Monsoon predictability during the 15 years (1979-1993) of the European Centre for Medium-Range Weather Forecasts (ECMWF) re-analysis (ERA), have been studied using 120-day ensemble integrations of the ECMWF model. These 9-member ensemble integrations are part of the European Union PROVOST project. Observed sea surface temperatures (SSTs) were updated daily at the model's lower boundary. There are two major and three moderate El Nino Southern Oscillation (ENSO) events occurred during the ERA period. Using the existing monsoon indices in the literature, the potentially predictable component of monsoon interannual variability is assumed to be that due to oceanic SST forcing, and using "analysis of variance," this is separated from the unpredictable internal component. Significant levels and confidence intervals are calculated. Both are essential for a more meaningful interpretation of predictability estimate. The potential predictability of the summer monsoon circulation and convective activities over the East Asia is higher than the India subcontinent. The estimates of seasonal predictability for each year of the dataset were also made to discriminate between years of high and low predictability. It generally indicates high predictability in years when ENSO was strong. The dependence of regional monsoon predictability on the different-types of forcing and forecast lead time is examined. For the seasonal-forecasting skill of precipitation anomalies over different monsoon systems, estimated by the anomaly correlation, the dependence of the ensemble mean skill on the magnitude and phase of the SST forcing is discussed. The spread of the skill of individual integrations from the ensemble mean and their relationship to the ensemble mean skill is assessed. The probabilistic skill estimated by the relative operating characteristics will also be highlighted.