In a field experiment on deep, yellow, sandy soil near Badgingarra, Western Australia, the residual value of superphosphate applied one and two years previously was measured relative to freshly-applied superphosphate using yields of narrow-leafed lupin (Lupinus angustifolius), barley and wheat. In addition, soil samples were collected for measurement of bicarbonate-extractable soil P. This was also used to estimate the residual value of the superphosphate. For lupins and wheat, and for bicarbonate-extractable soil P, the residual value decreased with increasing level of application. For barley grain, the residual value was not significantly affected by the level of application. The decrease in residual value of superphosphate with increasing level of application is attributed to increased leaching of applied phosphorus (P) down the profile of the sandy soils as the level of application increases. This may reduce subsequent plant yields due to the delay in seedling roots reaching the P in the soil during the crucial early stages of plant growth. For lupins, the relationship between yield and the level of superphosphate applied was markedly sigmoidal. The relationship for wheat and barley was exponential. Consequently, at suboptimal levels of P application, lupins required about two to three times more P than wheat or barley to produce the same yield. However, lupins required less P to achieve near-maximum yield.