In order to standardise biological assessment of surface waters in Europe, a standardised method for sampling, sorting and identification of benthic macroinvertebrates in running waters was developed during the AQEM project. The AQEM method has proved to be relatively time-consuming. Hence, this study explored the consequences of a reduction in sample size on costs and bioassessment results. Macroinvertebrate samples were collected from six different streams: four streams located in the Netherlands and two in Slovakia. In each stream 20 sampling units were collected with a pond net (25×25 cm), over a length of approximately 25 cm per sampling unit, from one or two habitats dominantly present. With the collected data, the effect of increasing sample size on variability and accuracy was examined for six metrics and a multimetric index developed for the assessment of Dutch slow running streams. By collecting samples from separate habitats it was possible to examine whether the coefficient of variation (CV; measure of variability) and the mean relative deviation from the “reference” sample (MRD; measure of accuracy) for different metrics depended only on sample size, or also on the type of habitat sampled. Time spent on sample processing (sorting and identification) was recorded for samples from the Dutch streams to assess the implications of changes in sample size on the costs of sample processing. Accuracy of metric results increased and variability decreased with increasing sample size. Accuracy and variability varied depending on the habitat and the metric, hence sample size should be based on the specific habitats present in a stream and the metric(s) used for bioassessment. The AQEM sampling method prescribes a multihabitat sample of 5 m. Our results suggest that a sample size of less than 5 m is adequate to attain a CV and MRD of ≤ 10% for the metrics ASPT (Average Score per Taxon), Saprobic Index and type Aka+Lit+Psa (%) (the percentage of individuals with a preference for the akal, littoral and psammal). The metrics number of taxa, number of individuals and EPT-taxa (%) required a multihabitat sample size of more than 5 m to attain a CV and MRD of ≤ 10%. For the metrics number of individuals and number of taxa a multihabitat sample size of 5 m is not even adequate to attain a CV and MRD of ≤ 20%. Accuracy of the multimetric index for Dutch slow running streams can be increased from ≤ 20 to ≤ 10% with an increase in labour time of 2 h. Considering this low increase in costs and the possible implications of incorrect assessment results it is recommended to strive for this ≤ 10% accuracy. To achieve an accuracy of ≤ 10% a multihabitat sample of the four habitats studied in the Netherlands would require a sample size of 2.5 m and a labour time of 26 h (excluding identification of Oligochaeta and Diptera) or 38 h (including identification of Oligochaeta and Diptera).