Numerous studies have established strong linkages between acid deposition, soil and surface-water acidification, and toxicity to aquatic biota. Little is known however, about the effects of acidification on fish assemblages in headwater streams because they are highly variable, and pre-acidification data are often lacking. The primary purpose of this study was to describe spatial and interannual (temporal) variability of fish assemblages in headwater streams affected by acidification so that future recovery targets and monitoring strategies can be established. Fish communities and water chemistry were sampled at 48 headwater streams in the Western Adirondack Mountains of New York during the summers of 2014 to 2016 to characterize the present-day condition of water quality and local fish assemblages. Additionally, data from six Adirondack streams that were sampled annually from 2014-16 were combined with data from seven streams in the Catskill Mountain region sampled annually for three or more years for an analysis of temporal variability. Inorganic monomeric aluminum concentrations (Ali, the toxic form of Al) were less than 1.0 µmol L-1, between 1 and 2 µmol L-1, and greater than 2 µmol L-1 in 79%, 13%, and 8% of the 48 Adirondack streams. Richness, abundance, and biomass of fish assemblages were negatively related to Ali concentrations. In streams with Ali concentrations less than 1.0 µmol L-1, species richness, density, and biomass averaged 2.0 species, 444.2 fish/0.1 ha, and 1924.4 g/0.1 ha, respectively, and the density and biomass of Brook Trout populations averaged 280.8 fish/0.1 ha and 1384.0 g/0.1 ha. These values may provide a reasonable approximation of fish community condition prior to anthropogenic acidification and can be used as targets for assessing future recovery of acidified streams. A power analysis that considered 21 fish metrics indicated a strong negative relationship between interannual metric variability and statistical power for detecting change over time. Large differences were identified in the sample size necessary to achieve adequate power (0.8) depending on the metric utilized. In general, greater statistical power was obtained from metrics based on entire fish communities and from metrics standardized by reach length or sampling effort. Given the variability observed in our dataset, most metrics could detect a change of 30% with moderate effort, suggesting this may be an appropriate goal for future monitoring. Together, knowledge of biological recovery targets and the statistical power obtained from various fish metrics can be used to develop the most effective strategies for monitoring and assessing biological recovery in New York streams.