explain how EEG works to me, and how difference references in EEG work and how changing references changes the data

Welcome! I'm excited to help you master EEG. Before we dive into the math and mechanics of electrodes and reference schemes, I'd love to know a bit more about your background so I can tailor my analogies perfectly to you.

Are you studying EEG for a specific context, like clinical neurology, neuroscience research, or biomedical engineering, and how familiar are you with basic electrical concepts like voltage?

I'm in research/engineering and understand voltage well.

Because of your engineering background, you already know that voltage doesn't exist in a vacuum at a single point; it's inherently a difference in electrical potential between two places. EEG is no exception. Every single squiggly line you see on an EEG monitor is actually a differential measurement comparing an "active" electrode to a "reference" electrode.

How the Brain Generates the Signal

When we record an EEG from the scalp, we aren't picking up the sharp, rapid action potentials of individual axons. Instead, we are looking at the slower, combined electrical activity of hundreds of thousands of cortical lying right beneath the skull. When neurotransmitters bind to the dendrites of these neurons, they cause local ions to flow across the cell membrane, creating tiny local currents. Because these large cells are aligned parallel to one another and perpendicular to the brain's surface, their individual postsynaptic potentials summate. This creates a coordinated electrical dipole strong enough to project through the meninges, cerebrospinal fluid, and bone, finally reaching our scalp electrodes.

The Differential Amplifier: G1 vs G2

To capture this tiny, microvolt-level signal, the EEG hardware uses a differential amplifier for each channel. By historical convention in neurophysiology, the two input terminals of this amplifier are labeled G1 (Grid 1, the active site) and G2 (Grid 2, the reference site).

V_{measured} = V_{G1} - V_{G2}

In practice, if a neural discharge makes the active electrode (G1) more negative relative to the reference electrode (G2), the amplifier output deflects upward on the screen. If the reverse is true, the wave deflects downward. Without a reference electrode (G2) to subtract from G1, we would have no baseline to measure against, and our amplifiers would be overwhelmed by environmental electromagnetic noise. By subtracting G2 from G1, any noise common to both sites is canceled out.

Continue to referential montages and how we choose a reference site

Now that we have established that voltage is always G1 minus G2, let's explore our options for where to place that G2 reference electrode. In a referential montage, we choose one single, fixed location on the body as our common G2, and we compare every single active scalp electrode (G1) to that same point.

The Referential Montage

In a referential setup, you have many active channels, but they are all looking at the exact same baseline. Because of this, referential montages are fantastic for showing the true, absolute amplitude and morphology of a signal at each individual electrode site.

referential

adjective

An EEG recording configuration where multiple active scalp electrodes are compared to a single, common reference electrode.

To make this setup work, we need a reference site that is as electrically quiet as possible. If the reference site is picking up brain activity of its own, it will contaminate every single channel. Because of this, clinicians typically choose a or relatively inactive point on the head. Let's look at the most common reference locations.

Reference Site	Description	Key Advantage	Key Disadvantage
A1 & A2 (Earlobes/Mastoids)	Electrodes placed on the bony process behind the ears or on the lobes.	Close to the head but physically away from major cortical generators.	Can still pick up temporal lobe activity or muscle artifact from jaw clenching.
Cz (Vertex)	The very top-center electrode of the scalp.	Located at the high midpoint; highly standard.	Located right on top of the brain; highly active during sleep, which contaminates sleep recordings.
Linked Ears/Mastoids	Physically linking A1 and A2 with a wire or mathematically averaging them offline.	Balances out lateralized hemispheric bias.	Can act as a low-resistance path that distorts the natural scalp potential fields.

If your reference electrode starts recording a big, active brain signal, that signal gets subtracted from all your active channels. This makes it look like there is a sudden, synchronous electrical change happening everywhere across the brain, when in reality, it was just the reference electrode fluctuating. This is known as reference contamination. To solve this, or to get highly localized details, we can ditch the single common reference point entirely and use either a bipolar montage or a calculated average. Would you like to see how those physical setups compare to this referential one?