Thomas F. Collura, Ph.D., MSMHC, QEEG-D, BCN, LPCC-S
(Note on construction: this post was drafted with Claude.ai under the author’s direction, from the BrainMaster research library and the published studies cited below. The selection, the interpretation and the opinions are the author’s.)
“When you can measure what you are speaking about, and express it in numbers, you know something about it.”
— William Thomson, Lord Kelvin, Electrical Units of Measurement, 1883
This essay looks back over a quarter century of published research carried out with BrainMaster instruments, and asks what that body of work tells us – not about a company, but about how a field comes to trust its tools. I want to make a case here that an instrument earns its standing in the same way a scientist does: slowly, in public, one study at a time, including the studies that did not come out the way anyone hoped.
An instrument earns its standing one study at a time
How do we know that a measurement is real? How do we know that the number on the screen belongs to the brain, and not to the amplifier, the electrode, the room, or the hopes of the person running the session? How would we know if two laboratories on two continents were measuring the same thing? And how would we know if they were not? These are not new questions. When Hans Berger recorded the first human EEG in 1924, he spent five years doubting his own traces before he published them, because he could not yet rule out that the “alpha” he was seeing was something his galvanometer was doing to itself. Galileo had the same problem with the telescope. His critics did not dispute what he reported; they disputed whether the instrument could be trusted to report anything at all. The answer, then as now, was not an argument. It was replication – other people, with their own copies of the instrument, looking and seeing the same thing.
That is the frame in which I would like the numbers in this post to be read. BrainMaster equipment has now been used in more than 500 published studies, by investigators in more than 40 countries. Those are not testimonials; they are something closer to a long-running inter-laboratory comparison that nobody organized. A carpenter’s rule that has been borrowed by five hundred carpenters, in forty countries, building everything from cabinets to bridges, has been tested in a way no single carpenter could ever test it. Each study used the instrument for its own purposes, under its own protocol, with its own skeptical reviewers. The instrument was never the subject of the study. It was the thing everybody had to be able to rely on so that the study could have a subject at all.
When we began, in the 1990s, the idea that a clinician might own a research-grade EEG amplifier, and that a university might use the same amplifier for a randomized trial, was not obvious. EEG was a hospital instrument, and neurofeedback was a practice instrument, and the two did not talk to each other. I do not think our field has fully absorbed how much changed when they began to share hardware. It is a bit like the moment the same thermometer began to be used in the kitchen and in the chemistry laboratory: the kitchen got better thermometers, and the laboratory got a great many more people taking temperatures.
Measuring against the typical brain
The single idea that has done the most to connect assessment with training is the live Z-score. The concept is simple to state and was not simple to build. A quantitative EEG gives us hundreds of numbers – power in each band at each site, the ratios among them, coherence and phase between every pair of sites, and so on. A normative database tells us, for each of those numbers, where a typical person of the same age would fall. Turning each measurement into a Z-score expresses it as a distance from typical. Doing that in real time, every fraction of a second, and feeding it back, lets the brain train toward the center of its own reference population (Collura et al., 2009, 2010).
I have often compared this to a GPS. The older neurofeedback protocols were like giving a traveller a single compass heading – raise this band, lower that one – and trusting that the destination lay in that direction. A Z-score system is more like a map with your current position on it. It does not tell you which road to take; it tells you how far you are from where you are likely to function most flexibly, along every dimension at once. And it is worth repeating, because it is so often misunderstood, that a Z-score of zero is not “normal” in any moral or diagnostic sense. It is typical. The center of the distribution is simply the region where a brain has the most room to move in every direction – the middle of the playground, not the fence.
What has struck me, watching this literature accumulate, is how consistently independent groups have taken the method somewhere new. Wigton and Krigbaum (2019) applied 19-channel, whole-head Z-score training in a clinical setting and reported significant improvements in attention, executive function and behavior, together with measurable movement of the electrocortical measures themselves toward typical. Just as important, Krigbaum and Wigton (2015) published a method for tracking that movement session by session in a single-subject design – the equivalent of keeping a ship’s log rather than only reporting the port of arrival. At the Defense and Veterans Brain Injury Center, Hershaw and colleagues (2020) used live Z-score training with service members who had persistent post-concussive symptoms, found it feasible and well tolerated, and saw symptom improvements that held at three months. A follow-up analysis then showed that the EEG changes during training predicted the symptom changes (Hershaw & Hill-Pearson, 2022) – which is exactly the relationship one would hope to see if the training is acting on the measurement and the measurement means something. Im (2023) reported that live Z-score training was feasible and acceptable for adults with PTSD.
The normative side has grown as well. Wu and Lin (2023) used the 24-channel Discovery amplifier to build a Taiwanese resting-state EEG normative database and cross-validated it against patients with major depressive disorder. That matters more than it might first appear. A normative database is a portrait of a population, and a portrait drawn in one country does not automatically resemble the people of another. Every new database built on the same instrument is another reference map drawn with the same ruler, and the more such maps exist, the more honestly we can say where any one brain actually stands.
What the trials have taught us
A field matures when its tools are put into the hands of people trying to prove it wrong. The randomized, placebo-controlled trial is the scientific equivalent of a load test on a bridge. Nobody builds a bridge hoping it will fail the test, but nobody trusts one that has never been loaded. BrainMaster amplifiers have been used in a good number of these trials, and I want to be candid about what they found, because the honest record is more valuable than a flattering one.
In ADHD, Meisel and colleagues (2013) compared neurofeedback with standard pharmacological treatment in a randomized trial with six-month follow-up and found the two comparable on the primary outcomes, with neurofeedback showing additional academic gains. Van Dongen-Boomsma and colleagues (2013), in a carefully blinded, placebo-controlled trial in children, did not find the active training superior to placebo on their primary measure. I count both of those as contributions. An instrument that can only produce good news is not an instrument; it is a mirror. The value of a well-made amplifier is precisely that, when a trial is negative, nobody has to wonder whether the hardware was the reason. A null result you can trust is worth more to the field than a positive result you cannot.
The applications have been remarkably broad. Stokes and Lappin (2010) followed 37 migraine patients treated with neurofeedback and biofeedback, and Walker (2011) reported on QEEG-guided neurofeedback for recurrent migraine. More recently, Dinc and colleagues (2025) examined alpha/theta training and its effects on mood, anxiety, emotion regulation and impulsivity. And the chronic pain work from the University of Otago, which I will return to below, has brought sham-controlled designs to infraslow and connectivity training. Taken together, these studies read less like a sales list and more like a geological core: layer after layer, laid down by different hands over different years, recording what the instrument was asked to do and what it found.
Safety deserves a sentence of its own, because it is the first question any responsible clinician or ethics board asks. In our own review of the 36 studies in the research library that met the Oxford evidence-based-medicine Level 2 standard – 2,292 participants in all – there were no adverse events and no device or performance issues reported, against seventeen transient side effects of the kind one expects from any training that asks the brain to do something new (BrainMaster Technologies, 2022). The 2025 randomized placebo-controlled trial in fibromyalgia likewise reported no adverse events (Anderson et al., 2025). An instrument, like a good anesthesiologist, is most valuable when nobody notices it at all.
Closing the loop
So where are we headed? For most of its history, neurofeedback has been a conversation in which the brain does the talking and a person does the listening, with a feedback display in between. What is changing now is that the instrument is becoming able to answer back – to compute features of the EEG in real time and use them to decide, moment by moment, what to do next. Norbert Wiener called this cybernetics: control by feedback, the governor on a steam engine that opens the throttle when the engine slows and closes it when the engine races. A thermostat does not heat a room. It decides when the furnace should heat it. Closed-loop neuromodulation is the same move, made with brain signals.
The Otago group has shown what this can look like in a disciplined research program. Mathew and colleagues (2022) ran a randomized, double-blind, sham-controlled feasibility trial of source-localized infraslow neurofeedback in people with painful knee osteoarthritis, and a secondary analysis (Mathew et al., 2024) showed that the closed-loop training altered cortical activity and connectivity in the targeted regions, with connectivity changes tracking the pain measures. Adhia and colleagues (2022) reported that infraslow training altered effective connectivity in chronic low back pain. Anderson and colleagues (2025) went a step further and trained directed alpha-band connectivity – from the pregenual anterior cingulate toward somatosensory cortex – in fibromyalgia. Here the thing being fed back is not the loudness of any one region, but the direction of traffic between two of them. It is the difference between listening to a single instrument in an orchestra and listening to whether the strings are following the conductor.
The same real-time machinery opens two further doors. The first is cross-frequency coupling. We have known for some time that the brain’s rhythms are not independent; a slow rhythm can set the timing for bursts of a fast one, as a drummer sets the beat on which the guitarist places each phrase. Phase–amplitude coupling measures that relationship directly, and computing it in real time makes it possible, at least in principle, to train the coupling itself rather than either rhythm alone. The second is stimulation. Pulsed electromagnetic fields, photobiomodulation, and haptic, auditory and visual stimulation can all be synchronized with the EEG, gated by it, or evaluated against it. Lee and Tramontano (2025), for example, used EEG with eLORETA source analysis to measure the effects of 40 Hz transcranial alternating current stimulation over the left frontoparietal network. Without a measurement in the loop, stimulation is a sprinkler on a timer – watering the lawn whether or not it rained. With the measurement in the loop, it becomes irrigation that knows the weather.
I want to be careful here. These stimulation and closed-loop applications are research and wellness territory, and they are early. It is not inconceivable that, in ten years, the most important thing an EEG amplifier does will be to tell a stimulator when not to fire. But that case has to be built the way every other case in this post was built – study by study, with sham arms, with negative results reported alongside positive ones, and with an instrument that nobody has to second-guess.
What a standard is for
A standard is not a boast. It is a promise that a measurement made today, in one place, will mean the same thing as a measurement made tomorrow, somewhere else. The meter was once a platinum bar in a vault outside Paris; the point was never the bar. The point was that a surveyor in Chile and a machinist in Ohio could argue about a length and settle it. Our field has needed the same thing for its signals – amplifiers that behave identically from unit to unit and year to year, files that open the same way in every laboratory, and normative references that can be checked against one another. Twenty-five years and five hundred studies later, that is what I believe the BrainMaster literature represents: not a single result, but a shared ruler.
Imagine a field in which a finding in Dunedin can be tested in Barcelona the following year on the same hardware. Imagine a clinician in a small practice reading a trial from Fort Carson and knowing that the instrument on her own desk recorded the same kind of signal. Imagine a graduate student designing a closed-loop protocol that has never been tried, and spending her doubt on the hypothesis rather than on the amplifier. That field is largely here. What remains is the harder and more interesting part – deciding what to measure next, and what to do with what we find.
Kelvin was right that measurement is where knowledge begins. He did not say where it ends.
The instrument is ready. The questions are ours.
BrainMaster systems are cleared for their indicated uses as EEG, qEEG, biofeedback, and normative database systems. Citing published articles does not imply clinical claims. Stimulation and closed-loop applications are described for research and wellness use.
References
Adhia, D. B., Mani, R., Turner, P. R., Vanneste, S., & De Ridder, D. (2022). Infraslow neurofeedback training alters effective connectivity in individuals with chronic low back pain: A secondary analysis of a pilot randomized placebo-controlled study. Brain Sciences, 12(11),
Anderson, L., De Ridder, D., Glue, P., Mani, R., van Sleeuwen, C., Smith, M., & Adhia, D. B. (2025). A safety and feasibility randomized placebo controlled trial exploring electroencephalographic effective connectivity neurofeedback treatment for fibromyalgia. Scientific Reports, 15, 209. https://doi.org/10.1038/s41598-024-83776-8
BrainMaster Technologies. (2022). BrainMaster literature search database: Evidence-based medicine level 2 studies, summary statistics [Internal post-market clinical follow-up report].
Collura, T. F., Guan, J., Tarrant, J., Bailey, J., & Starr, F. (2010). EEG biofeedback case studies using live Z-score training and a normative database. Journal of Neurotherapy, 14(1), 22–46. https://doi.org/10.1080/10874200903543963
Collura, T. F., Thatcher, R. W., Smith, M. L., Lambos, W. A., & Stark, C. R. (2009). EEG biofeedback training using live Z-scores and a normative database. In T. H. Budzynski, H. K. Budzynski, J. R. Evans, & A. Abarbanel (Eds.), Introduction to quantitative EEG and neurofeedback: Advanced theory and applications (2nd ed., pp. 103–141). Academic Press.
Dinc, L., Rybski, D., & Dineen, J. (2025). The effects of alpha/theta neurofeedback on mood, anxiety, emotion regulation and trait impulsivity. Brain Research, 1866, 149943.
Hershaw, J. N., & Hill-Pearson, C. A. (2022). Changes in EEG activity following live Z-score training predict changes in persistent post-concussive symptoms: An exploratory analysis. Frontiers in Neurology, 13, 714913. https://doi.org/10.3389/fneur.2022.714913
Hershaw, J. N., Hill-Pearson, C. A., Arango, J. I., Souvignier, A. R., & Pazdan, R. M. (2020). Semi-automated neurofeedback therapy for persistent postconcussive symptoms in a military clinical setting: A feasibility study. Military Medicine, 185(3–4), e457–e465. https://doi.org/10.1093/milmed/usz335
Im, S. (2023). Live Z-score neurofeedback training for PTSD: A feasibility and acceptability study. Trauma Care, 3(3), 154–160. https://doi.org/10.3390/traumacare3030015
Krigbaum, G., & Wigton, N. L. (2015). A methodology of analysis for monitoring treatment progression with 19-channel Z-score neurofeedback (19ZNF) in a single-subject design. Applied Psychophysiology and Biofeedback, 40(3), 139–149. https://doi.org/10.1007/s10484-015-9274-0
Lee, T.-W., & Tramontano, G. (2025). Neural changes to transcranial alternating current stimulation in the gamma range over the left frontoparietal network: A preliminary eLORETA EEG study [Preprint]. medRxiv. https://doi.org/10.1101/2025.01.24.25321077
Mathew, J., Adhia, D. B., Smith, M. L., De Ridder, D., & Mani, R. (2022). Source localized infraslow neurofeedback training in people with chronic painful knee osteoarthritis: A randomized, double-blind, sham-controlled feasibility clinical trial. Frontiers in Neuroscience, 16, 899772. https://doi.org/10.3389/fnins.2022.899772
Mathew, J., Adhia, D. B., Smith, M. L., De Ridder, D., & Mani, R. (2024). Closed-loop infraslow brain–computer interface can modulate cortical activity and connectivity in individuals with chronic painful knee osteoarthritis: A secondary analysis of a randomized placebo-controlled clinical trial. Clinical EEG and Neuroscience. Advance online publication. https://doi.org/10.1177/15500594241264892
Meisel, V., Servera, M., Garcia-Banda, G., Cardo, E., & Moreno, I. (2013). Neurofeedback and standard pharmacological intervention in ADHD: A randomized controlled trial with six-month follow-up. Biological Psychology, 94(1), 12–21. https://doi.org/10.1016/j.biopsycho.2013.04.015
Stokes, D. A., & Lappin, M. S. (2010). Neurofeedback and biofeedback with 37 migraineurs: A clinical outcome study. Behavioral and Brain Functions, 6, 9. https://doi.org/10.1186/1744-9081-6-9
van Dongen-Boomsma, M., Vollebregt, M. A., Slaats-Willemse, D., & Buitelaar, J. K. (2013). A randomized placebo-controlled trial of electroencephalographic (EEG) neurofeedback in children with attention-deficit/hyperactivity disorder. Journal of Clinical Psychiatry, 74(8), 821–827. https://doi.org/10.4088/JCP.12m08321
Walker, J. E. (2011). QEEG-guided neurofeedback for recurrent migraine headaches. Clinical EEG and Neuroscience, 42(1), 59–61. https://doi.org/10.1177/155005941104200112
Wigton, N. L., & Krigbaum, G. (2019). Attention, executive function, behavior, and electrocortical function, significantly improved with 19-channel Z-score neurofeedback in a clinical setting: A pilot study. Journal of Attention Disorders, 23(4), 398–408. https://doi.org/10.1177/1087054715577135
Wu, Y.-C., & Lin, I.-M. (2023). The resting state of Taiwan EEG normative database: Z-scores of patients with major depressive disorder as the cross-validation. Brain Sciences, 13(2), 351. https://doi.org/10.3390/brainsci13020351
Tom Collura Ph.D., MSMHC, QEEG-D, BCN, NCC, LPCC-S, Founder BrainMaster Technologies, Inc.
