We investigated the fish community approaching the Veazie Dam on the Penobscot River, Maine, prior to implementation of a major dam removal and river restoration project. Multibeam sonar (dual-frequency identification sonar, DIDSON) surveys were conducted continuously at the fishway entrance from May to July in 2011. A 5% subsample of DIDSON data contained 43 793 fish targets, the majority of which were of Excellent (15.7%) or Good (73.01%) observation quality. Excellent quality DIDSON targets (n = 6876) were apportioned by species using a Bayesian mixture model based on four known fork length distributions (river herring (alewife,Alosa psuedoharengus, and blueback herring, Alosa aestivalis), American shad, Alosa sapidissima) and two size classes (one sea-winter and multi-sea-winter) of Atlantic salmon (Salmo salar). 76.2% of targets were assigned to the American shad distribution; Atlantic salmon accounted for 15.64%, and river herring 8.16% of observed targets. Shad-sized (99.0%) and salmon-sized (99.3%) targets approached the fishway almost exclusively during the day, whereas river herring-sized targets were observed both during the day (51.1%) and at night (48.9%). This approach demonstrates how multibeam sonar imaging can be used to evaluate community composition and species-specific movement patterns in systems where there is little overlap in the length distributions of target species.