Do rare species have any place in multivariate analysis for bioassessment?

Cao et al. (2001) have recently published an assessment of the place of rare species in multivariate analyses associated with bioassessment. This assessment follows their earlier comments on this subject (Cao et al. 1998, Cao and Williams 1999) and criticism of their viewpoint (Marchant 1999). In their earlier comments Cao et al. (1998) claimed that elimination of rare species from data sets unduly biases the similarity measures used in multivariate analyses of community composition and results in an ‘‘unacceptable loss of ecological information’’. Marchant (1999) refuted this claim by noting that when rare taxa are eliminated from data sets the trends displayed by multivariate analyses are unaltered and that benthic data sets, both from freshwater and marine habitats, commonly contain much redundant information. If, as Cao et al. (1998) claimed, rare species are vital for multivariate analysis, then logically the less rare species and the common species must also be vital. In other words, all species are vital and, as Marchant (1999) noted, this amounts to saying that no redundancy exists. Such a claim is contrary to common experience. Cao et al. (2001) admitted that rare species may not be important when environmental gradients are strong. However, they suggested that rare species respond to different gradients, perhaps less important and subtle ones or those related to human impacts. They provided little evidence from the literature that such situations exist but they quoted a paper by Faith and Norris (1989) as supporting this view. Faith and Norris (1989) reanalyzed a study conducted in the La Trobe catchment of eastern Victoria, Australia (Metzeling et al. 1984, Marchant et al. 1985). Using all 269 taxa (mostly at genus or species level) Faith and Norris (1989) compared ordinations with and without rare taxa, as well as ordinations based solely on rare taxa (those with total abundances ,0.5% of the total abundance of all taxa). They showed that inclusion of rare taxa resulted in a greater number of sig-