Sorption of phosphate by soils is an important determinant of fertiliser efficiency. Its extent and rate differ between soils. This paper considers both how these differences can be described and how they can be explained. Published data on the rate of reaction of soil with phosphate are assembled. There are 88 soil samples from five continents, as reported in six publications. In almost all cases, sorption (S) was related to solution concentration (c) and time (t) by: S = a cb1tb2 – qtb3. The term after the minus sign describes the rate of desorption; its value is affected by soil phosphate status. The “a” parameter reflects the amount of adsorbing surfaces present plus their affinity for phosphate. The b1 parameter, which describes the curvature of the sorption plots, reflects the composition of the adsorbing surfaces and their phosphate status. Large values are obtained for soils with a small range of electric potentials of their reacting surfaces and a low phosphate status, as is common for Australian virgin surface soils. The greatest influence on the b2 parameter, which describes the curvature of the plots against time, is the method of measurement. Large values were observed when soil and solution were shaken together and are probably an artefact caused by mutual abrasion of the soil particles. When this problem was avoided, large values of b2 were associated with very phosphate-deficient soils; small values of b2 often occurred when b1 was also small, suggesting surfaces of varying potential were present. If the substances present included components such as kaolin, that might explain both the range in potentials and limited penetration. Highlights: The effect of solution concentration (c) and time (t) on sorption (S) can be described by S = acb1tb2 – qtb3 and the other terms are parameters. Small values of the “curvature” term b1 are associated with large values for heterogeneity of the reacting surfaces. Mutual abrasion of the soil particles causes large values of rate parameter b2 when soil and solution are shaken. When this problem is avoided the value of b2 depends on the rate of penetration.