Summary. A selection of commonly used soil phosphorus (P) tests, which included anion and cation exchange resin membranes, were compared in a glasshouse experiment using subterranean clover, and evaluated in the field at 19 sites from the National Reactive Phosphate Rock Project in 1993 and at 6 sites in 1995. The ability of the soil P tests to predict plant response was used to evaluate the tests. In the glasshouse experiment the resin test was less effective than the Bray 1 and Colwell tests in its ability to assess the level of plant-available P from the different fertiliser treatments. Seventy-one percent of the variation in total P content of the subterranean clover shoots was explained by resin-extractable P values, whereas the Colwell procedure accounted for 81% and the Bray 1 procedure accounted for 78%. Water and CaCl2 extracts were poor predictors of P content. In the field experiments all tests evaluated performed poorly in describing the relationship between soil test P and the level of P applied and relative yield and soil test P over a wide range of soil types and environments. The Bray 1 procedure performed best but the relationship was poor.