Accurate critical levels must be accompanied by predictive models on the amount of P required to increase P availability to a target value to obtain reliable P recommendations. We estimated, based on information on soil properties, the increase in soil available P after the addition of a unit of P ( b coefficient) in an area of homogeneous although geographically distant soils. All 71 selected soils were noncalcareous, belonging to the Mollisol order and located in the Pampean Region (Argentina). Samples (0–20 cm) were incubated for 45 d after the addition of five doses of P. Soils were characterized for parameters related to soil P availability: pH, particle size distribution, organic C, total P, initial soil available P, and two P retention indices differing in the interaction period between the added P and the soil matrix (1 or 18 h). Obtained b values ranged from 0.27 to 0.74. Soils located at the northern area of the study region had higher b values (mean = 0.58) than southern ones (mean = 0.42). The best multiple regression model for estimating the b coefficient ( R 2 = 0.70) included initial soil P, a binary variable that accounts for the location of the soil, and the P retention index with the shorter interaction period. The strong effect of the geographic position of the soil on the estimation of the b coefficient led us to hypothesize about a possible role of the mineralogic characteristics of the parent material on P dynamics. Our results indicate that single independent variable models may not be enough to predict the b coefficient in homogeneous soils.