The EEG Signal

What You Are Looking At

The picture is a slice through the head. Under the pink scalp and the thick gray skull sits a patch of cortex drawn as a row of pyramidal neurons: a triangular cell body deep in the tissue and a long apical dendrite reaching toward the surface. The EEG is not made of action potentials. Spikes are brief, biphasic, and not synchronized within microseconds across neurons, so their currents cancel before they reach the scalp. What reaches the scalp is the postsynaptic potential. When an excitatory synapse on the apical dendrites opens, positive ions flow into the cell there (a current sink), travel down the dendrite, and leave again near the soma (a current source). The current returns through the extracellular fluid to close the loop. Seen from outside, the neuron is briefly a tiny battery with a negative end near the surface and a positive end deeper down, a current dipole (Luck, 2014, Chapter 2).

One dipole is far too weak to register. The meter on each electrode only moves when thousands of neurons receive the same input at the same moment and point the same way. The alignment slider shows what synchrony alone cannot do. When the dipoles point in random directions their fields cancel (a closed field) and the meter barely moves no matter how many cells are active. Cortex is a layered sheet in which pyramidal cells line up in parallel (an open field), so it dominates the EEG, while deep nuclei with jumbled orientations are nearly silent. The shading shows volume conduction: the potential spreads instantly through the brain, is smeared sideways by the poorly conducting skull, and arrives at the scalp as a broad blur. Any electrode reads a weighted sum of every dipole active anywhere in the head. This is the superposition problem, and it is the main reason localization from the scalp is hard.

Try This

1. Keep the gyrus geometry and slide the number of neurons from 1 to 100,000. Below a few thousand cells the middle meter stays near zero; near a hundred thousand it reads several microvolts, the size of a real ERP component.
2. Pull alignment down to 20%. The neurons rotate into a jumble and the reading collapses even though just as many cells are active.
3. Switch the input from excitatory at the apical dendrites to inhibitory at the same place. The arrows reverse and the scalp goes from negative to positive. Now move the excitatory input to the soma and basal dendrites. The sign flips again, for a different reason, and the box under the readout explains which.
4. Switch to the sulcal wall. The dipole lies flat, the electrode right above reads almost nothing, and the two electrodes to either side show opposite signs.
5. Watch the ticker in Panel B. Each burst of synchronized input appears as a deflection in the ongoing EEG, whose background rhythms (alpha around 10 Hz, theta, beta) come from other populations doing the same thing.

Why It Matters for the Pipeline

Polarity at the scalp depends on four things at once: whether the input is excitatory or inhibitory, where on the cell it lands, how the cortical sheet is oriented under the electrode, and which reference site the voltage is measured against. A negative wave at Cz therefore says very little on its own about what the neurons were doing. The same physics sets the plan for the rest of the pipeline. Because tens of thousands of cells must act together to be seen at all, EEG is a signal of populations. Because it is only a few microvolts, it must be averaged over many trials to stand out from the rest of the brain's activity. When you load a dataset with EEG = pop_loadset('sub-001_N400.set') and scroll through the channels with pop_eegplot(EEG, 1, 1, 1) (Delorme & Makeig, 2004), the 20 to 100 µV traces you see are the summed field of countless dipoles like these, blurred by the skull and referenced to one site. Buzsáki et al. (2012) give the full account of how synaptic currents, cell geometry, and tissue conductivity combine into the recorded field.

References