A water‐extraction procedure which is specific for a pool of P defined by a sorption model based on the Langmuir equation was evaluated in a glasshouse experiment using perennial ryegrass grown in three contrasting soils. For each of three harvests, water‐extractable P values determined at the beginning of each period of growth gave a very much better relationship with dry matter yield responses than did Olsen and Truog P values. Also, the initial water‐extraction values were as useful as the Olsen or Truog values in predicting responses over three harvests. Water extraction appears to have potential as a soil‐testing procedure for determining the P status of soils which have received fertilizer P.