Experiments on dynamical extended range forecasting (DERF) conducted at the National Meteorological Center (NMC) are utilized to obtain estimates of the upper and lower bounds of dynamical predictability for the NMC operational medium-range forecasting systems (MRFS). Owing to the extended range of the integrations (up to 30 days) the upper bound can be estimated without resorting to an empirical model of error growth such as that used by Lorenz (1982). At the root-mean-square error level that equals the climatological standard deviation (about 100 m), the average limit of dynamical predictability is about 14 days, in good agreement with the recent results of Lorenz using the operational forecasting model of the European Centre for Medium Range Weather Forecasts (ECMWF). The current level of forecast skill is also evaluated, indicating loss of useful skill at about 6 days, in agreement with criterion of Saha and Van den Dool. This skill level is much below its potential, thereby leaving considerable room for improvement. The possible contributions to extending the forecast skill from model improvements, as well as initial data error reductions, are also assessed.