Long time series of high quality ocean surface topography, sea surface temperature, subsurface temperature and salinity information allow a thorough comparison between two distinct El Ni¿o events. The 1997 El Ni¿o showed a strong ocean-atmosphere coupling which propagated from west to east. The weaker 2002 event developed later in the calendar year and showed strongest anomalies near the dateline. In this study, observational data of sea level and sea surface temperature from satellite, and subsurface temperature and salinity (from all available sources including Argo) are used along with data assimilation to improve the initial state estimation of the ocean. Ocean model experiments are used to isolate differences between initial states of the system and forcing in the development of the two events. A statistical atmospheric model is utilized to highlight the atmospheric response to surface temperature anomalies. Different initial conditions for the month of November (climatology, 1996 and 2001) are used to initiate model experiments forced by climatological European Centre for Medium-Range Weather Forecasts wind stress for the two El Ni¿o periods (Nov 96--Dec 97, Nov 01--Dec 02). By differencing two experiments with the same forcing, but different initial conditions, the role of the ocean state at the start of the two events (Nov 96 versus Nov 01) can be examined. These results show that both initial conditions are predisposed to induce a subsequent El Ni¿o. However, for the first half of the 1997 event (Nov 96--Jun 97) the initial state of the ocean induces an eastward propagating wind field. For the 2002 event, the initial state served to limit the eastward propagation of the wind field by inducing destructive interference near the dateline. For the second half of the event (Jul 97--Dec 97) the initial state contributed little. However, the role of initial conditions contributed nearly half the total signal for the peak of the 2002 event. Similarly, the role of the different forcing can be identified. These results show that for both events forcing played a very weak role over the first half of the year. (In fact the statistical wind anomalies for these periods showed anomalous easterlies.) However, for the second half of the 1997 event the air/sea coupled forcing accounts for nearly all the temperature and wind anomalies. For the 2002 event, initial conditions and forcing contribute approximately equally to the temperature and resulting wind anomaly fields.