Twentymile Creek, a sand-bed stream draining a 450 km2 catchment in northeast Mississippi, was channelized prior to 1910, in 1938, and in 1966. Straightening and enlargement in 1966 was followed by channel instability—rapid bed degradation (2-4 m) and cross-section enlargement by 1.4 to 2.7 times. Grade control structures (GCS) (weirs with stoneprotected stilling basins) and various types of streambank protection were constructed along the channel in the early 1980s to restore stability. Other investigators have suggested that habitat recovery in incised, channelized streams is facilitated by construction of GCS because they create stable scour holes and promote natural formation of a low-flow channel flanked by vegetated berms. Effects of restabilization of Twentymile Creek on aquatic habitats were assessed in four ways. The fraction of the bank line covered by woody vegetation was mapped from aerial photographs taken in 1981 and 1985; physical habitat (depth, velocity, substrate, and cover) and fishes were sampled at base flow; and the existence and size of a low-flow channel was ascertained from cross-section surveys taken in 1980 and 1989. Woody vegetation, physical aquatic habitat, and fishes were also sampled from Mubby-Chiwapa Creek, a similar-sized unstable channel with no GCS. Physical habitat variables and fishes were sampled concurrently at five stations on Twentymile Creek, and four stations on Mubby-Chiwapa. Four of the five Twentymile stations were either above or below a GCS. Bank-line woody vegetation cover increased 8 per cent between 1981 and 1985 along Twentymile Creek but was stable along Mubby-Chiwapa. Reaches above and below GCS were deeper with slower current velocities than elsewhere. Mean Shannon diversity indices based on physical data were similar for both streams, but were 58 per cent higher for stations immediately above and below GCS than for other stations. Since construction of the GCS and bank protection measures, longitudinal berms have formed within the enlarged Twentymile Creek channel, creating a low-flow channel. Low-flow channel capacity was equivalent to a mean daily discharge equalled or exceeded 30 per cent of the time, and was considerably lower than the effective discharge. Differences in aquatic habitat diversity among the stations sampled were primarily due to the scour holes below the GCS and the low-flow channel. Thirty-nine fish species were collected from Twentymile Creek, but only 22 from Mubby-Chiwapa. Fourteen species were collected exclusively at GCS. Principal component analyses of the abundance of the eight numerically dominant fish species indicated similar faunas at most stations, but Twentymile Creek GCS stations were faunistically distinct. Abundance of several of the numerically dominant species was positively influenced by greater depths and lower velocities found near Twentymile GCS. The mean fish diversity index for Twentymile Creek was 29 per cent higher than for Mubby-Chiwapa, and fish diversity was positively correlated with substrate diversity and mean depth.