Equatorial westerly (easterly) wind anomalies, phase-locked to the seasonal cycle, typically ‘propagate’ from the eastern equatorial Indian Ocean and into the western Pacific immediately before an El Niño (La Niña). A space-time integration of this Indo-Pacific signal yields an index τ that, for 11 out of 12 months of the calendar year, leads the El Niño index NINO3.4 with a correlation of 0.5 or greater for at least some lead in the range 10–15 months. Cross-validated hindcasts suggest that a linear combination of this atmospheric index and the ocean indices NINO3.4 and (t), the anomalous equatorial Pacific upper ocean heat content, is an excellent predictor of El Niño. It can predict across the nearest spring persistence barrier, but not the one after that. The present El Niño should die over the spring, leaving near neutral conditions for the rest of 2003.