Long-term removal of grain P and soil test data suggested that the Colwell phosphorus (P) extraction from the surface 0.10 m of a Black Vertosol from south-eastern Queensland was a poor indicator of run-down of soil P pools. We proposed that plants were also accessing P from layers below 0.10 m or from surface soil P pools not extracted by the Colwell extraction. Both topsoil and subsoil samples in 1994 and 2003 were collected from nil and 20 kg P/ha per crop treatments in a long-term N × P field experiment established in 1985 for detailed P fractionation. An uncropped reference soil was also taken in 2003 from an adjacent area. The long-term effect of the field treatments on soil P fractions was evaluated by comparing the reference site, which was assumed to represent the original soil condition, to the 2003 samples. Without addition of P fertiliser, 55%, 35%, and 10% of total P removal were from 0 to 0.10, 0.10 to 0.30, and 0.30 to 0.60 m, respectively, compared with the uncropped reference soil. Labile fractions comprising resin, bicarbonate, and hydroxide pools in the top 0.10 m decreased by approximately 60% and accounted for 15% of the total P decrease from 0 to 0.60 m depth. Acid and residual-P fractions decreased by 50% and 20%, respectively, and accounted for ~20% and 15% of the total P decrease. In contrast, P addition at 20 kg P/ha per crop over 18 crops doubled the resin and bicarbonate inorganic P (NaHCO3-Pi) pools in the surface 0.10 m. Hydroxide (NaOH-Pi) and acid extracted inorganic P increased by 25% and 10%, respectively, while the residual-P pool decreased by about 15%. Below 0.10 m, very little P was removed by the first 3 extractants. Most of the P was present in the acid and residual fractions irrespective of fertiliser application. The acid and residual-P dropped by 30% and 12%, respectively, at 0.10–0.30 m and 12% and 8% at 0.30–0.60 m. When comparing the experimental soil samples in 2003 with those in 1994, similar trends were observed in the changes of each soil P fraction. In the surface 0.10 m, acid and residual-P pools decreased greatly and explained almost all of the total P decrease in the surface soil without P input. With P addition, labile pools acted as the main sink for P. The acid pool increased by 7%, while the residual-P showed a decrease in the topsoil. Total P level was elevated noticeably in this soil layer. However, at 0.10–0.30 m depth, acid and residual pools were the dominant fractions and decreased significantly irrespective of P fertiliser addition. Below 0.30 m, no significant changes were detected for each fraction and total P. The results suggest that crops had accessed significant amounts of P at 0.10–0.30 m depth irrespective of P fertiliser application, and that subsoil sampling (0.10–0.30 m) should be considered in order to improve the monitoring of soil P status. However, choice of appropriate extractants for monitoring subsoil P reserves is yet to be undertaken.