One of the often-stated functions of wetlands is their ability to remove sediments and other particulates from water, thus improving water quality in the adjacent aquatic system. However, actual rates of suspended sediment removal have rarely been measured in freshwater wetland systems. To address this issue, suspended sediment dynamics were measured in a 85-km2 bottomland hardwood (BLH) wetland adjacent to the highly turbid Cache River in eastern Arkansas during the 1988-1990 water years. A suspended sediment mass balance was calculated using depth-integrated, flow-weighted daily measurements at wetland inflow and outflow points. Over the three-year period, suspended sediment load decreased an average of 14% between upstream and downstream sampling points. To test the idea that the suspended sediments were retained by the adjacent wetland and to determine what portion of the BLH forest was most responsible for retaining the suspended sediments, concurrent measurements of sediment accretion were made at 30 sites in the wetland using feldspar clay marker horizons, sedimentation disks, the 137cesium method, and dendrogeomorphic techniques. Sedimentation rates exceeding 1 cm/yr were measured in frequently flooded areas dominated by Nyssa aquatica and Taxodium distichum. Maximum sedimentation rates did not occur on the natural levee, as would be predicted by classical fluvial geomorphology, but in the "first bottom," where retention time of the water reached a maximum. Multiple regression was used to relate sedimentation rates with several physical and biological factors. A combination of distance from the river, flood duration, and tree basal area accounted for nearly 90% of the variation in sedimentation rates.