Persistence forecast of the Niño 3 index show a sharp drop in correlation skill during boreal spring. This drop in skill serves to define the spring predictability barrier of ENSO (SPBE). The predictability barrier is studied with NASA’s Seasonal-to-Interannual Prediction Project forecasting system. The NSIPP-1 coupled general circulation model simulates the ENSO phenomenon realistically in multi-decadal free integrations and a SPBE occurs in the seasonal autocorrelation function of the Niño 3 index. Analysis of the mechanisms of the simulated ENSO suggests that the SPBE can be significantly reduced in forecast with initialized comprehensive coupled models. This suggestion is supported by comparison of ENSO hindcast produced by the NSIPP-1 coupled forecasting system with and without ocean data assimilation during the 1993 – 2002 decade.