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 7
Chapter 7 · 1 h · 8 quiz items · pass at 80%
BCIA Domain III requires the practitioner to place electrodes accurately and to understand what a montage does to the signal. This module gives the 10-20 arithmetic, the naming convention, the single-channel sites for common protocols, and the montage comparison that determines how a coherence value should be read. The quiz proves the learner can measure a site and choose a reference defensibly.
Where you put the electrode determines what you measure. A site a centimeter off sits over different cortex. A montage chosen differently makes the same brain look like a different brain. This is the chapter where the abstractions of the previous one (impedance, amplification, the differential pair) become a map on a real scalp, and where two skills separate competent practitioners from the rest: measuring a head accurately by the International 10-20 system, and understanding how the choice of reference and montage shapes everything you then read. Both are testable, and both are things you do with your hands before any software runs.
This chapter covers the 10-20 system in full (the landmarks, the arithmetic, the naming convention), the standard arrays you will actually use, the single-channel placements that anchor common neurofeedback protocols, and then the montages (linked ears, average reference, bipolar, Laplacian) with an honest account of what each emphasizes, what each hides, and when to reach for it. The recording-procedure depth lives in the companion QEEG Field Guide. Here the focus is the geometry and the referencing logic a BCN candidate must own.
The problem the 10-20 system solves is reproducibility. Heads differ in size and shape, so a placement measured in fixed centimeters would land on different cortex in different people, and a recording made at one clinic would not correspond to a recording made at another or to the normative databases everything is compared against. The system's answer, introduced by Herbert Jasper for the international EEG federation in 1958, is to place electrodes at fixed percentages of the distance between bony landmarks rather than at fixed absolute distances. Because the percentages scale with each head, a site lands over the same cortical region regardless of head size, and the name "10-20" refers to those percentages: electrodes sit at 10 percent or 20 percent intervals along the measured lines.
Four bony landmarks anchor the whole system, and finding them accurately by palpation is the foundational manual skill.
From those four landmarks the placement is built by measurement.
The anteroposterior (front-to-back) line. Measure the distance from nasion to inion straight over the top of the head along the midline. This total distance is divided into the canonical intervals. Stepping back from the nasion: the frontal-pole midline (Fpz, the reference point for Fp1 and Fp2) sits at 10 percent of the nasion-inion distance up from the nasion; then 20 percent intervals carry you to Fz, Cz, and Pz in turn; and the occipital-pole midline (Oz) sits at 10 percent up from the inion. The pattern over the midline is therefore 10-20-20-20-20-10, summing to the full 100 percent. Cz, the vertex, lands at exactly 50 percent of the nasion-inion line, the midpoint.
The coronal (ear-to-ear) line. Measure from the left preauricular point to the right preauricular point straight over the top, and crucially this line must cross the midline measurement exactly at Cz. The intersection of the two 50-percent midpoints defines the vertex, and getting Cz right is the linchpin of the whole placement: an error in Cz throws off every other site. Along this coronal line the lateral sites step out at the same 10 and 20 percent intervals, placing T3 (now also called T7) and T4 (T8) over the temporal regions and C3 and C4 over the sensorimotor strip, with Cz between them.
The circumferential line. A horizontal circle around the head through Fpz, the temporal sites, and Oz carries the remaining lateral electrodes (Fp1/Fp2, F7/F8, T3/T4, T5/T6, O1/O2) at their 10 and 20 percent positions around the circumference. The intermediate frontal and parietal sites (F3/F4, P3/P4) are then placed to fall evenly between the midline and the lateral chains.
Throughout, symmetry is verified: each left-right pair must sit equidistant from the midline, because an asymmetric placement manufactures a laterality finding that is pure measurement error, not brain. The required accuracy for clinical QEEG is placement within about 5 millimeters of the correct position. Soft caps and variable anatomy make perfect placement impossible, but the discipline of measuring carefully, anchoring on a correct Cz, and checking symmetry keeps you inside the tolerance the databases assume.
The 10-20 labels are not arbitrary. They encode location, and reading them fluently is exam-essential.
A nomenclature note the exam may probe: the older labels T3, T4, T5, T6 were partially revised in the modified combinatorial system to T7, T8, P7, P8 to make the spacing logic consistent when higher-density arrays are added. You will see both conventions in the literature and in software, and you should recognize that T3 and T7 name the same site, as do T4/T8 and (with the relabeling) the posterior temporal positions. For routine clinical work either label set is understood. Just be consistent within a recording.
Clinical QEEG is built on a 19-electrode array at the standard 10-20 positions, plus reference(s) and a ground. The 19 sites give whole-head coverage adequate for valid spectral analysis, topographic mapping, and comparison against the normative databases, which are themselves overwhelmingly built on 19 channels. Laid out by region, the array is:
Each site reports activity from several square centimeters of underlying cortex, not a pinpoint, because volume conduction spreads the signal as it passes through brain, cerebrospinal fluid, skull, and scalp. This spatial blurring is why surface EEG localizes coarsely and why source-analysis methods like LORETA exist to sharpen it (Chapter 12). The functional shorthand for the sites tracks the cortex beneath: frontal-pole and frontal sites over executive, attentional, and motor-preparatory cortex; central sites over the motor strip; temporal sites over auditory, language, and memory regions; parietal over sensory integration; occipital over visual cortex. Chapter 5 carried the full functional neuroanatomy; here the point is that the 10-20 label tells you, at practitioner resolution, what you are sitting over.
Not every job needs 19 channels. Much single-channel and two-channel neurofeedback runs on a handful of electrodes placed for a specific protocol (the next section), and reduced clinical sets of a few sites are common for training. The opposite extreme, high-density arrays of 32, 64, 128, or 256 electrodes, buys finer spatial resolution and better source localization and is used in epilepsy presurgical work and cognitive-neuroscience research, but it costs far longer setup, a heavier artifact burden, more client discomfort, and a mismatch with the 19-channel normative databases. For routine clinical QEEG the 19-channel array remains the standard precisely because it balances coverage against setup time and database compatibility. More electrodes add complexity without proportional clinical value for the everyday assessment.
Most clinical neurofeedback, especially early in a practitioner's development, is single-channel training: one active site, a reference, and a ground, chosen because the presentation and the protocol point to a particular piece of cortex. Knowing the standard sites and what they are for is both exam content and daily practice.
The pattern to carry from this list is that site selection follows function and protocol: you place where the cortex you intend to train lives, you default to Cz or the sensorimotor sites when the picture is mixed or uncertain, and you remember that some sites (the temporal ones especially) carry artifact risks that shape how cleanly you can train them.
Here is the conceptual core of the chapter, and the part most likely to trip a candidate who has only ever clicked a default in software. Every EEG measurement is differential: you never record the absolute voltage at a site, only the difference between that site and something else. That something else is the reference, and the arrangement of active sites and references across a recording is the montage. The same brain activity looks different under different montages because the reference is part of the measurement, not a neutral backdrop. Re-referencing a recording can make a finding appear or vanish, which is exactly why the field requires reviewing data across multiple montages before trusting any pattern.
The linked-ears reference takes the two earlobes (or mastoids), A1 on the left and A2 on the right, and uses their average as the reference for every channel. It has been the traditional default for clinical QEEG and is the reference most normative databases were built on, which is its great practical virtue: recording in linked ears keeps you compatible with the databases you will compare against. The ears are reasonably electrically neutral because they sit far from the cortical generators, so they add relatively little brain signal of their own to the reference.
The cost is twofold. First, physically linking the two ears creates an electrical bridge between the hemispheres, which can smear genuine lateralized activity and, in some configurations, manufacture artificial symmetry or asymmetry. Second, the ears are not perfectly neutral: strongly lateralized cortical activity does spread to the nearer ear, biasing the reference, and a contracting temporalis or jaw muscle near an ear can contaminate the reference and, through it, every channel at once. Linked ears remains a sound default for database-compatible recording, but you read it knowing it links the hemispheres and that anything contaminating an ear contaminates the whole record.
The average reference computes the mean of all the electrodes and uses that average as the reference for each channel. Its appeal is that it commits to no single physical site as "neutral," sidestepping the question of where a truly quiet reference could even be. With adequate whole-head coverage, the average approximates a reference free of any one electrode's local bias and can aid source identification.
Its characteristic weaknesses follow from the averaging itself. Because every channel feeds the average, strong global activity gets subtracted from all channels, attenuating widespread rhythms. More dangerously, a single bad electrode contaminates the average and therefore every channel: jaw-clench artifact at one temporal site, fed into the average reference, smears across the whole montage as a spurious diffuse increase. The average reference is also coverage-dependent, working best with the full array and degrading when sites are missing or unevenly distributed. Use it for the perspective it gives and for source-oriented work, but exclude or repair bad channels first, because the average inherits their problems.
A bipolar montage abandons a common reference entirely. Instead, each channel is the difference between two neighboring electrodes, and the electrodes are arranged in chains. A sequential (longitudinal) bipolar montage runs front-to-back chains (for example Fp1-F3, F3-C3, C3-P3, P3-O1 down the left side), the arrangement old EEG hands call the "double banana." A transverse bipolar montage runs the chains left-to-right across the head. Because each derivation measures the gradient between two adjacent sites, bipolar montages emphasize local differences and suppress activity that is shared across the whole head, which is the opposite bias from a referential montage.
What this buys you is localization. A focal generator, whether a real epileptiform spike or a focal artifact, produces a phase reversal in a bipolar chain: the electrode directly over the source reads as the maximum, and the derivations on either side of it show opposite-polarity deflections, so the polarity "reverses" at the source site. Tracking where the phase reversal sits pinpoints the source, which is why bipolar montages are the standard clinical-EEG tool for detecting focal slowing, asymmetries, and epileptiform activity, and why they are central to localizing artifact (Chapter 8). What bipolar montages suppress is large-scale shared activity and global connectivity information, and they are harder to read than referential montages because every channel is a difference between two moving sites rather than a site against a fixed reference. They are used more in clinical EEG review than in routine QEEG spectral analysis, but a QEEG practitioner inspects in bipolar montages precisely because they reveal focal features and phase reversals invisible to a referential view.
The surface Laplacian, also called current source density, references each electrode to a weighted average of its immediate neighbors. Mathematically it is a spatial high-pass filter: it emphasizes activity that is local and unique to a site and removes activity shared with the surrounding electrodes. The benefit is sharpened spatial resolution and a strong reduction of the volume-conduction blur that smears the surface EEG, so superficial, focal cortical generators stand out clearly and a site's signal is freed from contributions spreading in from elsewhere. The cost is the mirror image: because the Laplacian removes the broadly shared component, it discards large-scale and long-range connectivity information, so it is the wrong montage for coherence work between distant sites even as it is an excellent one for pinning down local activity and for cross-checking whether a putative focal finding survives spatial sharpening.
The honest summary is that no single montage is correct, because each is a different question put to the same data. Linked ears keeps you database-compatible and shows global power patterns. Average reference gives a coverage-wide perspective free of one electrode's bias, at the price of inheriting any bad channel. Bipolar montages localize focal features and reveal phase reversals while suppressing shared activity. The Laplacian sharpens local sources and deblurs the surface while discarding connectivity. The practical doctrine, codified in the 2025 standard, is that a qualified clinician must visually inspect the raw record across multiple montages (linked ears, longitudinal bipolar, transverse bipolar, average reference, and Laplacian) before any quantitative analysis (Collura et al., 2025). The reason is exactly the chapter's theme: a pattern that appears in one reference and vanishes in another is reference-dependent and suspect, while a pattern that persists across montages (though it looks different in each) is more likely to be real brain activity. You record in one reference, commonly linked ears for database compatibility or Cz for simplicity, then re-reference mathematically after the fact to check that your findings are not artifacts of the reference you happened to choose.
One consequence deserves singling out because it bites practitioners reading QEEG reports. Connectivity and coherence measures, which quantify how synchronized two sites are, are acutely sensitive to the reference and montage. A linked-ears reference that bridges the hemispheres can inflate apparent interhemispheric coherence; an average reference that subtracts a shared component can alter coherence in either direction; the Laplacian, by stripping out the shared activity, suppresses long-range coherence almost by construction. The same pair of electrodes can therefore yield meaningfully different coherence values under different montages, none of them "wrong" but none of them interchangeable. When you read a coherence finding in a report, or compute one yourself, you must know which reference produced it, because the number is partly a statement about the montage and only partly a statement about the brain. This is the single most important reason a candidate cannot treat montage as a cosmetic default.
When you measure a head, the sequence is fixed: palpate the four landmarks (nasion, inion, both preauricular points), measure the nasion-inion midline and the ear-to-ear line, mark their crossing as Cz at the 50 percent midpoint of each because every other site depends on it, step out the 10 and 20 percent intervals, and verify left-right symmetry so you do not invent a laterality finding. Place to within about 5 millimeters. Choose the active site from the protocol and presentation: Cz as the default, C3 or C4 for SMR, Fz or F3 to move attention work frontally, Oz for alpha-theta, the temporal sites with awareness of their muscle-artifact risk. Record in a database-compatible reference, then re-reference across montages to confirm any finding survives the change.
For the BCN exam, fix the system and the montage logic. The 10-20 system places electrodes at 10 and 20 percent intervals between nasion, inion, and the preauricular points so that sites scale with head size and stay database-compatible. Cz is the 50 percent vertex and the linchpin. Naming encodes location: the letter names the region (Fp, F, C, T, P, O), odd numbers are left, even are right, z is midline. The standard clinical array is 19 channels plus reference and ground. On montages, know that every measurement is differential and the reference shapes it: linked ears is the traditional, database-compatible default that bridges the hemispheres and is vulnerable to ear contamination; average reference gives a whole-head perspective but spreads any bad channel everywhere; bipolar montages take differences between neighbors, localize focal activity, and show phase reversals at a source while suppressing shared activity; the Laplacian sharpens local sources and deblurs the surface while discarding long-range connectivity. Know that you inspect across multiple montages before quantifying, that a finding that disappears under re-referencing is suspect, and that coherence values depend on the montage they were computed in. Place the electrode correctly and reference it knowingly, and the data you hand to the next chapter is data worth cleaning.