Results from long-term field experiments in south-western Australia are presented in the form of relationships between yield, expressed as a percentage of the maximum yield, and soil test for phosphorus (P) values. Maximum yields were not always indicated by well defined yield plateaux. Different methods have been used to estimate the maximum yield value which is used to calculate yield as a percentage of the maximum yield so as to remove interseasonal variation. For all of these methods and for the same site, the same P fertilizer (superphosphate), and the same plant species, the relationship between yield and soil test P differed for different years. Consequently fertilizer recommendations based on the assumption that this relationship is constant are likely to be incorrect. We therefore question the validity of the common practice in soil testing programmes of using percentage yield values to remove interseasonal variation.