Agricultural runoff contributes with significant amounts of nitrogen (N) to rivers and lakes causing water quality problems. Constructed wetlands (CWs) in first- and second-order streams reduce downstream loading of nutrients through mechanisms such as sedimentation, uptake by vegetation and microbial denitrification. (A stream with no tributaries (headwater stream) is considered a first-order stream. A segment downstream of the confluence of two first-order streams is a second-order stream [1].) Norwegian CW's are often too small to easily achieve high N-retention. It is therefore important to improve and optimize the N-retention processes in the CW's where the CW area cannot be increased due to local restrictions. The main aim of this project was to compare N-retention in an experimental wetland including eight different types of organic and mineral CW-filters, one of which was a standard Norwegian CW (depth 0.5 m). The average total-N retention through the experimental wetland was 17% for the period of May-September in 2003, but only 2% in the same period in 2004. Converted to mass, the retention in 2003 was 168 kg and 26 kg in 2004. Lower retention in 2004 was probably mainly caused by higher hydraulic load that year. The organic filters performed better than the mineral filters and the standard CW.