Four commonly used clustering methods (UPGMA, Ward Linkage, Complete L
inkage and TWINSPAN) were compared in their ability to recognise the s
tructure of three river macroinvertebrates datasets which were pre- de
termined based on habitat and biological characteristics or chemical w
ater quality of sampling sites. DCA, NMDS and ANOSIM were applied to t
he same datasets to provide further information about data structure,
and nonparametric tests were also undertaken on major chemical variabl
es to justify the predeterminations. The modified Rand Index was used
to measure the agreement between a particular solution and the pre-det
ermined classification. The results showed that Ward Linkage performed
best when its use was broadened and used with the CY Dissimilarity Me
asure, followed by TWINSPAN and Complete Linkage with UPGMA being leas
t successful. There was evidence to suggest that the effectiveness of
some clustering methods (e.g. UPGMA) may vary at different clustering
levels, and simulation techniques which have been used to assess clust
ering methods could leave some properties of clustering methods unexam
ined.