Methods of Classification

It will always be the case, in a study of such a large group as the bees, that some parts of the study are based on relatively recent investigations made with modern methods, whereas other parts, by necessity, are based on antiquated data and methods of analysis. The family-level taxa (including subfamilies and tribes) of some families have been analyzed using cladistic methods (Roig-Alsina and Michener, 1993; Alexander and Michener, 1995). Likewise, the genus- and species-level taxa of certain groups have been analyzed using morphology, molecular characters (usually DNA), or both. In some cases one's confidence in the phylogenetic results is limited, because the number of characters employed was small, as was the case in a cladistic analysis of the Meliponini (Michener, 1990a), or the methods were outdated, as in Michener's analysis ofthe Melittidae (Michener, 1981a), or the number of species analyzed was too small to represent the diversity of the group. More seriously, most groups have not yet received such treatment.

When a relevant phylogenetic analysis has been made, I have usually based classificatory decisions on that analysis. One must remember, however, that the results of such an analysis are hypotheses, not facts. Moreover, an analysis commonly yields many different cladograms. Authors usually offer reasons for preferring some of their clado-grams and rejecting others, but decisions necessary for making a classification are often arbitrary in spite of phy-logenetic analysis and various consensus and character-weighting methods. The use of additional characters, e.g., larval, molecular, or behavioral, or the introduction of new internal or external morphological characters, may clarify these matters. Remarkable changes in phylo-genetic results can arise simply from including additional taxa. But the classification one settles on at a given time must be made on the basis of the information then available. It is therefore often tentative.

I do not, in fact, believe that classifications should always be based exclusively on phylogenetic hypotheses. The proper functions of classifications include those that facilitate data storage and retrieval. Decisions based on the less reliable parts of a cladogram are likely to be corrected by later work, and thus run counter to these functions. I see nothing wrong with paraphyletic taxa as practical or temporary expedients, provided they are labeled as such and are justified by distinct differentiating characters, even if plesiomorphic. The Melittidae is such a "taxon of convenience," probably not monophyletic.

Users of phylogenetic information should always base their work on cladograms and associated information on the strength of each clade, not on often imperfect and usually incomplete summaries of phylogenies in the form of classifications. It is a disservice to users ofsystematic information to offer them classifications instead ofproperly documented cladograms. In nearly all cladograms some parts are supported by numerous strong characters and other parts are less certain; users should have such information at hand, so that they know which parts of a clado-gram are reliable and which are not. They cannot glean such information from classifications.

In preparing and trying to improve the classification of bees, I have found the information in cladograms very useful. An example concerns bumble bees. The genus Bombus consists of phenetically similar species; it has been broken up into many subgenera, but they are notoriously similar and difficult to separate. A preliminary cladistic analysis (Williams, 1985) indicated that the parasitic genus Psithyrus arose from within a paraphyletic Bombus, not from an ancestor of all Bombus. Williams therefore divided the Bombini into three genera showing the following phylogenetic relationships: Mendacibom-bus (Psithyrus, Bombus). Williams' decision was based on few characters; and Mendacibombus is similar to certain subgenera included in Bombus. If Psithyrus did not exist, one would not think of giving Mendacibombus generic rank. A derived group, Psithyrus, would dictate the classification of the group from which it arose. Recognition or identification of Bombus would be made difficult because one would have to separate it from Mendaci-bombus.

This particular problem became moot when more comprehensive phylogenetic analysis (Williams, 1994) showed that Mendacibombus was itself paraphyletic, and that Psithyrus is closest to other subgenera (Fig. 119-7). It seemed best, then, to place the whole clade, including Psithyrus, in the genus Bombus. I follow Williams (1994) in this decision because of morphological similarity and the apparent behavioral similarity to Psithyrus of parasitic species in two other mostly nonparasitic subgenera.

For the many groups that have never received any cladistic analysis, I have usually used the traditional taxa, largely phenetically based, and modified as seemed appropriate. Often it is possible to make an informed judgment about phylogeny, and such judgments influenced my taxonomic decisions. Ofcourse the classifications will be improved as more analyses are made.

Was this article helpful?

0 0

Post a comment