Portal de Periódicos da CAPES

Ceci n'est pas une pipe: names, clades and phylogenetic nomenclature

2003; Wiley; Volume: 41; Issue: 3 Linguagem: Inglês

10.1046/j.1439-0469.2003.00236.x

1439-0469

Fredrik Pleijel, Greg W. Rouse,

Mediterranean and Iberian flora and fauna

An introduction is provided to the literature and to issues relating to phylogenetic nomenclature and the PhyloCode, together with a critique of the current Linnaean system of nomenclature. The Linnaean nomenclature fixes taxon names with types, and associates the names with ranks (genus, family, etc.). In phylogenetic nomenclature, names are instead defined with reference to cladistic relationships, and the names are not associated with ranks. We argue that taxon names under the Linnaean system are unclear in meaning and provide unstable group–name associations, notwithstanding whether or not there are agreements on relationships. Furthermore, the Linnaean rank assignments lack justification and invite unwarranted comparisons across taxa. On the contrary, the intention of taxon names in phylogenetic nomenclature is clear and stable, and the application of the names will be unambiguous under any given cladistic hypothesis. The extension of the names reflects current knowledge of relationships, and will shift as new hypotheses are forwarded. The extension of phylogenetic names is, therefore, clear but is associated to (and thus dependent upon) cladistic hypotheses. Stability in content can be maximized with carefully formulated name definitions. A phylogenetic nomenclature will shift the focus from discussions of taxon names towards the understanding of relationships. Also, we contend that species should not be recognized as taxonomic units. The term 'species' is ambiguous, it mixes several distinct classes of entities, and there is a large gap between most of the actual concepts and the evidence available to identify the entities. Instead, we argue that only clades should be recognized. Among these, it is useful to tag the smallest named clades, which all represent non-overlapping groups. Such taxa – LITUs (Least Inclusive Taxonomic Units) – are distinguished from more inclusive clades by being spelled with lower-case initial letter. In contrast to species, LITUs are conceptually straightforward and are, like other clades, identified by apomorphies. Die Arbeit ist eine Einführung in die Kernpunkte der phylogenetischen Nomenklatur und des PhyloCodes mit einer Zusammenstellung der dazu vorliegenden Literatur sowie einer Kritik am gültigen System der Linnéschen Nomenklatur. Die Linnésche Nomenklatur macht Taxonnamen an Typen fest und verbindet diese Namen mit einem Rank (Gattung, Familie etc.). In der Phylogenetischen Nomenklatur werden die Namen dagegen auf der Basis von Cladogrammen festgelegt und erhalten keinen Rank. Wir vertreten den Standpunkt, dass Bedeutung und Umfang der Linnéschen Namen nicht wirklich klar sind und zu vagen Verbindungen zwischen Gruppen Veranlassung geben, ungeachtet, ob verwandtschaftliche Beziehungen zwischen ihnen akzeptiert werden oder nicht. Darüberhinaus fehlt den Linnéschen Rankzuordnungen die Begründung, was zu ungerechtfertigten Vergleichen über die Taxa hinweg verleitet. Im Gegensatz dazu ist die Intention eines Taxonnamens in der Phylogenetischen Systematik klar und gleichbleibend; sein Gebrauch ist eindeutig, welche cladistische Hypothese auch gerade angenommen wird. Der Umfang eines derartigen Namens spiegelt das aktuelle Wissen über die Verwandtschaftsbeziehungen wider; er wird sich verändern, wenn neue Hypothesen dazu vorliegen. Der Inhalt phylogenetisch definierter Namen ist daher klar, allerdings auch abhängig von der jeweiligen unterliegenden cladistischen Hypothese. Inhaltliche Stabilität kann am besten durch sorgfältig formulierte Namen erreicht werden. Eine phylogenetische Nomenklatur lenkt so den Blick von Diskussionen über Taxonnamen hin auf das Verstehen verwandtschaftlicher Beziehungen. Wir setzen uns auch stark dafür ein, dass Arten nicht als taxonomische Einheiten angesehen werden. Der Begriff ,,Art" ist unscharf und vieldeutig; er vermischt verschiedene Wesensklassen. So existiert zwischen den meisten aktuellen Artkonzepten und dem Erkennen dieser Einheiten eine große Kluft. Dagegen plädieren wir dafür, dass nur Kladen anerkannt werden sollten, wobei es sinnvoll erscheint, die jeweils kleinsten bekannten Kladen, die für sich nicht überschneidende Gruppen stehen, zu bezeichnen. Derartige Taxa – LITUs (Last Inclusive Taxonomic Units) – unterscheiden sich von übergeordneten Kladen durch kleine Anfangsbuchstaben. Im Gegensatz zu den Spezies sind LITUs begrifflich eindeutig und werden wie andere Kladen durch Autapomorphien charakterisiert. Phylogenetic nomenclature (PN) has been developed as an alternative system of nomenclature (e.g. de Queiroz and Gauthier 1990, 1992, 1994; Cantino and de Queiroz 2000). Briefly stated, it differs in two major respects from the traditional and current, 'Linnaean' nomenclature (LN) (as detailed in the nomenclatural codes for bacteria, Lapage et al. 1992, plants, Greuter et al. 1999, and animals, International Commission on Zoological Nomenclature – hereafter ICZN – 1999): Taxon names are fixed with reference to cladograms under PN, whereas names are fixed by associating taxa with name-bearing individuals (types) under LN. PN is rankless, whereas taxa must be referred to ranks (genera, families, etc.) under LN. Although the basic components of PN are simple, there certainly exists differing views on the effects of the introduction of such a nomenclatural system. Some recent studies have provided strongly negative criticisms of PN (Dominguez and Wheeler 1997; Benton 2000; Nixon and Carpenter 2000; Platnick 2001). Nevertheless, a new code of nomenclature (the PhyloCode; Cantino and de Queiroz 2000) will be established in the not-too-distant future, and a draft version is available at http://www.ohiou.edu/phylocode. We here provide an introduction to PN and the PhyloCode, and give a brief review on the literature for and against. We explain the reasons for our support for PN, and then present our views on species. We do not claim our views to be representative for PhyloCode adherents in general, and, on issues relating to species, we know for certain that they are not. Two notes on terminology are warranted. First, several definitions exist for the word 'taxon' in the systematic literature. Whereas the word taxonomy ('taxonomie') was introduced by de Augustin-Pyramus de Candolle (1813), 'taxon' is of a much later origin, and was coined in 1926 by Meyer. Today, it refers, either to named as well as unnamed groups, or to named groups only. The ICZN (1999, 118) gives the definition 'a taxonomic unit, whether named or not', whereas for Wiley (1981) a taxon is 'a grouping of organisms given a proper name, or a grouping that could be given a proper name but is not named as a matter of convention', and for de Queiroz and Gauthier (1992)'taxa are named entities, generally named groups of organisms'. We apply the term in the latter sense. Second, in a number of previous papers (e.g. de Queiroz and Gauthier 1992; Sundberg and Pleijel 1994; Pleijel 1999), the term 'phylogenetic taxonomy' was applied more or less synonymously for what is now termed PN. In agreement with present usage, we here apply the term PN for the rules and conventions related to the naming of clades (see also Bryant and Cantino 2002). Two issues are central for a system of nomenclature: (1) what to name, and (2) how to name it. As for (1) we here assume that there is general agreement among most biological taxonomists that only clades and species should be recognized (but see below regarding our view on species). There are, however, differing opinions on what clades and species actually represent. PN has been presented by most authors within a framework where taxa, in a philosophical sense, are regarded as ontological individuals (e.g. de Queiroz and Gauthier 1990; Pleijel 1999; Bryant and Cantino 2002), a view which has been criticized by several authors (e.g. Mahner and Bunge 1997; Nixon and Carpenter 2000; see also de Queiroz 1995a). We regard the taxa as individuals as a possible but not a requisite context. Pleijel and Härlin (in preparation) in a forthcoming study suggest that PN is equally applicable whether taxa are treated as classes or as individuals. Another issue is that some authors, for example, Kripke (1972), Mahner and Bunge (1997), Härlin (1998a, 1999), Härlin and Sundberg (1998), argue that names of individuals cannot be defined, only assigned. Be this applicable or not to PN (depending on ones view on how names are associated to objects and on the ontological status of taxa), we for simplicity and for agreement with previous literature here employ the term 'definition' for the procedure of attaching meaning to taxon names. There is also another issue relating to the nature of the named objects, namely that of ranks. LN attempts to include information about the taxa by making the rank assignments part of the name; the nature of the included information, however, is elusive (to say the least) and discussed further below. The main subject of this paper (point (2)) is how to name taxa. The title of this paper may require some explanation. It refers to the well-known painting, 'La Trahison des images' (The treachery of images), by the Belgian artist René Magritte, which shows a picture of a pipe, together with the text 'Ceci n'est pas une pipe'. It can be interpreted as commentary on the confusion between signs and symbols (in this context, taxon names), and the entities to which they refer (in this context, taxa). Apart from the binomial species names, not much which originally can be referred to Linnaeus actually remains in the present nomenclatural codes (Cain 1958); many of the ranks he employed were in use already by 17th century systematists such as Ray and Magnol, and the concept of nomenclatural types appears to have been introduced during the 20th century. Nevertheless, the term 'LN' is justified by Linnaeus' impact on applied taxonomy, and by the fact that the botanical and zoological codes uses his Species Plantarum (Linnaeus 1753) for (most) plants, and the 10th edition of Systema Naturae (Linnaeus 1758) for animals, as their starting points; names introduced prior to publication of these studies are not considered available (an elegant and unorthodox solution was provided for the important spider study by Clerck, which was published in 1757; it was deemed to have been published after the 10th edition of Systema Naturae, and thereby also nomenclaturally valid). The first nomenclatural code for zoology, the 'Strickland code' appeared in 1843 (Strickland et al. 1843) and was heavily influenced by the Linnaean system. Like the current International Code on Zoological Nomenclature (hereafter 'the Code'), it presents rules and recommendations. Although, there is no mention of mandatory ranks in the Strickland code, species-, genus- and family-group names are treated. Species- and genus-group names are covered by the rules, whereas the formalization of the names of subfamilies and families are referred to the recommendations. Rules for homonymy, synonymy and the principle of priority are clearly expressed (although the latter is based on the appearance of the 12th edition, rather than the 10th, edition of Systema Naturae). However, one major conceptual difference from today lies in name definitions and in that the 'nomenclatural type concept' (types as name-bearers, not as 'typical' types; see further below) is not formulated. Instead, it is stated that 'definition properly implies a distinct exposition of essential characters, and in all cases we conceive this to be indispensable, although some authors maintain that a mere enumeration of the component species, or even of a single type, is sufficient to authenticate a genus' (Strickland et al. 1843, 9). Thus, the characters and the essences of taxa were emphasized in the meaning of names (but note the mention of definition by extension). The 19th century saw a number of different, and partly competing, codes (see, e.g. Melville 1995), and the International Commission on Zoological Nomenclature was formed by the end of that century in order to unify the co-existing zoological codes into a single one. Not entirely successful, there were still disagreements on various issues and the continuing development of parallel codes. Long work of unification before and after the Second World War finally resulted in the first edition of the International Code of Zoological Nomenclature, which appeared in 1961. As far as types and ranks go, no important conceptual changes have been introduced in the subsequent editions of this code. During the 1990s, several drafts of a BioCode were circulated (Greuter et al. 1996), although this project has been abandoned. The aim of the BioCode was to replace the different codes of plants, animals, bacteria, viruses and cultivated plants with a single one. One notable difference from the zoological code was that it was not restricted to species- to family-group names, but applied also to ranks above the family-group. Assignment of ranks is a fundamental component of LN. The rank assignment in zoology is announced in the names of species- to family-group taxa. Species-group names are binomials, genus-group names are italicized uninomials and family-group names are uninomials with special suffixes (-inae, -idae, etc). Nevertheless, the actual information conveyed by these ranks is ambiguous. In the latest edition of the Code (ICZN 1999, 114) rank is vaguely defined as the 'level, for nomenclatural purposes, of a taxon in a taxonomic hierarchy'. Although this sketchiness may be understandable in view of the problems with ranks that we will shortly outline, it is still surprising considering that it relates to one of the central tenets in the LN. The Code (Introduction, p. xix) explicitly states that one of the underlying principles is to only regulate the correct application of names, and that it 'refrains from infringing on taxonomic judgement' or the determination of 'inclusiveness or exclusiveness of any taxon'. Things, however, are not so simple. Rank assignments are not taxonomically innocent (see, e.g. Stevens 1994; Minelli 2000). For one thing, the current rules forces a taxonomist to place anorganism within both a species and genus, and also recommends family, order and class assignments, notwithstanding the knowledge we may, or may not, have about the relationships of the taxon. And although the intention of ranks, as seen from the above quotation, is to give a hierarchical meaning to taxon names, the systematists' choice of rank assignment is a complex issue and is based on compromises between many different objectives. In the present day, these include the representation of monophyly and sister-group relationships, but also mnemonic criteria and concerns to minimize name changes in related taxa. This means that the subjective rank assignments, in turn, are directed by the subjective assignments made by previous systematists! Therefore, a basic argument for the elimination of ranks is that taxa of the same rank are not comparative entities in a cladistic sense (the only exception being sister taxa, but sister group relationships are not advertised in the Linnaean names). Simply put, a genus of beetles is not equivalent to a genus of green algae. While some of those that favour the retention of LN and the rejection of PN rightly recognize this problem (e.g. Nixon and Carpenter 2000), they state that it is trivial and that the benefits of keeping ranks outweigh the costs. We suggest that the opposite is the case, that the benefits of ranks are, in fact, trivial and that the costs are substantial. So what are the contemporary defences offered in favour of retaining ranks, beyond that of maintaining the status quo? Platnick (2001) suggests that 'the Linnaean hierarchy is inherently pre-adapted to presenting cladistic information, for the hierarchy can mirror a cladogram in as much, or as little, detail as one would like'. A simple example shows that much detail is not possible. Figure 1 illustrates the relationships of Vestimentifera, as proposed by Jones (1988), and uses the commonly used LN ranks of phylum, class, order, family, genus and species. Ignoring the problem with redundant names, which will be discussed below, one can see that within only a few bifurcations the most commonly used ranks of the Linnaean hierarchy to represent a cladogram are exhausted. When one considers that Vestimentifera represents but a small clade of animals, then it is abundantly clear that there can never be enough Linnaean ranks as they are currently used. In fact, it would appear that the current Linnaean hierarchy is inherently incapable of mirroring even a fairly depauperate cladogram, let alone the 'tree of life'. It is possible to make a large number of new ranks that would allow an accurate representation of cladistic hypothesis being used and such a system was outlined by Farris (1976). One is, however, left wondering what utility lies in ranks such as supersuperfamily, infrainfraphylum or supersuborder. Also, the endings of some taxon names today convey the rank meaning (e.g. idae, inae, ida) and this would clearly not be possible with a proliferation of ranks. Classification of Vestimentifera based on Jones (1988), illustrating how quickly Linnaean ranks are exhausted in mirroring a cladogram, as well as the use of redundant taxon names in LN Nixon and Carpenter (2000, 313) suggested that 'ranks convey very broad information about set exclusivity' and that they should be retained. Similar arguments were presented by Benton (2000) and Platnick (2001). Hence, on being presented with the family-ranked animal names Sabellidae and Flabelligeridae, one can infer that, on current hypotheses of relationships, Sabellidae and Flabelligeridae are not nested, whereas under PN this is not apparent from the names. However, the only factual information about the exclusivity of two Linnaean names referred to the same rank is the trivial statement that their nomino-typical taxa, and, in the end, their type specimens, are non-overlapping. Any further information requires additional knowledge relating to the delineation of the taxa, and is not directly accessible from the names. Names, to obtain non-trivial meanings, must under both systems be attached to hypotheses of relationships. So from the viewpoint of information retrieval, the introduction of a rankless nomenclature does not represent such a radical step. Platnick (2001) also made an argument emphasizing the non-overlapping nature of equivalent ranked taxa. He gave an example of identifying a spider as part of Salticidae, and produced various calculations designed to illustrate that the family rank of the taxon Salticidae provides information that is lost when the rank connotation is removed. In fact, his calculations require the knowledge (that he provides) that there are 4834 species taxa within Salticidae and 32 752 spider species not in Salticidae. Linnaean ranks play no part at all in his calculations. Benton (2000, 644) provided an example of a supposed benefit of ranks in LN, when he suggested that 'families and higher taxa prove to be useful proxies for species', and a number of authors have suggested that species numbers need not be assessed directly and that higher ranks can serve as surrogates in biodiversity studies (e.g. Warwick and Clarke 1995), or for measuring changes in the diversity of life through time (Raup 1979). Forey (2001, 89) noted that 'the abolition of ranks can have some unfortunate consequences for many people who compile' diversity indices 'based on generic counts or family counts'. Such arguments are manifestly flawed since, as pointed out above, taxa of the same rank are not comparable entities, a point accepted by anti-PN authors such as Nixon and Carpenter (2000). The erroneous reasoning behind such 'taxonomic surrogacy' exemplifies how the meaning of ranks is repeatedly misinterpreted, not only by users of taxonomies, but also by systematists. The elimination of LN will enforce the use of more rigorous measures of biodiversity with reference to cladistic hypotheses instead of classifications. These issues are further explored by Bertrand et al. (in preparation). Another problem with ranks is represented by redundant names, which refers to a single group with several names of different ranks. An example can be found in Funch and Kristensen's (1995) recent erection of a new phylum, class, order, family and genus for one new species. Another is seen in the establishment of a new insect order established for two new described species (Klass et al. 2002). Vestimentifera, including the 'giant tube-worms' from marine hydro-thermal vents, provides another example, where Riftia pachyptila is placed in a monotypic family Riftiidae, a monotypic order Riftiida, and a monotypic order Axonobrachia (Fig. 1). All these names refer to a single group, and the redundancy is a reflection, not of the relationships of Riftia pachyptila, but of the rank assignments of the neighbouring taxa. Jones (1988, 154) stated that 'in order to separate Riftia from the others I here propose two vestimentiferan classes'. As is common with redundant names, no justifications were provided for the order Riftiida, nor for the family Riftiidae or the genus Riftia, which were introduced in a previous paper (Jones 1981). Perhaps Jones was thinking that the Linnaean ranks provided the 'ease of communication', as was suggested by Nixon and Carpenter (2000, 315)? In the context of the relationships proposed by Jones (1988) the classes Axonobrachia and Basibrachia are sister taxa. But when we examine other taxa that are referred to the same rank, we see that the order Riftiida (which refers to the same group as does the names Riftiidae, Riftiida and Axonobrachia), is not sister to either of the two other orders Tevniida or Lamellibrachia. Similarly, the family Riftiidae is not sister to any of the other family-ranked taxa. The 'ease of communication' is not obvious. In addition to this story, it may be noted that recent analyses shows the former phylum-ranked group Vestimentifera to belong within annelid polychaetes (Rouse 2001 and references therein). For nomenclatural reasons it is now instead referred to as Siboglinidae, and virtually all the names discussed above are without current application. Benton (2000, 64) suggested in regard to the use of redundant names that it is now 'accepted practice not to do such things'. The above examples show that this is not the case. Benton (2000) also referred to solutions to the redundant names problem in an LN context, and outlined an example where the genus Archeopteryx is listed in a position to make it clear it is sister group to a taxon that has Infraclass rank. This convention then means that a genus 'equals' an infraclass and so makes a mockery of any statements about ease of communication and Linnean ranks being suitable to represent cladograms. A fundamental component of the LN is that all species names consist of two parts; a genus name plus a species epithet. Within the realm of each particular Code the combination to form a species name must be unique. Benton (2000, 644) stated that 'the most scary consequence of phylogenetic nomenclature is the proposal to abandon the Linnaean binomen'. Yes, the PhyloCode does indeed intend to alter the current binomial system. This means that existing species names will be converted into PhyloCode names, and their current binomial form must be altered. The various ways in which this can be implemented are detailed in Cantino et al. (1999) and discussed further below. But first we should see why such a change is necessary. As discussed above, taxa of the same rank are for the most part not comparable entities. Hence, genera are not equivalent and represent subjective constructs, and the mandatory attachment of a 'genus' name to the species epithet is obviously incompatible with the PhyloCode. Beyond the information of a more inclusive clade (which can easily be provided without forming a part of the name) when one is presented with species names such as Fabricinuda trilobata and Novafabricia infratorquata, there is tremendous cost associated with binomina, namely that of name changes. As currently applied the binomial nomenclature adds extra instability, as each generic reallocation results in an alteration of the species name. Considerable number of such changes may follow from re-evaluations of the relationships of type species of larger genera. These changes have no relationships whatsoever to the concept or to the inclusivity of the taxon in question (Cantino 1998). What has changed is the hypothesis of where it belongs relative to other taxa. As an example, the two taxa mentioned above were once known as Fabricia trilobata and Fabricia infratorquata. While the content of these taxa remains unchanged, the current Code rules mean that the names have had to shift when the hypotheses of their relationships has resulted in their moving to another part of the tree. And in other cases the name changes may be brought about, not by differences in view of their relationships, but as an effect of disagreements on rank assignments. To make matters worse, under the current Codes the species epithet sometimes has to change owing to problems with secondary homonymy, or to the fact that a specific epithet has a different gender to that of the new 'genus' (Cantino et al. 1999). This binomial system, in fact, is one of the worst possible imaginable, given an aim of name stability. Why should a name change when a taxon remains the same entity but simply moves from one part of the tree to another? And why should a name change because a taxonomist arbitrarily decides two sister 'genera' should be lumped into one? The grand majority of taxon names are in fact 'species' names and hence binomials. Changes in species names in association with taxonomic revisions represent the greatest source of instability in current nomenclature, and constitute a significant source of frustration for taxonomists as well as users of taxonomies. LN applies type-based definitions of taxon names. However, the term 'type' is quite misleading in this context. Whereas many earlier systematists (see Mayr 1982 for a review) tended to view types, in one sense or another (e.g. Stevens 1994), as central within a taxon and equipped with essential properties, few contemporary taxonomists would embrace this view. Types of today are not representing a taxon in any other sense than being a part or member of it. Their only purpose is to serve as devices for fixing names: nomenclatural types equal name-bearing specimens or taxa, 'onomatophores'sensuSimpson (1961). Once a group has been identified, the type belonging to that group will determine the name, or a new name and type is introduced and designated. Species-group, genus-group and family-group taxa have their own types and typification rules, and any individual type can serve as type only at a single one of these three levels. For instance, the holotype is the objective name-bearer of the species Phyllodoce laminosa, the type-species Phyllodoce laminosa is the objective name-bearer of the genus Phyllodoce and the genus Phyllodoce is the objective name-bearer of the family Phyllodocidae. The type–rank association has the important effect that any delineated group must be associated with a specific rank before it can be properly named. Within zoology each taxon must minimally be referred to species and genus, and the Code further recommends assignment to family, order and class. Homonyms appear whenever the same names are based on different nominal type taxa. Synonyms appear if there are several competing names for a taxon referred to the same rank-group. If the different names are based on the same type (which is rare) they constitute objective synonyms, and if the names are based on different types (which is common), they constitute subjective synonyms. Taxa of different rank-groups are deemed different and can thereby never be treated as synonymous. For example, Phyllodoce and Phyllodocidae can never be synonyms, even if both names refer to a single group with identical composition. Selection among homonyms and synonyms is generally based on priority of publication. The problem with the current type-based definitions lies in how name fixations and group delineations are linked to each other. The nomenclatural types are attached to taxa by ostensive definitions, but how the names should be applied to groups remains unclear (Fig. 2). The type by itself is (almost) devoid of information for the actual meaning of taxon names. The association between types and groups in LN is instead accomplished by rank assignment. For instance, in the example of types used above, Phyllodoce laminosa is the type species of the genus Phyllodoce, and Phyllodoce is the type genus of the family Phyllodocidae. But, as ranks lack any objective or empirical content, the actual delineation of the group's boundaries is unsettled and arbitrary. In the previous example, the only unquestionable member of these taxa is the holotype for P. laminosa. More inclusive membership of each of the group names Phyllodoce laminosa, Phyllodoce and Phyllodocidae are up to users discretion, no matter whether there is agreement on the relationships among the taxa or not. As noted by, for example, Griffiths (1976) and de Queiroz and Gauthier (1994), the result is that two systematists may delineate taxa differently, but still apply the same name (compare the application of both Phyllodoce and of Phyllodocinae in trees I and II in Fig. 3a). And they may delineate taxa identically, but apply different ranks and thereby also different names (compare Phyllodoce with Phyllodocinae, and Phyllodocidae with Phyllodocinae, respectively, in trees I and II in Fig. 3a). Another, hypothetical, example of this provided in Fig. 3b. Typ