Precise P fertilizer recommendations depend on the concentration of P removed by a given soil test extractant and the effect of other soil properties. Soybeans were grown several years on five Oxisols with a number of established soil P levels. A model was developed to maximize net return as a function of the P extracted by three methods and the optimum P fertilizer rate at each site. To account for site effects, other soil properties were then evaluated in individual combinations with extractable P by multiple regression. Mehlich‐3, Bray‐1, and resin extractable P accounted for 67, 57, and 62% of the variation in optimum P fertilizer rate, respectively. When clay content or surface area was included, the percentages increased to 84, 84, and 95%.In general, other soil properties reflecting P sorption, such as P adsorption maximum, P buffering coefficient, and measurements of clay type, were less effective than percentage clay or surface area. Two additional studies were conducted for one season to validate this approach. The results were excellent on one soil, but underestimated the optimum P rate on another. There was a difference in clay type between the two sites, indicating that perhaps the approach developed should be restricted to kaolinitic soils which predominate in the region