Invited Review Blackwell Publishing Ltd. The scientific status of metazoan cladistics: why current research practice must change RONALD A. JENNER Accepted: 14 August 2003 Jenner, R. A. (2004). The scientific status of metazoan cladistics: why current research practice must change. — Zoologica Scripta, 33, 293– 310. Metazoan phylogenetics is bustling with activity. The use of comprehensive morphological data sets in recent phylogenetic analyses of the Metazoa indicates that morphological evidence continues to play a key role in the reconstruction of metazoan deep history. In this paper I review the scientific status of morphological metazoan cladistics from the perspective of cladistic research cycles. Each research cycle consists of three main steps: (1) the compilation of a data matrix (2) the simultaneous evaluation of all possible cladograms in a character congruence test, and (3) the assessment of the relationship between evidence and hypothesis after finding the optimal tree. I identify a striking discrepancy between the sophistication of the analysis of given data sets (Step 2), and their compilation and the interpretation of the results (Steps 1 and 3). The latter two steps deserve far greater attention than is current practice. Uncritical and nonexplicit character selection, character coding, and character scoring seriously compromise Step 1. Careful comparative morphological study prior to data matrix construction is necessary to remedy this problem in future cladistic analyses. Step 2 is the locus of most recent advances in metazoan cladistics through the increasing availability of computing power, and the development of increasingly efficient phylogenetic software that allows analysis of large data sets. Failure to identify problems and errors generated in Step 1 of the research cycle is testament to the general failure of Step 3. Consequently, recent progress in metazoan cladistics is primarily analytical, while the only empirical anchor of the discipline receives surprisingly little attention. Not surprisingly, the first generation of modern metazoan phylogeneticists used computers principally as a relatively quick and easy means to generate abundant phylogenies from morphological data. The next phase should build on this foundation by critically testing these alternative hypotheses by a thorough qualitative reassessment and elaboration of morphological data matrices, and a more critical approach to data selection. A rigorous research program for metazoan cladistics can only be established when the cladistic research cycle is properly completed, and when subsequent research cycles are effectively linked to previous efforts. Ronald A. Jenner, University Museum of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, UK. E-mail: [email protected] Introduction Whereas the first blossoming of phylogenetic zoology in the second half of the 19th century was boosted by a wholesale conceptual change (the advent of evolutionism; Bowler 1996), it appears that current enthusiasm for metazoan phylogenetics derives largely from technological advances. The ability to discover phylogenetic signal in molecular sequence data, the availability of increasingly powerful computers, and the development of efficient software suitable for the analysis of large molecular and morphological data sets has lifted metazoan phylogenetics to a level unprecedented in the history of biology. © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 The current state of metazoan cladistics stands in sharp contrast with that prevailing just 15 years ago. The first generation of molecular phylogeneticists was hard-pressed to find reliable morphology-based hypotheses with which to compare their findings. Typically they either resorted to presenting unreferenced ‘traditional views’, or they presented (or misrepresented) the alleged views of Libbie Hyman ( Jenner 2000). The new millennium began with an explosion of new phylogenetic analyses of the animal phyla (throughout this paper I use the term ‘phylum’ as a general descriptor of higher-level taxa without any Linnaean rank connotations). If only those studies are counted that analyse the entire Bilateria 293 Scientific status of metazoan cladistics • R. A. Jenner or Metazoa, using morphological data, no fewer than seven have been published over the last 3 years (Giribet et al. 2000; Sørensen et al. 2000; Nielsen 2001; Peterson & Eernisse 2001; Zrzavy et al. 2001; Zrzavy 2003; Brusca & Brusca 2003). If we also count analyses of smaller scope, purely molecular papers, and papers published from the late 1980s onward, we are confronted with a veritable forest of metazoan phylogenies. In addition, the phylogenetic relationships within many different phyla are currently being analysed as well. This makes the phylogenetic systematics of the Metazoa a very active research area. In contrast, far fewer papers have been expressly concerned with the methodology of morphological metazoan cladistics in order to identify the nature of progress in this field ( Jenner & Schram 1999; Jenner 2000, 2003), while some further methodological comments can be found in Zrzavy (2001) and Giribet (2002). The wave of publications in the new millennium invites another detailed look at the practice of metazoan cladistics. Consensus and controversy in metazoan cladistics The easiest way to assess the degree of consensus in metazoan phylogenetics is by assessing topological congruence among the results of different studies; this is how all recent papers sketch established consensus and identify remaining controversy. For example, in a recent review of metazoan phylogenetics Giribet (2002: 352) outlines the beginning of a ‘stable hypothesis’ of metazoan relationships that is purported to be in agreement with at least the most recent comprehensive analyses. The split of the Bilateria into a monophyletic Protostomia and Deuterostomia, the division of the protostomes into the clades Ecdysozoa and Lophotrochozoa, and the division of the latter into the clades Trochozoa and Platyzoa are nominated as the strongest areas of consensus (see also Zrzavy 2001, 2003). Indeed, a glance at the current literature leaves no doubt that these patterns of deep divergences within the Metazoa are generally accepted as current consensus. Recent phylogenetic studies also show consensus on a finer scale, especially if one considers only one type of evidence, for example morphological data. There is at least agreement among morphological cladistic analyses published over the last 3 years about the monophyly of the Cephalorhyncha sensu Nielsen (2001) (= Scalidophora sensu Lemburg 1995), which comprises the Kinorhyncha, Loricifera, and Priapulida (Giribet et al. 2000 retrieved a paraphyletic Cephalorhyncha, but excluded the loriciferans from the analysis). The monophyly of the Panarthropoda (Tardigrada, Onychophora, and Arthropoda) is also supported, as is the sister-group relationship of the Phoronida and Brachiopoda (Zrzavy et al. 2001 resolved the phoronids and the brachiopods as successive sister taxa to a clade of Echinodermata, Chordata, and Hemichordata). The monophyly of the Syndermata (the rotifers and acanthocephalans) is also supported. However, beyond 294 the monophyly of these clades, their positions within the Metazoa and placement of other phyla within the Bilateria are still very much uncertain. Taking molecular and total evidence analyses into consideration as well does not greatly clarify the picture as relationships within the Ecdysozoa, and even more so in the Lophotrochozoa, remain uncertain. In fact, I would argue that the previously mentioned consensus of even the deepest divergences within the Metazoa is much more tentative than recent review papers depict. First, the most recent cladistic analyses frequently disagree with respect both to the placement of single phyla and to the monophyly of larger supraphyletic clades. Topologies of molecular and total evidence analyses are sensitive to parameter changes and differences in taxon selection. This becomes especially clear when the total evidence analyses of Zrzavy et al. (1998) and Giribet et al. (2000) are compared. Although they employ the same morphological data set, differences in taxon sampling, molecular sequence alignment procedure and analysis parameters lead to a remarkable degree of incongruence. More generally, a comparison of the most recent cladistic analyses shows that the phylogenetic positions of several phyla are far from certain. These include the ectoprocts, chaetognaths, gastrotrichs, brachiopods + phoronids, entoprocts, cycliophorans, and platyhelminths, to mention just some of the most prominent examples. Second, the appearance of consensus concerning even some of the most basal animal divergences may be misleading. The sister-group relationship between the monophyletic deuterostomes and protostomes is heralded as one of the clearest signs of current consensus (Giribet 2002; Zrzavy 2003). However, various of the most comprehensive phylogenetic analyses of 18S rDNA and total evidence (using either maximum parsimony or maximum likelihood methods) show instead a paraphyletic Protostomia with the deuterostomes variously related to the lophotrochozoans, the ecdysozoans, or both (Peterson & Eernisse 2001; Jondelius et al. 2002; Zrzavy 2003). The fact that similar topologies with a paraphyletic Protostomia are found with other molecules as well, such as mitochondrial gene sequences ( Jondelius et al. 2002) or myosin II sequences (Ruiz-Trillo et al. 2002) indicate that the monophyly of the protostomes is not yet firmly established. These findings cannot simply be dismissed as artifacts of cladogram rooting because a paraphyletic Protostomia is found in analyses with widely accepted basal metazoan relationships, viz. the acoels (or acoelomorphs) as sister group to the remaining bilaterians, and the nonbilaterians as sister group(s) to the Bilateria. Furthermore, morphological study indicates that various characters that have previously been taken to suggest a fundamental division between the protostomes and the deuterostomes are unreliable. This holds true, for example, for the sources of mesoderm, the relationship between the larval apical organ and the adult brain, and the possession of cerebral ganglia (Jenner, in press). Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters R. A. Jenner • Scientific status of metazoan cladistics Third, and of particular relevance to phylogenies based upon morphological evidence, even identical cladogram topologies do not necessarily reflect an agreement of underlying data. For example, many comprehensive morphological cladistic analyses published during the last decade have yielded apparent support for a monophyletic Plathelminthomorpha (Platyhelminthes + Gnathostomulida) (e.g. Schram 1991; Eernisse et al. 1992; Zrzavy et al. 1998; Giribet et al. 2000; Peterson & Eernisse 2001). However, as discussed in detail in Jenner (in press), the morphological support for Plathelminthomorpha in these different studies may not even be in agreement due to problems in character scoring. Similarly, the morphological phylogenies of both Zrzavy et al. (1998) and Peterson & Eernisse (2001) supported the nesting of the acoels within the Platyhelminthes rather than at the base of the Bilateria as had been suggested before (Haszprunar 1996). However, the power of these analyses properly to test the position of the acoels within the Metazoa differed markedly as a result of differences in character selection (Jenner, in press). Consequently, the results of these studies cannot be meaningfully interpreted as support for the same conclusion. In the remainder of this paper I will focus on morphological cladistic analyses of the Metazoa. Morphological evidence continues to play a key role in the reconstruction of metazoan deep history, as is witnessed by the analysis of morphological data sets in all recent cladistic analyses of the animal phyla. Despite the fact that molecular data, in particular 18S rDNA sequences, have come to play a major role in metazoan phylogenetics, these data offer little reliable finescale resolution within the major bilaterian clades. Here morphology still has a lot to offer. In any case, reliably compiled morphological data sets provide invaluable information for studying the evolution of animal body plans, if only to flesh out a molecular phylogenetic skeleton. Specifically, I want to address the question of whether recent advances in metazoan cladistics are producing a deeper understanding of metazoan phylogeny and character evolution, or merely a stochastic change of opinion. It should be noted that several of the general points made in this paper have been elaborated elsewhere (Jenner & Schram 1999; Jenner 2000, 2001, 2002, 2003). This contribution integrates all aspects into a single coherent framework, that of the cladistic research cycle, that can function as the basis from which to reform current practice into an efficiently operating research program. The science of cladistics: research cycles and hypothesis testing The goal of cladistics is to gain an understanding of phylogeny and synapomorphies. To gain such knowledge, cladistic research proceeds through different steps that can be related as parts of a research cycle (Kluge 1998). Three steps may be © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 distinguished. The first involves the selection of terminal taxa and characters, which leads to the construction of the data matrix. This crucial step (hereafter referred to as Step 1) involves all the science of selecting monophyletic taxa and determining primary homologies, i.e. studying organismic variation, and identifying, coding, and scoring of characters and character states. When higher-level taxa are used as terminals, many difficult issues need to be addressed, such as how to deal with character variation within terminals, and how to infer or assume ground pattern states for characters. The second step (Step 2) is the simultaneous testing of all possible cladograms during the character congruence test (the cladogram building phase). This is for the most part an automated process involving computers and specialized software, and it results in one or more most parsimonious cladograms. These cladograms provide the best explanation of observed similarity as homologies that are the result of descent with modification. During the third step (Step 3) the relationship between evidence (data matrix) and hypothesis (cladogram) is assessed, and it is in this phase that an increased understanding of phylogeny and character evolution may be achieved. This stage may involve the performance of clade support analysis, the reappraisal of character state identity and character state polarity, or the reevaluation of homoplasies incongruent with the favoured hypothesis. It may also include the performance of a test of consilience with independent data, such as comparing the phylogenies of hosts and parasites, or phylogenies derived from molecular and morphological data sources (Lee & Doughty 1997; Kluge 1998). The conclusions reached in Step 3 may subsequently be fed into a new cladistic research cycle, for example as modifications of character coding and scoring. In this manner taxonomic relationships and character homologies can be further assessed, while increasing our understanding of the relationship between evidence and hypothesis. Because the results of one research cycle need not provide a complete understanding of relationships and homologies, they may be used as a starting point for an emended analysis. This is, of course, especially relevant for a highly diverse taxon such as the Metazoa, for which a consensus of relationships is still elusive. A healthy cladistic research program therefore consists of repeated rounds through the research cycle in which new analyses take previous efforts explicitly into account as the goal of research shifts from providing an intial phylogenetic estimate of a group to the effective testing of alternative phylogenetic hypotheses. One can test previously proposed hypotheses by introducing incongruent synapomorphies, adding additional taxa, or reappraising homologies (e.g. Kluge 1997; Rieppel & Kearney 2002; Jenner 2003, in press). In order to qualify as good science, cladistic studies have to pay due attention to all steps of the cladistic research cycle. This means that for Step 1 the decisions that feed into the 295 Scientific status of metazoan cladistics • R. A. Jenner construction of the data matrix need to be made explicit. This is done by explicitly justifying the criteria used to select characters and taxa, and by performing a rigorous study of comparative morphology to justify character coding and scoring. For taxa where one research cycle has already been performed, previous work needs to be explicitly incorporated into the new analysis, that is, older data sets need to be taken into account when compiling a new one. A detailed comparison of data selection and character coding and scoring are necessary ingredients. In Step 2 a proper computer-assisted phylogenetic analysis needs to be performed to find the optimal tree under the adopted optimality criterion (as in maximum parsimony), or under the adopted model of evolution (as in maximum likelihood). In Step 3 the robustness of the results needs to be assessed with a variety of methods, such as quantitative clade support measures, bootstrap analysis, parsimony jackknifing, and Bremer support analysis. However, this final step also provides the opportunity to explore the relationship between hypothesis and evidence from the viewpoint of a biologist. For example, we may analyse what the character state transformations tell us about body plan evolution, and we may reevaluate character state identity, character independence, and homoplasies. A survey of current practice in metazoan cladistics leads to a striking conclusion: in general, only Step 2 receives adequate attention. Many serious shortcomings of Steps 1 and 3 can be identified that cannot simply be explained as occasional errors. As I will argue in this paper, the practice of metazoan cladistics should be revised significantly by improving Steps 1 and 3 if we are to deepen our understanding of metazoan phylogeny and character evolution. Hypothesis testing and selection of characters and taxa It is an acknowledged dictum of cladistics that all known pertinent information should be included within an analysis, lest the results be unjustifiably biased towards certain conclusions. Discovering incongruent synapomorphies can refute cladistic hypotheses, and the severity of a cladistic test is therefore correlated with the empirical content of the data set because a larger data set contains a greater number of potential falsifiers of a given cladistic hypothesis (Kluge 1997; Siddall & Kluge 1997). In order to find the most corroborated and most severely tested cladogram, it thus becomes critical to establish ‘how honestly the relevant data are surveyed for those synapomorphies that actually have the potential to refute a cladistic hypothesis, those synapomorphies that can count as independent ad hoc hypotheses of homoplasy’ (Kluge 1998: 350). Strikingly, recent phylogenetic analyses of the Metazoa or Bilateria (for simplicity I will only refer to Metazoa hereafter) differ markedly in the number of included characters, and 296 consequently in the amount of organismic variation represented by their data matrices. For example, recent cladistic analyses have data sets ranging from 64 characters (Nielsen 2001) to 276 (Zrzavy et al. 1998). This difference indicates that analyses differ significantly in the power to test competing phylogenetic hypotheses. Unfortunately, although some studies justify the exclusion of several characters from their analyses, explicit justification for the inclusion and exclusion of all previously proposed and potentially informative characters is never given. How are we to choose between contending hypotheses based on different data matrices? With high empirical content as a desired epistemic value, it appears reasonable to use sheer size of the morphological data sets as one criterion to select among available cladistic hypotheses. Zrzavy et al. (1998) compiled the largest morphological data set to date, comprising 276 morphological characters. However, the quality of the included data leaves much to be desired ( Jenner 2001, 2002, 2003, in press). The myriad hidden details in which published metazoan cladistic analyses differ from each other prevent at this time any straightforward conclusion about the relative merit of the different cladogram topologies generated by these studies. To start with the logic of character selection in recent studies, I will discuss one example of how insufficient attention to the relationship between character selection and hypothesis testing has hampered our understanding of the phylogenetic position of a phylum. A case study of arbitrary character selection: placing Entoprocta The entoprocts are notoriously difficult to place in the animal kingdom (Nielsen 2000, 2002). A broad phylogenetic assignment, however, can relatively easily be made. They possess spiral quartet cleavage, mesoderm derived from a 4d mesentoblast, and trochophore larvae, which are characters that at least appear to indicate a close relationship with other spiralian protostomes such as the neotrochozoans (Mollusca, Echiura, Sipuncula, Annelida), and the nemerteans. This conclusion is compatible with all recent cladistic studies. However, if we want to be able to test different hypotheses proposed to explain the evolution of the entoproct body plan (the adult plan has, for example, been derived through loss of the coelom in a sessile ectoproct-like ancestor, or by paedomorphosis from a coelomate or noncoelomate motile ancestor: Bergström 1997; Cavalier-Smith 1998; Nielsen 2001) we need a better idea about the sister group of the entoprocts. Recent morphological phylogenetic studies have suggested various different candidates, including Ectoprocta (Nielsen 2001), Mollusca (Bartolomaeus 1993; Haszprunar 1996, 2000; Ax 1999), Cycliophora (Zrzavy et al. 1998; Sørensen et al. 2000; M. Obst, pers. comm.), Lobatocerebromorpha (Zrzavy et al. 2001), Rotifera + Gnathostomulida Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters R. A. Jenner • Scientific status of metazoan cladistics (Nielsen 2001), Neotrochozoa (Rouse 1999), or the entoprocts be placed in an unresolved position in a clade of varying membership including coelomate and noncoelomate protostomes (Giribet et al. 2000; Peterson & Eernisse 2001). So far, 18S rDNA sequence data have not been able to resolve the issue. The phylogenetic position of the entoprocts is highly sensitive to parameter variation and method of analysis (maximum parsimony or maximum likelihood) in molecular studies, and in molecular analyses entoprocts have been united with a variety of phyla or members thereof, including Nemertea, Annelida, Sipuncula, Echiura, Mollusca, and Ectoprocta ( Zrzavy et al. 1998, 2001; Giribet et al. 2000; Peterson & Eernisse 2001; Jondelius et al. 2002). Analyses of combined molecular and morphological evidence suggest that the entoprocts are at least closely related to the neotrochozoans and nemerteans, but their exact position with respect to these taxa is uncertain (Zrzavy et al. 1998, 2001; Giribet et al. 2000; Peterson & Eernisse 2001). If we focus on the most recent generation of comprehensive phylogenetic analyses (Giribet et al. 2000; Sørensen et al. 2000; Nielsen 2001; Peterson & Eernisse 2001; Zrzavy et al. 2001; Zrzavy 2003), can we determine whether these cladistic analyses are effective tests of competing hypotheses using the criterion of character selection? For example, let us consider whether the sister-group relationship of Entoprocta and Mollusca, which was proposed by Bartolomaeus (1993) and named Lacunifera by Ax (1999), and Sinusoida by Haszprunar (2000), is properly tested in the latest studies. Haszprunar (1996: 22) considered the existence of ‘quite an impressive number of possible synapomorphies between molluscs and kamptozoans [entoprocts]’ if both larval and adult morphology are taken into account. The six chief potential synapomorphies of the molluscs and the entoprocts suggested by this and several other studies are the presence of a dorsal chitin-containing proteinaceous cuticle, a pedal sole with anterior compound cilia, pedal glands, a glandular midgut, a similar position of the anus, and a circulatory system of sinusoidal spaces in a largely noncoelomate body (Bartolomaeus 1993; Haszprunar 1996, 2000; Ax 1999). These were all proposed in the context of manual or computerized cladistic analyses of only a subset of the animal phyla. Subsequent more comprehensive cladistic analyses are therefore ideally suited for testing the phylogenetic significance of these characters. It should be noted that the primary homology of several of these characters is uncertain at best, in particular the latter two, but here I want to focus only on the logic of character selection. Giribet et al. (2000) used the data set compiled by Zrzavy et al. (1998), which included most of the potential synapomorphies supporting Lacunifera, with the exception of a glandular midgut and a sinusoidal circulatory system. Although a character on glands in the midgut was only introduced in Haszprunar (2000), a study that was published after the © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 analysis of Zrzavy et al. (1998), the sinusoidal nature of the circulatory system was left out of the analysis without argument. Neither Zrzavy et al. (1998) nor Giribet et al. (2000) recovered a monophyletic Lacunifera. The analysis by Sørensen et al. (2000) omitted most potential lacuniferan synapomorphies: pedal sole, pedal gland, anus position, sinusoidal circulatory system, and glandular midgut. A monophyletic Lacunifera was not recovered by this cladistic analysis. The same conclusions hold true for the cladistic analyses of Nielsen (2001) and Peterson & Eernisse (2001). The cladistic analysis by Zrzavy et al. (2001) excluded characters coding for a pedal sole, anus position, glandular midgut, and sinusoidal circulatory system without explicit justification. A monophyletic Lacunifera was not found. The phylogenetic analysis by Zrzavy (2003) excluded characters coding for anus position and midgut glands, again without argument. The entoprocts were resolved as the sister group to the Cycliophora instead. The only potential lacuniferan synapomorphy included in all recent analyses is that coding for a chitinous cuticle. However, the exact coding and scoring may differ among studies so that the initial primary homology proposal may not be maintained. For example, Zrzavy et al. (2001) and Zrzavy (2003) code a character for the presence of a cuticle with α-chitin, while the characters defined by Bartolomaeus (1993) and Haszprunar (1996) disregard the specific chemical composition of chitin in the cuticle. This leads to different character scorings, and consequently, the phylogenetic significance of the originally defined character remains untested. This case study makes clear that recent phylogenetic studies have not selected characters with the express purpose of constructing an effective test of available hypotheses for the position of the entoprocts within the Metazoa. Or, if they have done so, the decisions underlying character selection remain entirely hidden. The arbitrary selection of characters becomes even more strikingly clear when we compare the choice of characters in subsequent analyses of the same research group. Zrzavy et al. (1998) identified nine potential synapomorphies for Entoprocta + Cycliophora which coded for such diverse features as anus position, frequency of asexual reproduction, nervous system morphology, fate of larval apical organ, and muscle cytology (characters 34, 73, 128, 130, 132, 148, 234, 261, 266). In contrast, Zrzavy et al. (2001) merely included two characters comparable to characters 148 and 234 of the first analysis, whereas Zrzavy (2003) only included character 266 as part of a more complex character. Why the other seven characters most relevant for placing the entoprocts in Zrzavy et al. (1998) were excluded from the analysis of Zrzavy et al. (2001), and why eight of the nine characters are excluded from the analysis of Zrzavy (2003) is totally unclear as they are neither mentioned nor discussed. Consequently, the change in the phylogenetic placement of the 297 Scientific status of metazoan cladistics • R. A. Jenner entoprocts and Cycliophora from Zrzavy et al. (1998) to Zrzavy et al. (2001) and Zrzavy (2003) does not reflect any increased understanding of phylogeny and synapomorphies. Moreover, most of the potential synapomorphies of Entoprocta + Cycliophora found by Zrzavy et al. (1998) were not included or explicitly rejected in other recent cladistic analyses such as Peterson & Eernisse (2001). The power of cladistics to test alternative phylogenetic hypotheses breaks down when potential synapomorphies incongruent with the obtained results are left out of the analysis without justification. The largely arbitrary selection of characters without apparent link to the testing of available phylogenetic hypotheses is by no means unique for the placement of Entoprocta. I discuss additional examples for different phyla in Jenner (2003, in press). In these studies I show that recent attempts at the cladistic placement of Acoelomorpha, Platyhelminthes, Myzostomida, Nemertea, and Gnathostomulida suffer from problems caused by unjustified character selection. A survey of the literature leaves no doubt that uncritical character selection contributes significantly to the existence of multiple competing hypotheses for the phylogenetic position of phyla ranging from the gastrotrichs and brachiopods to the arthropods and tardigrades. These findings are in agreement with the study of Poe & Wiens (2000), who concluded that phylogeneticists seldom make their criteria for character selection explicit. I consider uncritical selection of characters in recent phylogenetic analyses of the Metazoa as an important factor crippling the objectivity and testing efficacy of these studies. Of course, it would not be fair to criticize workers for not considering all possible evidence that could bear on a phylogenetic problem. However, good scholarship necessitates inclusion, or at least discussion, of available data treated in previous studies. In this sense, recent phylogenetic hypotheses are no better than the older precladistic scenarios that cladistics was intended to supersede. Only when criteria for character selection are made explicit with specific reference to the hypotheses to be tested can we hope to fully exploit the rigour of the cladistic approach to phylogeny reconstruction. Selection of taxa Generally, criteria for the selection of terminal taxa in analyses of metazoan cladistics vary considerably between different studies. Some authors do not justify the selection of taxa at all, such as Nielsen et al. (1996). Peterson & Eernisse (2001) simply stated that taxa such as Myzostomida and Lobatocerebromorpha, which had been included in previous analyses, were excluded from the analysis. However, no reasons were given. Others, such as Giribet et al. (2000) and Zrzavy et al. (2001), explicitly justify taxon selection with respect to the availability of molecular sequence information so that total evidence analyses may be carried out. Others, such as 298 Haszprunar (1996), base the exclusion of taxa on the difficulty of relating certain of their features to those in some of the included taxa. For example, Haszprunar (1996: 5) left the aschelminth phyla out of his analysis because their cleavage ‘is difficult to relate to typical spiralian development.’ He excluded other taxa, such as the arthropods, in order to ‘restrict the scope’ of the analysis, even though several of the included taxa were previously proposed to be related to the arthropods. The recent studies by Rouse (1999, 2000a,b) are a particularly clear example of problematic taxon selection (Jenner 2003). Rouse used a cladistic analysis to elucidate the evolution of trochophore larvae and the homology of larval ciliary bands in the protostomes. It is particularly interesting to note how Rouse treats taxa that lack trochophore larvae and ciliary bands. These potentially pertinent phyla that have previously been grouped with several of the included taxa are all excluded from the analysis. This is justified by reasoning that even if the excluded taxa were ‘placed among the taxa selected here for study, they will have either lost the features in question (e.g. a prototroch) or be primitively lacking them.’ But how can we uncover convergent evolution and character losses in this case? Obviously, such a procedure entirely removes any power of the cladistic analysis to test the homology and evolutionary dynamics of the features in question. Astonishingly, several recent authors (Nielsen 2001: 496; Zrzavy 2003: 66) have excluded taxa from their phylogenetic analyses because their relationships are uncertain! For example, Zrzavy (2003) excludes Xenoturbella from his analysis because its position is ‘still contentious.’ But what is phylogenetics all about, if not the testing of the phylogenetic placement of taxa? Unjustified differences in data selection in recent phylogenetic studies hinder progress in understanding, and foster the existence of multiple opinions of uncertain merit. Character coding and scoring: the quality of cladistic data matrices ‘In reality the most intractable barrier to settling disputes about systematic conclusions of a particular group of organisms usually centres on the coding of characters rather than methods of analysis’ (Forey & Kitching 2000: 55). Disagreement among workers on metazoan phylogenetics is rife on many topics, but one issue stands out as universally agreed upon: the difficulty of character coding. Character coding and scoring are the most challenging steps in the cladistic research cycle. Character coding in metazoan cladistics is generally perceived as posing ‘enormous problems’ (Nielsen 2001: 499). The chief difficulty is that different workers can, and do, perceive and define characters in different ways (e.g. Forey & Kitching 2000; Hawkins 2000). Importantly, different coding decisions, such as to use nonadditive binary (absence/ presence), multistate, or conventional (see Hawkings et al. 1997) codes, may result in different cladogram topologies Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters R. A. Jenner • Scientific status of metazoan cladistics (Rouse & Fauchald 1997; Hawkins et al. 1997; Hawkings 2000; Forey & Kitching 2000; Donoghue et al. 2000; Rouse 2001), and different coding strategies have their own specific strengths and weaknesses. Therefore, coding and scoring decisions need to be carefully explained. Because different coding strategies may produce different results, it may come as a surprise to learn that in recent studies of metazoan cladistics the most comprehensive and detailed justification for the choice of a coding method is merely this: ‘[we] acknowledge that these coding issues are contentious but feel that at the moment this [binary absence/presence coding] is the most conservative coding scheme available’ Peterson & Eernisse (2001: 173). Because the use of binary absence/presence coding in phylogenetic parsimony analysis has recently been strongly criticized, this short statement falls far short of a valid justification (Hawkins et al. 1997; Strong & Lipscomb 1999; Hawkins 2000; Forey & Kitching 2000; Jenner 2002). In this regard it is important to note that more than 92% of the total number of characters in the seven most recently published metazoan cladistic analyses (Giribet et al. 2000; Sørensen et al. 2000; Nielsen 2001; Peterson & Eernisse 2001; Zrzavy et al. 2001; Zrzavy 2003; Brusca & Brusca 2003) are coded as absence/presence characters without any justification. Similarly, character scoring typically receives equally short shrift. Discussion of the morphological data used to build the data matrix is typically relegated to a marginal summary in an appendix at the end of the paper. I consider this a rather peculiar attitude towards the only empirical lifeline of metazoan cladistics. Perhaps data matrix compilation is, after all, more straightforward than we initially imagined. Nothing could be further from the truth. Before I illustrate my concerns with a variety of examples, it is important to recognize the strong effect that single changes in a data matrix may have on the resulting phylogeny. One might argue that as long as the percentage of errors in a given matrix remains low, the cladistic analysis may not be fatally weakened. I think this claim cannot be upheld. The overall reliability of a cladistic analysis is not simply inversely correlated with the percentage of errors because a matrix does not contain a single phylogenetic signal. A data matrix is a mosaic of characters with distinct phylogenetic signals at different levels. The hierarchical structuring of phylogenetic signal implies that certain characters disproportionately determine the placement of certain taxa rather than others. Any errors in the set of features most important for placing a particular phylum (or supraphyletic clade), even when they constitute only a small percentage of the overall information content of the matrix, will mislead us about the relationships of these taxa. This is also to be expected given the frequently small number of robust synapomorphies that can be identified for uniting different phyla. Experiments in character selection, and character coding and scoring in different data matrices © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 confirm that simple changes in data sets may have far reaching consequences for the placement of individual taxa, clades, or even for the overall topology and resolution of the cladogram (e.g. Rouse & Fauchald 1997; Hawkins et al. 1997; Hawkings 2000; Forey & Kitching 2000; Donoghue et al. 2000; Rouse 2001; Jenner 2002, 2003, in press; Turbeville 2002). With this understanding, let us consider the current state of character coding and scoring in metazoan cladistics in more detail. Character analysis in metazoan cladistics: where is the comparative morphology? In a series of papers I identified many problems with the quality of character coding and scoring in recent analyses of metazoan cladistics ( Jenner 2001, 2002, 2003, in press). Rather than representing occasional lapses of judgement, most of the identified errors are symptomatic of a generally cavalier attitude towards character study. A major aim of future cladistic analyses of the Metazoa must therefore be the correction of the many errors through a more detailed and explicit approach to character study. To this end I critically evaluated the coding and scoring of more than 70 characters across all published morphological cladistic analyses of the Metazoa to identify the limits in our understanding of their phylogenetic significance (Jenner, in press). I also tried to correct misscorings, and to resolve the many scoring conflicts for characters that are shared between different studies. These efforts are necessary steps towards a foundation for a new cladistic analysis of metazoan relationships. It is a striking observation that none of the recent cladistic studies of the Metazoa comprehensively support all data matrix entries with source citations. The ground patterns that are scored in the published matrices are frequently very uncertain due to scarcity of observations within a terminal, variation of a feature within a terminal, or uncertainty about the internal phylogeny of the terminal and its bearing on the reconstructed or assumed ground pattern states. This does not become obvious from the unambiguous ‘0’s and ‘1’s that fill the data matrices. For example, the unambiguous scoring of protonephridia for phyla such as the Acanthocephala, Echiura, and Mollusca in recent cladistic analyses does not reflect the ambiguity of these ground patterns. Protonephridia have only been described for some archiacanthocephalans within Acanthocephala, the larva of Echiurus and the dwarf male of Bonellia within Echiura, and not for the basal molluscan taxa Solenogastres and Caudofoveata (Conway Morris & Crompton 1982; Schuchert 1990; Dunagan & Miller 1991; Bartolomaeus & Ax 1992; Salvini-Plawen & Steiner 1996; Rouse 1999; Haszprunar 2000). In extreme cases, as is shown below, the coding and scoring of characters are not rooted in observations, but instead embody unsupported a priori assumptions about character evolution. The requirement to cite the source for each character scoring would force researchers to consider 299 Scientific status of metazoan cladistics • R. A. Jenner carefully each data matrix entry, and supply useful information about the robustness of the evidence underlying each character scoring. This would constitute a marked improvement of current practice. Here I provide additional examples of incorrect data matrix entries that are clear illustrations that the study of comparative morphology should be significantly improved in future analyses of metazoan cladistics. The examples will focus on the study of Zrzavy (2003) because it represents the most recently published analysis of metazoan phylogeny and had maximum opportunity to scrutinize and incorporate evidence from previous analyses. Only several characters coding for the morphology of nervous systems will be discussed. In the following the numbers of the characters in question will be given in parentheses after their source. Zrzavy et al. (1998, 2001) (238, 46), Sørensen et al. (2000) (55), and Zrzavy (2003) (92) scored a dorsal brain or cerebral ganglion as present in the echinoderms. However, echinoderms lack a centralized ganglion altogether (Harrison & Chia 1994). A basic error such as this indicates that authors do not explicitly think about each data matrix entry. Even a casual check of a completed matrix would easily allow one to pinpoint such errors. Peterson & Eernisse (2001) (104) scored the tardigrades as lacking a circumoesophageal nerve ring. However, the tardigrades do have a circumoesophageal nerve ring (Dewel & Dewel 1996). Nevertheless, Zrzavy (2003) (95) copied this character, including this misscoring. In addition, a circumoesophageal nerve ring is present in many other taxa, including the nematodes, nematomorphs, gastrotrichs, priapulids, kinorhynchs, and loriciferans (Harrison & Ruppert 1991; Nielsen 2001), but neither Peterson & Eernisse (2001), nor Zrzavy (2003) score these taxa correctly. Notably, Zrzavy (2003) introduces logical conflict into his data set for these phyla as he scores them at the same time as having a circumoesophageal brain (93: collar-shaped circumpharyngeal brain), and as lacking it (95). Even though both Peterson & Eernisse (2001) and Zrzavy (2003) score the echinoderms and hemichordates as lacking a circumoesophageal nerve ring, they do have circumoesophageal nerves (Hyman 1955; Harrison & Chia 1994; Benito & Pardos 1997; Nielsen 2001). Zrzavy (2003) codes a character (96) for the presence or absence of a ventral longitudinal nerve cord. Sørensen et al. (2000) (54) and Nielsen (2001) (46) are cited as the sources for this character, and the three studies exhibit identical character scoring for the taxa that are shared between them and that possess a ventral nerve cord. However, the character definition adopted in these latter two studies is not identical to the one in Zrzavy (2003). Sørensen et al. (2000) and Nielsen (2001) present a more restricted character definition, viz. ventral nerve cords that are paired or secondarily fused during ontogeny. Consequently, the less restrictive 300 definition used in Zrzavy (2003) should be reflected in a revised scoring, which is not the case. Many phyla are misscored as lacking ventral nerve cords. These include the hemichordates (based upon enteropneust anatomy, which is considered to be representative of the hemichordate ground pattern by Zrzavy), cycliophorans (all motile stages of the life cycle have ventral nerve cords, Funch 1996; Funch & Kristensen 1997; Obst & Funch 2003), micrognathozoans (Kristensen & Funch 2000), and rotifers (Clément & Wurdak 1991). Moreover, if the homology between midventrally and somewhat more laterally placed longitudinal nerve cords is allowed (as is indicated by Zrzavy’s scoring of Lobatocerebrum, which has more laterally placed longitudinal nerve cords; Rieger 1980), then there is no reason to score phyla such as Acanthocephala, Platyhelminthes and Nemertea as lacking ventral nervous systems (Dunagan & Miller 1991; Rieger et al. 1991; Turbeville 1991; see Jenner, in press for additional discussion). A unique type of brain organization is present in the introvertans sensu Nielsen (2001) (= cycloneuralians sensu Ahlrichs 1995), which comprise the nematodes, nematomorphs, priapulids, kinorhynchs, and loriciferans. Their circumoesophageal brains are tripartite with anterior and posterior pericarya regions separated by a middle neuropil region (e.g. SchmidtRhaesa 1996, 1997/1998; Nielsen 2001). The exception is the brain of the nematomorphs, for which the only available evidence indicates an equal distribution of pericarya around the brain (Schmidt-Rhaesa 1996). Nevertheless, Zrzavy (2003) (94) scored the nematomorphs as ‘?’ for a tripartite brain. This scoring might be explained by Zrzavy’s assumption (p. 80) that the equally distributed pericarya in the nematomorphs may be derived from a tripartite ancestral state. However, rather than letting an unproven scenario of evolutionary change influence character scoring, instead the only available morphological evidence suggests the lack of tripartite brains in the nematomorphs. Peterson & Eernisse (2001) (102) incorrectly score the nematomorphs as possessing a tripartite brain. Zrzavy (2003) (93) scored the onychophorans as having a collar-shaped circumpharyngeal brain comparable to that characterizing the cycloneuralians sensu Nielsen (2001) (Priapulida, Kinorhyncha, Loricifera, Nematoda, Nematomorpha, and Gastrotricha). However, the tardigrades are scored as ‘?’ and the arthropods as lacking this brain configuration. Zrzavy (2003) cites the work of Eriksson & Budd (2000) in support of this scoring. Eriksson & Budd (2000) compared the onychophoran and cycloneuralian nervous systems because they believed that onychophorans provided evidence for the derivation of panarthropod nervous systems from less complex cycloneuralian precursors, consistent with the Ecdysozoa hypothesis (Dewel et al. 1999). However, the primary homologies scored in a cladistic data matrix should be strictly grounded in morphological study, and should not be infused Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters R. A. Jenner • Scientific status of metazoan cladistics with unsupported assumptions about evolutionary pathways. Strictly adopting morphological evidence reveals that the onychophoran anterior nervous system is not more similar to the cycloneuralian condition than that found in the arthropods, tardigrades, annelids, or other bilaterians with circumoesophageal nerve rings. Accordingly, this would necessitate the rescoring of the Onychophora. The character scoring for the presence of an orthogonal (ladder-like) nervous system in Zrzavy (2003) (97) shows an even greater divorce between comparative morphology and data matrix entry as is discussed in Jenner (in press). The scoring of an orthogonal nervous system as present in phyla such as Nematoda, Nematomorpha, and Priapulida can only be explained by assuming that Zrzavy based his scoring upon an unsupported theory of nervous system evolution that was accepted prior to data matrix construction. These phyla do not exhibit any signs of a ladder-like nervous system (Harrison & Ruppert 1991). The above examples clearly show the unwarranted influence of unsupported evolutionary assumptions, which creates a chasm between morphological evidence and data matrix entry. The diversity and complexity of metazoan morphology and development poses a huge challenge for systematic biologists. In order to use these data to reconstruct metazoan phylogeny we should try to record organismic variation as completely and as accurately as possible in our cladistic data matrices. Therefore detailed comparative morphological study is an absolute requirement for properly conducted cladistic research. Shortcuts are simply not available, and the potential sensitivity of phylogenetic results to small changes in a data matrix should inspire careful matrix construction. Instead, a surprising degree of authoritarianism characterizes many of the recent analyses in that opinions of other authors are frequently uncritically accepted. That is not necessarily a problem for certain unproblematic features, but the many difficulties that plague the scoring of ground pattern states for higher-level taxa make critical character evaluation essential for progress in cladistics. The foregoing examples leave no doubt that morphological scholarship must be considerably improved in studies of metazoan cladistics (Step 1). I will end this section by addressing one general aspect of character coding in metazoan cladistics that has so far not received any explicit attention at all: the definition of absence character states. Unspecified character states: the collapse of phylogenetic parsimony Sometimes a habit is so ingrained within a discipline that potential weaknesses become invisible. I believe this applies to an integral part of standard practice in metazoan cladistics: the coding of characters, and specifically the definition of absence character states. A full account of this issue is given © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 in Jenner (2002), so here I only summarize the most important conclusions, and provide some new examples. If we focus on the character coding observed in the seven most recently published metazoan cladistic analyses (Giribet et al. 2000; Sørensen et al. 2000; Nielsen 2001; Peterson & Eernisse 2001; Zrzavy et al. 2001; Zrzavy 2003; Brusca & Brusca 2003), we see that more than 96% of the 819 characters are coded as binary, with over 92% coded as binary absence/ presence (a / p). Only 35 are multistate. Strikingly, the absence states in more than 40% of the total number of a/p characters are not properly defined (up to more than 50% in individual studies). That is, taxa scored for the same unspecified absence state actually do not exhibit any special similarity in morphology or development. That means that the character state is empirically empty, and this conflicts with some major requirements of standard cladistic analysis. First, in standard cladistic analysis character states are alternative conditions (‘the same but different’) of the same thing (character), that can evolve or transform into each other. Consequently, careful morphological study is used to ensure that character states embody precisely defined hypotheses of primary homology. If only one state of binary a/p characters is defined, outgroups lose their function in polarizing character state transformations. This furthermore implies that informative character states are known prior to the analysis, which can only lead to circular reasoning. One of the two main goals of cladistic analysis is to reconstruct character evolution, and parsimony analysis provides a straightforward way to do this. A state shared between two or more terminal taxa reflects homologous similarity if optimization attributes that state to all nodes (stem species) lying between the terminals. However, when character states are not defined, cladistic analysis may teach us very little about body-plan evolution. For example, it is impossible to determine the exact morphology of taxa lacking the presence state for character four in Zrzavy (2003), which codes for an annelid cross. The annelid cross is a supposedly unique configuration of cleavage blastomeres at a certain stage during the spiral cleavage of the annelids and echiurans. The phyla scored as ‘absent’ for an annelid cross in Zrzavy (2003) are a mix of animals for which the cleavage pattern is: (a) unknown (placozoans); (b) known, but not spiral and may be very different between the scored taxa (from acoels with duet cleavage to phoronids with radial cleavage); (c) spiral, but apparently not showing an annelid cross (molluscs, nemerteans) (Henry et al. 2000; Nielsen 2001). The same holds true for character 5, which codes for a molluscan cross (see Jenner 2003, for in-depth discussion of cleavage cross patterns). The coding and scoring for these taxa of the same character state scarcely reflects any shared similarity. To code variations of cleavage geometries either as parts of a multistate character, or to adopt conventional coding with the proper use of ‘inapplicability’ scoring 301 Scientific status of metazoan cladistics • R. A. Jenner (Hawkins et al. 1997; Hawkins 2000), would allow one to remove unspecified character states and resulting meaningless character state transformations. Second, phylogenetic parsimony analysis permits evolution between character states in both directions (Farris et al. 1970; Farris 1983; Omland 1999). This assumption of symmetrical transformation allows for character reversals to be phylogenetically informative. The failure to delimit absence character states in binary a / p characters a priori denies the value of reversals, and may thus lead to meaningless transformations. For example, character 97 in Zrzavy (2003) codes for the absence or presence of an orthogonal (ladder-like) nervous system. One of the unambiguous synapomorphies that unites the phylogenetically problematic brachiopods and phoronids with the deuterostomes is the reversal of this character. However, this character reverses to a meaningless unspecified absence state. The disparate nervous system configurations in the sessile brachiopods and phoronids, and in the free-living echinoderms, hemichordates and chordates, share no features that convincingly indicate their unique homology as an alternative to an orthogonal nervous system. This logical inconsistency is especially pertinent under the widely used ACCTRAN (accelerated transformation) character optimization, which favours reversals over convergence. Failure to properly specify all states within a character eliminates the prime objective of coding information: to translate original data into a coded format as faithfully as possible without loss or gross distortion of information. This is as much true for the coding of cladistic characters as for the writing of codes for computer programs. With knowledge of the coding process it should be possible to recreate the original data from the coded information. When character state identity is not properly specified the decoding process becomes impossible. That is, the original data are not accurately represented by the coded data, and accordingly can no longer be retrieved from the coded data set. To resolve these problems, characters will have to be recoded and / or rescored, for example either by properly using inapplicability scoring if binary a / p coding is preferred, or by recoding as multistate. Because all available coding strategies have their own strengths and weaknesses, and because not all phylogenetic software can deal with all coding methods, it may not be possible to nominate one method as superior to all others. However, nonadditive binary coding is the most problematic method for standard parsimony analysis, and its dominance in recent analyses should therefore be reassessed. Other methods are available, and should be explored in future studies. One can, of course, use available programs to analyse improperly coded characters, but that does not necessarily generate meaningful results. 302 Completing the cladistic research cycle: qualitative assessment of phylogenetic hypotheses After the compiled data sets are subjected to a character congruence test, it is essential to evaluate the resulting phylogenetic hypotheses in terms of both topology and supporting synapomorphies. During Step 3 of the cladistic research cycle the relationship between evidence (data matrix) and hypothesis (cladogram) is assessed, and it is therefore only in this phase that an increased understanding of phylogeny and character evolution is achieved (Kluge 1998). In published morphological cladistic analyses of the Metazoa it typically consists of comparing the topology of the most parsimonious cladogram(s) with that of previously published cladograms, and clade support is only sometimes (Giribet et al. 2000; Peterson & Eernisse 2001) assessed in terms of quantitative support measures such as bootstrap percentages or Bremer support values. Additionally, some authors may formalize the established clades with new names, but unfortunately, this is where published analyses terminate the analytical process. A qualitative study of the character state transformations that underlie the clades is not carried out, and if conflicting topologies between studies are identified, no attempt is made to determine the cause of this difference. As is shown above in relation to character selection, new topologies are simply accepted, even when character selection has biased the results. The succession of recent attempts to place the phylogenetically problematic myzostomids in the Metazoa is emblematic of this problem. Their fluctuating placement in different studies can be precisely correlated with differences in character selection in different studies (Jenner 2003). That in itself may not be a problem if character selection is explicitly justified on the basis of detailed character study. However, that is not the case. For example, the traditional idea that the myzostomids are highly modified polychaetes seems to have been refuted by the recent analyses of Zrzavy et al. (2001) and Zrzavy (2003). However, this conclusion cannot hold because neither of these two studies included several important characters uniquely shared between (certain) polychaetes and the myzostomids, including a nectochaete larva (postmetamorphic juvenile with long chaetae), aciculae (internalized chaetae functioning as support rods inside the parapodia), and cirri (assumed sensory organs) with a characteristic pattern of innervation (see Jenner 2003 for references). Clearly, biased character selection cripples the testing power of these two studies. In addition, recent studies make no attempt to evaluate their cladograms in terms of character state transformations. The most telltale sign of this problem is the prevalence of unspecified character states in recent analyses. If only one of the character states of a binary character is defined it is impossible to meaningfully interpret steps on a cladogram. Studying character changes on a cladogram makes this clear Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters R. A. Jenner • Scientific status of metazoan cladistics at once. The resulting breakdown of phylogenetic parsimony is most clearly brought home by the existence of several character reversals to unspecified states in the studies of Giribet et al. (2000), Peterson & Eernisse (2001), Nielsen (2001), and Zrzavy (2003). The value of character reappraisal after a cladistic analysis is easily illustrated. For example, it could help provide a better understanding of the homoplasies on a preferred cladogram. In the section on character coding above I discussed character 96 in Zrzavy (2003), which codes for the absence or presence of a ventral longitudinal nerve cord. Optimizing this character on either the morphological or total evidence topologies of Zrzavy (2003) unambiguously indicates two independent origins of ventral nerve cords in two clades of protostomes. This is a potentially significant result because the presence of ventral nerve cords has been regarded as one of the most reliable characteristics of the protostomes (Nielsen 1994, 2001). We may wonder whether this pattern of character evolution is trustworthy. Separating the two protostome clades characterized by ventral nerve cords are four phyla that are scored for the plesiomorphic state, viz. absence of the cords. These are the cycliophorans, micrognathozoans, ectoprocts, and entoprocts. The first two taxa are misscored as they do have ventral nerve cords (see above), and rescoring them changes character optimization such that ventral nerve cords are now unambiguously homologous throughout the protostomes. Furthermore, the scoring of the same character state for such phyla as the sessile ectoprocts and entoprocts, and the cnidarians and chordates badly misrepresents the comparative morphology of their nervous systems. Consequently, reanalysis of this homoplastic character shows that the favoured optimization is based upon incorrect character scoring and coding, and the convergent evolution of ventral nerve cords is therefore not supported. In conclusion, by not properly completing the research cycle, recent cladistic analyses of the Metazoa forego a precious opportunity to gain a deeper understanding of the phylogeny of animals and the evolution of their body plans. Strengthening and elaborating morphological data sets Future progress in metazoan cladistics is dependent upon the consolidation and elaboration of available morphological data sets. First, the continual reevaluation of existing data sets in the light of new information is an essential ingredient of effective cladistic research cycles (Kluge 1998). In view of the less than optimal quality of prevailing data matrices, this approach should take centre stage in future work. Second, the continuing search for new characters may add valuable phylogenetic information to our data sets, which will allow more effective tests of competing hypotheses. The ongoing development of new techniques to study animal © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 morphology and development allows older data to be rechecked. For example, studies based on light microscopy can now be reassessed with the use of more powerful electron microscopical methods, and the tracing of embryonic cell lineages performed with specific blastomere injections which allow much more reliable fate maps to be drawn (e.g. Freeman & Martindale 2002; Gerberding et al. 2002). It is encouraging to note that recent cladistic studies such as Peterson & Eernisse (2001), Zrzavy et al. (2001), and Zrzavy (2003) are incorporating new information, such as the absence or presence of particular Hox genes, differences in the mitochondrial genetic code, and expression patterns of developmental genes. Furthermore, new methods to visualize the details of the early development of organ systems, such as the muscular and nervous system, generate valuable information of potential phylogenetic significance (e.g. Friedrich et al. 2002; Wanninger & Haszprunar 2002). One area where such studies have particular significance is in the assessment of problematic absence characters that may reflect secondary loss rather than primitive absence (Purschke et al. 2000). For example, adult echiurans and sipunculans are unsegmented coelomate worms, and are undoubtedly related to the molluscs and annelids. However, so far their exact relationships remain contentious. The absence of obvious segmentation in these two phyla could indicate a primitive absence, or potentially a secondary loss. Recent immunohistochemical investigations of nervous system development in the echiurans and sipunculans indicated the metameric organization of nerve cells along the ventral nerve cords, and this could be taken to indicate that they are derived from segmented ancestors (Hessling 2002; Hessling & Westheide 2002; E Edsinger Gonzales, pers. comm.). Similarly, the myzostomids have long been considered a phylogenetic problematicum (Zrzavy et al. 2001). A recent immunohistochemical study of their nervous system development (Müller & Westheide 2000) revealed detailed similarities with the nervous system of polychaetes, which is in agreement with other available morphological evidence that suggest an annelidan affinity (Jenner 2003). Nevertheless, the significance of new morphological and developmental studies can only be exploited when new cladistic analyses take new information into account. Unfortunately, a worrying degree of authoritarianism and historical burden remains detectable in recent cladistic analyses. Entire data sets, or parts thereof, that are compiled by other authors may be taken over into new analyses without attention to new information, and earlier interpretations may be uncritically accepted by later workers (Jenner 2001, 2003, in press). For example, Kristensen (1991) showed that traditional reports of the presence of a spacious pseudocoel in Kinorhyncha were erroneous, and were instead the results of fixation artifacts. Yet phylogenetic studies subsequently published over more than a decade failed to incorporate this new information 303 Scientific status of metazoan cladistics • R. A. Jenner (Eernisse et al. 1992; Wallace et al. 1996; Zrzavy et al. 1998; Brusca & Brusca 2003). Admittedly, the literature on animal morphology and development is intimidatingly large, but we should try our best to reflect available evidence in our data matrices as accurately as possible. The quest for new phylogenetic data is not easy. Apart from being very time consuming I think that the most important reason for this difficulty is that we find it difficult to adopt a broad focus. We may find a distinctive similarity in a few phyla, but our a priori preconceptions about their relationships may lead us to believe, perhaps unconsciously, that the similarity is convergent, and further study is thus discouraged. However, in such cases a comprehensive comparative study could yield useful information. For example, Ruppert (1991) noted that the gastrotrichs show a distinctive type of neuromuscular contact in which the muscle cells send extensions to the nerves, a situation claimed to be similar to that found in the cephalochordates, nematodes, and nemerteans. A preliminary survey of the literature shows that the entoprocts and nematomorphs exhibit similar neuromuscular contacts. This shows that a comprehensive focus is necessary if the aim is to elucidate the phylogenetic significance of a character. That, however, is not always done. Examples of not carrying out a sufficiently wide-ranging comparative study after finding a potentially promising character can be found throughout recent data matrices. Noteworthy examples are related to clades that were recently proposed, or recently received a new name. For example, Zrzavy et al. (2001) proposed the name Prosomastigozoa to house the problematic myzostomids mentioned above. They were placed in a new clade together with Cycliophora and Syndermata. Prosomastigozoa refers to the supposed unique possession of sperm with flagella that attach anteriorly to the sperm body, and which curve posteriad. However, similar sperm have been described for the chaetognaths and phoronids, as well as for the aberrant echinoderm Xyloplax, but this information was not mentioned in Zrzavy et al. (2001) or Zrzavy (2003) (Alvario 1983; Rowe et al. 1991; Herrmann 1997). In similar vein, Ax (1999) proposed the new name Lacunifera for a clade composed of Entoprocta + Mollusca (see also section on character selection), which was initially proposed by Bartolomaeus (1993). Haszprunar (1996) suggested the presence of a circulatory system composed of sinuses and lacunae as a unique synapomorphy of the molluscs and entoprocts. This conclusion was supported by a later study, in which the clade was christened Sinusoida (Haszprunar 2000). However, the studies of Haszprunar only included a restricted set of phyla, and the subsequent study of Zrzavy (2003) included both this character (character 76) and a much broader sampling of taxa so that the true phylogenetic significance of a lacunar circulatory system could now be tested. Unfortunately, Zrzavy evidently did not look beyond the morphology of the molluscs and entoprocts. 304 A lacunar system is composed of widened interstitial spaces between the cells and organs that fill most of the space between the gut and the epidermis. It may be contrasted with a circulatory system with well-defined vessels that are typically located between the basement membranes of epithelia, for example between coelomic epithelia in the mesenteries of coelomate animals, or between the coelomic epithelium and the gut epithelium. These distinctions are grounded in ultrastructural data, but they have been referred to for a long time by their textbook names, viz. ‘open’ or ‘closed’ circulatory systems (Hyman 1951). Any system of interstitial spaces between internal organs not directly bounded by the basement membranes of epithelia answers the definition of a lacunar system. Therefore one can label all interstitial spaces in noncoelomate bilaterians, as well as particular regions of the body in coelomate animal lacunar systems. Yet Zrzavy (2003) only scores a lacunar system present in the molluscs and the entoprocts, and although both Haszprunar (1996) and Zrzavy (2003) included Lobatocerebrum in their analyses, neither of them properly scored it as having a lacunar system, which was clearly described in the first paper on this taxon (Rieger 1980). In conclusion, I believe that no matter how hard we try to be open minded, preconceived notions of what we think could be important and what is not, will always affect our choice of characters and taxa, consciously or not. To make the study of character selection, coding, and scoring explicit is therefore a first prerequisite of a scientific approach to cladistic analysis. Especially when a character shows a distribution that conflicts with a previously proposed phylogeny, it is likely to be excluded from the analysis. The effect of such a priori constraints on data selection are clearly shown in recent debates about the phylogenetic position of the turtles and the snakes (Lee 1995; Lee & Doughty 1997, 1998; Rieppel & Reisz 1999). Here, important breakthroughs in the understanding of the evolution of snake and turtle body plans are correlated with the inclusion of previously ignored characters and taxa into new cladistic analyses. The vigour of cladistics as a falsificationist research program is inextricably linked to finding incongruent characters (Kluge 1997). If available disconfirming evidence is excluded from cladistic analyses without justification, cladistics loses its testing power. The search for characters incongruent with proposed hypotheses will help progress in cladistic research as much as finding confirming evidence of established hypotheses. Metazoan cladistics as a research program: reconsidering current research strategies Cladistic analysis is a powerful method to tap the extensive body of morphological and developmental information about the Metazoa that has been compiled over more than a century. Molecular systematics is a major driving force of recent progress, but robust morphological data sets can offer a lot of Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters R. A. Jenner • Scientific status of metazoan cladistics additional information, especially for resolving relationships within the large clades of the Bilateria where molecular signal is weaker. Obviously, if we want to understand the dynamics of animal body-plan evolution at all levels of the biological hierarchy, then a reliably compiled morphological data set is invaluable. To conclude, let us summarize recent progress in metazoan cladistics for each of the steps in the cladistic research cycle. To start with Step 2, the automated testing of all possible cladograms for a given data set, there is no doubt that this is where progress in metazoan cladistics is the most visible. The availability of increasing computing power, and the continuing development of increasingly efficient phylogenetic software make phylogenetic analysis of large morphological, molecular, or combined data sets possible, and these are obviously positive developments. Technological advances have allowed metazoan phylogenetics to become a blossoming discipline at the heart of modern integrative biology. However, although technological advances have repeatedly proven to be positive driving forces of scientific progress throughout the history of science, this relationship hides an inherent danger. For example, Lewontin (1991) called the development and widespread adoption of gel electrophoresis in evolutionary genetics from the mid-1960s onward both a ‘milestone’ and a ‘millstone.’ ‘The introduction of protein gel electrophoresis as a tool to investigate the standing variation within and between species almost totally depauperized evolutionary genetics for 20 years’ (Lewontin 2002: 7). The enormous outpouring of data based on gel electrophoretic studies did not, however, resolve the many problems that earlier research in population biology dealt with. This illustrates that the uncritical embrace of a technological advance may lead to an unwarranted ‘find ’em and grind ’em’ mentality, yielding plentiful results, the significance of which may remain uncertain at best. Recently, concern has been aired about whether the increasing use of computers in different fields of biological research may be diverting our attention away from attempts to really understand the generated data (Allen 2001). Likewise, we should be vigilant not to overemphasize Step 2 at the expense of the other essential steps of the cladistic research cycle. Computers have assisted greatly in harvesting phylogenetic signal from available data sets, yielding many phylogenetic hypotheses. However, it is now obvious that we need to redirect our attention to the primary data of our trade if we are to continue the momentum of this discipline. Computers are an aid to thinking; they should not replace it. Step 1, or all the science that goes into the construction of a cladistic data matrix could benefit from much more explicit attention. As pointed out above, the selection, coding, and scoring of characters are never explicitly justified, or discussed in any comprehensive way. As an unfortunate result, we can see that even the most recent morphological and total evidence © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 studies exhibit many disagreements; at this time it is simply impossible to make a reasoned choice between these competing hypotheses. For that, we first need to address the many problems identified here and elsewhere. Likewise, Step 3, or the assessment of the relationship between evidence and hypothesis after the most optimal tree(s) is found, is hardly visible at all in recent studies. As a result, many problematic character state transformations remain hidden, thus removing the logical foundation upon which cladistic analysis rests. I think it is a striking finding that exactly those steps of cladistic research that require the greatest intellectual effort are the ones that are in the most pressing need of explicit attention. The late Sir Peter Medawar had the following general insight: ‘the scientific paper is a fraud in the sense that it does give a totally misleading narrative of the processes of thought that go into the making of scientific discoveries’ (Medawar 1996: 38). This characterization is particularly apt for recent papers on metazoan cladistics, where most steps that involve thought and decision making are neatly concealed in the black box of the data matrix. The empirical data underlying the analysis are typically presented only within a short table or appendix. This attitude towards the only empirical data of our trade is simply incomprehensible. Because we all agree on the methods used in Step 2, all prevailing disagreements have their origins in differences residing in the data sets that we employ. Therefore we have no choice but to explicitly confront all the difficulties of Steps 1 and 3 if we are to make any real progress. A cladistic research program can be visualized as a succession of research cycles. The goal of the first research cycle is to generate an initial phylogenetic estimate of a group of organisms. Because at this stage we are not yet confronted with competing hypotheses, the first rotation through the research cycle is a straightforward process. However, as we enter into subsequent research cycles, the process becomes more difficult, for now we have to take previous efforts explicitly into account. That is, at this moment the role of cladistic analysis shifts from proposing a phylogeny, to the testing of alternative hypotheses. Cladistics is particularly suited for the testing of alternative phylogenetic hypotheses, but testing can only be efficiently achieved if care is taken in the compilation of the data set. Data must be selected very carefully with respect to the hypotheses to be tested so as not to bias the results of the analysis, and the differences between previous and new data matrices have to be carefully compared. Therefore I think that the transition between the first and second research cycles is the most crucial point in a cladistic research program, and so far we have not yet fully completed this transition. Encouragingly, there is evidence of progress in that recent studies identify and resolve some previous scoring mistakes, and weed out some bad characters, although this essential 305 Scientific status of metazoan cladistics • R. A. Jenner process is never the explicit focus. I think that current efforts constitute a more or less parallel series of first cycles that often generate conflicting results, and which are not efficiently linked to each other even though the analyses have been published in sequence over more than a decade. This situation is perhaps not too surprising as computers fostered the quick and easy exploration of multiple data sets. Now we need to refocus on the primary data of our trade, which is unfortunately a much more time-consuming process. The published studies show a plethora of differences that are not made explicit. Consequently, it becomes virtually impossible to separate the arrow of true progress in understanding from random change of opinion. However, we should not be discouraged. The incomplete research cycles offer plentiful frayed ends to start a new round of research. Now is the time to pick them up and tie them together into a powerful research program. Seeds for revision I here provide a barebones recipe that may be helpful for improving future cladistic analyses of the Metazoa. For the three steps of the cladistic research cycle I minimally propose the following: Step1 1 Select terminal taxa so that all previously proposed sister groups of all the included terminals are included. Any exclusions from this minimal set should be explicitly justified. Selection should also be considered with respect to the domain of definition of each included character. I adopt a modified version of the domain of definition as proposed by Dayrat & Tillier (2000): all taxa for which the character is applicable. A failure to fulfil the domain of definition of a character, i.e. restrictive taxon sampling, compromises the reliability of the phylogenetic significance ascribed to that character, so that it may be a symplesiomorphy or homoplasy rather than a unique synapomorphy. 2 Select all characters that have previously been proposed in support of all alternative sister-group hypotheses for all included terminals. Any exclusion from this set of characters should be explicitly justified. In addition, new characters may be included. 3 Explicitly define all characters and character states so that character state transformations can be meaningfully interpreted on the resulting cladograms. The nature and number of character states within a given character will have to be based upon a careful comparative study of morphological variation. The source of each data matrix entry should be supplied. Step 2 The computer-assisted simultaneous assessment of all possible relationships between the selected terminals will yield one or 306 more most parsimonious cladograms for the given data set. These provide the best explanation of similarities as due to common descent with modification with a minimal requirement for ad hoc explanations of homoplasy. Step 3 1 Perform quantitative clade support analysis (e.g. bootstrap and Bremer). 2 Provide qualitative evaluation of clade support, such as the reappraisal of character state identity and character state transformations, and reevaluation of homoplasies. A sensitivity analysis can be performed in which contentious character codings and scorings are changed to assess the effect of different input assumptions on the result of the cladistic analysis. The sensitivity analysis is an appropriate point to start a new cladistic research cycle. Epilogue: lessons from history, from sociobiology to evolutionary morphology A comparison of current practice in metazoan cladistics with the operation of several other evolutionary disciplines during the history of biology can teach us something important about the scientific status of recent efforts. The mainstay of all science, including evolutionary science, is testability. The efficient operation of a falsificationist research program depends on the formulation of alternative hypotheses, which can be tested against each other by the performance of an experiment or study with alternative possible outcomes. However, it is not enough to merely state the possibility of testing. For cladistic research this means that all proposed sistergroup hypotheses of the taxa under consideration have to be identified, and then a data matrix that includes all available potentially corroborating and disconfirming evidence for all these hypotheses has to be compiled. The simultaneous assessment of all these data will then yield the best-supported phylogeny. Another strategy that is sometimes used in science is verificationism. At the heart of the verificationist approach is the use of evidence only to make a persuasive case for a particular hypothesis. Wilson (2000: 28) called it the ‘advocacy method of developing science’ as it depends upon the careful selection of only that evidence that corroborates a particular hypothesis. For example, in sociobiology myriad theories have been proposed to explain the evolutionary origins of many aspects of social behaviour in different animal groups. By focusing on different restricted sets of evidence, authors have come to radically different conclusions, without it becoming clear which of the proposed hypotheses is a superior summary of available evidence. At its worst this practice amounts to nothing more than speculative storytelling, in which different ‘schools’ of thought proceed to talk past each other. Consequently Wilson (2000: 28) claimed that ‘the greatest snare in Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters R. A. Jenner • Scientific status of metazoan cladistics sociobiological reasoning is the ease with which it is conducted’ and concluded that ‘sociobiologists of the past have lost control by their failure to discriminate properly among the schemes.’ In similar vein, Gould & Lewontin (1979) took issue with the prevalence of uncritical adaptive storytelling in evolutionary biology. They pointed out that often the sole criterion for accepting an adaptive explanation for the origin of a given organismic feature was the mere plausibility of the hypothesis. The general weakness of such reasoning identified by Wilson, Gould, and Lewontin is that no attempt is made to make a clear distinction between the possible or plausible alternative hypotheses. Too often, workers are willing to uncritically accept a hypothesis erected on the basis of only part of the available evidence. Current practice of metazoan cladistics shows some remarkable similarities to early sociobiological reasoning and adaptive storytelling. The first generation of metazoan cladistics saw the erection of many alternative phylogenetic hypotheses as independent research groups analysed data sets which differed in many details that were not made explicit, notably character selection. Subsequent analyses were not expressly designed to test previous efforts, creating a forest of phylogenies of uncertain merit. Of course, I do not imply that workers consciously massage their data to generate preferred results, but an unconsciously introduced bias in input data has the same result. This is akin to early sociobiological and adaptive scenarios, which were proposed on the basis of part of the available evidence, and without the intention to critically test alternative views. A cladistic hypothesis based on an uncritically compiled data set is just one possibility among many other plausible outcomes. Only critical simultaneous assessment of all pertinent information will allow one to test alternative phylogenies. Moreover, the uncritical attitude towards character coding and scoring seriously undermines the quality of our data matrices. If this practice does not change, we are in danger of recapitulating the fate of the first generation of phylogeneticists. The reconstruction of large-scale phylogenies was the focus of the first evolutionary biologists in the period immediately following the publication of Darwin’s On the Origin of Species. However, ‘by the last decade of the nineteenth century, their approach was already seen as a wasted effort, since the evidence stubbornly refused to admit clear evaluation of the various schemes which had been proposed’ (Bowler 1996: 443). Nyhart (1995, 2002) concluded that the first phylogenetic zoology largely dissolved from within because the conflicting results from comparative anatomy and embryology could not be reconciled. The evolutionary morphologists were simply unable ‘to account satisfactorily for contradictory evidence’ (Nyhart 1995: 276). The result was a marked decline in the attraction of promising young investigators to the phylogenetic enterprise. To prevent the same situation © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 recurring today, we have to abandon the use of cladistics as an easy tool to generate ‘new’ hypotheses and instead employ it more critically in the testing of existing ones. In the end, we may not be able to fully resolve all relationships within the Metazoa, but only by a proper use of cladistics will we be able to tell whether we have pushed the morphological evidence to its limits. Acknowledgements I want to thank Eric Edsinger Gonzales for kindly sharing unpublished information about nervous system development in sipunculans, and Matthias Obst for discussing cycliophorans. Drs Malte Ebach, Ariel Chipman, Max Telford and an anonymous reviewer provided incisive comments on the manuscript. This work is supported by a Marie Curie Individual Fellowship of the European Community program ‘Improving Human Potential’ under contract number HPMF-CT2002–01712. References Ahlrichs, W. H. (1995). Ultrastruktur und Phylogenie von Seison Nebaliae (Grube 1859) und Seison Annulatus (Claus 1876). Hypothesen zu Phylogenetischen Verwandtschaftsverhältnissen Innerhalb der Bilateria. Göttingen: Cuvillier Verlag. Allen, J. F. (2001). Bioinformatics and discovery: induction beckons again. Bioessays, 23, 104 –107. Alvario, A. (1983). Chaetognatha. In K. G. Adiyodi & R. G. Adiyodi (Eds) Reproductive Biology of Invertebrates (pp. 531–544). Chichester: John Wiley & Sons. Ax, P. (1999). Das System der Metazoa II. Ein Lehrbuch der Phylogenetischen Systematik. Stuttgart: Gustav Fischer Verlag. Bartolomaeus, T. (1993). Die Leibeshöhlenverhältnisse und Verwandtschaftsbeziehungen der Spiralia. Verhandlungen der Deutschen Zoolologische Gesellschaft, 86, 42. Bartolomaeus, T. & Ax, P. (1992). Protonephridia and metanephridia — their relation within Bilateria. Zeitschrift für Zoologische Systematik und Evolutionsforschung, 30, 21– 45. Benito, J. & Pardos, F. (1997). Hemichordata. In F. W. Harrison & E. E. Ruppert (Eds) Microscopic Anatomy of Invertebrates, Vol. 15. Hemichordata, Chaetognatha, and the Invertebrate Chordates (pp. 15 –101). New York: Wiley-Liss. Bergström, J. (1997). Origin of high-rank groups of organisms. Paleontological Research, 1, 1–14. Bowler, P. J. (1996). Life’s Splendid Drama. Evolutionary Biology and the Reconstruction of Life’s Ancestry, 1860–1940. Chicago: The University of Chicago Press. Brusca, R. C. & Brusca, G. J. (2003). Invertebrates, 2nd edn. Sunderland, MA: Sinauer Associates. Cavalier-Smith, T. (1998). A revised six-kingdom system of life. Biological Reviews of the Cambridge Philosophical Society, 73, 203–266. Clément, P. & Wurdak, E. (1991). Rotifera. In F. W. Harrison & E. E. Ruppert (Eds) Microscopic Anatomy of Invertebrates, Vol. 4. Aschelminthes (pp. 219–297). New York: Wiley-Liss. Conway Morris, S. & Crompton, D. W. T. (1982). The origins and evolution of the Acanthocephala. Biological Reviews of the Cambridge Philosophical Society, 57, 85–115. 307 Scientific status of metazoan cladistics • R. A. Jenner Dayrat, B. & Tillier, S. (2000). Taxon sampling, character sampling and systematics: how gradist presuppositions created additional ganglia in gastropod eutyneuran taxa. Zoological Journal of the Linnean Society, 129, 403– 418. Dewel, R. A., Budd, G. E., Castano, D. F. & Dewel, W. C. (1999). The organization of the subesophageal nervous system in tardigrades: insights into the evolution of the arthropod hypostome and tritocerebrum. Zoologischer Anzeiger, 238, 191–203. Dewel, R. A. & Dewel, W. C. (1996). The brain of Echiniscus viridissimus Peterfi, 1956 (Heterotardigrada): a key to understanding the phylogenetic position of tardigrades and the evolution of the arthropod head. Zoological Journal of the Linnean Society, 116, 35– 49. Donoghue, P. C. J., Forey, P. L. & Aldridge, R. J. (2000). Conodont affinity and chordate phylogeny. Biological Reviews of the Cambridge Philosophical Society, 75, 191–251. Dunagan, T. T. & Miller, D. M. (1991). Acanthocephala. In F. W. Harrison & E. E. Ruppert (Eds) Microscopic Anatomy of Invertebrates, Vol. 4. Aschelminthes (pp. 299 –332). New York: Wiley-Liss. Eernisse, D. J., Albert, J. S. & Anderson, F. E. (1992). Annelida and Arthropoda are not sister taxa: a phylogenetic analysis of spiralian metazoan morphology. Systematic Biology, 41, 305 –330. Eriksson, B. J. & Budd, G. E. (2000). Onychophoran cephalic nerves and their bearing on our understanding of head segmentation and stem-group evolution of Arthropoda. Arthropod Structure and Development, 29, 197–209. Farris, J. S. (1983). The logical basis of phylogenetic analysis. In N. I. Platnick & V. A. Vunk (Eds) Advances in Cladistics, Vol. 2 (pp. 7–36). New York: Columbia University Press. Farris, J. S., Kluge, A. G. & Eckardt, M. J. (1970). A numerical approach to phylogenetic systematics. Systematic Zoology, 19, 172– 191. Forey, P. L. & Kitching, I. J. (2000). Experiments in coding multistate characters. In R. Scotland & R. T. Pennington (Eds) Homology and Systematics. Coding Characters for Phylogenetic Analysis (pp. 54 – 80). London: Taylor & Francis. Freeman, G. & Martindale, M. Q. (2002). The origin of mesoderm in phoronids. Developmental Biology, 252, 301–311. Friedrich, S., Wanninger, A., Brückner, M. & Haszprunar, G. (2002). Neurogenesis in the mossy chiton, Mopalia muscosa (Gould) (Polyplacophora): evidence against molluscan metamerism. Journal of Morphology, 253, 109–117. Funch, P. (1996). The chordoid larva of Symbion pandora (Cycliophora) is a modified trochophore. Journal of Morphology, 230, 231–263. Funch, P. & Kristensen, R. M. (1997). Cycliophora. In F. W. Harrison & R. M. Woollacott (Eds) Microscopic Anatomy of Invertebrates, Vol. 13. Lophophorates, Entoprocta, and Cycliophora (pp. 409 – 474). New York: Wiley-Liss. Gerberding, M., Browne, W. E. & Patel, N. H. (2002). Cell lineage analysis of the amphipod crustacean Parhyale hawaiensis reveals an early restriction of cell fates. Development, 129, 5789 –5801. Giribet, G. (2002). Current advances in the phylogenetic reconstruction of metazoan evolution. A new paradigm for the Cambrian explosion? Molecular Phylogenetics and Evolution, 24, 345 –357. Giribet, G., Distel, D. L., Polz, M., Sterrer, W. & Wheeler, W. C. (2000). Triploblastic relationships with emphasis on the acoelomates and the position of Gnathostomulida, Cycliophora, Plathelminthes, and Chaetognatha: a combined approach of 18S rDNA sequences and morphology. Systematic Biology, 49, 539 –562. 308 Gould, S. J. & Lewontin, R. C. (1979). The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proceedings of the Royal Society of London B, 205, 581–598. Harrison, F. W. & Chia, F.-S. (1994). Microscopic Anatomy of Invertebrates, Vol. 14. Echinodermata. New York: Wiley-Liss. Harrison, F. W. & Ruppert, E. E. (1991). Microscopic Anatomy of Invertebrates, Vol. 4. Aschelminthes. New York: Wiley-Liss. Haszprunar, G. (1996). The Mollusca: coelomate turbellarians or mesenchymate annelids?. In J. D. Taylor (Ed.) Origin and Evolutionary Radiation of the Mollusca (pp. 1–28). Oxford: Oxford University Press. Haszprunar, G. (2000). Is the Aplacophora monophyletic? A cladistic point of view. American Malacological Bulletin, 15, 115 –130. Hawkins, J. A. (2000). A survey of primary homology assessment: different botanists perceive and define characters in different ways. In R. Scotland & R. T. Pennington (Eds) Homology and Systematics. Coding Characters for Phylogenetic Analysis (pp. 22–53). London: Taylor & Francis. Hawkins, J. A., Hughes, C. E. & Scotland, R. W. (1997). Primary homology assessment, characters and character states. Cladistics, 13, 275 –283. Henry, J. Q., Martindale, M. Q. & Boyer, B. C. (2000). The unique developmental program of the acoel flatworm, Neochildia fusca. Developmental Biology, 220, 285 –295. Herrmann, K. (1997). Phoronida. In F. W. Harrison & R. M. Woollacott (Eds) Microscopic Anatomy of Invertebrates, Vol. 13. Lophophorates, Entoprocta, and Cycliophora (pp. 207–235). New York: Wiley-Liss. Hessling, R. (2002). Metameric organisation of the nervous system in developmental stages of Urechis caupo (Echiura) and its phylogenetic implications. Zoomorphology, 121, 221–234. Hessling, R. & Westheide, W. (2002). Are Echiura derived from a segmented ancestor? Immunocytochemical analysis of the nervou system in developmental stages of Bonellia viridis. Journal of Morphology, 252, 100 –113. Hyman, L. H. (1951). The Invertebrates, Vol. II. Platyhelminthes and Rhynchocoela. The Acoelomate Bilateria. New York: McGraw-Hill. Hyman, L. H. (1955). The Invertebrates, Vol. IV. Echinodermata. New York: McGraw-Hill. Jenner, R. A. (2000). Evolution of animal body plans: the role of metazoan phylogeny at the interface between pattern and process. Evolution and Development, 2, 208 –221. Jenner, R. A. (2001). Bilaterian phylogeny and uncritical recycling of morphological data sets. Systematic Biology, 50, 730 –742. Jenner, R. A. (2002). Boolean logic and character state identity: pitfalls of character coding in metazoan cladistics. Contributions to Zoology, 71, 67–91. Jenner, R. A. (2003). Unleashing the force of cladistics? Metazoan phylogenetics and hypothesis testing. Integrative and Comparative Biology, 43, 207–218. Jenner, R. A. (in press). Towards a phylogeny of the Metazoa: evaluating alternative phylogenetic positions of Platyhelminthes, Nemertea, and Gnathostomulida, with a critical reappraisal of cladistic characters. Contributions to Zoology, 72, in press. Jenner, R. A. & Schram, F. R. (1999). The grand game of metazoan phylogeny: rules and strategies. Biological Reviews of the Cambridge Philosophical Society, 74, 121–142. Jondelius, U., Ruiz-Trillo, I., Baguñà, J. & Riutort, M. (2002). The Nemertodermatida are basal bilaterians and not members of the Platyhelminthes. Zoologica Scripta, 31, 201–215. Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters R. A. Jenner • Scientific status of metazoan cladistics Kluge, A. G. (1997). Testability and the refutation and corroboration of cladistic hypotheses. Cladistics, 13, 81–96. Kluge, A. G. (1998). Sophisticated falsification and research cycles: consequences for differential character weighting in phylogenetic systematics. Zoologica Scripta, 26, 349 –360. Kristensen, R. M. (1991). Loricifera. In F. W. Harrison & E. E. Ruppert (Eds) Microscopic Anatomy of Invertebrates, Vol. 4. Aschelminthes (pp. 351–375). New York: Wiley-Liss. Kristensen, R. M. & Funch, P. (2000). Micrognathozoa: a new class with complicated jaws like those of Rotifera and Gnathostomulida. Journal of Morphology, 246, 1– 49. Lee, M. S. Y. (1995). Historical burden in systematics and the interrelationships of ‘parareptiles’. Biological Reviews of the Cambridge Philosophical Society, 70, 459 –547. Lee, M. S. Y. (1997). The phylogeny of varanoid lizards and the affinities of snakes. Philosophical Transactions of the Royal Society of London, Series B, 352, 53 –91. Lee, M. S. Y. (1998). Convergent evolution and character correlation in burrowing reptiles: towards a resolution of squamate relationships. Biological Journal of the Linnean Society, 65, 369 – 453. Lee, M. S. Y. & Doughty, P. (1997). The relationship between evolutionary theory and phylogenetic analysis. Biological Reviews of the Cambridge Philosophical Society, 72, 471– 495. Lemburg, C. (1995). Ultrastructure of the introvert and associated structures of the larvae of Halicryptus spinulosus (Priapulida). Zoomorphology, 115, 11–29. Lewontin, R. C. (1991). Electrophoresis in the development of evolutionary genetics: milestone or millstone? Genetics, 128, 657–662. Lewontin, R. C. (2002). Directions in evolutionary biology. Annual Review of Genetics, 36, 1–18. Medawar, P. (1996). The Strange Case of the Spotted Mice and Other Classic Essays on Science. Oxford: Oxford University Press. Nielsen, C. (1994). Larval and adult characters in animal phylogeny. American Zoologist, 34, 492–501. Nielsen, C. (2000). The phylogenetic position of Entoprocta and Ectoprocta. Proceedings of the 11th International Bryozoology Association. Conference, 66–73. Nielsen, C. (2001). Animal Evolution. Interrelationships of the Living Phyla. Oxford: Oxford University Press. Nielsen, C. (2002). The phylogenetic position of Entoprocta, Ectoprocta, Phoronida, and Brachiopoda. Integrative and Comparative Biology, 42, 685–691. Nielsen, C., Scharff, N. & Eibye-Jacobsen, D. (1996). Cladistic analyses of the animal kingdom. Biological Journal of the Linnean Society, 57, 385–410. Müller, M. C. & Westheide, W. (2000). Structure of the nervous system of Myzostoma cirriferum (Annelida) as revealed by immunohistochemistry and cLSM analyses. Journal of Morphology, 245, 87–98. Nyhart, L. K. (1995). Biology Takes Form. Animal Morphology and the German Universities, 1800–1900. Chicago: The University of Chicago Press. Nyhart, L. K. (2002). Learning from history: morphology’s challenges in Germany ca 1900. Journal of Morphology, 252, 2–14. Obst, M. & Funch, P. (2003). Dwarf male of Symbion pandora. Journal of Morphology, 255, 261–278. Omland, K. E. (1999). The assumptions and challenges of ancestral state reconstructions. Systematic Biology, 48, 604 – 611. © The Norwegian Academy of Science and Letters • Zoologica Scripta, 33, 4, July 2004, pp293– 310 Peterson, K. J. & Eernisse, D. J. (2001). Animal phylogeny and the ancestry of bilaterians: inferences from morphology and 18S rDNA gene sequences. Evolution and Development, 3, 170 –205. Poe, S. & Wiens, J. J. (2000). Character selection and the methodology of morphological phylogenetics. In J. J. Wiens (Ed.) Phylogenetic Analysis of Morphological Data pp. 20 –36. Washington: Smithsonian Institution Press. Purschke, G., Hessling, R. & Westheide, W. (2000). The phylogenetic position of the Clitellata and the Echiura — on the problematic assessment of absent characters. Journal of Zoological Systematics and Evolutionary Research, 38, 165 –173. Rieger, R. M. (1980). A new group of interstitial worms, Lobatocerebridae nov. fam. (Annelida) and its significance for metazoan phylogeny. Zoomorphology, 95, 41– 85. Rieger, R. M., Tyler, S., Smith, J. P. S. III & Rieger, G. E. (1991). Platyhelminthes: Turbellaria. In F. W. Harrison & B. J. Bogitsh (Eds) Microscopic Anatomy of Invertebrates, Vol. 3. Platyhelminthes and Nemertinea (pp. 7–140). New York: Wiley-Liss. Rieppel, O. & Kearney, M. (2002). Similarity. Biological Journal of the Linnean Society, 75, 59– 82. Rieppel, O. & Reisz, R. R. (1999). The origin and early evolution of turtles. Annual Review of Ecology and Systematics, 30, 1–22. Rouse, G. W. (1999). Trochophore concepts: ciliary bands and the evolution of larvae in spiralian Metazoa. Biological Journal of the Linnean Society, 66, 411– 464. Rouse, G. W. (2000a). Bias? What bias? The evolution of downstream larval-feeding in animals. Zoologica Scripta, 29, 213–236. Rouse, G. W. (2000b). The epitome of hand waving? Larval feeding and hypotheses of metazoan phylogeny. Evolution and Development, 2, 222–233. Rouse, G. W. (2001). A cladistic analysis of Siboglinidae Caullery, 1914 (Polychaeta, Annelida): formerly the phyla Pogonophora and Vestimentifera. Zoological Journal of the Linnean Society, 132, 55– 80. Rouse, G. W. & Fauchald, K. (1997). Cladistics and polychaetes. Zoologica Scripta, 26, 139 –204. Rowe, F. W. E., Anderson, D. T. & Healy, J. M. (1991). Echinodermata: Concentricycloidea. In A. C. Giese, J. S. Pearse & V. B. Pearse (Eds) Reproduction of Marine Invertebrates, Vol. VI. Echinoderms and Lophophorates (pp. 751 – 760). California: The Boxwood Press. Ruiz-Trillo, I., Paps, J., Loukota, M., Ribera, C., Jondelius, U., Bagunà, J. & Riutort, M. (2002). A phylogenetic analysis of myosin heavy chain type II sequences corroborates that Acoela and Nemertodermatida are basal bilaterians. Proceedings of the National Academy of Sciences USA, 99, 11246 –11251. Ruppert, E. E. (1991). Gastrotricha. In F. W. Harrison & E. E. Ruppert (Eds) Microscopic Anatomy of Invertebrates, Vol. 4. Aschelminthes (pp. 41–109). New York: Wiley-Liss. Salvini-Plawen, L. V. & Steiner, G. (1996). Synapomorphies and plesiomorphies in higher classification of Mollusca. In J. D. Taylor (Ed.) Origin and Evolutionary Radiation of the Mollusca (pp. 29–51). Oxford: Oxford University Press. Schmidt-Rhaesa, A. (1996). The nervous system of Nectonema munidae and Gordius aquaticus, with implications for the ground pattern of the Nematomorpha. Zoomorphology, 116, 133 –142. Schmidt-Rhaesa, A. (1997/1998). Phylogenetic relationships of the Nematomorpha — a discussion of current hypotheses. Zoologischer Anzeiger, 236, 203 –216. 309 Scientific status of metazoan cladistics • R. A. Jenner Schram, F. R. (1991). Cladistic analysis of metazoan phyla and the placement of fossil problematica. In A. M. Simonetta & S. Conway Morris (Eds) The Early Evolution of Metazoa and the Significance of Problematic Taxa (pp. 35– 46). Cambridge: Cambridge University Press. Schuchert, P. (1990). The nephridium of the Bonellia viridis male (Echiura). Acta Zoologica, 71, 1– 4. Siddall, M. E. & Kluge, A. G. (1997). Probabilism and phylogenetic inference. Cladistics, 13, 313 –336. Strong, E. E. & Lipscomb, D. (1999). Character coding and inapplicable data. Cladistics, 15, 363–371. Sørensen, M. V., Funch, P., Willerslev, E., Hansen, A. J. & Olesen, J. (2000). On the phylogeny of the Metazoa in the light of Cycliophora and Micrognathozoa. Zoologischer Anzeiger, 239, 297–318. Turbeville, J. M. (1991). Nemertinea. In F. W. Harrison & B. J. Bogitsh (Eds) Microscopic Anatomy of Invertebrates, Vol. 3. Platyhelminthes and Nemertinea (pp. 285–328). New York: Wiley-Liss. Turbeville, J. M. (2002). Progress in nemertean biology: development and phylogeny. Integrative and Comparative Biology, 42, 692–703. 310 Wallace, R. L., Ricci, C. & Melone, G. (1996). A cladistic analysis of pseudocoelomate (aschelminth) morphology. Invertebrate Biology, 115, 104 –112. Wanninger, A. & Haszprunar, G. (2002). Chiton myogenesis: perspectives for the development and evolution of larval and adult muscle systems in molluscs. Journal of Morphology, 251, 103–113. Wilson, E. O. (2000). Sociobiology. The New Synthesis. Cambridge: The Belknap Press of. Harvard University Press. Zrzavy, J. (2001). The interrelationships of metazoan parasites: a review of phylum- and higher-level hypotheses from recent morphological and molecular phylogenetic analyses. Folia Parasitologica, 48, 81–103. Zrzavy, J. (2003). Gastrotricha and metazoan phylogeny. Zoologica Scripta, 32, 61– 81. Zrzavy, J., Hypsa, V. & Tietz, D. F. (2001). Myzostomida are not annelids. Molecular and morphological support for a clade of animals with anterior sperm flagella. Cladistics, 17, 170 –198. Zrzavy, J., Mihulka, S., Kepka, P., Bezdek, A. & Tietz, D. (1998). Phylogeny of the Metazoa based on morphological and 18S ribosomal DNA evidence. Cladistics, 14, 249 –285. Zoologica Scripta, 33, 4, July 2004, pp293 – 310 • © The Norwegian Academy of Science and Letters
© Copyright 2026 Paperzz