Total P (TP), total particulate P (TPP), dissolved reactive P (DRP), and dissolved organic P (DOP) were determined in waters from pipedrains (at 65‐cm depth) from the Broadbalk Experiment at Rothamsted. Soils that have received either no P, P in farmyard manure (about 40 kg P ha−1) or superphosphate (up to 35 kg P ha−1) annually for >150 yr, now contain 0.5 M NaHCO3‐extractable P concentrations (Olsen‐P) in the plow layer (0‐ to 23‐cm depth) between 5 and 100 mg kg−1 soil. Our aim was to determine if significant quantities of P could be detected in the drainage water and their relationship to soil P concentrations. On five occasions between October 1992 and January 1994, both TP and DRP from plots receiving superphosphate frequently exceeded 1 mg L−1 and were high compared with literature data. Ranging between 66 and 86% of TP, DRP was the largest fraction in drainage water. It remained low (−1) from plots below 60 mg Olsen‐P kg−1. There was then a rapid increase in DRP up to the maximum Olsen‐P concentration. A simple linear Split‐Line Model described this relationship very well for all drainage events. This implies that up to 60 mg Olsen‐P kg−1 soil (the change point), P was retained strongly in the plow layer. Above this, P losses in the drainage water were much more closely related to Olsen‐P than commonly suggested. The mechanisms could either be preferential flow or rapid transport of P in forms less susceptible to sorption but finally measured as DRP. The results suggest enhanced P losses through subsurface runoff on heavy soils, once a certain plow layer concentration of Olsen‐P is exceeded.