An EEG electrode does not detect a single neuron firing. What reaches the scalp is the summed activity of millions of cortical pyramidal neurons whose postsynaptic currents line up in time and in orientation. When enough of them act together, the field they produce passes through the brain, the cerebrospinal fluid, the skull and the scalp, and shows up as a voltage of a few to roughly a hundred microvolts.
Hans Berger recorded the first human EEG in 1924, and the principle has not changed since then. What has changed is everything around it: amplifiers, electrode counts, and above all the analysis. Modern EEG work is mostly signal processing and statistics applied to a very small, very noisy signal.
What EEG is good at, and what it is not
| Property | EEG | Why it matters |
|---|---|---|
| Temporal resolution | Milliseconds | Fast enough to follow evoked responses, oscillations and their timing within a trial. |
| Spatial resolution | Centimetres, blurred | Volume conduction smears each source across many electrodes, so one channel is not one brain region. |
| Depth | Mostly cortex | Deep structures contribute little that reaches the scalp. |
| Cost and portability | Low cost, wearable | This is why EEG dominates non-invasive BCI research. |
| Signal size | Microvolts | Eye blinks, muscle tension and line noise can be orders of magnitude larger than brain signal. |
Electrodes and the 10-20 system
Electrode positions follow the international 10-20 system and its higher-density extensions. Each name is a region letter and a number: Fp for frontopolar, F frontal, C central, P parietal, O occipital and T temporal. Odd numbers are on the left, even numbers on the right, and z marks the midline. So C3 sits over the left motor cortex, C4 over the right, and Oz over the visual cortex at the back.
The names matter for analysis because they carry anatomy. A motor imagery study looks at C3 and C4, and an SSVEP study looks at the occipital electrodes. BCILattice reads 10-20 names from your montage to build region features over frontal, central, parietal, occipital and temporal areas.
Every EEG channel is a difference
An EEG voltage is always measured against something: a reference electrode, linked mastoids, or the average of all channels. Changing the reference changes every waveform, which is why re-referencing is a standard preprocessing step and why the choice belongs in the methods section.
Artifacts: the signal you did not want
How to deal with each one is covered in EEG artifact removal.
Eye blinks and movements
Large, slow deflections strongest over frontal electrodes. ICA is the standard way to separate and remove them.
Muscle (EMG)
Broadband high-frequency activity from jaw, neck and face muscles, often over temporal sites.
Line noise
50 or 60 Hz interference from mains power, removed with a notch filter.
Bad electrodes
Flat, railing or noisy channels from poor contact, which should be detected and dropped before decomposition.
From raw EEG to a result
A typical analysis filters the recording, removes artifacts, cuts it into trials around events (epoching), and then measures something per trial: an evoked response, power in a frequency band, connectivity between sites, or features for a classifier. The order and settings of those steps change the result, which is why they need to be recorded.
The step-by-step version is on EEG preprocessing. The software that runs it end to end, offline on your own machine, is BCILattice.
Questions
What does an EEG measure?
EEG measures voltage differences on the scalp produced by the synchronised activity of large populations of cortical neurons, mainly their postsynaptic currents. It does not record individual neurons.
How does EEG work in simple terms?
Electrodes on the scalp pick up tiny electrical signals that millions of brain cells produce when they are active together. An amplifier boosts those signals and a computer records them as waves over time.
Why does EEG have poor spatial resolution?
Electrical fields spread as they pass through the brain, skull and scalp (volume conduction), so each source shows up on many electrodes and each electrode mixes many sources.
Is EEG safe?
Yes. Scalp EEG only records; it sends no current into the head. It is used routinely in clinics, including with newborns.
How many EEG channels do I need?
It depends on the question. Some BCI paradigms work with a handful of well-placed electrodes, while source analysis and connectivity work benefit from 64 or more. More channels also means more preparation time and more to clean.
What is the difference between EEG and fNIRS?
EEG measures electrical activity directly with millisecond timing. fNIRS measures changes in blood oxygenation with near-infrared light, which is slower, lagging the neural activity by seconds, but less disturbed by movement.
Read the reference
Related
EEG Frequency Bands: Delta, Theta, Alpha, Beta, Gamma
EEG is described in frequency bands, from slow delta to fast gamma. The bands are a useful shorthand for a continuous spectrum, and the popular "one wave, one mental state" story oversimplifies them.
EEG Preprocessing Pipeline: Steps in Order
Preprocessing turns a raw recording into clean, labelled trials. There is a conventional order, a few places where it genuinely matters, and many where labs reasonably disagree.
What Is fNIRS? Functional Near-Infrared Spectroscopy
Functional near-infrared spectroscopy (fNIRS) shines near-infrared light into the head and measures how much comes back, which tracks changes in oxygenated and deoxygenated haemoglobin in the cortex.
What Is a Brain-Computer Interface (BCI)?
A brain-computer interface (BCI) is a system that measures brain activity, decodes it, and turns it into a command for a computer or device, without using the muscles or nerves that normally carry that command.