Riparian buffers can be effective at removing phosphorus (P) in overland flow, but their influence on subsurface P loading is not well known. Phosphorus concentrations in the soil, soil solution, and shallow ground water of 16 paired cropland-buffer plots were characterized during 2004 and 2005. The sites were located at two private dairy farms in Central New York on silt and gravelly silt loams (Aeric Endoaqualfs, Fluvaquentic Endoaquepts, Fluvaquentic Eutrudepts, Glossaquic Hapludalfs, and Glossic Hapludalfs). It was hypothesized that P availability (sodium acetate extractable-P) and soil-landscape variability would affect P release to the soil solution and shallow ground water. Results showed that P availability tended to be greater in crop fields relative to paired buffer plots. Soil P was a good indicator of soil solution dissolved (<0.45 microm) molybdate-reactive P (DRP) concentrations among plots, but was not independently effective at predicting ground water DRP concentrations. Mean ground water DRP in corn fields ranged from < or =20 to 80 microg L(-1), with lower concentrations in hay and buffer plots. More imperfectly drained crop fields and buffers tended to have greater average DRP, particulate (> or =0.45 microm) reactive P (PRP), and dissolved unreactive P (DUP) concentrations in ground water. Soil organic matter and 50-cm depth soil solution DRP in buffers jointly explained 75% of the average buffer ground water DRP variability. Results suggest that buffers were relatively effective at reducing soil solution and shallow ground water DRP concentrations, but their impact on particulate and organic P in ground water was less clear.