Phylogenetic reconstruction

The basic principles and methods of phylogenetic systematics have first been outlined by the German entomologist Willi Hennig (summarized in HENNIG, 1950, 1966). His ideas were readily picked up by anglosaxon zoologists who subsequently transformed them into a field

of research called "pattern cladistics" (e. g., NELSON & PLATNICK, 1981). The advent of powerful personal computers allowed them to analyze complex character matrices using newly developed computer programs. However, in Hennig´s homeland the recognition of the computer as a useful tool for phylogenetic analysis was delayed. The majority of German systematists adhered to Hennig´s methods of constructing cladograms "by hand"; largely, because of the suspicion that these programs represent yet another form of phenetics.

In the following both methods are briefly compared. A convenient summary of the current methods of pattern cladistics can be found in KITCHING et al. (1998). Below, some justification is given for the author´s preference for this latter approach. The specific settings for each analysis are outlined separately in the respective chapters as they vary somewhat among the different analyses.

If parsimony is accepted as the ultimate criterion for constructing phylogenies, the cladograms obtained by manual reconstruction and by the respective computer programs

(e. g. Hennig86 or PAUP) are identical. A simple test of these programs can show that they

are in fact cladistic and not phenetic. If characters are traced on a rooted cladogram their states are divided into plesiomorphic and apomorphic ones. However, the greatest difference between manual and computerized reconstruction concerns the method of how the characters are polarized. While traditional phylogenetists have to polarize the characters a

priori, a computer program allows to take this step a posteriori. At first, only unpolarized

networks are calculated treating the members of the ingroup and one or several outgroup representatives the same way. The resulting most parsimonious topology is then "rooted" by selecting one taxon as an outgroup. If more than one member of the outgroup was included in the analysis it serves as a convenient test of the monophyly of the ingroup. As another consequence of this procedure, characters of initially unknown polarity can be used in the analysis, too(MEIER, 1995).

A second important difference between manual and computerized reconstruction concerns the weighting of characters (HASZPRUNAR, 1998): while weighting is principally possible in a cladistic analysis, it is rarely done by pattern cladists. Their claim is to exclude the subjectivity involved in weighting. However, the decision if a character is included in an analysis, or excluded from it is just as subjective. Thus, character selection per se is a weighting processs. Traditional phylogenetists estimate the probability of homology of apomorphic states a priori using specific homology criteria. The tree is then constructed beginning with characters of the highest a priori probability of homology, subsequently adding those with lower probabilities. There are clear advantages of this method: the entire knowledge of the phylogenetist can be used when assigning probability values to potential apomorphies, including ecological data of the taxa involved. This helps to avoid being misled

by so called concerted homoplasies. These are nonrandom homoplasies of independent characters that evolve under certain constraints, such as specific habitat conditions.

It must be emphasized that the result of a manual reconstruction based on maximum parsimony and the outcome of a computer analysis will be the same if characters are weighted equally. This fact is obvious if the data set contains no character conflict. The manual reconstruction of cladograms by successively joining sister taxa is relatively easy under such favourable conditions. This changes when character conflicts occur. It was noted

by ARNOLD (1981) that parallelisms occur especially common in phylogenetic analyses of

species groups - the prime purpose of this study. The enormous number of theoretically possible cladograms is often overlooked. While there are only 15 possible cladograms for four taxa, there are 8,2 x 1021 cladograms for the still moderate number of 20 taxa

(FELSENSTEIN, 1978). If there is a number of homoplasies contained in the data set (which is

usually the case) it is almost impossible to choose the most parsimonious solutions out if this huge number of possible cladograms without the aid of a computer.

A frequent argument of traditional phylogenetists against the use of computer programmes based on parsimony is that the course of evolution does not need to be parsimonious. However, this concern rests on a misunderstanding of the reasons why to follow parsimony. Parsimony is simply the most robust criterion for choosing between competing scientific hypotheses. It is not a statement about how evolution may or may not have taken place (FARRIS, 1983; RIEPPEL, 1999). In fact, most of the traditional phylogenetists make use of parsimony when it concerns selecting the most plausible of various cladograms that had been reconstructed manually: this is stated most explicitly by AX (1984: p.67); it is also implicit in the auxiliary principle of HENNIG (1966), which demands that homology should be presumed in the absence of evidence to the contrary.

It must be remembered that every cladogram - be it manually reconstructed, or by the help of a computer programme - is nothing more than a hypothesis. A good scientific hypothesis should be comprehensible to others. Pattern cladists usually provide matrices with their reconstructions, something that can rarely be found along manual reconstructions. Character list and matrix allow not only a thorough examination; they also guarantee that the data can be updated and improved easily. Because of this transparency a greater heuristic value can be attributed to cladistic reconstructions than to manual ones. However, the quality of a cladogram depends most importantly on the quality of the characters and on the correct identification of homologies. This should not be forgotten when performing a phylogenetic analysis, be it a traditional one, or a computerized one.

3. Results

This chapter is intended as a synoptic treatment of the Euops species of the Papuan region. It covers the known species, i. e. those having been formally described in literature. The described species which have not yet been treated in my previous publications are herein redescribed in a way to conform with the new standards. The remaining undescribed species at hand are listed as "morphospecies". Under present constraints on both time and funding this was the only way to present an overall picture of the Papuan Euops fauna. I consider the use of "morphospecies" a provisionary arrangement which was unavoidable under present circumstances. See the "Introduction" for more comments on accurate taxonomy and on "morphospecies".