Soil samples from 179 field experiments testing phosphate fertilizers on potatoes, swedes and grass were analysed for total phosphorus, and for phosphorus soluble in the following solutions: 0·3 N-HCl, 0·002 N-H2SO4, 1% citric acid (H3Ci), 0·5 N acetic acid (HAc), acetic acid-sodium acetate buffer (HAc-NaAc) at pH 4·8, 0·5 M-NaHCO3, and 0·01 M CaCl2.Average values for soluble P were closely related to average crop responses to superphosphate in the experiments on swedes, but not in the grass and potato experiments. The extractants that differentiated best between responsive and unresponsive groups of experiments were HAc, HAc-NaAc, and NaHCO3 for potatoes, and HCl, H2S04, HAc–NaAc, NaHCO3 and CaCl2 for grass.For the experiments as a whole 0·5 M-NaHCO3 was the ‘best’ extractant. The HCl, H2SO4, and HAc-NaAc buffer solution methods were roughly equally effective, though inferior to NaHCO3; the other three extractants (HAc, H3Ci, CaCl2) were of little general use. Total P in soil was also related to response to superphosphate, though less well than values for soluble P obtained by the better methods.Estimates of soluble P by different solvents were often related. Estimates by HCl and H2SO4 methods were most closely related; values for P soluble in H2SO4 and in HAc-NaAc were also often significantly correlated, as were estimates by HAc–NaAc and CaCl2. The H3Ci and CaCl2 methods gave results that were least related to those with other methods.The use of soil analyses in advising on P-manuring is discussed and a tentative method is proposed of establishing the analytical limits for soluble P that define ‘deficient’ soils. If the confidence attached to the limiting values that separate ‘deficient’ and ‘non-deficient’ soils is stated, farmers will be able to assess the risk entailed in accepting advice based on soil analysis.