Phosphorus transport from agricultural soils contributes to eutrophication of fresh waters. Computer modeling can help identify agricultural areas with high potential P transport. Most models use a constant extraction coefficient (i.e., the slope of the linear regression between filterable reactive phosphorus [FRP] in runoff and soil P) to predict dissolved P release from soil to runoff, yet it is unclear how variations in soil properties, management practices, or hydrology affect extraction coefficients. We investigated published data from 17 studies that determined extraction coefficients using Mehlich-3 or Bray-1 soil P (mg kg(-1)), water-extractable soil P (mg kg(-1)), or soil P sorption saturation (%) as determined by ammonium oxalate extraction. Studies represented 31 soils with a variety of management conditions. Extraction coefficients from Mehlich-3 or Bray-1 soil P were not significantly different for 26 of 31 soils, with values ranging from 1.2 to 3.0. Extraction coefficients from water-extractable soil P were not significantly different for 17 of 20 soils, with values ranging from 6.0 to 18.3. The relationship between soil P sorption saturation and runoff FRP (microg L(-1)) was the same for all 10 soils investigated, exhibiting a split-line relationship where runoff FRP rapidly increased at P sorption saturation values greater than 12.5%. Overall, a single extraction coefficient (2.0 for Mehlich-3 P data, 11.2 for water-extractable P data, and a split-line relationship for P sorption saturation data) could be used in water quality models to approximate dissolved P release from soil to runoff for the majority of soil, hydrologic, or management conditions. A test for soil P sorption saturation may provide the most universal approximation, but only for noncalcareous soils.