The map on your head

Open any paper in this field and the methods section fills with names like Fp1, C3, T7. They look like spreadsheet cells. They are in fact a street address system for the scalp, agreed in 1958 and still the universal map: the ten-twenty system. Understanding it takes ten minutes and converts every electrode list you will ever read from noise into geography.

Letters are regions, numbers are streets

The letter says which lobe the electrode sits over: Fp for frontopolar, F for frontal, C for central, T for temporal, P for parietal, O for occipital. The number says how far from the midline, with odd numbers on the left, even on the right, and the letter z marking the centerline. C3 sits over the left central region, C4 its right mirror, Cz the crown. The numbers themselves come from anatomy: positions are placed at 10 and 20 percent of the distance between standard landmarks, the nasion between the eyes, the inion at the back, and the preauricular points in front of the ears. That is the name: ten-twenty. High-density caps subdivide further into the ten-ten and ten-five systems, but the addresses stay the same scheme.

C3 is not a channel. It is a place, and every recording ever taken there is a measurement of that place.

The reference problem

Here is the detail that surprises everyone. An EEG channel never measures one spot. Voltage has no absolute value; it is always a difference between two points. Every channel is really 'here minus the reference', and the choice of reference changes what you see. Mastoids behind the ears, the nose, the crown, or the average of all channels are all common choices, and the same recording re-referenced can shift amplitudes and topographies noticeably. This is why papers state their reference, and why comparing raw traces across labs without checking it is a category error. Clinical work additionally uses bipolar chains, each channel measured against its neighbor, which localizes artifacts sharply at the cost of blurring broad activity.

Where speech lives

The speech network is a perisylvian neighborhood, meaning the cortex around the Sylvian fissure that divides temporal from frontal lobe. Planning and production lean on inferior frontal areas, with F7 and F3 the nearest addresses on the left for most right-handed speakers. The mouth, lips and tongue occupy a generous stretch of the motor strip, whose nearest electrodes are the C neighborhood, C3 through C4 by way of Cz. Comprehension lives in posterior temporal cortex, T neighborhood. None of these electrodes reads 'words'; each hears the murmur of the cortex beneath it, and the model's job is to combine those murmurs into language. More electrodes do not add resolution the way more pixels do: the skull smears each square centimeter of cortex across a wide patch of scalp, and recovering sources from the surface is an ill-posed problem, one with many possible answers. Density helps; it does not see through.

For our study design this note is the map we will choose addresses from: enough coverage of the C and left frontal temporal neighborhoods to hear the speech network, a reference choice made once and stated everywhere, and no illusion that a number of channels alone decides anything.