State-Dependent EEG

State is the hidden variable

Every EEG is recorded in a particular brain state, and that state silently conditions the meaning of everything on the page. The transition from wakefulness to sleep is not a dimmer switch but a sequence of discrete reconfigurations of the thalamocortical system, each orchestrated by a specific change in ascending neuromodulation. During wakefulness, tonic drive from the cholinergic brainstem and basal forebrain, the noradrenergic locus coeruleus, the serotonergic raphe, the histaminergic tuberomammillary nucleus, and the orexin (hypocretin) neurons of the lateral hypothalamus keeps thalamic relay and cortical neurons depolarized and the cortex desynchronized. As these systems withdraw their tonic firing at sleep onset, thalamic relay neurons hyperpolarize and switch from the tonic single-spike firing of wakefulness to the rhythmic burst firing that generates spindles and slow waves. The switch itself is sharpened by a flip-flop architecture: the sleep-promoting GABAergic and galaninergic neurons of the ventrolateral preoptic area (VLPO) mutually inhibit the wake-promoting monoaminergic nuclei, so the system snaps between consolidated wake and consolidated sleep rather than dwelling in intermediate states, with orexin neurons stabilizing the wake side of the switch. Understanding state is therefore not a preliminary nicety - it is the precondition for interpretation, because you cannot call a rhythm abnormal until you know what state should have produced it.

Wakefulness and the treacherous descent into drowsiness

Alert wakefulness with eyes closed is defined by a well-organized posterior dominant alpha rhythm that blocks crisply on eye opening, with low-amplitude beta admixed frontally and abundant eye-blink and muscle artifact that themselves testify to an awake, active patient. This is the reference state against which everything else is measured, and confirming it - by demonstrating a reactive PDR - is the anchor of any normal read. The presence of lambda waves with the eyes open, and of normal eye-movement and muscle artifact, are corroborating signs of genuine wakefulness that an experienced reader notes almost unconsciously.

Drowsiness is the most treacherous state on the routine EEG because its normal features are precisely the ones that mimic pathology. As vigilance fades the alpha rhythm slows, fragments, and spreads anteriorly before dropping out; slow, lateral, roving eye movements appear on the frontal channels; and runs of diffuse or temporal theta emerge. In children and young adults, drowsiness can produce striking paroxysmal rhythmic slowing - hypnagogic hypersynchrony, high-amplitude rhythmic 3 to 5 Hz activity that appears abruptly and is entirely benign - as well as benign sharp-looking transients that the inexperienced reader mislabels as epileptiform. The benign epileptiform transients of sleep (BETS, also called small sharp spikes) and rhythmic mid-temporal theta of drowsiness (RMTD) both emerge in this state. The expert response is to recognize the constellation of drowsiness - anterior spread of alpha, roving eye movements, dropout of muscle artifact, and the appearance of vertex waves - and to grade subsequent findings against a drowsy, not an awake, baseline. The Bayesian point is sharp here: the prior probability that a temporal sharp transient is epileptiform falls substantially once the surrounding epoch is recognized as drowsy, because drowsiness is a strong generator of benign mimics.

The NREM stages: N1 through N3 and their graphoelements

N1 is the lightest sleep, the formal continuation of drowsiness in the standard staging framework: the alpha rhythm is now present less than half the epoch, the background is low-voltage mixed theta, and the stage is marked by vertex sharp waves (sharply contoured negative transients maximal at the vertex, Cz) and, especially in children, POSTS (positive occipital sharp transients of sleep). These are normal sleep transients; their sharp morphology is the classic source of false epileptiform calls, and vertex waves can be strikingly sharp, repetitive, and even slightly asymmetric in children without being pathological. The diagnostic discipline is to recognize that vertex waves are maximal at the midline, occur in the appropriate state, and lack the after-going slow wave and the regional field of a true central spike.

N2 is the workhorse stage of sleep and is defined by two cardinal graphoelements. Sleep spindles are waxing-and-waning bursts in the sigma band (roughly 11 to 16 Hz) generated by the thalamic reticular nucleus pacing relay neurons - the purest electrographic readout of the thalamus in action. Two spindle subtypes are now well established and worth knowing: slow spindles of roughly 11 to 13 Hz that predominate frontally, and fast spindles of roughly 13 to 15 Hz that predominate over centroparietal regions, with distinct generators and distinct roles in memory consolidation. K-complexes are large biphasic transients - a surface-negative sharp wave followed by a slower positive component - maximal frontocentrally, that arise spontaneously or in response to a stimulus and represent a stereotyped cortical down-state, a network-wide silencing that may serve both arousal regulation and sleep-dependent memory processing. Spindles and K-complexes often couple in a precise temporal sequence with slow oscillations, a coupling that is increasingly studied as a substrate of memory consolidation and that degrades in aging and in neurodegenerative disease. Recognizing spindles and K-complexes is what lets you confidently declare a patient asleep and reinterpret any accompanying slowing as state-appropriate; their absence, asymmetry, or marked attenuation can itself be a sign of thalamic or hemispheric pathology.

N3, slow-wave sleep, is defined by high-amplitude delta - conventionally measured as activity over 75 uV in the 0.5 to 2 Hz range occupying at least 20 percent of the epoch over the frontal derivations where it is maximal. This is the cortical slow oscillation writ large: the synchronized alternation of cortical up- and down-states across vast territories, sculpted by the withdrawal of cholinergic and aminergic tone, and it is the physiological state in which delta is not only normal but expected and even necessary. The amount of slow-wave activity is the principal electrographic index of homeostatic sleep pressure, and it is here that the contrast with pathology is sharpest: identical-looking delta in an awake patient signals encephalopathy, whereas here it signals nothing more than deep, healthy sleep. State, once again, inverts the meaning of the waveform - the same morphology that would prompt a stroke or encephalopathy work-up in an alert patient is the reassuring hallmark of restorative sleep.

REM sleep, the paradoxical state, and its clinical biomarkers

REM sleep earns its old name paradoxical sleep because the cortical EEG returns to a low-voltage, mixed-frequency, wake-like pattern even as the person is most deeply disconnected from the environment. The state is generated by a brainstem flip-flop of its own, in which cholinergic and glutamatergic REM-on populations in the pons reciprocally interact with aminergic REM-off populations; the locus coeruleus and raphe fall nearly silent in REM, while pontine cholinergic drive surges. The defining triad is the desynchronized EEG, rapid conjugate eye movements visible on the EOG channels, and profound skeletal muscle atonia on the chin EMG - the latter produced by active glutamatergic and glycinergic/GABAergic inhibition of spinal motor neurons via a sublaterodorsal/medullary circuit, a brainstem-mediated paralysis that prevents dream enactment. Sawtooth waves (frontocentral notched, serrated theta) often herald the eye-movement bursts. Because the cortical EEG alone resembles wakefulness, REM can only be identified by integrating the EOG and EMG channels, which is precisely why polysomnography records them and why no single scalp derivation can stage REM.

Circadian and homeostatic modulation, and state as an activation strategy

State is governed by the interaction of two clocks, the framework Borbely formalized as the two-process model. The homeostatic process (process S, sleep pressure, indexed by slow-wave activity) builds monotonically with time awake and discharges during sleep, so an EEG recorded after sleep deprivation shows faster, deeper transitions into sleep and richer slow-wave content. The circadian process (process C), driven by the suprachiasmatic nucleus and entrained by light, gates alertness across the 24-hour cycle, so the same patient is far more likely to drift into drowsiness during the mid-afternoon dip or in the early morning hours when the circadian alerting signal is at its nadir. These dynamics are not academic. The sleep-deprived EEG is a deliberate activation procedure precisely because the descent through drowsiness and light sleep dramatically increases the yield of interictal epileptiform discharges, and natural sleep - whether spontaneous or induced by deprivation - is among the most powerful activators available, complementing hyperventilation and photic stimulation. Certain epilepsy syndromes are tightly state-locked: the generalized spike-wave of idiopathic generalized epilepsies is markedly activated by drowsiness, sleep, and awakening; benign epilepsy with centrotemporal spikes (rolandic epilepsy) may show discharges almost exclusively in sleep; and the continuous spike-wave of slow-wave sleep is, by definition, a phenomenon of NREM. Knowing which state to capture is therefore a clinical decision with direct diagnostic consequences, and a normal routine study in a patient with a strong history is often an indication to obtain a sleep-deprived or prolonged recording rather than to reassure.

The unifying clinical message is that interpretation is conditioned on state at every step. The reader's first task is always to establish where on the wake-sleep continuum the recording sits, because the same delta, the same theta, the same sharp transient flips between normal and abnormal as the state changes beneath it - and the cognitive biases that ambush state interpretation are predictable. Anticipation bias leads a reader who expects sleep to under-call genuine slowing as drowsy, or one who expects wakefulness to over-call sleep transients as epileptiform. Anchoring on a single dramatic graphoelement - a sharp vertex wave, an asymmetric spindle - without integrating the surrounding state context produces both false positives and false negatives. The remedy is procedural: read the eye and muscle channels, identify the state explicitly before grading any abnormality, and ask at each suspicious feature whether a state-appropriate benign explanation exists. Mastery of state-dependent EEG is thus not a separate topic from abnormality - it is the lens through which all abnormality is judged, and it directly shapes the most consequential decisions in the laboratory: whether to extend a study, induce sleep, sleep-deprive a patient, or record overnight to capture the diagnostic state.