Mountaintop mining (MTM) affects chemical, physical, and hydrological properties of receiving streams, but the long-term consequences for fish-assemblage structure and function are poorly understood. We sampled stream fish assemblages using electrofishing techniques in MTM exposure sites and reference sites within the Guyandotte River basin, USA, during 2010–2011. We calculated indices of taxonomic diversity (species richness, abundance, Shannon diversity) and functional diversity (functional richness, functional evenness, functional divergence) to compare exposure and reference assemblages between seasons (spring and autumn) and across years (1999–2011). We based temporal comparisons on 2 sites that were sampled during 1999–2001 by Stauffer and Ferreri (2002). Exposure assemblages had lower taxonomic and functional diversity than reference assemblages or simulated assemblages that accounted for random variation. Differences in taxonomic composition between reference and exposure assemblages were associated with conductivity and aqueous Se concentrations. Exposure assemblages had fewer species, lower abundances, and less biomass than reference assemblages across years and seasons. Green Sunfish (Lepomis cyanellus) and Creek Chub (Semotilus atromaculatus) became numerically dominant in exposure assemblages over time because of their persistence and losses of other taxa. In contrast, species richness increased over time in reference assemblages, a result that may indicate recovery from drought. Mean individual biomass increased as fish density decreased and most obligate invertivores were apparently extirpated at MTM exposure sites. Effects of MTM were not related to physical-habitat conditions but were associated with water-quality variables, which may limit quality and availability of benthic macroinvertebrate prey. Simulations revealed effects of MTM that could not be attributed to random variation in fish assemblage structure.