The vertical variability of the salinity field, with emphasis on interannual timescales, is examined within the tropical Pacific Ocean (10¿N--10¿S) using a compilation of the conductivity-temperature-depth (CTD) casts for the period 1975--1998. Compared to the vertical dependence of temperature that exhibits a systematic maximum variability at the depth of the main thermocline, the salinity typically shows a maximum variability in the surface layers. One notable region where this rule is violated is the southwestern and central Pacific Ocean. Below the surface, salinity variability is correlated with the strong gradients in mean salinity above and below the subsurface salinity maximum. It is shown that using conventional mean T-S curves to estimate salinity profiles from temperature observations leads to strong biases because of the large scatter around the T-S relationships. From the surface to the bottom of the thermocline, the dispersion of the T-S diagrams is large regardless of whether the conditions are representative of El Ni¿o or La Ni¿a conditions. Error introduced in computing the dynamic height anomaly (DHA) is larger than 2 dyn. cm (dynamic centimeters) throughout the tropical Pacific if salinity variability is neglected. This error represents 30% of the total variability of the sea surface height in the western Pacific, and more than 50% in the south central Pacific. In order to estimate salinity variability when direct measurements are not available, an empirical orthogonal function analysis of existing temperature and salinity profiles was conducted. It is shown that a relatively low number of dominant modes, typically less than 6, are sufficient to explain 80% or more of the total variance. The separate contributions of the temperature and salinity fields to the DHA are then examined. The salinity contribution is generally smaller than the temperature contribution, though in some instances the two oppose one another, resulting in lowered dynamic height anomalies. These results confirm that salinity variability should not be neglected in ocean analyses that attempt to infer vertical changes in density from sea level fluctuations within the tropical Pacific Ocean.