Typology schemes are useful in partitioning natural spatial variability and thus helping implementing bioassessment methods. They have become widely accepted in Europe after the publication of the Water Framework Directive (WFD). The major objective of this study was to test if six Portuguese pre-defined physical stream-types (using System-B of the WFD) are concordant with macroinvertebrate assemblages collected in 31 undisturbed sites in a Portuguese river basin. A top-down (stream types defined by physical attributes) and a bottom-up approach (river classes defined by biological communities) were used. No significant differences were found in the mean number of taxa and abundance of different physical stream types and few taxa are characteristic of specific stream types. Discriminant analysis (DA) considering family-level composition revealed that all stream-types were significantly different and the cross-validation process showed that all stream-types had more than 50% of their samples correctly assigned. On the other hand, the bottom-up approach based on a correspondence analysis (CA) showed some overlap of the macroinvertebrate communities of pre-defined stream types, indicating that variations in the macroinvertebrate community structure was primarily related to altitude, mineralisation and a temporal gradient. DA cross-validation and CA results suggest that tested stream types do not account for natural temporal changes known to affect macroinvertebrate communities in this Mediterranean basin and that the WFD typology should account for these natural variations. The exclusion of natural variability could indicate impairment when it does not exist or no impairment when it does exist (type I and II statistical errors).