Cladistic controversies

Paraphyly vs. Holophyly
While few would contest the need for classification to reflect genealogy via use of monophyletic groupings, it is when cladistic methods are carried to their logical end result and a system of complete hierarchy is established that bitter debate erupts. A taxon which is united by common descent and yet which excludes all the groups derived from that shared ancestry, is not at all reflective of evolutionary reality&#8212;it is instead a relict of a pre-evolutionary system of typological classification (Patterson 1981, 1982). In other words it is a taxonomic artifact, which cuts evolution off in midstream by implying disparity where there is none. Thus, proponents of cladistic methodology have long argued for the need to discourage the use of paraphyletic taxa and delimit in their stead, holophyletic clades.

The implications of this reasoning are profound&#8212;and to some systematists, unsettling. If paraphyletic grades are not indicative of evolution, then pre-evolutionary taxa established using Linnean methods are by default artificial constructs. No better example can be found than that of Reptilia sensu Linne. Linne defined Reptilia such that it excluded birds, and later systematists employed his definition with little modification for the next two hundred years. And yet, it has been evident since 1861 that birds are derived from ancestral stock somewhere within Archosauromorpha, and are therefore nested within Reptilia. Under a typological system such as Linne used, however, a form like Archaeopteryx cannot be both reptile and bird; it can only be one or the other. Yet it is both. And this is the fundamental dilemma: to extricate birds from Reptilia denudes Reptilia of any phylogenetic reality&#8212;it makes the taxon paraphyletic.

Including birds and defining the avian lineage such that it is nested within Archosauromorpha and Reptilia as a whole can resurrect Reptilia as a valid taxonomic category. Thus emended, Reptilia can be defined as: &#8220;the common ancestor of extant turtles and saurians, and all its descendants&#8221; (Gauthier et al 1988). Some authors have protested that this sort of phylogenetic &#8220;semantics&#8221; is confusing both in its counter-intuitiveness and its rejection of taxonomic convention (Paul 2002). Yet neither of these reasons are satisfactory reasons to distort the phylogeny of reptiles

The principal arguments, which have been advanced against a strict application of holophyly to phylogenetic reconstruction, do not dispute that holophyletic clades are preferable to paraphyletic groupings&#8212;in that the former is superior in modeling phylogeny than the latter. Rather, arguments from convention and clarity (at the expense of accuracy) have been made, but moreover, some systematists insist that there are practical difficulties in redefining taxa (particularly extinct taxa) such that they meet the requirements of holophyly (Carroll 1988). Carroll, for example, uses a cascade analogy whereby the problem of delimiting a clade is continuously offset to the next earliest representative of the lineage, and so on. Carroll uses the phylogenetic affinities of birds and dinosaurs as his primary example of this cascade in action, arguing that if one makes theropods birds, it renders Saurischia paraphyletic, and so on.

However, in this case, it is by including both the sister clade of birds&#8212;Deinonychosauria&#8212;and Avialae proper (Aves sensu Linne, emended by Chiappe) within a holophyletic containing node, Eumaniraptora, that the paraphyly of more inclusive nodes is avoided. Similarly, delimiting containing clades, or nodes on a cladogram, to prevent paraphyly can be effected in systematic review of any given lineages. The concomitant drawback is the need to name each node on a cladogram, which necessarily results in an explosive proliferation of taxa&#8212;cluttering and obfuscating the phylogenetic map and rendering it largely esoteric. This undesirable result is avoided through assigning names only to major nodes. Critics of cladistics have derided this process as arbitrary, and in so doing forget that all phylogenetic reconstruction is to a degree arbitrary. Latent arbitrariness notwithstanding, it is a general convention to consider major nodes as those robustly underwritten by synapomorphies (Paul 2002).

A remedy to this nomenclatural problem which has been suggested in some circles, namely amongst vertebrate paleontologists, is the use of apomorphy + clade names for taxa. Gauthier (1999) and Paul (2002) have outlined the methodology and protocol for employing apomorphy + clade names in phylogenetic reconstruction. Paul (2002) has extensively employed such a system in his review of the systematics of Dinosauria. Despite claims to the contrary, apomorphy + clade names are in no particular way more convenient than erecting new node-based names for containing clades, in that they do the same thing and while innovative, I can see no pressing argument for their wide-spread introduction in phylogenetic reconstruction.

Ultimately, critiques such as those offered by Carroll (1988) and Carroll & Dong (1991) have asserted that paraphyly is largely inevitable in our systematic analyses, not the least of the reasons being the limit of our taxonomic nomenclature to articulate the dynamic processes of evolution. This objection is correct to an extent, but it is paramount to bear in mind that paraphyletic taxa do not accurately reflect phylogeny. Thus, it should remain a goal of systematists to avoid as much as possible, describing and naming such taxa.

Convergence vs. Cladistics
Critics of cladistic methodology present few arguments as vociferously as they do the argument that cladistics is unable to distinguish pseudo-homologies generated by homoplasy and parallelism from homologies derived through common ancestry (Mayr 1981, Olson 1985, Carroll 1988, Carroll & Dong 1991, Feduccia 1996). While some practitioners of cladistics have minimized the role of these processes in vertebrate evolution, it is well documented that both homoplasy and parallelism are powerful forces in that process (Simpson 1961, Mayr 1981, Cain 1982, Carroll 1982, Mayr 2001). The real question is whether or not homoplasy and parallelism occur at such a massive degree that they effectively overwhelm the statistical analysis of a data set&#8212;the process which cladistics rests upon&#8212;and thus preclude this methodology.

The proponents of this argument have advanced multiple cases of cladistic analyses, which upheld as synapomorphic, characters that in fact were not homologous (see mainly Mayr 1981, Olson 1985, Carroll 1988, Sibley & Ahlquist 1990 and Feduccia 1996, on Cracraft&#8217;s infamous 1982 attempt to reconstruct the phylogeny of Aves using cladistics). The examples most rigorously advanced are the convergence between Podicipediformes, Gaviiformes, and Hesperornithiformes, and the case of Hupehsuchus. Opponents of cladistics have cited these as cases of massive convergence, with which cladistic analysis is unable to cope.

Yet what do the facts of the matter indicate? To prevent this very problem cladistic analysis uses both the principle of parsimony, and attempts to collect as many characters as possible for analysis, to yield the largest data-set, following the logical conclusion that the number of homologous apomorphies will be greater than that of non-homologous apomorphies. The utility of both of these principles are tacitly or explicitly rejected in the critiques listed above, and the concomitant conclusion is drawn that even when using these methods, cladistics fails to tease out convergence.

Critical review of the sorry case of Cracraft&#8217;s &#8220;Gaviomorphae&#8221;&#8212;the putative clade allying Hesperornithiformes, Gaviiformes, and Podicipediformes&#8212;illustrates how this same taxon, seen as so emblematic of the shortcomings of cladistics by its opponents, is in fact a sterling example of poor cladistics. Cracraft asserted holophyly of Gaviomorphae on the basis of a handful of alleged synapomorphies, among which a sharply pointed and prominent cnemial crest was the principal character used to underwrite the validity of the &#8220;gaviomorph&#8221; assemblage. Yet the cnemial crest in the three constituent taxa of Cracraft&#8217;s &#8220;Gaviomorphae&#8221; is derived from entirely different elements, and thus is non-homologous, in each taxon (Storer 1971, Olson 1985, Feduccia 1996). It was upon this flaw, that Cracraft&#8217;s phylogeny collapsed. The question must be, is this indicative of an underlying flaw in cladistic methodology?

In their landmark review of avian phylogeny, Charles Sibley and Jon Ahlquist (1990), commented that: &#8220;the errors in Cracraft&#8217;s reconstruction of the phylogeny of the diving birds are due to the difficulties of interpreting morphological characters, not to the principles he used as the basis for his analysis.&#8221; Opponents of cladistics would do well to bear this in mind. It was Cracraft&#8217;s research, overlooking plain pseudo-homologies in the structure of the cnemial crest amongst the &#8220;gaviomorphs&#8221; which accounts for the rejection of that particular phylogenetic hypothesis. All this example illustrates is that cladistic practitioners, like all systematists, are human&#8212;and just as capable of errors and subjectivity as any other, which in and of itself is the more important point (i.e., it helps show that cladistics is not a panacea for systematics).

Yet some, (e.g., Feduccia 1996) have advanced this very same case as an example of &#8220;massive convergence.&#8221; It is nothing of the sort. It is in fact stereotypical convergence of elements associated with a particular biophysical function, in this case diving, and is no more a case of massive convergence than the independent derivation of a patella in varying tetrapod lineages. To maintain otherwise is at best mistaken, and at worst, specious.

The dubious "Gaviomorphae" aside, we can ask the question: are there any cases of bona fide &#8220;massive osteological convergence&#8221; which might be so thorough as to preclude accurate cladistic analysis? The answer may lie in the remains of a peculiar aquatic diapsid from the Triassic, Hupehsuchus nanchangensis, which Carroll & Dong (1991) have explicitly advanced as an example of extraordinary convergence, capable of both precluding the parsimony principle and fooling cladistic analysis.

Statistical evaluation of a character data set tabulated by PAUP produced a parsimonious cladogram inferring monophyly of Hupehsuchus, ichthyosaurs, and nothosaurs (Carroll & Dong 1991). However, the work of Rieppel (1989) and subsequent evaluation of this hypothetical phylogeny by Carroll & Dong has made it clear that the large number of synapomorphies used to support the monophyly of this aquatic diapsid assemblage, are in fact non-homologous and a function of significant convergence.

This much is not in dispute, what is in dispute is the contentious assertion that the case of Hupehsuchus is indicative of the evolutionary norm. Of Carroll & Dong&#8217;s three basic suppositions in their paper on Hupehsuchus, only one, that any given character is subject to homoplastic or parallel evolution, is correct. Their concomitant conclusions on the utility of parsimony, is entirely invalid and reflective of the convergence fallacy: that because homoplasy can foil cladistic analysis, it must. Hupehsuchus is a discrete case; an exception, which proves the rule: only in cases of large-scale convergence is the generally accurate principle of parsimony rendered suspect. The long-dead diapsid from China further emphasizes the far more meaningful conclusion which Carroll & Dong (and others who have parroted this example ceaselessly, such as Alan Feduccia) could draw were they not so concerned with discrediting cladistics: that this very discipline is by no means the infallible answer to all phylogenetic dilemmas.

The conclusion that parsimony combined with statistical analysis of characters is unreliable is a distortion; they can be unreliable. The greater the data set of characters, the more likely it is that convergent pseudo-homologies can be differentiated from synapomorphies, and this remains the single most effective procedure for distinguishing convergence or parallelism from common-descent. The convergence fallacy notwithstanding, cladistics remains a formidable tool for phylogenetic reconstruction.

Geological Time vs. Anatomy
Few debates in all of systematic biology are as acrimonious as the argument over which takes precedence in phylogenetic reconstruction: geological time, or anatomy? And while some disputes are esoteric quibbles between specialists, this particular dispute is a bona fide intellectual schism between systematists.

The former is the classical argument. The legendary Dean of vertebrate paleontology, Alfred Sherwood Romer, summarized the stratigraphic argument thusly, &#8220;In discussing fossils, some notion of the geological time scale in necessary&#8221; (1970). In general, time-dependent phylogenetic reconstruction asserts that a character, which appears earlier than another, especially in groups, which are well documented in the fossil record over long periods of time, is likely to be a basal character.

Robert Carroll, in his revision of Romer&#8217;s tome, Vertebrate Paleontology and Evolution (1988), further stresses that in establishing phylogeny, one must &#8220;emphasize the earliest known members&#8221; of a lineage, in that &#8220;they have had the shortest amount of time to evolve new characters since their initial divergence.&#8221; Carroll concludes with: &#8220;Hence, they should provide us with the best opportunity to identify the derived features that they share with their closest sister group.&#8221;

While these arguments are certainly coherent and intuitively appealing, it is an unfortunate fact that reality rarely conforms to our limited view of how it ought to behave. The cut and dry world of time-dependent phylogenetic reconstruction is just not the way of it&#8212;things are not that simple. As discussed earlier, relying on stratigraphic data exclusively to document polarity and ontogenetic change is fraught with problems, and these problems are only magnified as one considers a group whose fossil record is both spotty and of limited temporal distribution. Time-dependent arguments are prey to the vagaries of the fossil record and geology, among other factors, not the least of which being the inconsistent way in which lineages diverge, subsequent to cladogenesis (e.g., parental taxa outliving their descendants, see section 2).

This stratophenetic methodology is further weakened by the reliance on general similarity between taxa over time as a measure of phylogeny. Thus time-dependent phylogenies often chart the evolution of grades only, and do not reflect phylogeny accurately.

Recognizing these limitations, an opposing camp has stressed anatomy as the fundamental basis for phylogenetic reconstruction, arguing that generally speaking, character distribution is more indicative of phylogeny than mere stratigraphic occurrence, especially if one is dealing with a group whose fossil record is poor and distributed over a brief period of geological time. Considering that changes in ontogeny are the most concrete example of evolutionary progression in a lineage, the anatomy-dependent argument is entirely logical. Moreover, mere stratigraphic occurrence does not impute to fossils phylogenetic context or relevance. For this, one must refer to characters--anatomy. There is simply no rational way to make phylogenetic reconstruction dependendent upon stratigraphic context (Clark et al in Chiappe & Witmer 2002).

Logic notwithstanding, the assertion that anatomy generally takes precedence over stratigraphy, has been so distorted by opponents of this methodology, that flippant statements such as those of Alan Feduccia, to the effect that cladistics has &#8220;discarded geological time as a tool in deciphering evolution&#8221; (1996), are common. Such arguments are entirely specious&#8212;geological time has not been categorically dismissed, rather it&#8217;s utility has been minimized. Accurate phylogenetic reconstruction demands a critical view of stratigraphy as a criterion for establishing phylogenies, not dogmatic adherence.

Ultimately, one must advance a case where two phylogenies generated using each method can be compared side by side, their likelihood weighed against the data and our knowledge of evolutionary processes. There is no single example, which can more effectively fulfill that role, than the phylogeny of the acanthodian fishes.

Assuming that most readers are not familiar with so obscure a group as Acanthodia, an introduction of sorts is in order. Reader, meet Acanthodia&#8212;a curious group of extinct fishes, whose osteology and evolution was reviwed byMiles (1965, 1968, 1973) and Maisey (1996). Acanthodia represents a successful, but ultimately failed experiment in the gnathostome adaptive radiation, which culminated in the Lower Devonian&#8212;when the clade was at its zenith (Denison 1979, Carroll 1988).

The relationship of Acanthodia to other gnathostome clades has long been contentious in and of itself, and there remain arguments for a close affinity of Acanthodia and Chondrichthyes (Orvig 1973, Jarvik 1977). Without exhaustively cataloguing the data to support an alternative phylogenetic status for Acanthodia, I instead refer the reader to the work of Roger Miles, and summaries presented in Denison (1979), Carroll (1988) and Maisey (1996). These ongoing arguments aside, we can turn to the far more vitriolic debate over the intra-relationships of Acanthodia, a subject, which pits time-dependent phylogenetic reconstruction, against the anatomy-dependent school.

Stratophenetics isolates Climatiida, the earliest occurring acanthodian lineage, dating from the Middle Silurian, as basal to the entire taxon. Counter-intuitively, climatiid acanthodians display the most apomorphic morphology comparative to other members of the lineage, including the presence of two dorsal fins, and numerous paired intermediate spines on the ventral surface of the body&#8212;a condition autapomorphic of Acanthodia itself. The time-dependent phylogenies go on to isolate the Acanthodida, as the most derived acanthodian fishes, and yet they display the most plesiomorphic of all acanthodian morphologies, retaining a single dorsal fin, and with but one pair of intermediate spines. The two groups are separated by a mere 12 million years&#8212;which, to geology, is a blink of the eye.

The time-dependent phylogeny, therefore, has two significant flaws: it ignores the fact that the most basal members of a lineage, as Carroll himself would point out, are those which will display the most plesiomorphic anatomy comparative to other members of the ingroup, and will have more synapomorphies comparative to the outgroups, than will the most derived members of the lineage. Yet here we have the most derived members of the acanthodian lineage, from a morphological point of view, being advanced as the most basal. Furthermore, the time-dependent phylogeny must invoke massive reversal with no apparent causal factor, to explain why such a plesiomorphic form should have been recapitulated by the acanthodiids. All in all, this phylogeny is something of a mess.

An anatomy dependent phylogeny turns the matter upside down: the clearly plesiomorphic acanthodiids are labeled as the most basal Acanthodia, while the clearly apomorphic morphology of the climatiids is considered the most derived. This phylogenetic hypothesis, in addition to being far more congruent with what we know of evolution, has no need to invoke massive reversals to arrive at its conclusions, and is more parsimonious.

The acanthodian case, is exemplar of the difficulties in relying on stratigraphy alone, and stands as a warning to the prospective systematist who would naively adhere to idea that geological time must outrank all other considerations in phylogenetic reconstruction.

Acrimony, the Quest for Objectivity, and Cladistics
Cladistics being the immensely useful tool that it is, one must question naturally why it should be so contentious a subject. And contrary to the popular impression, there remains much debate on cladistics. While everyone agrees as noted that synapomorphic characters are the only traits which have phylogenetic relevance, nearly every other aspect of cladistic methodology and its implications are debated. Cladistics, quite simply, has been its own worst enemy, as there are practitioners thereof who fail to see cladistics as a tool only, however powerful and useful it is, and instead adhere to it with dogmatic tenacity, as great as that of any religious fundamentalist. For these systematists, cladistics has been elevated to the level of inerrant and purely objective, when in fact this is not the case. In their eyes cladistics has become the only objective and legitimate way of going about phylogenetic reconstruction, and all other methods are pseudoscientific, philosophical, or simply prejudiced on some personal grounds. This viewpoint is not only lacking in realistic corroboration, but it is immensely injurious to the science of phylogenetic reconstruction and will result in great damage to this very field. The ardent cladists have as warped a view of cladistic methodology as do those who categorically reject cladistics. This sort of religious deification of cladistic analysis is not science: those who advance cladistics as the "One True Way" are attempting to define nature, not describe it, and in so doing have taken up philosophy, and let science fall by the wayside.

Credits
This page incorporates material from EvoWiki and so is licensed under the Attribution-NonCommercial-ShareAlike Creative Commons License). EvoWiki url: http://wiki.cotch.net/index.php/Cladistics