Subterranean clover was grown on five different soils in the glasshouse with superphosphate treatments, which provided factorial combinations of previously and of currently applied phosphorus. For most treatments, powdered superphosphate was used but on two of the soils there was an additional set of treatments using granular superphosphate. In a further set of treatments on these two soils, Wimmera ryegrass was grown. Values for yield and for uptake of phosphate were used to test measures of residual value. The measure of residual value was the slope of the yield-fertilizer curve for the previous application divided by slope for the current application. Where only the rapidly ascending part of the response surface was available, linear equations satisfactorily described the response. Where the region in which yields approached the maximum was also available, a modified form of the exponential or Mitscherlich equation was found to be preferable. For linear equations, methods for calculating the confidence limits of the ratio were available. For the non-linear equations it was found that the joint confidence regions were generally symmetric and confidence regions derived by assumin g the functions were approximately linear were adequate. It was shown that the confidence limits were wide and that effects of treatments which may be small are difficult to prove. No significant effects of granulation or of level of application were detected, nor were there significant differences between the two species. With repeated clipping of the ryegrass, the previous application of phosphate became relatively less effective on a soil of low buffering capacity for phosphate and relatively more effective on a soil of high buffering capacity. Hence, residual value does not necessarily have a unique value through a season. It was suggested that these contrasting effects were due to progressive depletion of the phosphate in the soil of low buffering capacity, and to increasing root density in the, soil of high buffering capacity.