Linked Mastoids Reference in EEG

Every Measurement Needs a Baseline (and EEG's Baseline Problem Is Weirder Than You Think)

Here's something that might surprise you about EEG. When a researcher says they're "recording brain activity" from a spot on your scalp, that's not quite right. What they're actually recording is the difference in electrical voltage between that spot and some other spot.

This is a fundamental thing about electricity. Voltage is always relative. Always. You can't measure the voltage at a single point any more than you can measure how tall something is without defining where the ground is. When an electrician tests a wall outlet, one probe goes into the hot slot and the other goes into the neutral slot. The reading on the multimeter is the difference between those two points.

EEG works the same way. Every channel on every EEG device on the planet is measuring the voltage difference between an "active" electrode (the one over the brain region you care about) and a "reference" electrode (your baseline). And here's where things get interesting: there is no such thing as a perfect reference point on the human head.

Every spot on your body carries some electrical signal. Your heart is pumping. Your muscles are twitching. Your eyes are moving. Even the bones in your skull have measurable electrical activity flowing across them. The quest for a clean EEG reference is really a quest for the least noisy spot you can find.

For nearly a century, that quest has kept coming back to the same two lumps of bone behind your ears.

Those Bumps Behind Your Ears Have a Name (and a Very Important Job)

Run your fingers behind your earlobes. Feel those hard, bony bumps? Those are your mastoid processes, projections of the temporal bone that sit just behind and below each ear canal. They're filled with air cells and covered in a thin layer of skin and periosteum. Not much muscle. Not much soft tissue. Just bone.

This matters enormously for EEG. Muscles generate electrical signals called electromyographic (EMG) artifacts that are often ten to one hundred times larger than the brain signals you're trying to record. Put a reference electrode on your jaw, your forehead, or your neck, and you'll spend half your analysis time trying to separate brain activity from muscle noise.

The mastoids aren't perfect. They still pick up some cardiac artifact from blood flow in the nearby posterior auricular artery. They can catch some signal from the temporalis muscle if a person is clenching their jaw. But compared to almost anywhere else on the head, they're remarkably quiet.

And there's another advantage that's easy to overlook. The mastoid bones are anatomically consistent. Everyone has them. They're in the same place on every skull. You can find them by touch in about two seconds. In a field where electrode placement precision matters enormously, having a reference location that's both electrically quiet and anatomically obvious is a genuine gift.

One Mastoid or Two? That Question Took Decades to Settle

Early EEG researchers typically used a single mastoid as their reference, usually the one on the right side (called A2 in the international 10-20 electrode naming system, with the left mastoid being A1). This was simple. One reference electrode, one wire, done.

But single-mastoid referencing has a sneaky problem. If your reference is only on the right side, any electrical activity that happens to be stronger on the right side of your head will be partially cancelled out in your recording. Meanwhile, activity on the left side gets slightly amplified relative to the reference.

Think about it this way. Imagine you're trying to measure how tall different buildings are in a city, but your "ground level" baseline tilts slightly to the east. Buildings on the east side of the city would look shorter than they really are. Buildings on the west side would look taller. You'd get a systematically skewed picture of the skyline.

That's exactly what happens with a single-mastoid reference. It introduces hemispheric bias. And for any study looking at brain asymmetry (which includes research on emotion, language, attention, and dozens of other topics), that bias is a serious problem.

The solution seems obvious in retrospect: use both mastoids. Average the electrical signal from A1 and A2, and you get a reference point that doesn't favor either hemisphere. This is linked mastoids referencing, and it's been one of the most widely used reference schemes in EEG research since the 1980s.

The Wire Problem: Why Physically Linking Mastoids Backfired

The original approach to linked mastoids was straightforward. Place an electrode on each mastoid bone. Connect them with a wire. Use the junction point as your reference. The averaging happens through basic electrical physics: connect two voltage sources with a wire, and the junction naturally sits at their midpoint.

This worked, sort of. It did eliminate hemispheric bias. But it introduced a new problem that took years to fully understand.

When you connect two electrodes with a wire, you create an electrical bridge. Current can now flow from one side of the head to the other through that wire. This is called shunting, and it has two effects that are both bad.

First, it reduces the amplitude of any brain signal that's lateralized, meaning stronger on one side than the other. The bridge provides an alternative path for current to flow, which effectively dampens the signal. Studies have shown that physically linked mastoids can reduce the measured amplitude of lateralized activity by 20% to 50%, depending on the impedance of the electrodes and the wire.

Second, the shunting doesn't affect all electrode sites equally. Sites closer to the mastoids (like temporal electrodes) are affected more than sites further away (like frontal midline electrodes). This means that physically linked mastoids don't just reduce signal amplitude. They distort the topographic map of brain activity across the scalp.

For a technique that was supposed to clean up EEG data, physically linked mastoids had an uncomfortable habit of introducing its own set of distortions.

The Digital Fix: Math Instead of Wire

The solution arrived with digital EEG systems. Instead of connecting the two mastoid electrodes with a physical wire during recording, modern systems record each mastoid as its own channel. The reference during the actual recording might be any convenient electrode, often Cz (the vertex of the head) or one of the mastoids.

Then, after the data has been collected and digitized, software computes the average of the A1 and A2 channels and subtracts that average from every other channel. This is digitally linked mastoids, sometimes called "offline re-referencing to linked mastoids."

The result looks almost identical to what you'd get with physically linked mastoids, but without the shunting problem. No wire means no electrical bridge. No bridge means no current flowing between hemispheres. No distortion of the topographic map. No artificial reduction in lateralized signal amplitude.

This is one of those cases where moving a computation from hardware to software eliminated an entire category of artifact. It's a small thing, technically. But it's the kind of small thing that changes the quality of thousands of research studies.

The Crown captures brainwave data at 256Hz across 8 channels. All processing happens on-device. Build with JavaScript or Python SDKs.

Explore the Crown

Why Reference Choice Changes What You See in the Brain

Here's the part that surprises people who are new to EEG: the same brain, recorded at the same time, will produce different-looking data depending on which reference you use. Not slightly different. Visibly different.

This isn't a flaw. It's a mathematical consequence of how voltage measurement works. When you change the reference, you're changing the baseline that every other electrode is compared against. The underlying brain activity hasn't changed, but the way it appears in your data has.

Consider a simple example. Suppose there's a burst of alpha brainwaves activity concentrated over the right parietal cortex. With a left mastoid (A1) reference, this burst will appear as a large signal at right parietal electrodes and a smaller signal at left parietal electrodes. With a right mastoid (A2) reference, the same burst will appear smaller at the right parietal electrodes (because the reference is nearby and picking up some of the same activity) and relatively larger on the left side.

With linked mastoids, you get something in between. The right-side signal is somewhat reduced, the left-side signal is somewhat increased, and the overall topographic map is more balanced.

With average referencing (where the mean of all electrodes serves as the reference), the same alpha burst would look different yet again. And with a nose reference, or an earlobe reference, or a Cz reference, you'd see yet another variation.

This is why EEG papers always specify their reference. It's not a minor detail. It's essential information for interpreting the data.

The Great Reference Debate: Why Smart People Still Disagree

You might think that after nearly a century of EEG, there would be a consensus on the best reference scheme. There isn't. And the debate is surprisingly heated.

Advocates of average referencing argue that with enough electrodes (typically 64 or more), the average of all channels approximates a theoretical "zero potential" reference better than any single physical location can. They're mathematically correct. The average reference converges on zero as electrode density increases. The problem is that with lower channel counts, the average reference is biased toward whatever brain regions happen to have the most electrodes covering them.

Advocates of linked mastoids counter that their reference has clear anatomical definition, works well with any number of channels, and has decades of research behind it. They're practically correct. For the vast majority of EEG setups, especially those with fewer than 64 channels, linked mastoids produces reliable, interpretable data.

Then there are those who argue for the REST (Reference Electrode Standardization Technique), which uses mathematical models of the head to estimate what the data would look like with a truly neutral reference at infinity. It's theoretically elegant but computationally intensive and sensitive to the accuracy of the head model.

The honest truth is that there is no perfect reference. Each scheme makes different tradeoffs between practical convenience, mathematical elegance, and signal fidelity. And for most applications, the differences between well-implemented reference schemes are smaller than the differences between sloppy and careful electrode application.

What This Means for Consumer EEG (and for You)

If you're not a researcher, you might be wondering why any of this matters. Here's why: every EEG device you put on your head makes referencing decisions for you. Those decisions affect the quality of the data you see and the accuracy of any metrics derived from that data.

Consumer EEG devices face a particular challenge. They typically have far fewer channels than research systems (8 to 14 channels versus 64 to 256). This means they can't rely on average referencing, because the math doesn't work well with so few electrodes. They need to use some form of dedicated reference, and the quality of that reference directly impacts signal quality.

The Neurosity Crown, for example, uses 8 EEG channels positioned at CP3, C3, F5, PO3, PO4, F6, C4, and CP4. That's eight active channels spanning frontal, central, centroparietal, and parieto-occipital regions across both hemispheres. The Crown's on-device N3 chipset handles signal processing including referencing, filtering, and artifact rejection before the data ever reaches your application.

For developers working with the Crown's JavaScript or Python SDKs, the data you receive has already been referenced and cleaned. But if you're doing custom analysis and want to re-reference the raw 256Hz data to a different scheme (linked mastoids, average reference, or something else), that raw data is available.

This is actually one of the underappreciated aspects of modern consumer EEG. The device makes smart default referencing choices so casual users get clean data automatically. But it doesn't lock researchers and developers out of the raw signal. You get convenience and flexibility.

The Bigger Picture: Why Baselines Matter Beyond EEG

There's something deeper worth noting about the linked mastoids story. It's really a story about measurement philosophy. Every measurement in science is relative to some baseline, and the choice of baseline shapes what you can see.

Astronomers measure the brightness of stars relative to reference stars. Economists measure GDP growth relative to the previous year. Psychologists measure cognitive performance relative to population norms. In every case, a bad baseline leads to bad conclusions, and the baseline itself is often invisible in the final result.

EEG just makes this problem unusually tangible. You can literally see how different references change the shape of brainwaves on a screen. It's a useful reminder that the numbers produced by any measurement system are not raw truth. They're truth filtered through choices, some obvious, some hidden, about what counts as zero.

The next time you look at an EEG recording, whether it's a research paper or the output of a consumer device on your own head, remember that what you're seeing is not "brain activity." It's brain activity relative to a reference point. And that reference point, those quiet bones behind your ears or a mathematical average or something else entirely, is doing more work than you might think.

Understanding that distinction doesn't just make you a better reader of EEG data. It makes you a better thinker about measurement in general. And in an age where more aspects of human cognition are becoming measurable, that kind of thinking matters more than ever.