Central equatorial Pacific sea surface temperature (SST) anomaly is a critical predictor of basin-wide oceanographic and atmospheric effects of the El Ni¿o-Southern Oscillation (ENSO) phenomenon. We employ two geochemical thermometers measured on coralline aragonite to reconstruct an independent proxy-based measure of central equatorial Pacific sea surface temperature anomaly. In addition, we assess the observational error associated with extraction of large-scale SST anomalies from Δ18O and Sr/Ca measurements. On the basis of paired data for the 1981--1987 period, RMS error for the estimation of SST from Kiritimati coral Δ18O is about 0.4 ¿C, assuming no seawater Δ18O influence; for Sr/Ca, the observational error is about 0.5 ¿C. Singular spectrum analysis of the Δ18O time series suggests that 1/3 of the variance is explained by SST anomaly and that this variance may be separated from other signals in the frequency domain. The interannual component of the Δ18O record shares 70% variance with the interannual component of local SST anomaly estimates (&rgr;=-0.84) and correlates as highly with NINO3 region (150 ¿W-90 ¿W, 5 ¿N-5 ¿S) SST anomaly estimates; comparison of annual averages of the Δ18O data with analyzed SST for 55 years suggests that the error in deriving annually averaged SST anomalies at Kiritimati is about 0.4 ¿C. A global SST correlation analysis suggests significant correlation with the rest of the Pacific Basin and the tropical Indian and Atlantic sectors. The ocean-wide level of correlation achieved using the coral data is indistinguishable from that achieved using the NINO3 SST anomaly index and suggests that a few well-located coral reconstructions with low observational error may be sufficient to reconstruct the global SST anomaly field associated with ENSO activity. ¿ 1999 American Geophysical Union