Scientific Method, Opinion, Phylogenetic Reconstruction, and Nectar-feeding Bats: A Response to Griffiths and Warner

not ensure that their joint solutions do not interact. This is clearly an undesirable property that is not shared by methods that form groups based only on the spatial relationships of OTUs that are relatively close together. Classifications based on minimum spanning trees (e.g., by using the single-link clustering method) and the UPGMA clustering method are obvious examples of methods that do not have these-problems. ~he'problems discussed above may be part of the reason why a number of empirical studies (e.g., Sokal and Rohlf, 1981; Rohlf et al., 1983) have shown that phylogenetic classifications based on Wagner trees do not have the properties that many workers expected. l his is despite the various arguments for its theoretical advantage as a phylogenetic method. The theoretical advantages of classifications based on true phylogenetic trees may not be shared by classifications based on trees from particular statistical estimation procedures.