Normal EEG Rhythms

Frequency is a state variable, not a label

The convention of carving the EEG into delta, theta, alpha, beta, and gamma bands predates any mechanistic understanding of where those rhythms come from. Hans Berger named alpha and beta in the 1920s by inspection; the slower and faster bands were appended later, and the exact boundaries still vary modestly between laboratories and guidelines. A century of work has since shown that the boundaries, however historically arbitrary, track something real: each band corresponds to a distinct dynamical regime of the thalamocortical system, with its own cellular pacemakers, its own preferred topography, and its own relationship to arousal. When you read a frequency off the page you are not measuring an isolated number - you are inferring the state of the brain that generated it. A posterior 10 Hz rhythm that blocks with eye opening, an anterior 2 Hz rhythm in a comatose patient, and a frontal 20 Hz rhythm under sedation are three completely different physiological events that happen to be quantified in the same units.

The single most important principle is that the same frequency means opposite things depending on where it appears and in whom. Theta over the temporal regions of a drowsy young adult is normal; the identical frequency appearing focally and persistently in an alert patient is a localizing abnormality. Slowing that is appropriate for deep sleep is catastrophic in a patient who is supposed to be awake. Beta that is reassuring as a drug effect becomes a localizing sign when it is focally absent. Frequency interpretation is therefore always conditional on state, age, location, and reactivity - the four axes you will return to in every subsequent module. A useful way to formalize this is Bayesian: the prior probability of pathology is set by the clinical context, and each EEG feature is a likelihood that updates that prior. A 7 Hz posterior rhythm carries one meaning when the technologist has documented an alert, cooperative patient and quite another when the comments note a drowsy, post-call resident; the raw frequency is identical, but the posterior probability of disease diverges sharply because the conditioning variables differ.

Delta and theta: the slow bands and their two faces

Delta (0.5 to 4 Hz) is the slowest and, paradoxically, often the highest-amplitude activity, because slow oscillations recruit the most synchronous cortex. It has two mechanistically distinct origins. Cortical delta arises intrinsically: layer-5 pyramidal neurons possess intrinsic membrane conductances that let them oscillate slowly even when partially disconnected from the thalamus, and the cortical slow oscillation - the alternation between depolarized up-states and hyperpolarized down-states - paces deep NREM sleep. The down-state is essentially a network-wide disfacilitation, a collapse of synaptic activity, and its summed extracellular signature is the large surface delta wave. Projected or released delta in the waking adult is almost always pathological, generated when deep white-matter lesions or diffuse encephalopathy strip cortex of its normal high-frequency thalamic drive, releasing the intrinsic slow rhythmicity that arousal normally suppresses. Polymorphic, arrhythmic delta that is continuous and focal suggests a structural white-matter lesion beneath that region; intermittent rhythmic delta, classically frontal (FIRDA) in adults or occipital (OIRDA) in children, is a non-specific marker of diffuse dysfunction or deep midline pathology and, importantly, is a projected phenomenon - it does not localize a lesion to the frontal lobe.

Theta (4 to 8 Hz) occupies an ambiguous middle ground and is the band where normal and abnormal overlap most heavily. In the healthy adult waking record a modest amount of temporal theta is normal, more so with advancing age, and theta is expected during drowsiness as the alpha rhythm fragments and slows. Its generators are several: the classic hippocampal theta of the rodent, paced by a septohippocampal circuit, has no clean scalp correlate in humans, but frontal-midline theta - a roughly 6 to 7 Hz rhythm over the medial frontal region tied to sustained attention, working memory load, and mental effort - is a genuine cognitive signal generated in anterior cingulate and medial prefrontal cortex. The clinical challenge is precisely that drowsiness and early diffuse dysfunction both live in this band, so theta interpretation depends almost entirely on two questions: is the patient actually drowsy, and is the theta diffuse (favoring drowsiness or mild encephalopathy) or focal and persistent (favoring a localized lesion)? A run of left temporal theta that persists into clear wakefulness and does not migrate is a different animal from symmetric drowsy theta, even though a spectrogram would bin them identically.

Alpha, beta, and gamma: from idling cortex to interneuron timing

Alpha (8 to 13 Hz) is the signature rhythm of the relaxed, awake, eyes-closed posterior cortex. Its pacemaker is the thalamocortical loop: rhythmic burst-firing in thalamic relay and reticular neurons, gated by the calcium-dependent low-threshold spike, paces cortical pyramidal cells into the alpha range, and corticothalamic feedback sharpens and entrains the rhythm; high-order thalamic nuclei such as the pulvinar are particularly implicated in posterior alpha. The modern functional interpretation is that alpha is not mere idling but active inhibitory gating - a pulsed inhibition that periodically silences task-irrelevant cortex, so that alpha power rises over regions being suppressed and falls over regions being engaged. This reframes the classic observation neatly: alpha appears posteriorly when the eyes are closed because the visual cortex is being gated out, and it disappears the instant visual input must be processed. The same logic explains alpha lateralization during spatial attention, a robust finding in cognitive neuroscience that has become a staple of EEG-based brain-computer interfaces.

Beta (13 to 30 Hz) is low-amplitude, fast activity that predominates frontally and centrally. It marks active cortical engagement, attention, and the maintenance of the current sensorimotor or cognitive set; in the motor system, beta is high during steady posture and drops sharply before and during movement (the event-related desynchronization that BCIs exploit), rebounding afterward. Beta is dramatically enhanced by benzodiazepines and barbiturates, which potentiate GABA-A inhibition and drive cortex into fast, low-amplitude oscillation, so excessive diffuse beta is a pharmacological fingerprint as much as a physiological one. The clinically decisive beta finding, however, is often a focal one: a localized increase in beta amplitude - a breach rhythm - appears over a skull defect, where the missing bone no longer attenuates the underlying activity, and must not be mistaken for a focal abnormality; conversely, focal beta loss (a beta asymmetry) can localize a cortical lesion. The asymmetry is the signal in both directions.

Gamma (greater than 30 Hz) is the fastest band and the one least visible on routine scalp EEG, because its intrinsically low amplitude and the skull's low-pass filtering attenuate it severely, and because scalp muscle (EMG) activity contaminates exactly this frequency range. Mechanistically gamma is the product of fast-spiking, parvalbumin-positive (PV) GABAergic interneurons providing rhythmic perisomatic inhibition to pyramidal cells. The two canonical models - interneuron network gamma (ING), in which mutually inhibitory interneurons set the rhythm, and pyramidal-interneuron network gamma (PING), in which excitatory drive and feedback inhibition alternate - both place the time constant of GABA-A inhibition at the heart of the gamma cycle. Gamma is implicated in feature binding, local cortical computation, and the gating of information flow; the communication-through-coherence framework holds that gamma-band phase alignment between regions opens windows for effective signaling. Although gamma is largely a research and intracranial phenomenon, an appreciation of its PV-interneuron origin is essential, because the same interneurons generate the high-frequency oscillations (ripples at 80 to 250 Hz and fast ripples above 250 Hz) that are among the most specific intracranial biomarkers of the epileptogenic zone, and because PV-interneuron dysfunction is a leading hypothesis for the gamma abnormalities reported in schizophrenia.

The posterior dominant rhythm: quantifying the most informative feature

The posterior dominant rhythm (PDR) is the single most informative feature of the routine waking record, and quantifying it correctly is the first reflex of expert interpretation. In the healthy adult the PDR is an alpha-frequency rhythm, maximal over the occipital and parietal regions, present with eyes closed, of roughly symmetric amplitude, and exquisitely reactive - it attenuates (blocks) within a fraction of a second of eye opening and re-emerges with eye closure. Its frequency is strongly age-dependent: it begins as a slow 3 to 4 Hz occipital rhythm in infancy, reaches roughly 8 Hz by age three, approaches 9 to 10 Hz by late childhood and adolescence, and a normal adult PDR sits between roughly 8.5 and 13 Hz. The lower bound is the clinically load-bearing number: a PDR slower than 8 Hz in a fully alert, cooperative adult is abnormal and points toward diffuse cortical dysfunction. With normal aging the PDR drifts down only slightly and should remain at or above 8 Hz; a frequency in the 8 to 8.5 Hz range in an elderly patient is borderline and must be judged against state and reactivity rather than reflexively called.

Reactivity carries the most diagnostic weight, because it tests the integrity of the thalamocortical loop in real time. A well-formed but unreactive alpha is the hallmark of alpha coma, in which a comatose patient shows a widespread, monotonous, non-reactive alpha-frequency pattern that - despite looking superficially normal - signals severe brainstem or diffuse cortical injury and carries a grave prognosis. Several other features sharpen the read. A persistent amplitude asymmetry of the PDR greater than roughly fifty percent, or an interhemispheric frequency difference greater than about 1 Hz, is abnormal; the Bancaud phenomenon (loss of normal alpha reactivity over one hemisphere) is a classic sign of an underlying structural lesion on that side. The asymmetry rule is not symmetric, however, because the PDR is normally slightly higher in amplitude over the right hemisphere in many individuals: a modestly higher-voltage right-sided alpha is expected and not by itself pathological, whereas a left-predominant alpha clears the threshold for abnormality more readily, and the side of lower amplitude is generally the abnormal one. A useful convention holds that a right-greater asymmetry is tolerated up to roughly a two-to-one voltage ratio while a left-greater asymmetry is suspect at a lower ratio. The reader should also be alert to mimics: an unusually fast PDR may simply reflect a high-normal alpha, while a slow alpha variant (a subharmonic at roughly half the alpha frequency, around 4 to 5 Hz, with the same reactivity and topography as alpha) can masquerade as pathological theta until its reactivity and harmonic relationship to alpha are recognized.

Normal variants, mimics, and the artifacts that masquerade as rhythms

Several normal variants deserve explicit recognition so they are not misread as alpha or as pathology. The mu rhythm is an arciform (comb-shaped) alpha-frequency rhythm over the central regions (C3/C4) that reflects the idling sensorimotor cortex; crucially, mu does not block with eye opening - it attenuates with movement, with tactile stimulation, or even with the intention to move the contralateral limb. Its central topography and motor reactivity distinguish it from posterior visual alpha, and it may be asymmetric or unilateral in normal subjects, which should not be over-interpreted. Lambda waves are surface-positive occipital sharp transients evoked by scanning a complex visual scene with the eyes open; they are the waking analog of the sleep POSTS and are entirely benign. The benign variants of the temporal regions noted earlier - wicket spikes, RMTD, benign epileptiform transients of sleep, and the 14-and-6 positive bursts and 6 Hz spike-and-wave of drowsy adolescents - round out the catalogue of patterns that look ominous but are not, and their recognition is a defining competency of the experienced reader.

Equally important is the discipline of separating brain rhythms from artifact, because non-cerebral signals frequently impersonate physiological bands. Frontal eye-blink and lateral eye-movement artifact produces large slow deflections that can be mistaken for frontal delta until their phase reversal at the frontopolar electrodes and their correlation with the EOG channel are appreciated. Sustained scalp and especially temporalis muscle (EMG) activity fills the beta and gamma range with spiky, high-frequency potentials that vanish when the patient relaxes the jaw - a critical point given how completely EMG can swamp any genuine fast activity. The ECG and the pulse artifact (a slow wave time-locked to the heartbeat over electrodes lying near a scalp vessel) mimic rhythmic transients; sweat produces very-low-frequency baseline sway; 60 Hz line noise (50 Hz outside North America) contaminates a narrow band and signals an impedance or grounding problem. The cardinal rule is that a putative rhythm must have a plausible cerebral field and an electrophysiologically sensible distribution across the montage; a discharge that appears at a single electrode, or that does not respect any anatomical field, is artifact until proven otherwise.

The failure modes of frequency interpretation are therefore both technical and cognitive, and the expert guards against both. Technically, mistaking EMG for gamma, eye movement for frontal delta, or a breach rhythm for a focal abnormality are the recurring traps. Cognitively, the dominant biases are anticipation (reading the EEG to confirm the referral diagnosis - the patient referred for spells whose drowsy theta is then over-read as epileptiform), anchoring on the first striking feature and neglecting the state context that reframes it, and confirmation in the form of selectively weighting features that fit the expected answer. The Bayesian framing is the antidote: name the prior the clinical context sets, treat each waveform as a likelihood, and remain disciplined about features - reactivity, state, field, evolution - that would lower the probability of the favored interpretation, not only those that raise it. Quantitative descriptors anchor this discipline: report frequency in Hz; amplitude in microvolts (uV), conventionally graded low (under 20 uV), medium (20 to 50 uV), or high (over 50 uV) for the dominant background; symmetry across homologous regions; and reactivity and state. Quantitative EEG can formalize these as absolute and relative band power, spectral edge frequency, and the alpha-to-delta ratio used in ischemia monitoring - but the qualitative descriptors remain the language of the clinical report, and using them with discipline is what separates a precise read from a vague one.