Rapid bioassessments of stream health using macroinvertebrates are particularly useful when information is required quickly or when a large number of sites is to be investigated. The purpose of our study was to examine the effect of small-scale temporal variability (weeks) on rapid bioassessments using the New South Wales AUStralian RIVer Assessment System (AUSRIVAS) protocols. AUSRIVAS provides an assessment of stream condition based on the ratio of the number of taxa collected (observed) at a site, and the number predicted (expected) by a multivariate model. Repeated bioassessments were conducted in both riffle and edge habitats at 2 reference condition sites. Mean ratio of observed to expected number of taxa (O/E-Taxa value) was close to 1 for all sites and times, as expected for reference condition sites. O/E-Taxa values did not vary significantly over 4 to 6 wk, except for riffle samples from one site. Collecting 4 replicate samples increased the consistency of the allocation of sites to categories or bands of biological quality. However, the costs of further replication are likely to outweigh the benefits of using a rapid assessment method. Although tested using the AUSRIVAS rapid bioassessment protocol, these findings have application to other qualitative rapid bioassessment protocols.