As part of the extensive field sampling programme within the European Union STAR project, replicate macroinvertebrate samples were taken using the STAR-AQEM sampling method at each of 2-13 sites of varying ecological quality within each of 15 stream types spread over 12 countries throughout Europe. The STAR-AQEM method requires the sub-sampling and taxonomic identification of at least one-sixth of the sample and at least 700 individuals. Replicate sub-samples were also taken at most of these sites. Sub-sampling effects caused more than 50% of the overall variance between replicate samples values for 12 of the 27 macroinvertebrate metrics analysed and was generally greatest for metrics that depend on the number of taxa present. The sampling precision of each metric was estimated by the overall replicate sampling variance as a percentage Psamp of the total variance in metric values within a stream type. Average over all stream types, the three Saprobic indices had the lowest percentage sampling variances with median values of only 3-6%. Most of the metrics had typical replicate sampling variances of 8-18% of the total variability within a stream type; this gives rise to estimated rates of mis-classifying sites to ecological status class of between 22 and 55% with an average of about 40%. This suggests that the precision of such metrics based on the STAR-AQEM method is only sufficient to indicate gross changes in the ecological status of sites, but there will be considerable uncertainty in the assignment of sites to adjacent status classes. These estimates can be used to provide information on the effects of STAR-AQEM sampling variation on the expected uncertainty in multi-metric assessments of the ecological status of sites in the same or similar stream types, where only one sample has been taken at a point in time and thus there is no replication.