JonoThora: Phase N (01b): LaTeX restoration — promote Unicode display-math to $; lint-clean per tools/wiki_latex_lint.py$

2026-05-11T20:08:30Z

Phase N (01b): LaTeX restoration — promote Unicode display-math to <math>; lint-clean per tools/wiki_latex_lint.py

New page

= Wilson-Cowan Model =

{{Audience_Sidebar
| difficulty = Intermediate
| reading_time = 8 minutes
| prerequisites = ODEs; basic dynamical-systems (fixed points, Hopf bifurcation); some neuroscience.
| if_too_advanced_see = [[Could_the_Brain_Use_Quantum_Mechanics]]
| if_too_basic_see = [[Wilson-Cowan_Coupled_to_Psi]]
| if_you_want_the_math_see = This page; [[Neural_Field_Equations]]
}}

{{Notation
| psi_convention = ψ = scalar field amplitude (not used in baseline Wilson-Cowan).
| signature = Mostly-plus (irrelevant for this non-relativistic model).
| units = SI; time in seconds; population activities dimensionless ∈ [0,1].
}}

The '''Wilson-Cowan model''' (1972) is the foundational mean-field model of neuronal population dynamics. It describes the interaction of an '''excitatory''' (E) and an '''inhibitory''' (I) population of neurons through two coupled ordinary differential equations.

It is the simplest neural-population model that exhibits oscillations, multistability, and pattern formation — and it serves as the conceptual core of nearly all later population-level models, including [[Amari_Neural_Field|Amari]] fields, [[Jansen-Rit_Neural_Mass|Jansen-Rit]] neural masses, and the framework's [[Wilson-Cowan_Coupled_to_Psi|ψ-coupled extension]].

== Statement ==

The Wilson-Cowan equations for two interacting populations E(t) and I(t):

:<math>\tau_E\,\frac{dE}{dt} = -E + (1 - r_E\,E)\,S_E\!\bigl[\,w_{EE}\,E - w_{EI}\,I + P(t)\,\bigr]</math>

:<math>\tau_I\,\frac{dI}{dt} = -I + (1 - r_I\,I)\,S_I\!\bigl[\,w_{IE}\,E - w_{II}\,I + Q(t)\,\bigr]</math>

== Components ==

{| class="wikitable"
! Symbol !! Meaning !! Units / range
|-
| E(t) || Excitatory population activity || dimensionless ∈ [0,1]
|-
| I(t) || Inhibitory population activity || dimensionless ∈ [0,1]
|-
| τE, τI || Population time constants || ~ 5–20 ms
|-
| rE, rI || Absolute-refractory periods || ~ 1 ms
|-
| wEE, wEI, wIE, wII || Synaptic-weight matrix || dimensionless
|-
| SE, SI || Sigmoid activation, e.g. S(u) = 1/(1+exp(−(u−θ)/σ)) || dimensionless
|-
| P(t), Q(t) || External drives (sensory / top-down) || dimensionless
|}

== Derivation sketch ==

Start with a population of neurons whose firing depends on aggregated synaptic input. Coarse-grain over individual neurons:

# Each neuron fires when its summed synaptic input exceeds a threshold; the population fraction firing is a sigmoid function S of mean input.
# The mean input to an excitatory neuron is (positive contributions from other E neurons) minus (inhibition from I neurons) plus external drive.
# After firing, a refractory period (1 − r·E) reduces the fraction of cells available to fire.
# The membrane dynamics imposes a time constant τ.

Combining these gives the Wilson-Cowan form.

== Dynamics ==

Linearise about a fixed point (E*, I*); the Jacobian determines:

* '''Stable fixed point''' (both eigenvalues with negative real part) — resting / baseline state.
* '''Hopf bifurcation''' (complex-conjugate pair crosses imaginary axis) — emergence of α/β/γ rhythmic oscillations.
* '''Saddle-node''' (one eigenvalue passes through zero) — multistability; basis for perceptual bistability and working-memory models.

Wilson-Cowan with appropriate parameters reproduces:

* α (8–13 Hz) rhythms in cortex.
* β (13–30 Hz) sensorimotor oscillations.
* γ (30–80 Hz) high-frequency synchrony.
* Slow oscillations and up/down states.

== Sanity-check limits ==

* '''No coupling''' (wEE= wEI= wIE= wII= 0): pure exponential relaxation E(t) → 0; I(t) → 0. ✓
* '''Linear regime''' (σ → ∞ in sigmoid): linear coupled ODEs; classical damped oscillator. ✓
* '''r → 0''': no refractoriness; population activity unconstrained above. (Limit not biologically realistic; mathematically clean.)
* '''Single population, wEE > 1''': bistability between low and high firing.

== Connection to ψ ==

The framework's [[Wilson-Cowan_Coupled_to_Psi|ψ-coupled extension]] adds:

* '''Brain → ψ''' source: Jψ(x,t) = κJ · f(u(x,t)).
* '''ψ → brain''' feedback: extra term + β · ψ(x,t) in the activation argument.
* '''ψ field equation''': □ψ − m2ψ − λψ3 = α FμνFμν + Jψ.

Wilson-Cowan provides the baseline neural dynamics; ψ coupling enters as small additive terms with parameters κJ, β.

== Experimental status ==

Wilson-Cowan is a '''pillar of mainstream computational neuroscience'''. It is taught in every computational-neuroscience curriculum, used as the population-level component of most large-scale brain models (TVB, Spaun, others), and validated by:

* Direct fit to electroencephalographic (EEG) and magnetoencephalographic (MEG) spectra.
* Match to local-field-potential recordings in cortex.
* Reproduction of bistability, hysteresis, and oscillatory regimes in cortical slice preparations.

The Wilson-Cowan equations are not in dispute. The framework adds ψ-coupling on top of standard Wilson-Cowan.

== See Also ==

* [[Amari_Neural_Field]]
* [[Hodgkin-Huxley_Equations]]
* [[FitzHugh-Nagumo_Equations]]
* [[Jansen-Rit_Neural_Mass]]
* [[Neural_Field_Equations]]
* [[Wilson-Cowan_Coupled_to_Psi]]
* [[Effective_Field_Theory_of_Consciousness]]

== References ==

* Wilson, H. R., Cowan, J. D. (1972). "Excitatory and inhibitory interactions in localized populations of model neurons." ''Biophysical Journal'' 12: 1–24.
* Wilson, H. R., Cowan, J. D. (1973). "A mathematical theory of the functional dynamics of cortical and thalamic nervous tissue." ''Kybernetik'' 13: 55–80.
* Coombes, S., Beim Graben, P., Potthast, R., Wright, J. (eds.) (2014). ''Neural Fields: Theory and Applications.'' Springer.

[[Category:Psionics]]
[[Category:Equations]]
[[Category:Neuroscience]]

Wilson-Cowan Model - Revision history

JonoThora: Phase N (01b): LaTeX restoration — promote Unicode display-math to ; lint-clean per tools/wiki_latex_lint.py

JonoThora: Phase N (01b): LaTeX restoration — promote Unicode display-math to $; lint-clean per tools/wiki_latex_lint.py$