Dams are ubiquitous in coastal regions and have altered stream habitats and the distribution and abundance of stream fishes in those habitats by disrupting hydrology, temperature regime and habitat connectivity. Dam removal is a common restoration tool, but often the response of the fish assemblage is not monitored rigorously. Sedgeunkedunk Stream, a small tributary to the Penobscot River (Maine, USA), has been the focus of a restoration effort that includes the removal of two low-head dams. In this study, we quantified fish assemblage metrics along a longitudinal gradient in Sedgeunkedunk Stream and also in a nearby reference stream. By establishing pre-removal baseline conditions and associated variability and the conditions and variability immediately following removal, we can characterize future changes in the system associated with dam removal. Over 2 years prior to dam removal, species richness and abundance in Sedgeunkedunk Stream were highest downstream of the lowest dam, lowest immediately upstream of that dam and intermediate farther upstream; patterns were similar in the reference stream. Although seasonal and annual variation in metrics within each site was substantial, the overall upstream-to-downstream pattern along the stream gradient was remarkably consistent prior to dam removal. Immediately after dam removal, we saw significant decreases in richness and abundance downstream of the former dam site and a corresponding increase in fish abundance upstream of the former dam site. No such changes occurred in reference sites. Our results show that by quantifying baseline conditions in a small stream before restoration, the effects of stream restoration efforts on fish assemblages can be monitored successfully. These data set the stage for the long-term assessment of Sedgeunkedunk Stream and provide a simple methodology for assessment in other restoration projects. Copyright © 2011 John Wiley & Sons, Ltd.