Surface water contamination can often be reduced by passing runoff water through perennial grass filters. Research was conducted in 2006 to 2008 to evaluate the size of cool season grass filters consisting primarily of tall fescue (Festuca arundinacea Schreb) with some orchard grass (Dactylis glomerata L.) relative to drainage area size in reducing runoff sediment and phosphorus (P). The soil was Pohocco silt loam Typic Eutrochrepts with a median slope of 5.5 %. The grass filters occupying 1.1 and 4.3 % of the plot area were compared with no filter with four replications. The filters were planted in the V-shaped plot outlets which were 3.7x11.0 m in size. The filter effect on sediment and P concentration was determined from four natural runoff events when nearly all plots had runoff. Filter effect on runoff volume and contaminant load was determined using total runoff and composites of samples collected from 12 runoff events. Sediment concentration was reduced by 25 % with filters compared with no filter (from 1.10 to 1.47 gL(-1)), but P concentration was not affected. The 1.1 and 4.3 % filters, respectively, compared with having no grass filter, reduced: runoff volume by 54 and 79 %; sediment load by 67 and 84 % (357 to 58 kgha(-1)); total P load by 68 and 76 % (0.58 to 0.14 kgha(-1)); particulate P (PP) load by 66 and 82 % (0.39 to 0.07 kgha(-1)); and dissolved reactive P (DRP) load by 73 and 66 % (0.2 to 0.07 kgha(-1)), respectfully. A snowmelt runoff event had 56 % greater DRP concentration compared with rainfall-induced runoff events. Grass filters reduced sediment and P load largely by reducing runoff volume rather than reducing concentration. Well-designed and well-placed grass filters that occupy 1.0 to 1.5 % of the drainage area and intercept a uniform flow of runoff from a drainage area can reduce sediment and nutrient loss in runoff by greater than 50 %.