Phenotypic resistance of salinity is expressed as the ability to survive and grow in a salinised medium. Some subjective measure of overall performance has normally been used in plant breeding programmes aimed at increasing salinity resistance, not only to evaluate progeny, but to select parents. Salinity resistance has, at least implicitly, been treated as a single trait. Physiological studies of rice suggest that a range of characteristics (such as low shoot sodium concentration, compartmentation of salt in older rather than younger leaves, tolerance to salt within leaves and plant vigour) would increase the ability of the plant to cope with salinity. We describe the screening of a large number of rice genotypes for overall performance (using an objective measure based on survival) and for the aforementioned physiological traits. There was wide variation in all the characters studied, but only vigour was strongly correlated with survival. Shoot sodium concentration, which a priori is expected to be important, accounted for only a small proportion of the variability in the survival of salinity. Tissue tolerance (the cellular component of resistance reflecting the ability to compartmentalise salt within leaves) revealed a fivefold range between genotypes in the tolerance of their leaves to salt, but this was not correlated positively with survival. On the basis of such (lack of) correlation, these traits would be rejected in normal plant breeding practice, but we discuss the fallacies involved in attempting correlation between individual traits and the overall performance of a salt-sensitive species in saline conditions. We conclude that whilst overall performance (survival) can be used to evaluate the salt resistance of a genotype, it is not the basis on which parents should be selected to construct a complex character through breeding. It was the norm for varieties which had one good characteristic affecting salt resistance to be unexceptional or poor in the others. This constitutes experimental evidence that the potential for salt resistance present in the rice genome has not been realised in genotypes currently extant. The results are discussed in relation to the use of physiological traits in plant breeding, with particular reference to environmental stresses that do not affect a significant part of a species' ecological range.