Sign in to Peak Brain Path
Sign in to access your courses, books, and progress tracker. New here? Signing in creates your account automatically.
Want to explore courses first?
Browse courses and booksSign in to access your courses, books, and progress tracker. New here? Signing in creates your account automatically.
Want to explore courses first?
Browse courses and booksModule 5
Chapter 5 · 2 h · 8 quiz items · pass at 80%
BCIA Domain II requires the practitioner to connect EEG sites and rhythms to the brain structures that generate them. This module gives the practitioner-resolution map: which lobe and Brodmann area sits under each 10-20 site, what the thalamocortical engine does, why the SMR lives at the motor strip, and why coherence and plasticity matter clinically. The quiz proves the learner can read an electrode site as a hypothesis about the cortex beneath it.
A protocol says to train SMR at C4. Why there, and not somewhere else? Answering that question is the move that turns neuroanatomy into something you use at the cap, and it is the work of this chapter. An electrode site is not arbitrary. It is a bet about which functional cortex sits underneath, paced by structures the electrode cannot see directly, and the quality of that bet depends on knowing what each region does, what drives the rhythms over it, and how loosely a scalp site maps to the tissue below.
This is a condensation. The physiology book spends seven chapters on this material, with cellular and channel-level detail you do not need to read a map well. Here you get the practitioner's working set: the cortical landmarks and the structures beneath them, the thalamocortical engine that paces the major rhythms, the arousal systems that set the volume on the whole recording, the lobe and Brodmann functions that let a site stand for a function, the sensorimotor rhythm and why central sites matter, connectivity and the networks, and the plasticity that makes training work at all. Each block ends with the "so what" for a practitioner. For the full mechanistic treatment, forward to Neurophysiology for Neurofeedback.
The cerebral cortex is the wrinkled outer sheet, folded into ridges (gyri) and grooves (sulci). Two landmarks orient everything. The central sulcus runs across the top of each hemisphere and divides the frontal lobe in front from the parietal lobe behind. The lateral sulcus runs along the side and marks the upper border of the temporal lobe. At the back sits the occipital lobe. These four lobes are the addresses you use constantly: frontal for executive control and movement, parietal for spatial and bodily integration, temporal for hearing, language, and memory, occipital for vision. A fifth region, the insula, sits buried inside the lateral sulcus out of reach of any scalp electrode; it maps the body's internal state and anchors the salience network, a structure the scalp cannot see directly that still shapes what the surface reports.
The cortex is gray matter, the thin surface layer of cell bodies where the dipoles of Chapter 4 are generated. Beneath it lies white matter, the bundled, myelinated axons that carry signals between regions. These tracts are the brain's long-distance wiring and the physical substrate of the connectivity and coherence measures later in this chapter. The cortex is also layered, conventionally into six laminae, and the layering matters for one practical reason: incoming sensory information arrives mainly in the middle layer, output to other regions leaves from the deep layers, and the superficial layers carry the cortico-cortical cross-talk between regions and sit closest to the electrode. Because the scalp signal falls off with distance, the EEG is weighted toward this superficial activity, the cortex coordinating with cortex, more than toward the deep output layers.
Under the cortex sit structures that shape and pace cortical activity even though their closed-field geometry keeps them nearly silent at the scalp. Know them as the hidden hands behind the surface rhythms, because certification blueprints ask for the major functions of the major subcortical structures:
So what: you reason about these structures by their consequences on the cortical signal, not by reading them directly. When arousal shifts, suspect the brainstem and thalamus. When emotion drives the picture, suspect the amygdala and its autonomic train. When movement or habit is the theme, suspect the basal ganglia. The scalp shows the cortex. These are the hidden hands moving it. A durable habit from QEEG-aligned teaching is to hold every finding along three axes at once: cortex versus subcortex (surface versus depth), left versus right (the hemispheric asymmetries), and anterior versus posterior (front to back). Reading a finding on all three at once organizes interpretation better than any single number.
Close your eyes and the back of your head fills with a rhythm near ten cycles a second. Open them and it drops away. Posterior alpha is the most reliable feature in human EEG, and the cortex on its own does not keep such steady time. The metronome is the thalamus, working in a loop with the cortex it drives.
The loop is the engine. Thalamic relay cells project up to the cortex, the cortex projects back down to the thalamus, and the inhibitory reticular nucleus regulates the traffic. A circuit with excitation one way, inhibition another, and feedback between them can oscillate, and thalamic relay cells add a second ingredient: membrane properties that make them prone to rhythmic bursting on their own. The grouping of brain rhythms by these corticothalamic systems was worked out across decades of study (Steriade, 2006). The key idea for a practitioner is that relay cells have two modes. In transmission (tonic) mode, when relatively depolarized, the cell relays inputs faithfully, the waking mode. In burst mode, when it has been hyperpolarized, it answers the next input with a burst, then primes again. Arousal sets which mode dominates: alert and engaged, the cortex runs fast and desynchronized; as arousal falls, the cells slip toward burst mode, the loop tightens, and the rhythms grow slower and larger, alpha giving way to spindles and then to the slow waves of deep sleep. The channel-level detail (the T-type calcium channel that enables the burst, the HCN/Ih pacemaker channel that sets the interval; McCormick & Pape, 1990) is in the physiology book; the consequence is that a slowing rhythm is a loop sliding toward its lower-arousal modes, and your first question is why.
Each band has a generator, and knowing the origin changes how you read the band:
Alpha is the worked example. It sits over occipital and parietal cortex, is strongest with the eyes closed, and blocks the moment the eyes open and the visual cortex engages. Eyes closed, the posterior cortex idles, its cells fall into the common thalamocortical rhythm, and the aligned dipoles summate into a tall wave. Eyes open, the population desynchronizes and amplitude collapses. A clean, reactive alpha that blocks crisply is a working thalamocortical loop reporting in. The frequency at which an individual's alpha peaks, the individual alpha peak frequency, is itself a meaningful number: generally eight to thirteen hertz, stable across sessions like a trait, reflecting the speed of the thalamocortical clock, and it slows with age (Klimesch, 1999; Scally et al., 2018). A slow peak has been associated with nonresponse in some neurofeedback and medication contexts (Arns, 2012), and rhythmic sensory stimulation works best tuned to a person's own alpha rather than a fixed ten hertz (Gulbinaite et al., 2017).
So what: too much alpha power and a slow alpha peak are different findings with different meanings. Excess power is a question about how much a region is idling. A slow peak is a question about how fast the underlying clock is running. A practitioner who collapses the two into "alpha problem" has lost the more useful reading. This distinction also explains a common artifact you met in the protocol chapters: a slow individual alpha peak can bleed into the theta band and inflate an apparent theta excess, so check the individual alpha frequency before treating theta as too high. Sleep spindles, finally, are the same engine running in another state, brief twelve-to-fifteen-hertz bursts from the same relay-and-reticular machinery, which is why daytime SMR training (the same frequency and circuitry) plausibly strengthens the spindle-generating system that stabilizes sleep.
The same brain looks different at ten in the morning and at two in the afternoon after a large lunch, not because its architecture changed but because its arousal level did. Arousal is the volume knob on the whole recording, and it is the single largest source of confusion in a resting record, because a shift in arousal mimics findings that look like stable traits.
Running up through the brainstem is the ascending arousal system, classically the reticular activating system, now resolved into chemically defined populations (cholinergic, noradrenergic, serotonergic, histaminergic, and others) that promote wakefulness, balanced against sleep-promoting circuits, with the hypothalamus acting as a switch between the two states (Saper et al., 2005). When the system is active it desynchronizes the cortex into alert, low-voltage activity; when it quiets, the cortex synchronizes and slows. The chemistry, acetylcholine for the alert cortex of waking and REM, norepinephrine rising with engagement and salience, orexin holding the waking state stable, is the territory of the neurotransmitter chapter in the physiology book. The system-level point is that a small set of deep nuclei sets the arousal of the entire cortex, so anything that shifts those systems (fatigue, caffeine, a sedating medication, a stimulant) moves the whole EEG with it. A map is always a map of a particular chemical state of arousal.
The hypothalamic switch behaves differently from the smooth dial of arousal within waking. The wake-promoting and sleep-promoting populations inhibit each other, so the circuit is bistable: it tends to flip between awake and asleep rather than linger in the middle. A drowsy client therefore does not slide evenly from alert to asleep across a recording; the EEG flips, holding alert, dropping into drowsy intrusions, snapping back. Within waking there are gradations worth staging, because most of your recordings live here: full alertness is low-voltage and fast with a crisp reactive alpha; relaxed wakefulness (eyes closed) is the tall steady alpha most protocols want; and as vigilance drops the alpha begins to wax, wane, and slow, the first slow rolling eye movements appear, and theta rises. A fragmenting, slowing alpha with rolling eye movements is drowsiness wearing a resting record's clothes.
You should also recognize the sleep-stage signatures, because drowsiness and sleep intrusions are the commonest contaminants of a waking record, and because alpha-theta and sleep work sit at this end of the dial. N1 (lightest sleep): alpha dropout, theta, slow rolling eye movements, vertex sharp waves. N2: sleep spindles (twelve to fifteen hertz bursts) and K-complexes. N3 (slow-wave sleep): high-amplitude delta, the most synchronized state the cortex reaches. REM: low-voltage, mixed-frequency, wake-like EEG with rapid eye movements and muscle atonia, which is why it is called paradoxical sleep. Unstable vigilance regulation has been tied to conditions such as ADHD and insomnia, and stabilizing it through sleep-spindle and circadian mechanisms is part of how some neurofeedback approaches are understood to work (Arns & Kenemans, 2014). The dial is not only a confound; for some clients it is the target.
So what: read arousal state first. A practitioner who maps a new client, finds elevated frontal theta and a slowed posterior rhythm, and starts planning an arousal-raising protocol has skipped the one check that decides everything: was the client alert during the recording? A client who was up at five, drove through traffic, and recorded after lunch can produce exactly this picture from drowsiness alone. The same map means one thing in an alert brain and something entirely different in a sleepy one. Drowsy frontal theta is a state, not a trait, and reading arousal first is the precondition for trusting anything else on the map.
Now map the four lobes onto function precisely, because this is what lets a site stand for a function. The frontal lobe holds executive control: planning, working memory, impulse regulation, and the approach-versus-withdrawal balance that distinguishes the left frontal cortex (biased toward approach and positive engagement) from the right (biased toward withdrawal and avoidance). This asymmetry is the physiological basis of the frontal alpha asymmetry read in mood work. The parietal lobe handles spatial integration and attentional allocation, with a posterior-midline hub of the default mode network. The temporal lobe handles hearing, language on the left, prosody and social tone on the right, and, through medial structures the scalp cannot reach, memory. The occipital lobe handles vision and is where alpha is strongest, which is no accident: a visual cortex at rest is an idling cortex, and idling cortex synchronizes.
Facing each other across the central sulcus are the primary motor cortex (just in front) and the primary somatosensory cortex (just behind), together the sensorimotor cortex, organized as a body map (the homunculus) running from legs and feet at the top of the head down through trunk, arm, and hand to the face near the lateral sulcus. The map is distorted by use and innervation density: the hand and face occupy far more cortex than the trunk. Most of the cortex, though, is association cortex, integrating across the primary regions and holding the higher functions: the prefrontal cortex for executive control, the parietal association cortex for spatial integration, the temporo-parietal junction for social cognition. Hold primary-region mappings (motor strip, visual cortex) more firmly than association-region mappings, which are real but fuzzier because association functions are distributed and individually variable.
A century ago Brodmann divided the cortex into numbered regions by differences in their layering, and the numbering survives because those structural divisions correspond reasonably well to functional ones. You do not memorize all fifty-some areas, but a working handful repays knowing. The table below pairs each standard 10-20 site with the cortex and the Brodmann area or areas most likely beneath it, drawn from studies that projected electrode positions onto the cortex by MRI (Homan et al., 1987; Okamoto et al., 2004; Koessler et al., 2009). Read it as a teaching prior, not a coordinate system: the correspondence is approximate, individually variable, and blurred by the volume conduction and gyral folding of Chapter 4.
| Site (10-20) | Cortex beneath | Likely Brodmann area(s) | Function |
|---|---|---|---|
| Fp1 / Fp2 | Frontopolar prefrontal | 10 | executive, prospective control |
| F3 / F4 | Dorsolateral prefrontal | 8, 9, 46 | working memory, executive control |
| F7 / F8 | Ventrolateral frontal | 45, 47 (44/45 left = Broca) | speech production (left), inhibition |
| Fz | Superior-frontal / pre-SMA | 6, 8 (24/32 deeper) | midline executive, frontal-midline theta |
| C3 / C4 | Pre- and post-central gyri | 1, 2, 3, 4 (6) | hand motor and somatosensory |
| Cz | Vertex sensorimotor / SMA | 4, 6 (5) | leg-foot motor, SMA, SMR |
| T3 / T4 (T7 / T8) | Mid-superior temporal | 21, 22 (42) | auditory association, language (left) |
| T5 / T6 (P7 / P8) | Posterior temporal / TPJ | 37, 39 (22) | comprehension (left), social-spatial (right) |
| P3 / P4 | Inferior-superior parietal | 7, 39, 40 | spatial integration, calculation (left) |
| Pz | Precuneus / posterior cingulate | 7, 31 (23) | default-mode hub, visuospatial |
| O1 / O2 | Occipital | 17, 18, 19 | primary and association vision |
The clearest way to learn what a region does is to see what breaks when it is damaged. The left angular gyrus (near P3) sits at the crossroads of language, number, and body-space; damage there can produce Gerstmann syndrome, a cluster of agraphia, acalculia, finger agnosia, and left-right confusion that travels together because the functions share that cortex. Hemispatial neglect is the right-hemisphere equivalent: damage to the right inferior parietal cortex and temporo-parietal junction (under roughly P4 and T6) can leave a person unaware of the left side of space, evidence for the right hemisphere's dominance in spatial attention. Along the midline, the anterior cingulate (near Fz) monitors conflict and error and generates frontal-midline theta (Nakamura-Palacios et al., 2023), and the posterior cingulate (near Pz) anchors the default mode network.
So what: when you choose a site, you are choosing a function, with the standing caveat that scalp-to-cortex correspondence is approximate and a site names a neighborhood, not a point. A deviation at T6 is a question about social processing and reading others' intentions, not a generic abnormality to normalize toward the population mean. A slowed alpha peak at O1 and O2 is a question about the speed of the thalamocortical clock, read against the client's age. Elevated theta at Fz is a question about cognitive-control engagement or, just as plausibly, drowsiness raising frontal theta, with the arousal check deciding between them. The shift from reading C4, Fz, and O1 as labels to reading them as motor cortex, cingulate, and visual cortex is what lets a protocol be reasoned about rather than merely followed.
The most trained rhythm in the history of the field is produced by a system holding still. Over the sensorimotor strip, when the body is still and the mind is alert, the cortex produces an oscillation around twelve to fifteen hertz, the sensorimotor rhythm (SMR), built by the same thalamocortical machinery that makes posterior alpha. A cortex with nothing immediate to do settles into a synchronized idle, and for the motor system that idle is SMR. The closely related mu rhythm is the same family of sensorimotor idling, slightly lower in frequency and shading into the alpha band. Both mark a sensorimotor system that is poised and quiet, ready to move but not moving, which is the specific state a protocol reaches for when it rewards SMR: not relaxation alone and not alertness alone, but a still body and an alert mind.
The rhythm entered the field through a discovery that was not looking for it. In work on the sleeping and waking cat, Sterman found that animals trained to produce this central rhythm later proved more resistant to chemically induced seizures, a durable change in excitability that followed from conditioning a rhythm (Sterman & Egner, 2006). That finding moved into human work, where enhancing SMR suppressed seizures in a person with epilepsy (Sterman & Friar, 1972) and was then extended to the restless, impulsive presentation of attention disorders, on the reasoning that a rhythm marking calm motor readiness might be worth strengthening in a nervous system that could not hold still (Lubar & Shouse, 1976). The lineage from the cat to the clinic is the historical spine of neurofeedback, and Chapter 2 carries the full history. What matters here is the physiology of why the rhythm exists and what state it marks.
The mirror image of the resting rhythm is what happens when the system engages. The moment a movement is prepared, executed, or even vividly imagined, the mu and SMR rhythms over the corresponding part of the strip desynchronize, their power dropping as the cortex shifts from idle to work. This event-related desynchronization is spatially specific, appearing over the hand area for a hand movement and the foot area for a foot movement, and it is released by motor imagery and by watching someone else move, which is why motor-imagery tasks drive sensorimotor rhythms and form the basis of brain-computer interfaces (Pfurtscheller & Lopes da Silva, 1999). The return is as informative as the drop: once the movement ends, beta over the same cortex briefly overshoots its resting level before settling, a post-movement beta rebound read as the motor cortex resetting to its idle, inhibited state.
The motor system is mostly invisible to the scalp except at this one well-built strip. The cerebellum, central to coordination and timing, contributes almost nothing to the surface recording because its tightly folded geometry cancels and it sits deep in the posterior fossa, and the basal-ganglia loop that gates which movements happen at all is likewise inferred rather than read.
So what: the central strip (C3, C4, Cz) means sensorimotor cortex and the rhythm it idles at, which is why C3 (left) and C4 (right) sit directly over the SMR generator and why a protocol targets them. But SMR is a state-dependent rhythm, so before you read a low value as a finding, establish that the client was actually still during the recording: a fidgeting client suppresses SMR for the plainest mechanical reason, and the low value is then a fact about the recording rather than the brain. The rhythm names a state, and the state has to be established before the number means anything.
A connectivity display lights up between two sites and the software reports high coherence. What is actually being claimed about the brain? Connectivity, phase, and coherence are among the most-used and least-understood words in QEEG, and the reason they confuse is that practitioners meet them first as numbers in software rather than as facts about tissue.
Start with phase: where an oscillation sits in its cycle at a given instant. When two oscillations rise and fall together they are in phase; when one peaks as the other troughs they are out of phase. Coordination between two regions shows up as a stable, consistent phase relationship over time, because two regions genuinely working together tend to hold a steady relationship while independent ones drift. Coherence asks exactly this: across a stretch of recording, how consistent is the phase relationship between two sites at a given frequency? High coherence is usually interpreted as the two regions communicating or being driven by a common source. In clinical reasoning, excessive coherence (a rigid coupling that cannot reconfigure) shows up in perseverative and obsessive patterns, while insufficient coherence (fragmented communication) shows up in some attention and autism presentations.
A serious caution belongs here, and it is the reason to learn the physiology rather than just the software. Two scalp electrodes can show high coherence simply because one deep source projects to both of them, the same field picked up in two places at once. That is volume conduction, not communication, and it produces spuriously high coherence at near-zero phase lag. Genuine coordination between two populations carries the conduction and synaptic delays of real communication, so it shows a phase lag rather than a zero-lag lock. This is why phase-lag measures, which discount the zero-lag coupling volume conduction creates, have become important in careful work (Stam et al., 2007). The methods for handling this belong to The QEEG Field Guide; the physiological point is that coherence is a claim about coordination the recording cannot always justify, so a bright pair of adjacent sensors over a strong common generator is the textbook setup for spurious coherence, and the burden of proof sits on calling it real.
Underneath the measures sits one organizing principle worth holding, because it makes sense of why both too much and too little coupling are problems. A healthy brain balances integration (letting distant regions combine information, which requires coordination) against segregation (letting regions work independently, which requires that they not be locked together). Excessive coupling is a failure of segregation, the rigid perseverative pattern. Insufficient coupling is a failure of integration, the fragmented pattern. The vocabulary for describing this at the system level is graph theory: regions are nodes, heavily connected regions are hubs, and the densely interconnected hub regions of the medial cortex and frontal and parietal association areas form a rich-club backbone of global integration that is also the network's most vulnerable component (Bullmore & Sporns, 2009; van den Heuvel & Sporns, 2011). You do not compute graph metrics at the cap, but the vocabulary is how the field now thinks about what coherence displays are sampling.
The most discussed network is the default mode network (DMN), a set of midline and parietal regions (prominently the posterior cingulate near Pz and the medial prefrontal cortex) most active during rest and internally directed thought, quieting when attention turns to an external task (Raichle et al., 2001). It does not act alone. A widely used model describes three interacting networks: the default mode network for internally directed thought, a central executive network for externally directed tasks, and a salience network, anchored in the anterior insula and anterior cingulate, that detects what matters and switches control between the other two (Menon & Uddin, 2010). Disruption of this triple-network balance recurs across many clinical presentations. The theta-to-beta ratio you track is not separate from this picture; it covaries with mind-wandering and with executive-network connectivity, which is one reason a familiar power measure carries network meaning.
So what: when a coherence map lights up, you are reading a claim that two populations are coordinating, a claim that is sometimes biology and sometimes volume conduction. Reading connectivity well means holding the network picture and the volume-conduction caution at the same time, and never letting a bright pair on a display stand in for an inference the recording cannot support. Whether a confirmed finding warrants attention, and in which direction, is the Field Guide's call and the protocol chapters'; the physiological humility, that these are estimates of coordination rather than photographs of it, is what you carry from here.
Reward a brain state a few hundred times in a session, twice a week, for months, and the resting EEG can shift. The entire premise of brain training rests on that possibility, and the cellular mechanism behind it is well established even though the exact pathway from a given protocol to a lasting change is still partly inferred.
The foundational idea is Hebb's: when one neuron repeatedly takes part in firing another, the connection between them strengthens. "Cells that fire together wire together" compresses a precise claim about coincidence detection, a synapse active at the same moment the receiving cell is active gets strengthened, uncorrelated input does not. This became physiology when the strengthening was demonstrated directly: brief high-frequency stimulation of a hippocampal pathway produced a durable increase in synaptic strength, long-term potentiation (Bliss & Lømo, 1973), with long-term depression as its weakening counterpart. The coincidence detector at many synapses is the NMDA receptor, which admits the calcium that triggers strengthening only when input and output coincide, so the Hebbian rule is built into the gating of a channel rather than imposed as a metaphor.
Timing is the whole of it. Whether a synapse strengthens or weakens depends not only on whether input and output coincide but on their order, inside a window of a few to a few tens of milliseconds: input just before the cell fires potentiates the synapse, input just after weakens it (Bi & Poo, 1998). This spike-timing-dependent plasticity sets a hard physical constraint that the design of neurofeedback has to respect: if the strengthening rule runs on a millisecond-scale window, then any system trying to recruit it has to land its reward inside that window. A coincidence the synapse can read is the only coincidence it can learn from. This is the credit-assignment problem, and it is why feedback that lags by a second, or fires whether or not the target occurred, teaches nothing or teaches the wrong thing. Some apparent protocol failures are timing failures: a client who is not progressing may not have the wrong target but feedback too slow, too noisy, or too loosely contingent for plasticity to find the coincidence it needs.
Two more mechanisms shape how training plays out over weeks. Plastic changes are stabilized over time, and sleep is central to that consolidation: waking experience drives a net strengthening of synapses across the day, and sleep renormalizes and consolidates those changes, preserving what matters and scaling back the rest (Tononi & Cirelli, 2014). What is rehearsed in a session is consolidated between sessions, which makes sleep part of the training apparatus, not separate from it: a client who trains well and sleeps poorly consolidates poorly. And homeostatic plasticity scales a neuron's overall sensitivity up or down to keep its activity in a workable range, pushing back toward the prior set point, which is part of why lasting change requires the nudge to be repeated often enough and consolidated well enough to move the set point rather than merely perturb it. This is why brain training takes many sessions over months rather than a few intensive ones, and why consistency and sleep matter more than heroic single efforts.
Within that honest frame, neurofeedback is best understood as operant conditioning of an internal state: the brain produces a target pattern, the pattern is detected and rewarded in close to real time, and plasticity does with that rewarded coincidence what it does with any rewarded coincidence. The mechanism is real; the cellular machinery above is not controversial. What remains partly inference is the precise step-by-step pathway from a given protocol to a specific lasting change in a particular feature of the resting EEG, which is why careful reviews have pressed the field to separate its clinical science from its fads, and why a practitioner who says so plainly is on firmer ground than one who promises more than the evidence supports (Garcia Pimenta, Brown, Arns & Enriquez-Geppert, 2021).
So what: when the resting EEG shifts over a course of training, the most defensible account is Hebbian strengthening of rewarded states, consolidated across sleep, with the standing caveat that the exact pathway is still being worked out. That account supports cautious optimism and forbids the guarantee: the honest promise to a client is that training makes a change more likely and that progress will be tracked, not that a specific outcome is assured. It also sets a clear operational demand, watch the quality and timing of the feedback as closely as the choice of target, because immediate, clean, contingent feedback is the only kind the Hebbian machinery can use.
Functional neuroanatomy, for a practitioner, comes down to reading a site as a function paced by structures the electrode cannot see. The thalamus, in a loop with the cortex and gated by the reticular nucleus, paces the major rhythms, so a slowing rhythm is a loop sliding toward lower arousal and a slow alpha peak is a slow clock, a different finding from excess alpha power. The arousal systems set the volume on the entire recording, so you read arousal state before any single band, and drowsy frontal theta is a state, not a trait. A 10-20 site stands for the cortex beneath it (C3/C4/Cz for sensorimotor cortex and SMR, Fz for anterior cingulate and frontal-midline theta, posterior sites for visual and parietal cortex and alpha, frontal sites for executive cortex and the approach-withdrawal balance), with the standing caveat that a site names a neighborhood, not a point. A coherence display is a question (real coordination or volume conduction?) before it is an answer. And neurofeedback works because training recruits the same plasticity that lets a brain learn anything, which is why it is slow, why sleep is part of it, and why contingent, well-timed feedback is the method rather than a refinement of it.
For the BCN exam, fix the anchors. The thalamocortical loop generates the major rhythms; T-type calcium and HCN channels enable thalamic bursting; posterior alpha blocks on eye opening and its peak frequency (eight to thirteen hertz) is a stable trait that slows with age. Arousal runs alert (low-voltage fast, reactive alpha) to drowsy (theta, vertex waves) to sleep (N1 theta, N2 spindles and K-complexes, N3 slow-wave delta, REM wake-like with atonia), set by a bistable switch that flips rather than drifts. SMR (twelve to fifteen hertz) and mu are sensorimotor idling rhythms over the central strip, desynchronized by movement and imagery, the founding neurofeedback target through Sterman and Lubar. Coherence measures phase-relationship consistency and is inflated by volume conduction at zero lag, which phase-lag measures discount; the triple-network model is DMN, central executive, and the salience network that switches between them. Plasticity runs on Hebbian coincidence (LTP/LTD, NMDA as detector, spike-timing windows), is consolidated by sleep, and is stabilized by homeostatic scaling, which together make neurofeedback a form of operant conditioning whose mechanism is established and whose per-protocol causal chain is still partly inferred. The full mechanistic depth behind every one of these lives in Neurophysiology for Neurofeedback when a case, or a client's question, sends you looking for it.