Fig. 4. For low degree of input K_ext, the network shows clear signatures of criticality beyond power laws.

Only for low values of K_ext, a the estimated branching ratio m tends toward unity, and b the estimated autocorrelation time τ_corr peaks. c The match of the τ_corr, and the ${\tau }_{\mathrm{branch}}~-1/\log \left(m\right)$ as inferred from m supports the criticality hypothesis (correlation coefficient of ρ = 0.998, p < 10⁻¹⁰). d Trial-to-trial Δ_VRD variations as well as the susceptibility χ increase for low K_ext.