expect.
The gear is small enough now that distance is rarely the obstacle. I use battery-powered Q amplifiers that fit in a hand. You need some internet here and there, not a constant live connection, although we provide live support when it is available.
Why is EEG harder to record in Europe?
The 50 Hz mains noise in Europe floods recordings in a way that 60 Hz in the US usually does not. Many European buildings were not wired with grounded circuits everywhere, so ambient 50 Hz bleeds across the signal. A notch filter from 45 to 55 Hz that rolls off at around 120 dB will handle a reasonable amount of it. Get too close to a transformer and no filter saves you, because you are picking up more noise than the filter can attenuate.
What does a successful brain change look like on a QEEG?
I do nine or ten direct consults on a typical day, going over change maps with clients and validating the electrophysiology against their reported experience. When you review pre and post recordings, the areas you trained are often the areas with maximum change. That local correspondence is what makes the data convincing.
One case from this week: a 75-year-old I had seen four months earlier came back with a clean 10 Hz alpha at the back of the head and beautiful topography. He is sleeping deeply, dreaming, and remembering his dreams, which he says he cannot recall ever doing before. He is a well-known sports figure, and the change was significant enough that he set up a foundation to fund neurofeedback for traumatic brain injury in former football players, including those who cannot pay.
That dominant posterior alpha at 10 Hz matters beyond one case. Individual alpha frequency tracks with cognitive performance and tends to slow with age (Klimesch, 1999). Training peak alpha has shown cognitive benefit in older adults in small studies (Angelakis et al., 2007), which is emerging evidence, not settled fact. I walk through what alpha does and why it matters in Decoding Alpha Waves: Your Brain's Idle and Its Brakes, and through the imaging itself in QEEG Brain Mapping: What It Is, What It Shows, and What to Expect.
What can AI actually do in neurofeedback right now?
Start with the problem you are trying to solve. There is a lot you can do with machine learning, but you have to frame the specific task before the tool is useful.
I do not trust current systems to do EEG analysis. I have watched colleagues run their data through custom models and get plausible-looking treatises that fall apart on inspection. One model produced text about "true sources" in EEG, which is a hallucination; there is no ground-truth source in scalp EEG. I would have given that output a C-minus as a student. If someone is working on my brain, I do not want a C-minus.
Where I do trust it is in the teaching role. I have close to 10,000 videos of me discussing people's brains with them over live data, 30 to 45 minutes each, walking through what a QEEG is, what a continuous performance test shows, and what their results mean. I built a tool that ingests those videos, identifies likely phenotypes and goals, pulls from my list of starting-point protocols, and produces a formatted summary. The analysis is mine. The AI reformats, reframes, and expands what I already taught it, and I can read the output and confirm it is valid. I do not let it flag transients or read the EEG itself, because that would not save me time. I would still have to look at every channel.
The technology is moving fast, and open-source models now run on inexpensive hardware. That speed cuts both ways. It compounds the potential for help and the potential for harm, which is a fair caution. The thing about scalp EEG is that interpretation rests on knowing what you are looking at, and that knowledge is exactly what these systems do not yet have.
Gamma, consciousness, and the so-called spiritual states
For the first 20 years I did this work, people would ask me to record their brain while they did something extraordinary, and I rarely saw a clean correspondence between the brainwaves and what they reported. In the last five years a handful of cases have shown the same phenomena repeatedly.
If you want to study these states, a standard medical-grade amplifier will fail you. DC to 70 Hz is the window the medical community uses, and it misses the gamma that matters for consciousness research. To catch gamma-2, between 80 and 100 Hz, you need to record up to at least 150 Hz. Cliff Saron's project recorded gamma stripes well above that. Some of that high-frequency content is genuine neural signal; some is the brain acting like a radio transmitter, and some likely reflects the oscillation of charged neurotransmitter structures such as dopamine.
The relationship I keep coming back to is between slow cortical potentials and gamma. When they are time-locked, you have consciousness. When they are not, you are deep enough for surgery or in a state that is hard to describe. Europe historically filtered out the high frequencies and looked only at slow cortical potentials; in the US we filtered out the slow cortical potentials and looked at the oscillatory EEG. You do not understand brain function until you put both back together.
The dissociative or ecstatic states I have recorded show something different from trait gamma. Long-term meditation produces acquired gamma traits (Lutz et al., 2004). Ecstatic states are absorptive and dissociative, and I see big bursts of focal delta, as if parts of the brain are switching off. When John Duca took salvia divinorum under observation at an ISNR meeting, he produced roughly a thousand microvolts of 10-second-long waves, sub-delta infraslow content, with massive gamma alongside it. The pattern was too organized to be a perspiration artifact, because sweat artifact has no coherent phase relationship and this did.
We also published work in the Society for Scientific Exploration journal on a healer and a "healee" at a distance. The healer produced standing waves that phase-synchronized the other person's EEG, hitting the Schumann resonances at 7.83 Hz and their harmonics, with the fourth harmonic landing in gamma. The bispectrum showed the harmonics clearly once you knew to look for the checkered cross-spectral coupling. That synchronization identifies a connection between two nervous systems. It does not demonstrate healing. A connection is the phone line; something still has to be delivered across it.
How should you read an EEG montage?
Calling a two-site placement "bipolar" is a common misnomer. Your amplifier is a push-pull differential amplifier; it has two inputs and a ground. You can run it sequentially across two active sites and read the difference between them, or you can assume one site is a neutral reference. No reference is truly neutral. The ears are active references and pick up signal off the temporal lobe, then smear it as though it came from everywhere.
You need more than one montage to know what you are looking at. A montage is your perspective, the way turning a stranger sideways tells you whether they have a big nose or a flat face. One view is not enough.
I favor the Laplacian montage. Paul Nunez, who is not selling you a machine, has argued since 1997 that Laplacian montages are the best way to see coherence (Nunez et al., 1997). The Laplacian is somewhat insensitive to global distortions, so medication and fatigue effects wash out a bit while local spatial precision sharpens. It does not eliminate medication effects, though. You will see theta in a medicated ADHD state more clearly in Laplacian than in linked ears, and a benzodiazepine intoxication shows up in any montage. If you want to see where activity actually originates, Laplacian shows local function better than the alternatives.
A note on protocols and sites
On my Monday live stream I ran a C4-minus-Pz protocol on myself at 11.5 to 14 Hz and noticed my voice tightening into a mild vocal stridor over half an hour, then resolving. A sequential C4-Pz placement tends to speed up alpha at Pz toward faster-than-usual frequencies, which feels like higher function. C4 and Pz will not synchronize, so you are effectively training two sites to produce SMR-type content. SMR training is worth understanding on its own; I cover it in SMR Neurofeedback: Train Sleep, Focus, and Self-Control.
What about consumer EEG devices?
Reliability and validity are relative to your purpose. If you want to float a ball up and down on a screen, most consumer devices are adequate. For an event-related potential study, the EEG itself is noise to the ERP, so the amplifier does not need to be excellent.
For real neurofeedback work, you cannot get into the EEG space usefully for under a grand or two. There are cheaper boards, like an OpenBCI Ganglion at around a hundred dollars, but you have to already be an EEG expert to use them. The opposite problem exists too: people spend thousands and still do not know what the device is doing or how to use it.
A quick physical check helps. Look at the electrode. A solid metal disc has contact jitter and is usually a poor design. A polymer-coated sensor holds a little moisture, which makes it a damp sensor and a partial salt bridge, and signals that someone thought about contact jitter. Shiny bare metal is a step down from a coated sensor.
This is a small field. There are maybe 15,000 providers including prosumers and perhaps 30,000 active software and hardware installs worldwide. That is why you do not see mature consumer products with the development pressure you would find in a larger market. The way to think about a device is simple: match the engineering to the task you actually have.
Infraslow training and the case for personalization
Infraslow content matters for brain function. Europe studied the slow cortical potentials and the US studied the oscillatory EEG, and consciousness lives in the relationship between the two. Infraslow training is not my modality, so I stay quiet on its specifics, but I have seen it work well for profoundly over-aroused nervous systems, including some autism cases, and trauma-informed practitioners report good results. It functions largely as forced quieting, and the field is still working out where to train and what frequency tuning to use.
The deeper point is that there is no head-to-head horse race telling you which technique beats another. When researchers study neurofeedback, they usually require one fixed protocol for the whole group. At that point you are studying a technique, not the personalized way neurofeedback is actually delivered, which is individualized by design. You can study a personalized approach if you define the personalization rules in advance, but funding for that kind of trial is hard to come by. The honest answer to "which training is most beneficial" is the one your brain needs, and the only way to know that is to evaluate the brain first. A good QEEG points to what to train up, what to train down, and which relationships between regions to change.
If you are weighing whether any of this is established science, I lay out the evidence base in Is Neurofeedback Legitimate? A Research Overview and the ADHD-specific data in Does Neurofeedback Work for ADHD?.
You can find my weekly live streams on YouTube at Dr Hill, where I demo neurofeedback and talk through biohacking the brain, and you can reach me at Peak Brain Institute.
References
- Klimesch (1999). EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. doi:10.1016/s0165-0173(98)00056-3
- Angelakis (2007). EEG neurofeedback: a brief overview and an example of peak alpha frequency training for cognitive enhancement in the elderly. doi:10.1080/13854040600744839
- Lutz (2004). Long-term meditators self-induce high-amplitude gamma synchrony during mental practice. doi:10.1073/pnas.0407401101
- Nunez (1997). EEG coherency. I: Statistics, reference electrode, volume conduction, Laplacians, cortical imaging, and interpretation at multiple scales. doi:10.1016/s0013-4694(97)00066-7