Brainwave Oscillations: Coupled Clocks Behind the EEG

A brain as a crowd of clocks

An EEG electrode does not read individual neurons -- it reads the summed electrical field of tens of thousands of pyramidal neurons firing near the scalp. What makes that sum look like a clean rhythmic wave instead of noise is synchrony: when many neurons fire in step, their tiny fields add constructively; when they fire independently, the fields mostly cancel. The simplest model that captures this is the Kuramoto model, originally built for chemical oscillators and now a standard toy model for neural populations. Each neural population is an oscillator with its own phase θ_i and natural frequency ω_i, and every oscillator nudges every other one toward its own phase:

d(theta_i)/dt = omega_i + (K/N) * sum_j sin(theta_j - theta_i)

theta_i   phase of oscillator i
omega_i   its natural (uncoupled) frequency
K         coupling strength
N         number of oscillators

Each term sin(theta_j - theta_i) pulls oscillator i toward oscillator j's phase, and averaging over all N oscillators keeps the pull independent of population size. Turn K up and the population locks into a shared rhythm; turn it down and each oscillator drifts at its own natural frequency, and the population looks like noise.

The order parameter: measuring synchrony in one number

Kuramoto's trick for quantifying synchrony is to treat every oscillator's phase as a point on the unit circle and average their positions as complex numbers:

r * e^(i*psi) = (1/N) * sum_j e^(i*theta_j)

r     synchrony ('order parameter'), 0 (incoherent) to 1 (perfectly in phase)
psi   mean phase of the population

If every phase is scattered uniformly around the circle, the vectors cancel and r ≈ 0. If every oscillator shares the same phase, the vectors stack and r = 1. As K rises past a critical value K_c (set by the spread of natural frequencies), the population undergoes a phase transition: r jumps from near zero to a growing positive value, and a coherent rhythm emerges out of disorder with no external conductor. This bifurcation is the mathematical skeleton behind why EEG traces sometimes show a strong, clean rhythm and sometimes look like flat noise -- it is literally the same order-parameter transition, driven by how strongly nearby cortical populations are coupled at that moment.

The five EEG bands

Clinical and research EEG carves the 0.5-100 Hz spectrum into named bands, each associated with a different brain state, though the boundaries are conventions rather than sharp physical cutoffs:

delta  0.5-4  Hz   deep dreamless sleep, slow-wave sleep
theta  4-8   Hz   drowsiness, meditation, memory encoding
alpha  8-13  Hz   relaxed wakefulness, eyes closed, idling visual cortex
beta   13-30 Hz   active thinking, focus, motor planning
gamma  30-100 Hz  binding of features into one percept, high-level integration

Alpha is the most famous: Hans Berger first recorded it in 1924, strongest over the occipital cortex when the eyes are closed and relaxed, and it attenuates the instant the eyes open or attention sharpens -- a phenomenon called alpha blocking. Higher frequencies generally correlate with smaller, more local populations synchronising (gamma binding is thought to link spatially separate feature detectors into one coherent perception), while lower frequencies correlate with large-scale, whole-cortex or thalamocortical synchrony, which is exactly why slow delta waves dominate during deep sleep when huge swaths of cortex fire together.

Noise, coupling and why real EEG is never a pure sine wave

Two knobs shape the simulated trace: coupling K sets how strongly oscillators pull each other into phase, and independent noise on each oscillator's phase pulls the other way, constantly trying to desynchronise the population. Real cortex sits in a similar tug-of-war -- neuromodulators, arousal state and task demands continuously retune the effective coupling between populations, which is why a real EEG trace is a shifting mixture of bands rather than one clean tone, and why the order parameter r fluctuates from moment to moment instead of settling at a fixed value.

Why this model, and where it breaks down

The Kuramoto model is deliberately a caricature: real cortical columns are not point oscillators with a single phase, coupling is not all-to-all with uniform strength, and axonal conduction delays matter for high-frequency synchrony over any real distance. Its value is that it is the simplest model that reproduces the actual phenomenon of interest -- a sharp synchronisation transition controlled by one coupling parameter -- with equations you can integrate by hand. More detailed neural-mass and biophysical models (Wilson-Cowan, Jansen-Rit, Hodgkin-Huxley networks) add realism at the cost of losing this clean analytic picture of how synchrony itself switches on.

Frequently asked questions

Try it live

Everything above runs in your browser — open Brainwave Oscillations and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.