Soil nutrient availability is known to vary across landscapes. Consequently, variable rate fertilization is being studied to improve fertilizer use efficiency. The objective of this study was to examine the long term effects of variable phosphorus (P) fertilization and constant P fertilization on variably responsive sites within a field using a predictive response model. A 15.5 x 155.4 m cultivated area was divided into five transects each containing 50 3.1 x 3.1 m cells. Soil samples from the 250 cells were collected in the summer of 1996. Mehlich III soil test P (pp2m) ranged from 10 to 88, over the 250 cells. Data from P analysis was then used to determine the long‐term effects of both constant and variable rate application of P fertilizer to continuous wheat production. Projections from the data were made using a model that considers crop yield and response to fertilizer P at different soil test P levels. Soil test P values were adjusted following each annual yield projection using a relationship of one soil test P unit equal to 9 kg ha‐1input or removal of P. Annual application of 23 kg P ha‐1resulted in increased soil test P, however variability remained unchanged over a projected 50 years. Variable rates were identified as the P fertilizer level required to remove crop deficiency for winter wheat(Triticum aestiviumL.) based on an existing soil test calibration. Annual variable rate application of P fertilizer reduced variability, as measured by standard deviation, which reduced by 50% after 20 years. Seasonal marginal profit of variable rate application over a constant‐rate application increased as the field element size (resolution or area to which different rates are applied) decreased. Furthermore, as field element size decreased total P fertilizer used and yield increased for the entire area.