Mg/Ca in foraminiferal calcite has recently been extensively used to estimate past oceanic temperatures. Here we show, however, that the Mg/Ca temperature relationship of the planktonic species Globigerinoides ruber is significantly affected by seawater salinity, with a + 1 psu change in salinity resulting in a + 1.6 degrees C bias in Mg/Ca temperature calculations. If not accounted for, such a bias could lead, for instance, to systematic overestimations of Mg/Ca temperatures during glacial periods, when global ocean salinity had significantly increased compared to today. We present here a correction procedure to derive unbiased sea surface temperatures (SST) and delta O-18(sw) from G. ruber T-Mg/Ca and delta O-18(f) measurements. This correction procedure was applied to a sedimentary record to reconstruct hydrographic changes since the Last Glacial Maximum (LGM) in the Western Pacific Warm Pool. While uncorrected T-Mg/Ca data indicate a 3 degrees C warming of the Western Pacific Warm Pool since the LGM, the salinity-corrected SST result in a stronger warming of 4 degrees C.