Natural temporal variation in biological assemblages is generally acknowledged, but rarely controlled for in existing bioassessment approaches. Many bioassessment methods ignore the variation of assemblage structure through time and, thus, lack a process to evaluate the effects of natural temporal variation on their outcomes. We used an 11-y data set from 24 near-pristine reference streams (validation sites) and a 5-y data set from 10 streams exposed to human disturbance (test sites) in northern Finland to examine: 1) the amount of interannual variation in macroinvertebrate communities in both types of streams measured as variation in taxonomic completeness (i.e., ratio of observed/expected taxa; O/E ratio), and 2) how temporal variability affects bioassessment outcomes. Most of the variation in O/E values was explained by stream type (validation or test). Interannual variation in O/E was low at both validation and test sites, a result supporting the assumption of high constancy of biological assemblages through time. Temporal variation in O/E at validation sites was related to climatic factors (regional precipitation) and to some local environmental variables, e.g., substratum heterogeneity. Communities in validation and test sites responded differentially to climatic variation. The low interannual variation in O/E suggests that 1-y data from reference sites can be extrapolated across years relatively safely, but climatically extreme years may be problematic. Temporal variability in O/E was low at both types of sites, but it was still high enough to result in variable ecological status assessments across years, mainly at test sites with lower O/E ratios. Thus, at test sites, use of only 1-y data may cause erroneous management decisions, and accurate assessment of the biological quality of test sites might require repeated sampling over several years.