Physiologic Artifacts

The scalp EEG is an exquisitely sensitive voltmeter, and it does not discriminate between the cerebral generators we wish to measure and the many non-cerebral generators that share the same head. Among these intruders, the physiologic artifacts are the most common, the most stereotyped, and paradoxically the most dangerous, precisely because they are biological rather than electrical. A line-frequency artifact announces itself with mechanical regularity; an eye flutter or a temporalis discharge wears the disguise of brain. The expert reader does not attempt to filter these signals away so much as to understand them. Every physiologic artifact obeys a predictable field, a predictable polarity, and a predictable temporal relationship to the patient's behavior or autonomic state. Once those rules are internalized, the artifact ceases to be a nuisance and becomes a source of information - and frequently a source of diagnosis in its own right, since the EEG is also a polygraph that records the eyes, the heart, the muscles, and the movements of the patient.

A unifying mental model anchors this entire module. Scalp electrodes do not see neurons; they see volume-conducted potentials, the smeared-out summation of any current source within the conductive head, whether that source is a sheet of cortical pyramidal cells, a battery-like eyeball, a contracting muscle, or a beating heart. Whether a source produces a recognizable artifact depends on three quantities: its strength, its distance from the electrode, and its orientation relative to the recording derivation. The corneoretinal dipole is millivolts strong; a cortical generator is microvolts strong but engages square centimeters of synchronized tissue. The art of artifact reading is the disciplined reverse-engineering of which generator, cortical or otherwise, best explains the field, the polarity, the frequency, and the timing that appear on the page.

The Corneoretinal Dipole and the Logic of Eye Movement

Ocular artifacts dominate the frontal channels of nearly every record because the eye is a powerful standing dipole. The cornea is electropositive relative to the retina by roughly 0.4 to 1.0 millivolts, an enormous potential by EEG standards where cortical signals are measured in tens of microvolts. This corneoretinal dipole is not generated by excitable tissue but is attributed to the high metabolic activity of the retinal pigment epithelium; it is a battery anchored to the globe that rotates with it. When the eyes move, the positive pole of each dipole swings toward or away from the frontopolar electrodes, generating a deflection far larger than any cerebral rhythm and governed entirely by geometry rather than by cortex. The EOG signal changes by approximately 20 microvolts per degree of rotation and behaves linearly across roughly thirty degrees of gaze. The single most useful insight in all of artifact reading follows directly: because the corneoretinal dipole is positive at the cornea, any electrode that the cornea turns toward becomes relatively positive, and any electrode the cornea turns away from becomes relatively negative.

Consider the eye blink, the prototypical vertical event. The classic teaching attributes the blink potential entirely to the globe rolling upward, the Bell phenomenon, but careful magnetic search-coil work has refined this account, and the modern reader should hold the more accurate picture. The dominant generator of the blink potential is the upper eyelid sliding downward across the positively charged cornea: the moving lid alters the electrical field around the standing corneoretinal dipole and drives the supraorbital electrodes Fp1 and Fp2 positive. The eyeball itself contributes a smaller, state-dependent share - an ordinary spontaneous blink is accompanied by a small globe movement directed downward and inward rather than upward, whereas the upward roll of the true Bell phenomenon emerges mainly with slow or forced eye closure. The clinically decisive point is unchanged across both mechanisms: a blink produces a synchronous, symmetric deflection that is positive at the frontopolar electrodes, broadly distributed, time-locked to the moment of lid closure and decaying rapidly as one moves posteriorly. The blink field is bilateral and frontally maximal with no phase reversal across the parasagittal chain, which immediately distinguishes it from a frontal cerebral discharge that would show a localizable field and a true phase reversal. Vertical eye and lid movements recorded from a dedicated infraorbital electrode show the same potential with reversed polarity relative to the supraorbital scalp pair, because the infraorbital lead sits on the opposite side of the eye. This out-of-phase relationship between a sub-eye electrode and Fp is the single cleanest confirmation that a frontal deflection is ocular and not cerebral, and it is the reason a vertical EOG montage resolves nearly every blink-versus-spike question instantly.

Horizontal, or lateral, gaze obeys the same dipole logic but maps onto the anterior temporal electrodes F7 and F8. When the eyes deviate to the left, the left cornea turns toward F7, driving it positive, while the right cornea turns away from F8, driving it negative, producing a characteristic out-of-phase pair: a positive deflection at F7 mirrored by a negative deflection at F8. This phase opposition between the two anterior temporal chains is the signature of a conjugate horizontal saccade and should never be mistaken for independent temporal activity. A second, subtler feature frequently rides on the leading edge of a lateral saccade: the lateral rectus spike, also called the saccadic spike potential. This is a brief myogenic potential generated by the contracting lateral rectus muscle, appearing as a small monophasic spike at F7 on leftward gaze or F8 on rightward gaze, in phase with the onset of the eye movement. Its temporal coupling to gaze, its restriction to the ipsilateral anterior temporal electrode, and its consistent association with the larger ocular deflection prevent it from being read as an anterior temporal sharp wave, a mimicry it superficially resembles. In sleep recordings, the slow rolling eye movements of N1 and the rapid, conjugate, often phasic saccades of REM are themselves the diagnostic content - here the ocular artifact is the signal, and recognizing its frontal out-of-phase fields is how one stages the record.

Two ocular patterns deserve special mention because of their diagnostic mimicry. Eyelid flutter produces rhythmic vertical potentials at Fp1 and Fp2 in the 4 to 8 Hz range that can resemble frontal rhythmic theta or, when slower, frontal intermittent rhythmic delta activity (FIRDA). The discriminators are reactivity and field: flutter is confined to the frontopolar electrodes, abolished by gentle eye closure or by asking the patient to hold the eyes lightly shut, and accompanied on the EOG channel by the unmistakable squared-off ocular waveform that inverts against a sub-eye lead. True FIRDA, by contrast, extends posteriorly to F3/F4 and beyond and persists with the eyes closed. The second pattern is the slow lateral eye movement of drowsiness, in which the eyes drift conjugately and slowly from side to side as the patient transitions toward sleep. These produce gentle, out-of-phase, low-frequency excursions at F7 and F8 that are a normal and useful marker of light drowsiness, but that can be misread as independent bitemporal slowing by a reader who has not noticed their reciprocal polarity. A genuine bitemporal slow process would not show the perfect mirror-image phase relationship that betrays a conjugate ocular origin.

A third ocular pitfall is the nystagmus artifact, in which the alternating slow and fast phases of a nystagmoid eye movement imprint a rhythmic, partly sawtooth pattern on the anterior temporal and frontopolar electrodes. Recognized through its out-of-phase horizontal field and, again, an EOG channel, it is occasionally the first objective sign of a vestibular or brainstem process. The recurring lesson is that the eye is simultaneously the commonest source of frontal artifact and a rich source of clinical information, and the discipline that turns one into the other is the dedicated EOG derivation.

Myogenic Contamination: The EMG Field

Muscle artifact is the second great physiologic contaminant and arises whenever a muscle within range of an electrode contracts. The motor unit action potentials of the cranial musculature are very brief events, and their summed spectral energy sits predominantly above 20 Hz, frequently extending past 70 to 100 Hz. This high-frequency, spiky, irregular discharge is most prominent over the electrodes that overlie the largest cranial muscles: the frontalis beneath Fp1/Fp2 and the frontal chain, and above all the temporalis beneath F7/F8/T3/T4/T5/T6. Because the temporalis is large and tonically active in an anxious, cold, or uncomfortable patient, persistent temporal EMG is one of the most common reasons a routine record is rendered uninterpretable over precisely the region where mesial temporal epileptiform discharges would appear. The clinical irony is acute: the artifact preferentially obscures the very territory most often interrogated in epilepsy.

The defining features of EMG are its frequency content and its morphology. A genuine cerebral spike has a duration of 20 to 70 milliseconds and a defined, repeatable field with an after-going slow wave; a muscle spike is shorter, typically well under 20 milliseconds, occurs in irregular trains, lacks an after-going slow wave, and shows a field that is maximal at a single electrode and drops off almost completely at the next, because the generator is millimeters away in the scalp rather than centimeters away in the brain. The danger lies in the single-motor-unit discharge or the brief myogenic burst that, on casual inspection, resembles a polyspike. The remedy is never to reach first for the low-pass filter. Aggressive low-pass filtering - for example dropping the high-frequency filter from 70 Hz to 15 Hz - will smooth a muscle spike into a rounded waveform that looks far more like a cerebral sharp wave than the original did. The correct response to muscle artifact is to relax the patient, to reposition, to warm a cold patient, to ask the patient to let the jaw hang loosely open, and only then to interpret what remains. A useful confirmatory maneuver is to ask the patient to clench the jaw deliberately: the burst that intensifies with clenching and abates with relaxation is myogenic by definition.

Several specialized myogenic patterns round out the picture. Glossokinetic artifact arises from movement of the tongue and appears as slow, frontally maximal delta-range activity that can wax and wane and even appear to evolve, mimicking the rhythmic theta-delta of a frontal seizure or anterior slow dysrhythmia. The traditional teaching modeled the tongue as a simple dipole with a negative tip and positive base, but careful intracranial work has shown that this is an oversimplification; the tongue most likely shunts other scalp and soft-tissue currents rather than acting as a clean dipole, which is why the polarity behavior of the artifact is more variable than the textbook account predicts. Whatever its precise generator, glossokinetic artifact is recognized by its time-locking to tongue movement, talking, chewing, or sucking, by its strictly anterior maximum at Fp1/Fp2, and by the absence of the orderly field and evolution of a true ictal rhythm. Asking the patient to say a repeated lingual consonant such as la-la-la reproduces it on demand. A related entity, chewing and sucking artifact, combines temporalis and masseter EMG with rhythmic glossokinetic slowing and is a classic mimic of an evolving temporal seizure in infants and uncooperative patients - resolved, like so much else, by synchronized video.

Tremor artifact deserves its own caution because it produces a rhythmic mechanical-plus-myogenic potential whose frequency falls in the same 4 to 6 Hz band as theta. A resting tremor in Parkinson disease or an essential tremor can imprint a remarkably regular rhythm on electrodes near the affected musculature, and without an accelerometer or EMG channel and synchronized video it can be misattributed to a cerebral rhythm. The discriminators are the lack of a cortical field, the lack of frequency evolution, and the tight correspondence to the observed tremor; a dedicated surface EMG or accelerometer channel converts the question into an immediate answer.

Cardiac and Pulse Contamination

The heart contributes two distinct artifacts. The first is the electrocardiographic (ECG) artifact, the volume-conducted QRS complex appearing in the EEG as a sharp, regular deflection recurring at the heart rate. It is usually small and most visible in derivations with long interelectrode distances and in patients with short, broad necks or high body mass, where the cardiac vector projects efficiently onto the scalp; it is also accentuated in referential montages using ear or mastoid references that lie close to the cardiac axis. The cardinal feature is its rhythmicity: a sharp transient that recurs with metronomic regularity at roughly 60 to 100 per minute, perfectly time-locked to the simultaneously recorded ECG channel, is cardiac until proven otherwise. This is the single most important reason a dedicated single-lead ECG channel must accompany every EEG; without it, a regularly recurring sharp transient can be misread as periodic epileptiform activity. The mimicry is most treacherous in the comatose or burst-suppressed patient, where the reader is primed to find periodic discharges and where ECG artifact can masquerade as generalized periodic discharges at approximately one per second.

A particularly instructive variant is pacemaker artifact, in which the sharp, narrow pacing stimulus appears as a needle-thin spike preceding the paced complex; because it is even briefer and more stereotyped than a QRS, it can be mistaken for an electrode pop or a repetitive spike until it is matched to the ECG and the patient's device history. A second special case arises in cardiac arrest and resuscitation, where chest compressions during cardiopulmonary resuscitation produce a rhythmic compression artifact at roughly 100 to 120 per minute that can dominate the trace; recognizing it as mechanical, and reading the EEG only in the brief pauses between compression cycles, prevents both false reassurance and false alarm during one of the highest-stakes interpretive moments in the intensive care unit.

The second cardiac artifact is the pulse artifact, generated when an electrode happens to sit over a scalp arteriole. Each arterial pulsation mechanically moves the electrode, producing a slow, smooth, dome-shaped wave that follows each QRS by approximately 200 to 300 milliseconds, the delay reflecting pulse transit time from heart to scalp. Pulse artifact is therefore not sharp like ECG but slow and rolling, confined to one or two electrodes, and locked to the ECG with a fixed lag. Because it is slow and focal, it can be mistaken for focal delta activity, and the discriminator is again the ECG channel: a slow wave that recurs at a fixed interval after every QRS, and that resolves or changes when the electrode is repositioned a few millimeters, is mechanical and not cerebral. A related entity is the ballistocardiographic artifact seen in simultaneous EEG-fMRI, where pulsatile head and blood motion within the static magnetic field produce large rhythmic contamination requiring dedicated removal algorithms, but in routine clinical recording the simple pulse artifact is what the reader must recognize at the bedside.

Field Logic and Bayesian Reasoning as the Unifying Tools

What unites every example above is that the discriminating evidence is rarely the morphology of the waveform itself and almost always its field, its polarity relationships, and its timing relative to a reference channel. A cerebral generator is a sheet of synchronized pyramidal cells centimeters below the scalp; its potential therefore spreads smoothly across several adjacent electrodes and, in a bipolar montage, produces a phase reversal that localizes the generator. A physiologic artifact violates one of these expectations: the ocular dipole produces fields that are too frontal, too symmetric, and reverse against a sub-eye lead; muscle produces fields too focal and too fast to be cortical; cardiac signals recur with a regularity no cortical rhythm sustains and lock to the ECG. The expert reading mind runs a continuous Bayesian check, asking of every suspicious transient whether its field, polarity, frequency, reactivity, and temporal coupling are those of brain or of a known impostor.

It is worth making the Bayesian structure explicit, because it is the antidote to overreading. The posterior probability that a transient is epileptiform depends not only on how spike-like it looks (the likelihood) but on the prior probability of artifact in that context. A perfectly regular sharp transient in a comatose patient on a ventilator has a high prior for ECG artifact; a sharp deflection confined to Fp2 in a blinking, anxious patient has a high prior for ocular or frontalis origin; a rhythmic 5 Hz run in a tremulous patient has a high prior for tremor. When the prior probability of artifact is high, that hypothesis must be actively excluded - by inspecting the reference channels, by trying a second montage, by manipulating the patient - before any abnormality is reported. The reader who attends only to morphology, ignoring the base rates that context supplies, is the reader most likely to manufacture disease.

The cost of the opposite error is concrete and asymmetric. Patients have been committed to antiseizure medication, and even admitted to the epilepsy monitoring unit, on the strength of eye flutter read as frontal spikes or ECG artifact read as periodic discharges. A missed equivocal sharp wave rarely changes management on its own, because the diagnosis of epilepsy is built from convergent evidence rather than a single transient; but a confidently miscalled artifact can cascade into a wrong diagnosis, a wrong drug, a lost driving license, and a durable label. The disciplined reader therefore treats the word epileptiform as a high bar, demands that artifact be excluded before it is cleared, and recognizes that the failure modes of EEG interpretation are not only electrical and physiologic but cognitive - confirmation bias toward the referral question, anchoring on the first impression, and the false sense of safety in overcalling.