The impact of temperature on bacterial activity and community composition was investigated in arctic lakes and streams in northern Alaska. Aquatic bacterial communities incubated at different temperatures had different rates of production, as measured by 14C-leucine uptake, indicating that populations within the communities had different temperature optima. Samples from Toolik Lake inlet and outlet were collected at water temperatures of 14.2°C and 15.9°C, respectively, and subsamples incubated at temperatures ranging from 6°C to 20°C. After 5 days, productivity rates varied from 0.5 to ∼13.7 µg C l−1 day−1 and two distinct activity optima appeared at 12°C and 20°C. At these optima, activity was 2- to 11-fold higher than at other incubation temperatures. The presence of two temperature optima indicates psychrophilic and psychrotolerant bacteria dominate under different conditions. Community fingerprinting via denaturant gradient gel electrophoresis (DGGE) of 16S rRNA genes showed strong shifts in the composition of communities driven more by temperature than by differences in dissolved organic matter source; e.g. four and seven unique operational taxonomic units (OTUs) were found only at 2°C and 25°C, respectively, and not found at other incubation temperatures after 5 days. The impact of temperature on bacteria is complex, influencing both bacterial productivity and community composition. Path analysis of measurements of 24 streams and lakes sampled across a catchment 12 times in 4 years indicates variable timing and strength of correlation between temperature and bacterial production, possibly due to bacterial community differences between sites. As indicated by both field and laboratory experiments, shifts in dominant community members can occur on ecologically relevant time scales (days), and have important implications for understanding the relationship of bacterial diversity and function.