The ability of artificial marshes to polish high quality pulp mill effluents was assessed on a pilot scale using areas of 2.6 m2, 15 to 40 cm in depth and employing marl substrate of 1 to 3 cm in size. Cordgrass, cattail and reeds were planted and there were controls without vegetation. Initial experiments showed little advantage in exceeding a retention period of 15 h. Average percentage removals of suspended solids, BOD and organic nitrogen were 57, 38 and 29 respectively, with no advantages arising from the use of plants. Removals of ammonia and phosphate were greatly increased by plants. Only 16 per cent of ammonia was removed by the control compared with a range of 63 to 82 by the plants. They also reduced phosphorus by around 20 per cent with the control having virtually no effect. Shallow beds exhibited the greatest efficiency. Annual reed biomass production was estimated at 30 dry tons per acre. A 10 per cent loss of interstitial volume per year was noted. The process should be considered where the reduction of algal nutrients was important. Some factors affecting the design of full scale units are discussed.