Hydroelectricity is often presented as a clean and renewable energy source, but river impoundment, regulation and fragmentation caused by dams have been reported to have varying effects on aquatic biodiversity and ecosystem functions. The effects of river impoundment on fish are often difficult to isolate because of the presence of confounding factors such as stocking, fishing, species introduction and other human activities. In our study, we examined changes in littoral fish communities over 20 years, using a network of sites located in remote boreal ecosystems (northern Quebec, Canada) with minimal confounding pressures. We found little evidence of divergent temporal trends in contemporary diversity metrics in reservoirs relative to reference sites across three spatial scales (i.e., sampling station, reservoir and hydroelectric complex). Using β-diversity analyses, we detected a high degree of stability in fish composition over time and space at the complex and reservoir scales. However, at the scale of the sampling station, we observed higher rates of species turnover coincident with the time of reservoir filling and shortly after. Likewise, species assemblage shifts that correlated with time since impoundment were detectable only at the sampling station scale. Our work shows that examining community data at different scales is key when trying to understand the anthropogenic impacts on fish biodiversity, and in designing impact assessment studies. Overall, the isolated effect of hydroelectricity production in these remote boreal ecosystems caused little change in fish diversity but resulted in species assemblage shifts.