A study was initiated to investigate the relationship between soil test P and depth of soil sampling with runoff losses of dissolved molybdate reactive phosphorus (DMRP). Rainfall simulations were conducted on two noncalcareous soils, a Windthorst sandy loam (fine, mixed, thermic Udic Paleustalf) and a Blanket clay loam (fine, mixed, thermic Pachic Argiustoll), and two calcareous soils, a Purves clay (clayey, smectitic, thermic Lithic Calciustoll) and a Houston Black clay (fine, smectitic, thermic Udic Haplustert). Soil (0- to 2.5-, 0- to 5-, and 0- to 15-cm depths) and runoff samples were collected from each of the four soils in permanent pasture exhibiting a wide range in soil test P levels (as determined by Mehlich III and distilled water extraction) due to prior manure applications. Simulated rain was used to produce runoff, which was collected for 30 min. Good regression equations were derived relating soil test P level to runoff DMRP for all four soil types, as indicated by relatively high r2 values (0.715 to 0.961, 0- to 5-cm depth). Differences were observed for the depth of sampling, with the most consistent results observed with the 0- to 5-cm sampling depth. Runoff DMRP losses as a function of the concentration of P in soil were lower in calcareous soils (maximum of 0.74 mg L(-1)) compared with noncalcareous soils (maximum of 1.73 mg L(-1)). The results indicate that a soil test for environmental P could be developed, but it would require establishing different soil test P level criteria for different soils or classes of soils.