To calculate nutrient requirement it is necessary to know how much of that nutrient must be applied to overcome any buffering effects and raise the test value to a desired level. A glasshouse study was conducted on a wide range of Natal soils to establish this information for phosphorus. The upper 200-mm depth at 54 sites was sampled. Soils varied greatly in texture (4–83% clay), organic carbon (0.2–9.2%) and clay mineralogy (kaolinitic-sesquioxic through to illitic and smectitic). Soil from each site was treated with four levels of P and taken through a number of wetting and drying cycles over a six-week period. Analysis for P using the Ambic, Bray no. 1 and Truog extradants showed that the quantity of applied P (kg ha−1) required to raise the soil test value by a unit amount (1 mg L−1), i.e. the P requirement factor, varied greatly for the different soils. The range in values for these three extradants was 2.5–37.9; 2.0–17.7; and 2.3–30.3, respectively. The level of P sorption was found to be strongly related to clay content, and soils rich in 2:1 minerals showed surprisingly high P sorption.