The Landscape of Neurofeedback Methods

Part 1 — A practitioner's introduction to the Types of Neurofeedback series

At nearly every workshop I teach, someone — usually a clinician three or four months into their first neurofeedback training — asks me a version of the same question. Sometimes, they don’t even know they’re asking it.

When they dont, the question looks something like: I was thinking of buying NeuroBlahBlah, can you teach me how to use it?

When they do: Which type of neurofeedback should I be doing?

They are good questions. They are also questions I find harder to answer than I used to.

Not because the answer isn’t clear… it’s really, really clear to me.

These questions are truly difficult to answre because the field has gotten more crowded. The lines between techniques can be blurry. (Sometimes this is because methods overlap, other times because some methods piggyback on other’s science. I’ll let you go ahead and guess which one drives me nuts.)

In the early years, "neurofeedback" meant one thing. It quickly grew to a small cluster of overlapping practices, all rooted in operant conditioning and/or voluntary self-regulation of EEG activity measured from the scalp. Today the term covers everything from a clinician slowly shaping SMR amplitude over a sensorimotor strip (just like Sterman did back in the day) with one or two electrodes to a closed-source consumer headband that promises to "balance your brain" while you watch a video.

These are not the same kind thing.

To clinicians coming into the field — and to referring professionals trying to figure out where to send a client — that landscape can feel less like a map and more like an overcrowded marketplace. There are confident voices on every side, vendors with strong opinions about whose method is "real" neurofeedback, and a steady drip of new devices and acronyms that arrive faster than the literature can evaluate them. (You can go read our piece on what Scientifically validated means and the one on How to read a scientific claim for a lot more sharpening of your critical thinking instincts in the context of Neurofeedback and Neuroscience in general.)

That gap is what I’m writing this series to close.

Neurofeedback is, at its core, a learning-based method: clients learn to modulate aspects of their own brain activity through real-time feedback, supported by reinforcement, attention, and a coaching relationship with a trained practitioner. (Biofeedback applies the same principle to peripheral physiology — heart rate, respiration, skin conductance, muscle tension.) That core does not vary across methods. What varies is everything around it — what gets measured, how the measurement gets mapped to feedback, how individualized the protocol is, how reproducible the procedure is, and how robust the supporting science is.

This first piece sets up the series by doing three things. It traces, briefly, how the field arrived at today's landscape. It explains why we need a unified, fair taxonomy — and why building one is harder than it looks. And it previews the eight deep-dives that follow, so you know what is coming and how to read each entry.

A short history of how we got here

Neurofeedback didn't begin as a method. It began as an experiment.

In the 1960s, Joe Kamiya at the University of Chicago demonstrated that healthy volunteers could learn to identify and modulate (as I conceptualize it: via conscious and voluntary self-regulation) their own alpha rhythm using auditory feedback. The finding was striking enough on its own — that conscious awareness of an EEG state was even possible — and it opened the door to a broader idea: if you can give the brain feedback about its own activity, the brain can learn.

Around the same time, Barry Sterman, working with cats at UCLA, discovered that operant conditioning of the sensorimotor rhythm (SMR, a 12–15 Hz activity over the central strip) raised the seizure threshold. When that finding was extended to humans with epilepsy in the early 1970s, it became the first robust clinical demonstration of EEG biofeedback. Joel Lubar and others extended the work into attention deficits soon after, and by the late 1970s the term "neurofeedback" was beginning to take on the clinical shape we recognize today.

For a long stretch, neurofeedback was largely a single practice somewhere along this spectrum of operant conditioning—conscious, voluntary self-regulation. Brain activity measured by surface EEG. One or two channels. Frequency bands chosen on the basis of clinical interview, behavioral profile, EEG observations (raw signal and spectral analysis), and a growing case-series literature.

Then several things happened in parallel.

Quantitative EEG matured into a clinically usable tool, allowing protocols to be guided by individual brain-activity profiles rather than chosen from a generic menu. Source-localization techniques — LORETA and its variants — made it possible to target activity in deeper or more distributed structures rather than at a single scalp location. Connectivity and coherence metrics opened up training of network-level properties rather than amplitude in isolation. The infra-low-frequency tradition, associated most prominently with the Othmer family, departed from amplitude-band thinking entirely. And a parallel ecosystem of closed-source and consumer-facing systems — automated devices, dynamic-feedback algorithms, home headbands — entered the conversation, often with marketing claims that outpaced the evidence supporting them.

By 2026, "neurofeedback" was no longer one method. It was at least eight, plus several emerging variants, plus a lay-facing market that often blurred all of them together.

Why a unified taxonomy is hard

If you spend an afternoon trying to write a clean taxonomy of neurofeedback methods, you discover the same thing I did the first time I tried: the methods do not sort along a single axis (anymore).

Some differences are about what is measured — amplitude in a frequency band, the ratio between two bands, the coherence between two locations, source-localized current density, hemodynamic signal in fMRI or fNIRS. Some are about how localized the measurement is — a single channel, a multi-channel cap, a network reconstruction. Some are about how the protocol gets chosen — from a normative database, from clinical reasoning, from an automated algorithm, from a one-size-fits-all default. And then there is delivery model — clinic-based, home-based with supervision, fully consumer.

Most attempts to "rank" types of neurofeedback flatten those dimensions into a single line. Best to worst. Traditional to advanced. Scientific to marketing-driven. That works as a rhetorical device. It does not work as a clinical tool.

A method that sits at the methodologically demanding end of the spatial-resolution dimension might still be wrong for a particular clinical question. A method that sits at the simpler end might be exactly right. The honest version of the taxonomy is messier than a leaderboard. Each method has a profile across several dimensions, and a practitioner's job is not to rank methods — it is to read those profiles against the clinical question in front of them, the infrastructure they can resource, the supervision they can access, and the evidence base relevant to the population they serve.

That kind of reading is what this series is meant to teach.

The methodological core

That said, not every method sits at equal distance from the center of the research literature.

When the field talks about what works — when a meta-analysis aggregates effect sizes for ADHD, when a regulatory body weighs whether to recognize a treatment, when a referring physician asks whether neurofeedback is "evidence-based" — the conversation is supposed to largely be about a particular kind of method. Operant conditioning of EEG activity. Electrode placement and frequency-band targets chosen on the basis of an individualized assessment. A transparent protocol that another practitioner could read and reproduce. Delivery by a trained practitioner in a clinical relationship, with progress tracked across sessions and reassessed periodically.

That is the core. Most of the published clinical evidence in neurofeedback — for ADHD, anxiety, PTSD, OCD, depression, epilepsy — comes from one of the methods that match that description. But even in that narrower range, there are a lot of techniques that all end up amalgamated into “neurofeedback” (without mentioning the growing list of neuroimaging tools used for neurofeedback: MEG, PiR, fNIRS, fMRI…)

Around that core sit several research-aligned extensions. LORETA-based source-localized training, coherence and connectivity protocols, infra-low-frequency training, and SCP each push the core into different mechanistic territory, and each carries its own published literature — usually thinner than the amplitude-training base, but not absent. These are best read as variants of the same paradigm rather than departures from it.

And then there is everything else.

Several products and services on the market today use the word "neurofeedback" without following the methodological discipline that the research literature describes. No individualized assessment. No transparent protocol. Sometimes no practitioner in the feedback loop at all. The marketing claims tend to be broad; the published evidence supporting those specific claims tends to be narrow or absent. This is not a moral failing on anyone's part — it reflects a real tension between what a clinical-research method requires and what a low-friction consumer product can deliver. But it does mean that "neurofeedback" as a marketing term and "neurofeedback" as a research-validated method have drifted further apart than is healthy for the field.

The deep-dives ahead (our next parts in the series) will sit each method on that map. Some will fall close to the core. Some will sit at one of the orbits. Some will sit further out, and the post will say so without pretending otherwise.

Equipment is not method-neutral

A practical point that deserves naming directly: the equipment a practitioner uses constrains the methods they can practice.

Open, professional-grade systems — research-tier amplifiers made by companies like Thought Technology and Mitsar (my preferred tools) and paired with software like BioGraph Infiniti or WinEEG, support a wide range of methods. Amplitude training, bipolar montages, coherence work, source-localized training, SCP, ILF — these can all be run within the same hardware-and-software footprint, depending on the practitioner's skill and the analytic add-ons. The same setup that runs a single-channel SMR protocol can run a 19-channel qEEG and feed a LORETA-based session.

Closed, consumer-facing systems are different. There are a few closed “clinical” grade systems too. (Pardon the quotation marks, but it is my opinion that any system targeting clinical practise must offer a high level of transparency and adaptability. Many are designed around one method (often a proprietary algorithm), one interaction model (passive viewing), and one feedback delivery scheme. The choice of equipment, in those cases, is also a choice of method — and often a choice to remove the practitioner (or at the very least their expertise) from the loop.

That last part matters more than it sounds.

Neurofeedback, in the research-validated form, is not just real-time data piped into a screen. It is a learning process structured around a therapeutic relationship. The practitioner sets thresholds, watches the client's behavior and physiology, adjusts the session in flight, debriefs afterward, and connects what happened in the session to what the client reports between sessions. Take the practitioner out of that loop — by automation, by design, by a delivery model that simply does not include them — and the intervention being delivered is no longer the same intervention the research literature has been studying.

That is one of the reasons the equipment-vs-method thread will run through several of the deep-dives. The hardware and software a practitioner buys is shaping what method they can practice, and what role they and their client are allowed to play inside that method. It is not a neutral choice.

What this series does

Across the next several entries, I'll work through eight core neurofeedback methods. For each method, I will use the same eight-section structure:

• Intro and brief history — where the method came from, who developed it, what problem it was originally trying to solve.

• Alternate names — what you'll see this method called in the literature, in marketing, and in other practitioners' notes. Many methods carry several names; sorting them out is the first practical step toward reading the field clearly.

• How the method works — the practical procedure: electrodes, signal extraction, feedback delivery, session structure.

• Mechanistic specifics — what is being trained at the level of brain activity, and what neural mechanism is being recruited or shaped. This is where method-vs-method differences become real.

• Overview of the science base — what the literature shows, in which conditions, with what effect sizes and confidence intervals where reported. Where the evidence is mature, where it is emerging, and where it is still thin.

• Strengths and weaknesses — a fair summary, in list or table form, of what the method does well and where its limits sit.

• Brendan's perspective — what I think clinically, where the method fits in my own practice, and what I have learned from using it (or, in some cases, from deciding not to use it).

• Would I do this method myself? In what context? — a direct, clinical answer. Not a vague "it depends," but a concrete picture of when this method earns a place in my decision-making and when it does not.

That structure is opinionated by design. Most existing overviews of neurofeedback methods either describe each method on its own terms, which makes comparison impossible, or rank them against each other on a single axis, which collapses real distinctions. Any graph style presentation would be oversimplifying to a degree below the standards I’d accept; I won’t do it. The eight-section structure is built to do something different — to give each method its fair turn under a consistent set of questions, so you can read across the series and form your own working view.

It is also designed to give working clinicians something to take into the next session. Methods are not interesting in the abstract. They are interesting because they answer a clinical question, and because a practitioner has to decide, in front of an actual client, whether this is the right tool today.

Preview of the eight deep-dives

Part 2 — Classic amplitude training. The foundational method. (Mostly) single-channel surface EEG, frequency-band reinforcement and inhibition (theta, beta, SMR, alpha-theta where appropriate), protocols guided by clinical reasoning and increasingly by qEEG. Where most of the published clinical evidence in neurofeedback actually lives. The implicit reference point against which every other method is, fairly or unfairly, compared.

Part 3 — Bipolar protocols. The montage choice that gets glossed over in many trainings. A bipolar reference shapes what the signal actually represents — a difference between two adjacent recording sites, rather than activity at a single location relative to a remote reference — and that shape determines which clinical questions the protocol can answer. Under-represented in the literature, often misunderstood.

Part 4 — Coherence and connectivity training. A different mechanistic claim altogether: targeting how brain regions coordinate rather than how active any single region is. The mathematical foundations differ across coherence, phase synchronization, and graph-based connectivity metrics, and those differences matter clinically. Hypothetically useful for network-level training. Cognitively demanding to interpret responsibly.

Part 5 — LORETA and source-localized neurofeedback. The spatial-resolution upgrade. By estimating activity in source space rather than at the scalp, LORETA and related methods let practitioners target deeper or more anatomically specific structures than surface EEG allows. The cost is acquisition complexity, normative-database dependence, and a steeper interpretive learning curve. Sounds great, right? Just keep the “inverse problem” in mind; we’re essentially training an estimated source rather than an inherently real one. I also think it’s worth asking the question: does the added spatial precision really make a significant difference?

Part 6 — Infra-low frequency training and the Othmer method. The lineage that departs from amplitude-band thinking. ILF training treats slow cortical processes as the relevant target rather than activity in a classical band, and the Othmer tradition wraps that target in a particular coaching style and clinical philosophy. Distinct rationale, distinct training experience, distinct conversations to have about evidence.

Part 7 — Slow cortical potentials (SCP) training. The Birbaumer lineage. SCP training works in the time domain rather than the frequency domain — clients learn to produce or suppress slow polarity shifts of cortical voltage on a trial-by-trial basis. Distinct paradigm, distinct training experience, and one of the better-evidenced methods in the field for specific indications, particularly in epilepsy and certain ADHD presentations. Often missed in "types of neurofeedback" surveys because it does not fit the band-amplitude template — but methodologically rigorous and clinically serious wherever it is properly resourced.

Part 8 — NeurOptimal and dynamical neurofeedback systems. The closed end of the transparency dimension. Automated, algorithm-driven feedback delivery with limited public methodological detail. This is the post that requires the most care to write fairly (I’m going to try really hard to both acknowledge my feelings and separate them from the facts)— the method does some things well (low-friction delivery, broad accessibility, low practitioner training threshold); these are things I think we can learn from. That said, methodological-transparency questions are real and need to be answered before anyone should take it seriously.

Part 9 — Home neurofeedback. Not really a method but a delivery model. Cuts across most of the methods above. The conditions under which home training is appropriate, the supervision structure that makes it defensible, the hardware tiers, and the practitioner's responsibility outside the clinic. Caps the series by widening from method-selection to practice design.

A few additional variants — alpha-theta, live z-score training, real-time fMRI neurofeedback, hemoencephalography — will appear inside the deep-dives where they fit naturally, and a small number of them may eventually earn their own entries. The eight above are the spine.

How to read the series

A few framing notes before we start.

First — this is not (really) a "what is the best type of neurofeedback" series. It is a "what does each type actually do, what is it good for, and when does it earn a place in your practice" series. If you are looking for a winner, you are going to find eight differently-shaped tools instead. I will unabashedly and unapolagetically give my opinion on each one. I invite any discussion on any method with anyone willing: I’m more than open to challenge my own ideas.

Second — every entry is going to include strengths and weaknesses, the things a method does well alongside the things it does less well. Discipline matters to me and is in NeuroLogic’s DNA. We position our work as transparent, individualized, and evidence-based, and that position is only credible if our public writing about other methods is fair. If a deep-dive ever reads like a hit-piece, the post has failed — even if every claim in it is true.

Third — while these posts are written for clinicians and for the professionals who refer to clinicians, I will do my best to make them accessible to all. They assume you have encountered neurofeedback before, even if you have not yet decided where it fits in your practice. They are not introductory content; if you get stuck on something, feel free to start back at the very beginning of NeuroBLOG, and come back here when you’ve at least skimmed through enough that you’re comfortable.

Closing

Neurofeedback became the field it is today because clinicians and researchers kept adding tools — different signals, different montages, different analyses, different delivery models — to a learning chassis that has stayed remarkably stable since Sterman's first SMR work.

That accumulation has produced real progress. It has also produced the confusion that brings someone to the back of a workshop room with the question I started this post with.

The aim of the series is not to declare a winner among the methods. It is to give you a clearer way of thinking about each one — its history, its mechanism, its evidence, its strengths, its limits, and its honest place in clinical work.

If we do that well, the next time a colleague asks you which type of neurofeedback they should be doing, you will have a better answer than I had three or four years ago.

Probably a longer one.

But also, I hope, a more honest one.

References

• Hammond, D. C. (2011). What is neurofeedback: An update. Journal of Neurotherapy, 15(4), 305–336.

• Marzbani, H., Marateb, H. R., & Mansourian, M. (2016). Neurofeedback: A comprehensive review on system design, methodology, and clinical applications. Basic and Clinical Neuroscience, 7(2), 143–158.

• Sterman, M. B., & Friar, L. (1972). Suppression of seizures in an epileptic following sensorimotor EEG feedback training. Electroencephalography and Clinical Neurophysiology, 33(1), 89–95.