[EN] We study the dependence of salinity on temperature in two by two degrees latitude-longitude boxes, for surface and upper-thermocline waters of the eastern North Atlantic subtropical gyre. The initial data set, from historical databases as well as from recent hydrographic cruises in the region, is carefully scrutinized to reject dubious measurements. We search for polynomial relations of variable degree between salinity and temperature, the optimal fit is selected as the polynomial with the lowest degree that satisfies several statistical criteria. An independent hydrographic cruise is used to confirm that the method performs substantially better than estimates from climatological data, and leads to relatively low deviations in geopotential anomaly and other derived quantities. An error propagation analysis using the Monte Carlo method shows equally good results.[ES] En este estudio caracterizamos la dependencia de la salinidad con la temperatura en cajas de dos por dos grados de latitud y longitud para aguas superficiales y de la termoclina superior en el margen oriental del giro subtropical del Atlántico Norte. El conjunto inicial de datos, procedente de bases de datos históricos, así como los procedentes de recientes campañas hidrográficas en la región, son cuidadosamente analizados para eliminar las medidas de calidad dudosa. Se analizan varias relaciones polinómicas de grado variable para la salinidad en función de la temperatura, eligiendo como ajuste óptimo aquel polinomio de menor grado que satisface diversos criterios estadísticos. Se utiliza una campaña hidrográfica independiente para confirmar que el método da resultados substancialmente mejores que estimaciones a partir de datos climatológicos, y que produce desviaciones relativamente bajas en la anomalía geopotencial y en otras magnitudes derivadas. Un análisis de la propagación del error usando el Método de Montecarlo muestra resultados igualmente buenos.We are grateful to the Institut für Meereskunde and IFREMER for making their data sets available. The Spanish cruises that gathered some of the data analyzed here were carried out through projects from the Spanish government (program CYTMAR) and the European Union (program MAST). This work has been carried out with the support of the Spanish Government through project FRENTES (AMB95-0731) and the European Union through projects CANIGO (MAS3-CT96-0060) and OASIS (EVK3-CT-2002-00073). Part of this work was written while JLP was at the University of Wisconsin-Madison with funding from the Secretaría de Estado de Educación y Universidades of the Spanish government.Peer reviewed