An identification methodology for nonlinear dynamic systems using Gaussian process prior models is presented. It combines equilibrium and off-equilibrium information. The advantages of so doing are highlighted.