The North Atlantic Oscillation (NAO) is the most dominant hemispheric variability affecting the winter climate. It is, however, difficult to predict the NAO on a seasonal time scale. Thus, a better understanding of the NAO is important for improving the accuracy of seasonal forecasts. We investigated the seasonal predictability of the NAO and the zonal mean fields of the stratosphere using hindcasts based on the operational seasonal prediction system of the Japan Meteorological Agency. We found that both the predictive skill and the potential predictability of the NAO index increased from late winter to early spring and that this seasonality relates to high predictability of the zonal mean geopotential height in the stratosphere and near the surface. Analysis of the convergence/divergence of the Eliassen-Palm flux indicates that the high predictability in the stratosphere in late winter results from the large signal of upward-propagating Rossby waves. The downward influence of the predictable signal from the stratosphere to the troposphere was also quantitatively estimated using the approach proposed by Ambaum and Hoskins (2002). The signal of the surface pressure estimated from that of stratospheric potential vorticity is roughly consistent with the actual signal, substantiating the stratospheric contribution to the near-surface seasonal predictability.(Citation: Saito, N., S. Maeda, T. Nakaegawa, Y. Takaya, Y. Imada, and C. Matsukawa, 2017: Seasonal predictability of the North Atlantic Oscillation and zonal mean fields associated with stratospheric influence in JMA/MRI-CPS2. SOLA, 13, 209−213,