EEG Recording Systems
Electrodes, the 10-20 system, montages, and amplifiers
The electrode: a chemical transducer
An EEG electrode is not a passive wire; it is an electrochemical transducer that converts ionic current in the tissue into electronic current in the wires. This conversion happens at the metal-electrolyte boundary, where a half-cell potential develops as ions exchange between metal and solution, and the quality of the recording depends critically on how stable and reversible that interface is. The field standard is the silver/silver-chloride (Ag/AgCl) electrode, whose reversible chloride exchange yields a low, stable half-cell potential and minimal drift, making it suitable for the low-frequency content that other electrode metals would distort. A poorly chosen or unstable interface behaves as a partially polarizable electrode that accumulates charge, blocks slow signals, and drifts; the Ag/AgCl interface is prized precisely because it is non-polarizable, passing direct and very-low-frequency current faithfully. A conductive electrolyte gel or paste bridges the electrode to the scalp, displacing the high-resistance stratum corneum and completing the ionic path.
Two mismatched electrodes, or an unstable interface, introduce slow drifts and offset potentials that the amplifier must reject; this is one reason consistency of electrode type across the array matters. A particularly important failure mode is the salt bridge: when gel from two adjacent electrodes runs together across sweaty skin, it short-circuits them, collapsing the potential difference between the sites and producing a spuriously flat or attenuated channel that can be misread as focal voltage loss. Modern systems also use active electrodes, which place a small buffer amplifier at the electrode itself to lower the effective source impedance before the signal travels down the cable, dramatically reducing susceptibility to movement and electromagnetic interference and relaxing the traditional demand for low skin impedance.
Click any electrode. Letters denote the region (Fp, F, C, T, P, O); even numbers lie on the right, odd on the left, and z marks the midline.
The International 10-20 System places electrodes at 10% and 20% intervals along measured arcs between the nasion, inion, and preauricular points — so placement scales to any head size.
Toggle the bipolar chains to see how adjacent electrodes are linked into the montages you read every day.
The international 10-20 system
Electrode positions are never eyeballed; they are measured as proportions of the distances between bony landmarks - the nasion (bridge of the nose), the inion (occipital protuberance), and the two preauricular points. Positions are placed at 10 percent and 20 percent of these distances along the measured contours, which is the origin of the name. Because every coordinate scales to the individual head, a given label marks approximately the same cortical territory across skulls of any size, which is what makes recordings comparable between patients and laboratories and across time in the same patient. This proportional logic, standardized by Jasper in 1958, is one of the quiet triumphs of clinical neurophysiology: it converted an idiosyncratic art into a reproducible measurement.
- Letters encode region: Fp (frontopolar), F (frontal), C (central), T (temporal), P (parietal), O (occipital), with A or M for the ear/mastoid references.
- Odd numbers lie over the left hemisphere, even numbers over the right; the subscript z marks the midline (Fz, Cz, Pz).
- Numbers increase with distance from the midline, so smaller numbers are more medial (e.g., C3 is nearer the vertex than T7/T3).
- The expanded 10-10 and high-density 10-5 systems interpolate additional sites and have largely standardized the modern modified combinatorial nomenclature (e.g., T7/T8 replacing the older T3/T4, and T9/T10 and FT9/FT10 adding inferior-temporal coverage).
Absolute (centimeter) placement would put the same physical coordinate over different gyri in a large versus a small head. Proportional 10-20 placement guarantees that Cz overlies the vertex and C3/C4 overlie the hand motor strip regardless of head size - the foundation of reproducible localization and of any meaningful comparison across patients or across serial studies.
An expert also knows the system's coverage gaps. The standard 10-20 array under-samples the inferior and basal temporal surfaces, which is precisely where mesial temporal epileptiform activity is most likely to project. This is why supplementary electrodes - inferior temporal chains, sphenoidal or true-temporal placements, and the T9/T10 sites of the 10-10 system - are added when temporal lobe epilepsy is the question. Failing to extend coverage in such cases is a recognized cause of false-negative recordings: the generator was real, but no electrode sat where its field reached the scalp.
Montages: bipolar, referential, and average reference
Because EEG is intrinsically a measurement of potential difference, what you see depends entirely on which two inputs each channel compares. A montage is a defined, ordered set of such derivations. In a bipolar montage, adjacent electrodes are chained in lines (for example the classic anterior-to-posterior temporal and parasagittal chains of the double-banana), so each channel shows the difference between two nearby scalp sites. Bipolar montages excel at pinpointing a focal generator through phase reversal: at the channel pair flanking the source, the deflections point toward (or away from) each other, and the electrode shared by those two channels marks the field maximum. Because each channel subtracts two nearby sites, bipolar chains also reject far-field and widely distributed activity, sharpening focal features at the cost of suppressing genuinely diffuse ones.
In a referential (monopolar) montage, every electrode is compared against a single common reference (an ear, the mastoid, or a designated site such as Cz). Referential montages preserve true waveform morphology and amplitude and reveal the spatial extent of a field better than bipolar chains, but they are only as good as the reference. An active reference - one that picks up cerebral or artifactual activity of its own - will inject that activity, inverted, into every channel, creating spurious widespread findings or, worse, cancelling the real one when the reference sits in the field of interest. The average reference computes the mean of all electrodes and subtracts it from each, approximating an inactive reference when coverage is dense and symmetric; it is excellent for broad fields but can be skewed by a single high-amplitude or artifact-laden channel that contaminates the common average and therefore distorts every derivation simultaneously.
| Montage | Reference | Strength | Principal limitation |
|---|---|---|---|
| Bipolar (chains) | Adjacent electrode | Localizes focal sources via phase reversal; rejects far-field noise | Distorts amplitude/morphology; attenuates broad or in-phase fields |
| Referential | Single common site | Preserves true morphology and amplitude; shows field extent | Vulnerable to a contaminated or 'active' reference that contaminates all channels |
| Average reference | Mean of all electrodes | Approximates an inactive reference; good for widespread fields | A single high-amplitude/artifact channel contaminates every derivation |
The expert practice is to read multiple montages and let them cross-check one another, because no single montage is sufficient. A finding that localizes by phase reversal on a bipolar montage should show a corresponding amplitude maximum on a referential montage; a putative focus that appears only in one montage and vanishes in another is suspect. The classic trap is the reference itself becoming the source - for example, a temporal discharge contaminating an ipsilateral ear reference, producing apparent activity at every other electrode and a confusing field that resolves only when the montage is changed. Disciplined reformatting between montages is not a stylistic preference; it is how one separates true cortical events from referencing artifacts and from the distortions each montage inevitably imposes.
Click to move the focus. Color shows the scalp potential (negative max in blue).
Find the phase reversal. In a bipolar chain, the electrode of maximal negativity sits where adjacent deflections point toward each other (an upgoing then downgoing pair, since EEG is plotted negative-up). That confluence localizes the source — here, T7. Switch to a positive focus and the reversal flips. Channels that don't cross the focus show little or no deflection.
The differential amplifier and common-mode rejection
The scalp signal is on the order of tens of microvolts, while the surrounding environment - mains power, lighting, monitors, the patient's own muscle and movement - imposes electrical noise that can be thousands of times larger. The instrument that rescues the signal is the differential amplifier, which amplifies the difference between its two inputs while rejecting whatever is common to both. Noise from a distant source (such as 50/60 Hz line interference) arrives at two nearby electrodes nearly identically; the differential amplifier subtracts this common-mode signal away, leaving the differential-mode brain activity that differs between the inputs. The figure of merit is the common-mode rejection ratio (CMRR), expressed in decibels; high-quality EEG amplifiers achieve CMRR well above 100 dB, meaning common-mode interference is attenuated by a factor exceeding one hundred thousand relative to differential signal.
Crucially, common-mode rejection only works if the two input impedances are low and well matched. If one electrode has high or unbalanced impedance, the common-mode noise divides unequally between the inputs (a voltage-divider effect against the amplifier's finite input impedance), converts into a differential signal, and is amplified along with the brain activity - the mechanism by which a single poorly applied electrode injects 60 Hz buzz or movement artifact into a channel. This is why technologists are taught to keep electrode impedances low (conventionally below about 5 kiloohms) and, just as importantly, balanced across the montage. It is the imbalance, more than the absolute value, that defeats rejection: two electrodes at 10 kiloohms matched to each other may record cleanly, while one at 5 and one at 40 will not. Impedance management is therefore not housekeeping; it is the precondition for the amplifier's noise rejection to function at all.
It is worth being precise about what impedance is. At EEG frequencies the electrode-skin interface is not a pure resistance but a frequency-dependent impedance with capacitive character, dominated by the stratum corneum. This is why impedance is checked with a small alternating current rather than direct current, and why abrading or hydrating the skin - lowering that capacitive barrier - is the technologist's chief lever. The rise of active electrodes and high-input-impedance amplifiers has changed this calculus: by buffering the signal at the source, active systems tolerate far higher and less perfectly matched skin impedances, which is what makes rapid-application and dry-electrode montages feasible at all.
By long-standing clinical convention, when input 1 (the first electrode of a derivation) is negative relative to input 2, the pen deflects upward. This 'negative-up' rule is the inverse of standard engineering plots and is the basis for reading phase reversals: at a surface-negative focus, the two channels sharing that electrode deflect in opposite vertical directions, and the shared electrode marks the maximum. Misremembering this convention inverts every localization you attempt.
Dry electrodes, density, and the 2026 recording landscape
The instrumentation chain is being reshaped on two fronts that an expert in 2026 must situate accurately. The first is the maturation of dry and semi-dry electrodes - pin, fabric, and polymer designs that contact the scalp without gel. Their appeal is obvious: vastly faster setup, suitability for ambulatory and home recording, and patient comfort over long monitoring. Their cost is higher and less stable contact impedance and greater motion sensitivity, mitigated but not eliminated by active buffering. The honest current position is that gel-based Ag/AgCl remains the reference standard for diagnostic-grade clinical EEG, while dry systems have become genuinely useful for screening, long-term and ambulatory monitoring, brain-computer interfaces, and high-volume applications where speed and tolerability outweigh the loss of low-frequency fidelity and the increase in artifact.
The second front is electrode density and miniaturization. High-density arrays of 64, 128, and 256 channels improve the spatial sampling that source localization depends upon, partially counteracting the blur of volume conduction by oversampling the field; they are now standard for research and presurgical source analysis. At the opposite extreme, miniaturized and even subscalp or ear-worn devices are enabling ultra-long-term seizure monitoring outside the laboratory, trading spatial coverage for the ability to capture rare events over weeks. Across all of these, the underlying physics of this module is unchanged: every system still rests on a stable electrode interface, proportional and reproducible placement, a reference whose limitations must be understood, and a differential amplifier whose rejection depends on balanced impedance. New hardware changes the trade-offs, not the principles.
Choice of recording system is itself a clinical decision. A rapid dry-electrode montage may be exactly right for ambulatory screening or ICU trend monitoring but inadequate for characterizing subtle interictal morphology or for source localization, where gel-based high-density arrays co-registered to MRI are appropriate. Knowing the fidelity limits of the system in front of you is part of reading its output honestly.
1. A single channel in an otherwise clean referential recording shows prominent 60 Hz noise. The most likely cause is:
2. In a longitudinal bipolar (double-banana) montage, a focal surface-negative spike produces deflections that point toward each other across two adjacent channels. This phase reversal indicates that:
3. A referential recording shows apparent epileptiform activity at nearly every electrode, with morphology that looks identical across the scalp. Before diagnosing widespread cerebral pathology, the expert first suspects: