Near bifurcation points, systems recover slower from perturbations — a universal early warning signal
Near a saddle-node bifurcation (tipping point), the dominant eigenvalue λ₁ → 0: the system's "spring constant" vanishes. Perturbations decay more slowly — variance increases and lag-1 autocorrelation AC₁ → 1. These generic statistical signals appear hundreds of time steps before the transition, independent of system details. They've been detected before: lake eutrophication, fisheries collapse, climate tipping elements (Scheffer et al. 2009 Nature). The model here is: dx/dt = r·x − x³ + σ·η (subcritical pitchfork). As r → 0, recovery time τ = 1/|r| → ∞.