Aims: Many soil scientists think that soil phosphate exists as discrete compounds of iron, aluminium and calcium and, accordingly, use chemical fractionation schemes to identify these compounds. Methods: We reacted a sample of goethite and a sample of aluminium oxide with a phosphate solution under conditions chosen to facilitate penetration of phosphate. Thus the sample of goethite had neither calcium nor aluminium present and similarly the sample of aluminium oxide had neither iron nor calcium. We included a sample of hydroxyapatite which had neither iron nor aluminium present. We subjected the samples to two fractionation procedures; the original Chang and Jackson (1957) method and a variant of it. Results: For the phosphated goethite and aluminium oxide, energy dispersive X-ray spectra did not detect any discrete aluminium or iron phosphates; dissolution studies were consistent with penetration of phosphate. Both fractionation procedures detected discrete compounds even though none were present. They also detected iron, aluminium and calcium phosphates for samples for which they were not present. We also critically discuss other evidence for the existence of discrete iron, aluminium and calcium phosphates in soils. Conclusions: Fractionation procedures designed to measure chemically specified phosphate fractions in soil are fallacious and should be abandoned.