In this paper, the conditional nonlinear optimal perturbation (CNOP) method is used to explore the optimal precursors that trigger the North Atlantic Oscillation (NAO) anomaly pattern with a triangular T21, three-level, quasi-geostrophic global spectral model based on a viewpoint that NAO is a nonlinear initial-value problem.With a three-dimensional winter climatological flow as the basic state, initially baroclinic localized optimal precursors on the northward flanks of the climatological Atlantic jet undergo wave breaking during their evolution into the NAO-like anomalies.Accompanied with the formation of NAO, the north-south variability of the zonal mean westerly anomaly has arisen.Analysis reveals that in the evolution of optimal precursors, the role played by the self-interaction of perturbations (namely the nonlinear process) in the onset of the negative phase NAO (NAO-) event is stronger than that in the onset of the positive phase NAO (NAO+) event.Both the perturbation/basic state interaction and self-interaction of perturbations determine whether the NAO-event occurs, whereas the nonlinearity process in the NAO+ onset only appears to modulate the structure of the perturbation to have a dipole mode over the north Atlantic at the optimization time, and meanwhile cause this dipole mode to become zonally extended.That is to say, the nonlinear process indeed plays an important role during the onset of an NAO event and CNOP method is a useful tool to identify the dynamics of the onset of NAO teleconnection patterns.