The relationship between isotope concentration and salinity provides important evidence on the origin of water masses. In polar regions it can be used to deduce the isotopic composition of melting ice. We discuss a source of error implicit in the use of isotope salinity diagrams for this purpose. We argue that conventional δ18O and δD salinity diagrams should be abandoned in favor of (1 ‐ S) δ18O and (1 ‐ S) δD salinity diagrams which are physically more meaningful and which introduce no significant error.