Part 2 — Types of Neurofeedback series

If neurofeedback had a default setting, this would be it.

When most clinicians picture a neurofeedback session — a client wearing one or two electrodes on the scalp, a screen showing a game or a movie that gets clearer (or quieter, or brighter) when the EEG behaves a certain way, a practitioner adjusting thresholds and watching the signal — the method they're picturing is classic amplitude training. Single channel, sometimes two. Surface EEG. A frequency band reinforced upward, one or two bands inhibited. Sessions of about an hour, with anywhere from twenty to thirty (more rarely up to forty-five) minutes, repeated several times a week, across a course that typically runs from twenty to forty sessions depending on the indication.

It is the oldest organized form of clinical neurofeedback, the one that lent the field its language (reward band, inhibit band, threshold, SMR, theta/beta ratio), and the one against which every newer method ends up implicitly compared. When a clinician asks me, is this thing real?, the answer the literature gives them most clearly is built on this method — on the cumulative weight of forty-plus years of case series, controlled studies, and a handful of methodologically serious randomized trials in conditions like ADHD, refractory epilepsy, depression, generalized anxiety, and PTSD. Newer techniques have their own evidence bases (some growing fast, some still thin), but classic amplitude training is the substrate beneath the field's claim to clinical legitimacy.

Which is also why this is, in some ways, the hardest post in the series to write, because writing fairly about the foundational method means acknowledging two opposite extremes, at least as neurofeedback is commercialised and marketed. The first is the pitch that sells classic amplitude training as if every system produced by every vendor were equivalent (they are not). The second is to the contrarian counter-pitch that paints it as outdated, low-tech, or somehow superseded by newer methods that have not yet matched its evidence base (beware anything that claims to be the “newest generation of neurofeedback”). Neither version is true. The classic method has aged remarkably well; it has also accumulated some clinical baggage along the way that deserves to be named honestly.

This entry walks through the method using the eight-section structure established in Part 1: the history and lineages, the names you'll see in the literature and on practitioner forums, how the method actually works in a session, what is being recruited mechanistically, what the science base does and does not show, the strengths and weaknesses I would put on the table for any colleague considering it, where it sits in my own clinical practice, and the direct question of whether — and when — I would do it myself.

A brief history

Classic amplitude training came out of a few different rooms.

Joe Kamiya's alpha-conditioning work at the University of Chicago, in the early 1960s, was the first demonstration that healthy volunteers could learn to identify and modulate their own EEG state through real-time feedback. The work was quiet, scientifically careful, and largely cognitive in its interests — Kamiya was interested in whether internal states could be brought into conscious awareness — but it opened the conceptual door for everything that followed. Brain activity could be felt, identified, and shaped, given the right feedback signal.

Barry Sterman's work at UCLA, with cats first and then with humans, gave the field its first robust clinical application. Sterman's group showed that operantly conditioning the sensorimotor rhythm (a 12–15 Hz oscillation over the central strip, distinguishable from mu and from alpha by site and behavioral correlate) raised the seizure threshold in cats exposed to a convulsant substance, and that the same conditioning protocol, transferred to humans with refractory epilepsy in the early 1970s, reduced seizure frequency, durating and/or intensity in patients who had not responded to medication. That was the first time neurofeedback had a target condition, a clear mechanistic story, and a published clinical demonstration in the same place at the same time.

Joel Lubar's group at the University of Tennessee took the same operant-conditioning logic into attention and learning. By the late 1970s and through the 1980s, Lubar's work — and the work of the practitioners and researchers he trained — had established theta/beta amplitude training as a clinically usable protocol for what was then called attention deficit disorder and is now ADHD. Inhibit theta, reinforce beta, train over the central or frontal strip, run several dozen sessions, and a meaningful proportion of children improved on attention measures, behavior ratings, and, in some studies, on academic outcomes. The Lubar lineage is where most of the field's ADHD evidence came from for the next thirty years.

The third major lineage runs through Eugene Peniston and Paul Kulkosky's alpha-theta work with veterans with PTSD and alcohol-use disorder, in the late 1980s, and into John Gruzelier's later cognitive-performance and creative-performance studies in healthy musicians, dancers, and surgeons. Alpha-theta training is mechanically still amplitude training — eyes closed, reinforcing alpha and theta amplitudes over (typically) posterior or central sites — but the state it recruits is different (a hypnogogic, low-arousal state at the edge of sleep) and the clinical use of that state is different (trauma processing, narrative work, performance integration). Gruzelier's work in particular pulled alpha-theta out of the addiction/PTSD context and into the cognitive and performance literature, where the published effect sizes are smaller but the methodological designs have generally been cleaner.

There are more lineages — Peter Rosenfeld's frontal alpha asymmetry work for depression, Joel Lubar's separate line on focused attention, the European tradition of theta/beta and SMR work coming out of Holger Gevensleben's group in Germany — but those four (Kamiya, Sterman, Lubar, Peniston/Gruzelier) are the spine. Every clinician working in classic amplitude training today is downstream of at least one of them, often of all four.

Alternate names

The method goes by several names in the literature and on practitioner forums, and the names overlap inconsistently. A short orientation:

• Classic neurofeedback, sometimes traditional neurofeedback or standard neurofeedback. Quantitative, spectral, frequency, and other similar terms are often used to denote this branch of neurofeedback training. Used most often to distinguish amplitude-band training from newer methods (LORETA, ILF, dynamical systems, etc.). The terms are mostly fine but a little freighted — classic can sound dismissive if you're a fan of one of the newer methods, and traditional can sound dismissive if you're a fan of amplitude training. I tend to default to classic amplitude training in writing, because it names what is actually being trained.

• Amplitude training, band-amplitude training, frequency-band training. All accurate descriptions of the method itself. The first is the cleanest.

• Surface EEG neurofeedback. Names the recording modality rather than the analytic target — but in practice, surface EEG and amplitude training are usually co-occurring (a few exceptions: bipolar protocols, which are also surface; coherence training, which is surface but not amplitude-based).

• Operant EEG conditioning. The mechanistic name. Accurate, and useful when writing for an audience that hasn't encountered the clinical vocabulary.

• Theta/beta training, SMR training, alpha-theta training. Sub-types named by their training bands. You will see these in the ADHD literature, the epilepsy and sleep literature, and the PTSD/addictions literature respectively.

• qEEG-informed and qEEG-based neurofeedback. These two terms get used interchangeably in the field, and they should not be. qEEG-informed describes a workflow where a quantitative EEG assessment is used to adapt a research-derived protocol — taking the standard Lubar theta/beta target and shifting it medially or laterally, or moving from a generic SMR site to the side where the assessment shows the relevant departure. The literature default is the anchor; the qEEG is the adjustment. qEEG-based describes a workflow where the protocol itself is derived from the assessment — site and band targets are determined by the individual's qEEG profile rather than adapted from a literature default. The literature is contextual; the qEEG is the anchor. The two are not equivalent, clinically or methodologically. Both have a place, and the choice between them should be a deliberate clinical decision. In practice, qEEG-guided is the umbrella that gets used for both — when I use it in writing, I am usually pointing at qEEG-informed, but the field's vocabulary is muddier than the underlying distinction deserves. I think the two terms, informed and based, more accurately describe a spectrum of protocol “individualisation” that is important to address for each part of each individual training process.

• Linear neurofeedback. This one is a (sarcastic, tongue-in-cheek) favourite. A term used almost exclusively by NeurOptimal marketing to describe everything that is not NeurOptimal. NeurOptimal is self-labelled as “dynamical” (which is honestly a word that makes my arm hair stand on end); so they need to paint other methods in a separate label to make the distinction clear; hence “linear”. The label does not survive examination — linear, in signal-processing or systems vocabulary, names a specific class of mathematical operation (one where output scales proportionally with input), and the methods being labeled linear here include some that are obviously nonlinear (the operant conditioning of an amplitude relationship over time is not a linear input-output mapping in any usual technical sense) and exclude nothing that is genuinely linear in the technical meaning. Even applying a simple Z-score calculation makes this work “non-linear”. The term is a marketing dichotomy dressed in engineering vocabulary. I do not use it; you should not either. If you encounter it elsewhere, that is the context to read it in.

When you see a method described in a paper or in a practitioner's notes, the first practical exercise is to translate it into the technical vocabulary — what band, at what site, with what reference, in what direction, with what threshold logic — rather than to accept the lineage label at face value. Two protocols both called theta/beta training can differ on every one of those parameters, and the differences are hugely important.

How the method works

In its simplest configuration: one or two active electrodes, placed on the scalp at a clinically chosen site (often C3, C4, Cz, F3, F4, Fz, or the central strip more broadly), referenced to a linked-ear or mastoid reference, with a ground at the forehead or another standard site. The EEG signal is amplified, filtered, and decomposed into its frequency components in real time. The clinician selects one or two reward bands (where amplitude in a target frequency range should go up) and one or two inhibit bands (where amplitude in a target frequency range should go down). Thresholds are set, usually to deliver reinforcement roughly 60–80% of the time in early training, and the feedback signal — a game element, a video that gets clearer, an audio tone, sometimes all three — responds to whether the moment-to-moment amplitude relationships are in or out of the target zone.

A session is typically twenty to thirty minutes of active training (more rarely up to fourty-five), broken down into trials or runs of (typically) 3 to 5 minutes, the whole training portion preceded by a brief baseline recording and each trial followed by a short debrief. Sessions are scheduled two to three times a week in most clinical practices, though research protocols have used everything from once a week to daily. A course of training for a clinical indication usually runs twenty to forty sessions, with periodic re-assessment (qualitative review of the EEG, behavioral measures, sometimes a follow-up qEEG) along the way. There is no concensus on what frequency/intensity of training is best. My experience tells me that twice a week is necessary and sufficient for most.

The variables a clinician adjusts, within and across sessions, are familiar to anyone who has trained: site, reference, reward and inhibit bands, threshold tightness, ratio of reinforcement, the type of feedback (visual, auditory, mixed), the cognitive task the client engages with during training (eyes open vs eyes closed, attentional task, narrative content), and the pacing and structure of the session itself. None of these variables are trivial. A protocol described in a paper as theta/beta training at Cz has at least a dozen implementation details underneath it, and two practitioners running theta/beta training at Cz are not necessarily running the same intervention.

That detail-level variability is one of the reasons the published evidence base, even within classic amplitude training, is harder to summarize than it looks. We will come back to this in the Science base section, because it matters both for how the literature is read and for what a clinician should expect when they bring the method into their own practice.

The three best-known sub-protocols are worth a sentence or two each:

SMR training sits over the sensorimotor strip — C3, Cz or C4 (or along the FC or CP bands) — and reinforces amplitude in the 12–15 Hz band, with inhibits on slower (theta, 4–8 Hz) and faster (high beta, 22–35 Hz) bands depending on the indication. The state recruited is one of relaxed but alert focus: physically still, sensorily quiet, cognitively engaged but not effortful. Used clinically for refractory epilepsy (the Sterman lineage), sleep onset and continuity, attentional and impulsivity components of ADHD, and certain performance applications.

Theta/beta training reinforces beta (typically 15–18 Hz or 16–20 Hz, with variants) and inhibits theta (4–8 Hz), usually over the central or frontal strip. The state recruited is one of sustained, engaged attention — the kind of cognitive activation that someone working through a math task or an interesting book is in. Used clinically for ADHD-inattentive and combined presentations, certain anxiety states with attentional disorganization, and post-concussive cognitive difficulties.

Alpha-theta training reinforces both alpha (8–12 Hz) and theta (4–8 Hz), most often at Pz or Oz, with eyes closed and with the client guided into a low-arousal, hypnogogic state at the edge of sleep. The state recruited is one of deep relaxation with emergent imagery — not sleep, not ordinary wakefulness, but a particular state in between. Used clinically for PTSD and trauma work (Peniston-Kulkosky), addiction recovery, and creative-performance training (Gruzelier).

The shared architecture across all three — surface EEG, amplitude in clinically chosen bands, operant conditioning of the relationship between those amplitudes — is what makes it accurate to call all three classic amplitude training. The state being recruited, and the clinical use of that state, is what makes the sub-protocols genuinely different interventions rather than cosmetic variants.

Mechanistic specifics

What is actually being trained, mechanistically?

At the level of brain activity: a particular relationship between amplitudes in selected frequency bands, at a particular cortical site, sustained over time. SMR reinforcement at C3 trains the client to spend more time in a state where the 12–15 Hz oscillation over the sensorimotor strip is elevated relative to its own baseline (and relative to neighboring bands, when those are inhibited). Theta/beta training at Cz trains the client to spend more time in a state where beta amplitude exceeds threshold while theta amplitude stays below threshold. Alpha-theta training at Pz trains a state where alpha and theta amplitudes both exceed threshold simultaneously — a configuration that, in healthy adults at rest with eyes closed, is associated with relaxed alertness sliding toward drowsiness.

At the level of mechanism: the operant-conditioning model — reinforcement of an emergent state, gradual shaping of the underlying neural configuration, eventual transfer of the trained state into voluntary control outside of training — is the most parsimonious framework, and it has the most empirical support. But the operant model is not the whole story. Classical conditioning components (paired associations between the trained state and the feedback environment), instructional learning (the client's mental rehearsal of the state, especially when given a behavioral strategy), and the broader context effects (relationship with the practitioner, expectancy, attentional engagement during sessions) all contribute in ways that are difficult to disentangle in a typical clinical course.

A few mechanistic specifics worth naming, because they shape the clinical work:

State-specific recruitment. Each sub-protocol is recruiting a different neural and physiological state, not a different amount of one thing. SMR is a state of sensorimotor inhibition. Theta/beta is a state of cognitive engagement. Alpha-theta is a state of low-arousal interoceptive openness. The reinforcement is shaping the brain's willingness to go to that state, not just the amplitude reading on the screen.

Learning curves are real, individual, and not always linear. Some clients show measurable changes in band amplitudes within the first few sessions. Others plateau for ten or more sessions before something shifts. A few never produce the expected amplitude change but show clinical improvement anyway, which is one of the open mechanistic questions in the field. The relationship between learning to produce the trained state in session and clinical outcome is not as tight as some early models assumed it would be — and that gap is something honest practitioners need to keep in mind when they read their clients' progress.

Transfer is not automatic. The state trained inside a session does not automatically generalize to the client's daily life. Transfer tasks — structured exercises that link the trained state to specific behavioral or cognitive contexts outside the clinic — are part of what makes the method work clinically. (NeuroLogic's Barry and Barry+ tools, (built inside of Thought Technology’s BioGraph Infiniti platform,) and most other practitioner-grade software platforms, structure transfer explicitly because the method is incomplete without it.)

Specificity is a serious empirical question. The question of whether the reinforced band is doing the clinical work, or whether the broader state and context effects are sufficient on their own, is one of the more sophisticated debates in the field. I have my own paper out on the question of control conditions in neurofeedback trials (Discover Neuroscience, 2026) and you can read more if you'd like — but the short version is that the specificity question is real, the answer depends on the indication and the protocol, and clinicians should hold both the trained band probably matters and some of the clinical effect is broader than the band alone in mind at the same time.

Overview of the science base

This is the longest section, by necessity, and the one with the most ground to cover. I can’t cover every application - the AAPB has a great publication to that end, or if you prefer just take a look through the NeuroBLOG and you’ll find what you’re looking for (at least in 90% of applications). I'll give the abbreviated tour, indication by indication, and flag where the evidence is mature, where it is emerging, and where it is still thin.

ADHD is the most studied indication. The Lubar-tradition theta/beta protocols, and SMR variants, have accumulated several dozen randomized and quasi-randomized studies and a series of meta-analyses over the past twenty-five years. The headline finding, across the meta-analyses by Arns and colleagues (2009, 2014), Cortese and colleagues (2016, with the European ADHD Guidelines Group), Riesco-Matías and colleagues (2021), and Van Doren and colleagues (2019, with the follow-up on standard protocols specifically): standard-protocol neurofeedback shows medium effect sizes on parent- and teacher-rated inattention and hyperactivity/impulsivity, with smaller effect sizes on blinded probably-blinded outcomes and on neuropsychological tests of attention. The Strehl group's comparison studies (e.g., Strehl et al. 2017) show effect sizes comparable to methylphenidate in some outcomes, with the persistence of effects looking better at follow-up. The methodological discussion is ongoing — control conditions, blinding, allegiance effects, learning-confirmation — but the underlying signal in the literature is robust, and the field's "ADHD has the best evidence base in neurofeedback" claim is fair.

Epilepsy. The Sterman SMR work has been replicated and extended over decades, with the meta-analysis by Tan and colleagues (2009) reporting a substantial seizure-frequency reduction across studies and a more recent review (Strehl et al. 2014) adding the SCP literature alongside. The evidence base is older, smaller in absolute number of studies, and methodologically less polished than the ADHD literature — but the effect sizes are large, the mechanistic story is clean, and the population is one where any meaningful response matters because alternatives are limited.

Anxiety and depression. The evidence base is more varied. SMR and alpha-theta protocols have evidence for generalized anxiety; alpha-asymmetry and frontal-amplitude protocols have a smaller but real evidence base for depression (Hammond's reviews, the Cheon and Choi group, more recent meta-analytic work). The methodological problems — small samples, heterogeneous protocols, mixed outcome measures — are real, but the signal across the literature is consistent with clinical effect. I would call this emerging and convincing evidence rather than mature.

PTSD and trauma. Alpha-theta training (the Peniston-Kulkosky tradition and its extensions) has a small but interesting literature in PTSD, particularly in veterans and addictions populations. Replications by Burkett and colleagues, the Bessel van der Kolk-affiliated work, and more recent randomized trials (Kluetsch et al. 2014, with rt-fMRI; van der Kolk et al. 2016, with EEG amplitude training) suggest meaningful clinical effects. The evidence base is thinner than ADHD's, the methodological discussion is more contested, and the active-control problem is sharper, but the clinical signal — combined with what practitioners and clients report — is hard to dismiss.

Sleep. SMR training has consistent effects on sleep onset and sleep continuity in both insomnia and ADHD populations (Hoedlmoser et al. 2008 in healthy participants is a clean methodological reference point; Arns et al.'s work on ADHD-sleep links extends the picture clinically).

Performance. The Gruzelier alpha-theta cognitive-performance literature, plus the SMR and beta-related sport and surgical performance work, gives a small but methodologically clean evidence base for peak-performance applications. Effect sizes are modest; the populations are usually healthy; the comparison conditions are usually reasonable. As a category, this is one of the more methodologically careful sections of the broader neurofeedback literature, partly because the absence of clinical desperation means researchers can take their time with the design.

A few honest caveats that apply across the whole evidence base:

Heterogeneity matters. The implementation detail variability we discussed in How the method works shows up in the literature as protocol heterogeneity across studies. Meta-analyses that pool all neurofeedback for ADHD are weaker than meta-analyses that restrict to standard-protocol amplitude neurofeedback (theta/beta and SMR specifically). The Van Doren and colleagues 2019 reanalysis, which restricted to standard protocols, recovered a substantially larger effect than the broader analyses had reported. That tells us something about the cost of methodological lumping.

Blinding is hard. (And if you read my recent article in Discover Neuroscience, you’ll see why I think it’s wholly inappropriate.) It is difficult to blind clinicians and clients to whether they are receiving real or sham neurofeedback in a way that survives a few sessions. The most common sham conditions — yoked feedback, random feedback, feedback from a different brain region — each have their own problems, and none of them produces a clean active-control comparison. The good literature is honest about this; clinicians reading the literature should be honest about it too.

Allegiance and expectancy. Studies run by enthusiastic clinicians produce, on average, larger effect sizes than studies run by independent groups. That is a real finding (Cortese et al. 2016 documents it), and it is not unique to neurofeedback. It is a reason to weight independent replications more heavily, not a reason to dismiss the literature as a whole.

The summary I would offer a colleague new to the field: ADHD and epilepsy have mature evidence bases that justify clinical use; sleep, anxiety, depression, PTSD, and performance applications have emerging-to-moderate evidence bases that justify clinical use with appropriate framing of expectations; some indications often listed in vendor marketing have evidence bases that do not justify the marketing claims, and clinicians need to read the literature themselves rather than trust the brochure.

Strengths and weaknesses

Set out fairly, the method has the following profile:

Strengths

• The most thoroughly studied form of neurofeedback. Several dozen randomized trials, multiple meta-analyses, replicated effects across decades and labs in the strongest indications (ADHD, epilepsy).

• Lowest hardware floor of any of the methods in this series. A two-channel professional system is sufficient for most clinical work. The same equipment that runs an SMR protocol can run theta/beta and alpha-theta — the choice of method is in the protocol, not in the hardware purchase.

• Mature training pathways. A practitioner trained well in classic amplitude training can build a sustainable clinical practice without further specialized equipment, and can branch out to coherence, source-localized, or other methods later if the clinical questions warrant.

• Transparent and reproducible. A protocol described well in a paper can be replicated by another practitioner; a colleague's session notes can be read and understood. This is not always true of other methods in this series.

• Compatible with individualized qEEG-guided protocol selection. The amplitude framework adapts cleanly when site and band targets are chosen from a quantitative assessment rather than from a menu — the underlying method does not change; the individualization gets layered on top.

Weaknesses

• The implementation-detail variability that makes the method flexible also makes literature interpretation hard. Theta/beta training is not one intervention; it is a family of related interventions, and the family resemblance is sometimes less tight than the literature presents it.

• The state-vs-band specificity question — is the reinforced band doing the work, or is the broader state and context doing the work? — is not resolved, and the answer probably depends on the indication.

• Single-channel surface EEG is, by construction, a limited window into brain activity. The method cannot directly target deeper structures, network-level coordination, or anatomically specific regions that surface EEG does not reach. For some clinical questions, that limit matters.

• The method is easy to mis-implement. Generic-menu protocols (for ADHD use C3 theta/beta, for anxiety use Pz alpha increase, for sleep use C4 SMR) deliver weaker results than individualized qEEG-guided protocols, and the same equipment running the same nominal protocol can produce very different outcomes depending on the practitioner's clinical reasoning. The method punishes shortcuts.

• The reputation problem: because classic amplitude training is the oldest method, it is sometimes treated in the field as the low-tech option, an impression that newer methods often trade on in their marketing. That framing is mostly unfair — newer methods are not, in most indications, demonstrably better — but it shapes how clinicians and clients think about the method, and a practitioner choosing classic amplitude training needs to be able to explain why this method, rather than treating the choice as a default.

Brendan's perspective

I have spent more clinical hours doing classic amplitude training than any other form of neurofeedback, and after twenty years of practice and teaching, I will say something more direct than the field's polite framing usually allows: classic amplitude training is the best neurofeedback method for the great majority of clinical questions a working practitioner will face. Ninety to ninety-five percent, in my own practice. Not because the newer methods are wrong or worthless — some have a place, and the next several entries in this series take each of them seriously — but because what most of those methods are doing, mechanistically, is a fancier version of what classic amplitude training is already doing more cleanly.

What the other methods are mostly doing

Look closely at LORETA, at ISF/ILF, and at connectivity training, and the family resemblance to classic amplitude training is striking once you stop being distracted by the new vocabulary.

LORETA and the source-localized methods are amplitude training estimated in source space rather than at the scalp. The operant target is still the amplitude of activity in a frequency band; the conditioning logic is the same; what has changed is the spatial transform that places the trained activity inside a cortical source rather than at an electrode. That transform earns its place in some clinical scenarios — but it also multiplies the number of parameters being trained simultaneously (often dozens to hundreds of voxels, each with its own targets and inhibits), reduces interpretive clarity, depends heavily on the assumptions baked into the inverse model and the normative database, and demands a level of practitioner skill that the field has not built training pathways for at scale. For most clinical questions, a well-chosen single-site amplitude protocol is doing the same mechanistic work with more transparency and less room for the practitioner's reasoning to get lost in the parameter space.

Infra-slow and infra-low frequency methods (ISF/ILF) are amplitude training at a different frequency. The operant logic is the same; the recording chain is similar with different filter settings; the practitioner's role inside the session has the same shape. What has changed is the frequency band being targeted, and the framing of what that band is doing physiologically. ISF/ILF practitioners often describe their work as a categorically different paradigm, but the mechanistic core — operant conditioning of cortical activity through real-time feedback — is the same one running underneath classic theta/beta or SMR work. The clinical case for ISF/ILF rests on the claim that the slow frequencies index something distinct that the classical bands miss, and there are presentations where that case is defensible — but framing the method as a separate paradigm overstates what is, structurally, a band-selection choice inside the same paradigm.

Connectivity and coherence training are, in most clinical applications, hitting the same networks that classic amplitude training already engages, indirectly, through the correlated activity at the channels being trained. When a clinician trains SMR amplitude at C3 and C4, they are already shaping the activity of a sensorimotor network whose constituent channels co-vary. A connectivity protocol that explicitly targets the coherence between C3 and C4 is operating on a derivative measure of the same underlying network state. There are clinical questions where the explicit connectivity target buys something — particularly when the clinical formulation points to a coordination problem rather than an activation problem — but in most applications, the additional methodological complexity is not buying additional clinical leverage.

None of this is an argument that these methods should not exist, or that practitioners drawn to them are wrong. It is an argument that the field's marketing language tends to overstate the novelty of what the newer methods are doing, and that a working clinician with a fluent grasp of classic amplitude training will already be reaching most of the mechanistic territory those methods stake out, more transparently. The five to ten percent of the clinical landscape where the newer methods earn their place is real. It is also smaller than the volume of the field's methodological discourse would lead a reader to expect.

A few other thoughts

The qEEG-informed vs qEEG-based distinction matters more than the method-name distinction. When colleagues ask me to compare classic neurofeedback to LORETA or to connectivity training, I find myself answering a different question than the one they asked. The question that actually shapes clinical outcome is whether the protocol was selected from an individualized assessment or from a menu — not whether the band reinforced was at the surface or in source space. A qEEG-based amplitude training protocol — site and band targets derived from the individual's quantitative assessment — is, in my experience, a stronger clinical intervention than a menu-driven LORETA protocol run without that interpretive context. A qEEG-informed amplitude protocol, where a research-derived target is anchored but adapted to what the assessment shows, sits between the two and is the right choice for many indications where the literature provides a strong starting point. The method matters; the assessment-and-individualization layered on top of the method matters more. This is one of the places where the anti-black-box discipline that NeuroLogic positions itself around is not a marketing claim — it is a clinical observation that has shaped my practice.

Equipment is doing more clinical work than the field tends to acknowledge. The point made in the Part 1 — that equipment is not method-neutral — applies most sharply here. The same nominal protocol delivered on a professional research-grade amplifier with practitioner-controlled session software produces a different clinical experience than the same protocol delivered on a closed system that the practitioner cannot fully see into. The signal quality, the feedback latency, the threshold-adjustment options, the ability to monitor artefacts in real time, the transfer-task integration — none of these are present equally across hardware tiers. When I recommend equipment to colleagues starting out, I am not just recommending a price point; I am recommending the methods they will be able to do honestly and the methods that will quietly drift out of their reach.

The state being trained is part of the clinical formulation, not a generic upgrade. SMR training is not a better version of relaxation; it is sensorimotor inhibition with sensory-quietude correlates. Theta/beta is not a better version of focus; it is sustained cognitive engagement under particular state conditions. Alpha-theta is not a deeper version of meditation; it is a hypnogogic state used for narrative and somatic work in trauma and addiction contexts. The clinical question — what state does this client need to access more of, in service of which outcome? — is what should drive sub-protocol selection. Default protocols selected by diagnostic label alone are leaving clinical leverage on the table.

The literature is more usable than the field's worst-case readings suggest. When critical reviews of neurofeedback land in JAMA or in the Cochrane Library, the headline often reads as weak evidence, mixed effects, methodologically problematic. That is not wrong, exactly — but it is not the full reading. The methodological problems are real and worth fixing. The effect sizes in the strongest indications (ADHD, epilepsy) are real and clinically meaningful. The structural biases in the meta-analytic literature (some of which Hill 2026 names directly for ADHD) cut against the method in ways that the headlines don't always surface. A clinician reading the literature for themselves, with attention to which protocols and which sub-populations were included in which analyses, ends up with a different and more accurate picture than the headlines deliver.

And the relational architecture matters more than I used to think. Early in my career, I thought of the practitioner's role as primarily technical — set the protocol, monitor the signal, adjust thresholds. Twenty years later, I would say the practitioner's role is at least half clinical-relational: framing what the session is for, holding the formulation steady across sessions, debriefing in a way that connects the session to the client's life, and being a stable presence while the client does work that is sometimes uncomfortable and often slow. The technical work is necessary, but the relationship is what carries the method's clinical effect over twenty or thirty sessions. The methods most likely to wear well in a clinical practice are the ones that leave room for that relationship. Classic amplitude training, done well, leaves a lot of room.

There is one thing I want to name about the alpha-theta lineage specifically, because it sits a little awkwardly in this post. Mechanically, alpha-theta is classic amplitude training — same architecture, same operant logic, same hardware. Clinically, it is its own animal: the state it recruits is sufficiently different, the integration with narrative and trauma work is sufficiently distinct, and the practitioner skills required are sufficiently specific that I have considered, more than once, whether alpha-theta deserves its own entry in this series rather than a sub-section here. I have left it here for this draft, in the interest of giving readers the full classical-amplitude landscape in one place. But if the alpha-theta material ends up earning more depth than it gets here, do not be surprised to see a separate post on it later in the series.

Would I do this method myself? In what context?

Yes. In roughly nine cases out of ten that a working practitioner is likely to face. The qualifications are narrower than the field's every method has its place framing suggests.

I would — and do — reach for classic amplitude training first across most of its strong-indication range (ADHD, refractory epilepsy, sleep, the more straightforward anxiety presentations, certain components of PTSD, performance work with healthy adults), when a qEEG-based or qEEG-informed protocol can be developed from a quantitative assessment, when the equipment available is professional-grade and the session software supports the kind of practitioner-in-the-loop control I think the method requires, and when the client's life allows for the session frequency the method needs to do its work (two to three sessions a week, across twenty to forty sessions).

The five to ten percent where I do not reach for classic amplitude training first is rhetorical. I’ve learned how to mimic some other forms of training with two-channels of amplitude training; why do connectivity when I can link two inhibit bands together in the same feedback modality? Why do ISF when I can modulate SMR, alpha or theta to better understand inhibitory mechanisms in my target network? Why lose specificity with a bipolar montage when I can use my thresholds to look at balancing two homologous sites?

What I would say to a colleague starting out, and what I say in NeuroLogic's trainings: classic amplitude training is where most working clinicians should start, where most of them should stay for most of their cases, and where a sustainable clinical practice can be built. It is not the floor of the field. It is the foundation upon which everything is built. For the great majority of clinical questions, it is also the ceiling — and the discipline of recognizing that, rather than chasing every new methodological vocabulary that arrives in the field, is part of what separates serious clinical practice from method-shopping and over-zealous marketing.

The next post in the series — Part 3 — Bipolar protocols — turns to a methodological decision that often gets glossed over in trainings: when bipolar montages serve clinical questions better than referential, and how to read what the bipolar signal actually represents. Like classic amplitude training, bipolar protocols are sometimes treated as obvious or default, and the discipline of choosing them deliberately rather than by habit is where the practitioner's clinical reasoning shows up.

References

• Arns, M., de Ridder, S., Strehl, U., Breteler, M., & Coenen, A. (2009). Efficacy of neurofeedback treatment in ADHD: The effects on inattention, impulsivity and hyperactivity — a meta-analysis. Clinical EEG and Neuroscience, 40(3), 180–189.

• Arns, M., Heinrich, H., & Strehl, U. (2014). Evaluation of neurofeedback in ADHD: The long and winding road. Biological Psychology, 95, 108–115.

• Cortese, S., Ferrin, M., Brandeis, D., Holtmann, M., Aggensteiner, P., Daley, D., … European ADHD Guidelines Group. (2016). Neurofeedback for attention-deficit/hyperactivity disorder: Meta-analysis of clinical and neuropsychological outcomes from randomized controlled trials. Journal of the American Academy of Child & Adolescent Psychiatry, 55(6), 444–455.

• Gruzelier, J. H. (2014). EEG-neurofeedback for optimising performance. I: A review of cognitive and affective outcome in healthy participants. Neuroscience & Biobehavioral Reviews, 44, 124–141.

• Hoedlmoser, K., Pecherstorfer, T., Gruber, G., Anderer, P., Doppelmayr, M., Klimesch, W., & Schabus, M. (2008). Instrumental conditioning of human sensorimotor rhythm (12–15 Hz) and its impact on sleep as well as declarative learning. Sleep, 31(10), 1401–1408.

• Kluetsch, R. C., Ros, T., Théberge, J., Frewen, P. A., Calhoun, V. D., Schmahl, C., … Lanius, R. A. (2014). Plastic modulation of PTSD resting-state networks and subjective wellbeing by EEG neurofeedback. Acta Psychiatrica Scandinavica, 130(2), 123–136.

• Lubar, J. F., & Shouse, M. N. (1976). EEG and behavioral changes in a hyperkinetic child concurrent with training of the sensorimotor rhythm (SMR): A preliminary report. Biofeedback and Self-Regulation, 1(3), 293–306.

• Parsons, B. (2026). Rethinking control conditions in clinical neurofeedback trials. Discover Neuroscience.

• Peniston, E. G., & Kulkosky, P. J. (1989). Alpha-theta brainwave training and beta-endorphin levels in alcoholics. Alcoholism: Clinical and Experimental Research, 13(2), 271–279.

• Riesco-Matías, P., Yela-Bernabé, J. R., Crego, A., & Sánchez-Zaballos, E. (2021). What do meta-analyses have to say about the efficacy of neurofeedback applied to children with ADHD? Review of previous meta-analyses and a new meta-analysis. Journal of Attention Disorders, 25(4), 473–485.

• 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.

• Strehl, U., Aggensteiner, P., Wachtlin, D., Brandeis, D., Albrecht, B., Arana, M., … Holtmann, M. (2017). Neurofeedback of slow cortical potentials in children with attention-deficit/hyperactivity disorder: A multicenter randomized trial controlling for unspecific effects. Frontiers in Human Neuroscience, 11, 135.

• Tan, G., Thornby, J., Hammond, D. C., Strehl, U., Canady, B., Arnemann, K., & Kaiser, D. A. (2009). Meta-analysis of EEG biofeedback in treating epilepsy. Clinical EEG and Neuroscience, 40(3), 173–179.

• Van Doren, J., Arns, M., Heinrich, H., Vollebregt, M. A., Strehl, U., & Loo, S. K. (2019). Sustained effects of neurofeedback in ADHD: A systematic review and meta-analysis. European Child & Adolescent Psychiatry, 28(3), 293–305.