EEG Fundamentals

Electroencephalography (EEG) is a technique for recording the brain’s electrical activity through sensors placed on the scalp. This page introduces the biological origin, signal characteristics, frequency structure, and practical aspects of EEG recordings used in the Neuroengineering & Brain–Computer Interfaces Research Infrastructure of the University of Patras.

The Biological Basis of EEG

Neurons and Electric Potentials

The brain consists of roughly 85 billion neurons, which communicate using tiny electrical impulses. When neurons fire together, they generate postsynaptic potentials—small voltage changes across cell membranes.

Most of the measurable EEG signal comes from pyramidal neurons in the cerebral cortex. These cells are elongated and aligned perpendicular to the cortical surface, so when many fire in synchrony, their small electrical fields sum and propagate through brain tissue, skull, and scalp.

Why Only Cortical Activity Is Seen?

Deep structures (brainstem, limbic system, cerebellum) also generate electric fields, but their neurons are not consistently aligned, and their signals tend to cancel each other out before reaching the scalp. Therefore, scalp EEG primarily reflects the synchronized activity of cortical pyramidal cells.

What EEG Measures

Each EEG electrode measures a voltage difference between two points: a recording electrode and a reference electrode. Because electrical potential is relative, there is no “absolute” voltage in the brain; only differences in potential across sites.

Typical Amplitudes and Units

Signal amplitude: 10–100 μV (microvolts)
Frequency content: 0.1–100 Hz, depending on filters and sampling rate
Sampling rate: typically 250–2000 Hz
Resolution: depends on amplifier bit depth (commonly 16–24 bits)

Temporal Resolution

EEG’s greatest advantage is temporal precision. Because it records voltage fluctuations directly from neural activity, it captures changes occurring within milliseconds.

EEG tells us “when” things happen in the brain, rather than exactly “where.”

The Nature of the EEG Signal

Continuous Oscillations

The EEG signal is a continuous, oscillatory time series of voltage values sampled across electrodes. It reflects both:

Endogenous oscillations (spontaneous brain rhythms), and
Event-related activity (responses to specific stimuli or tasks).

When large groups of neurons oscillate together, we observe rhythmic patterns called brain waves.

Frequency Bands

Band Frequency Range (Hz)	Associated Functions	Typical Location
Delta (1–4)	Deep sleep, unconscious processes	Frontal during deep sleep
Theta (4–8)	Working memory, navigation, sustained attention	Frontal, midline
Alpha (8–12)	Relaxation, reduced visual attention, eyes closed	Occipital, parietal
Beta (13–25)	Active thinking, motor preparation, concentration	Central, frontal
Gamma (>25)	Sensory binding, high-level integration (still debated)	Distributed

These bands overlap and vary by individual.
EEG analysis often examines power (signal energy) and phase (synchrony) within these frequency bands to infer mental states.

🎶 Imagine an orchestra: each brain rhythm plays its part, sometimes dominating, sometimes blending with others. EEG analysis in frequency bands separates those instruments to understand the composition.

While frequency analyses focus on rhythmic activity, ERPs capture the brain’s time-locked responses to specific events.

When a stimulus (image, tone, word) is presented repeatedly, each occurrence triggers a small neural response. Averaging many trials cancels random background activity and highlights consistent features, the event-related potential.

Common ERP Components

Component	Typical Latency	Interpretation
P1/N1	100 ms	Early sensory processing
P2/N2	200–300 ms	Attention and perception
P3 (P300)	~300 ms	Cognitive evaluation, decision-making
N400	400 ms	Semantic processing
ERN	50–100 ms post-error	Error detection

Signal Quality and Clean Data Principles

Electrode–Skin Contact

High-quality EEG requires low impedance at each electrode (< 10–15 kΩ). Impedance depends on skin preparation, electrode material, and conductive gel quality.

Sources of Noise

Artifact	Description	Mitigation
Muscle activity (EMG)	Jaw, forehead, neck tension	Ask participants to relax, avoid speech
Eye movements (EOG)	Blinks and saccades	Instruct to fixate; record EOG for correction
Line noise	50 Hz (EU) electrical interference	Shield cables, use notch filters
Electrode motion	Cap or cable movement	Secure setup; avoid head motion
Heart and sweat artifacts	Slow drifts, conductivity changes	Maintain stable temperature and dryness

⚠️ “Garbage in, garbage out” applies: no software can fully fix poor acquisition.

The 10–20 System and Sensor Placement

EEG sensors are typically arranged using the International 10–20 system, which ensures standardized electrode locations relative to anatomical landmarks:

Nasion (Nz): top of nose bridge
Inion (Iz): bump at back of skull
Preauricular points: in front of each ear

Electrodes are spaced at 10 % or 20 % intervals along these lines.
Positions are labeled by:

Lobe letter: F (frontal), C (central), P (parietal), O (occipital), T (temporal), Fp (frontopolar)
Number: odd = left, even = right, z = midline (e.g., F3 = left frontal, Cz = midline central)

Amplification, Digitization, and Sampling

EEG amplifiers boost tiny microvolt signals before digitization.

Sampling rate: the number of samples per second (Hz)
250 Hz → 4 ms resolution
500 Hz → 2 ms resolution
Nyquist criterion: the sampling rate must be at least twice the highest frequency of interest.
For example, if analyzing up to 40 Hz, use ≥ 100 Hz; most systems use 256–512 Hz as a minimum.
Analog-to-digital conversion: higher bit depth (≥ 16-bit) ensures fine voltage precision.

From Raw Data to Insight

Preprocessing

Before analysis, EEG data undergoes cleaning and structuring:

Filtering (0.1–40 Hz typical)
Artifact rejection or correction (blink removal, ICA)
Segmentation (epoching around events)
Baseline correction and averaging

Analysis Approaches

Time-domain: temporal response to stimuli, ERPs, statistical analysis in the time domain
Frequency-domain (FFT, wavelet): rhythmic power changes
Connectivity: coherence or phase-locking between electrodes
Spatial mapping: scalp topographies, source localization (with enough channels)

Typical Software

MNE-Python: open-source Python toolbox used in our infrastructure
EEGLAB: MATLAB-based classic environment
Brainstorm: GUI-based multimodal platform

Key Takeaways

EEG records rapid cortical dynamics through scalp electrodes.
It offers millisecond-level temporal resolution, ideal for studying real-time cognition.
The signal reflects synchronized neuronal activity, not thoughts themselves.
Clean data collection and ethical participant handling are essential for valid results.