Efficient use of phosphorus (P) by plants in agriculture relies on accurate estimation of the phytoavailable soil P. However, poor relationships are frequently observed between P availability indices, such as Olsen P, and P uptake by plants. We therefore studied which soil properties modify the accuracy of Olsen P as a P availability index. We conducted a soil P depletion experiment in pots with 17 soils. Two samples differing widely in Olsen P were used for each soil and named "low-P" and "high-P." Soil was mixed with siliceous sand to achieve 1 mg of Olsen P per pot, so that the ratio of non-readily available P, either inorganic P or organic P, to Olsen P in the pot was higher for low-P than for high-P samples. Results show that, in high-P samples, P uptake by cucumber is positively correlated with the affinity of the soil solid phase for P, with R 2 of 0.76. In high-P samples, P uptake by cucumber is positively correlated with total P adsorption capacity as estimated by Fe in the form of Fe oxides, with R 2 = 0.56. P uptake is positively correlated with the inorganic P released by NaOH and citrate-bicarbonate in low-P samples, with R 2 of 0.52. In low-P samples, total organic P and phosphatase activity in the rhizosphere explained 37 % of plant P uptake. In low-P samples, phosphatase activity and pH explained 50 % of plant P uptake. These findings reveal that organic P forms and hydrolytic activity by roots are explaining P availability to plants. Whereas physico-chemical processes controlling inorganic P dynamics play a major role for low organic P/Olsen P ratios.