Denitrification was measured in two grass and two forest vegetated filter strips (VFS) in Rhode Island. The grass plots were established on a well‐drained soil and were planted to either tall fescue (Festuca arundinacea) or reed canarygrass (Phalaris arundinacea). One forest site was on an excessively well‐drained soil and was dominated by oak (Quercus sp.), and the other was on a poorly drained soil and was dominated by red maple (Acer rubrum). Denitrification was measured using soil cores under aerobic and anaerobic conditions with a range of treatments: no amendment, acetylene, water, nitrate (NO−3), NO−3 plus C. Unamended rates of denitrification were low in all plots. Nitrate and NO−3‐plus‐C amended rates were consistently higher in the grass plots than in the forest plots. Nitrate‐plus carbon‐amended rates were higher than NO−3‐amended rates in all plots, but the differences were significant (P < 0.05) in the forest plots only. Denitrification enzyme activity (DEA) was measured in 14 additional forest sites of varying natural drainage classes and was related to soil moisture (r2 = 0.56, P < 0.01) and pH (r2 = 0.43, P < 0.01) at these sites. The results suggest that the ability of VFS to support denitrification varies strongly with vegetation, soil type and pH, and that denitrification in VFS may be amenable to management.