A pilot scale surface flow constructed wetland (~100 m²) in Truro Nova Scotia, Canada was loaded with dairy wastewater over a continuous 6 year period (November, 2000 through April, 2005). The wetland included both deep and shallow zones, and was planted with cattails. A long-term assessment of the phosphorous (P) treatment efficiency was performed. Mean daily loading rates for total phosphorus (TP) and soluble reactive phosphorus (SRP) were 1.5 ± 1.0 and 1.0 ± 0.9 kg P ha^-1 d^-1, respectively. The mean daily wastewater hydraulic loading rate was 2.5 ± 1.5 mm d^-1. Treatment efficiencies for both TP and SRP were strongly influenced by both mass and hydraulic loading. The majority of wetland outflow and effluent loads occurred during the spring and fall due to increased precipitation as well as ice and snowmelt. As the wetland aged the sensitivity to higher loading rates increased, while overall P treatment performance decreased. In the first year of operation monthly mass removals routinely exceeded 90% for both TP and SRP. In subsequent years mass and hydraulic loads were higher and a decrease in treatment performance coupled with an increase in treatment variability was observed. During the winter of 2005 wetland outflows were extremely high and mass reductions were negative indicating that the wetland was acting as a source of P. Linear regression analyses of monthly TP and SRP mass reductions and effluent hydraulic load (q_out) found mass reductions decreased as q_out increased. Mass reductions for both parameters were >50% when q_out was <100 mm month^-1. When q_out increased beyond 100 mm month^-1, mass reductions became much more variable making it difficult to predict system performance.