Palæos:		Cladistics
SYSTEMATICS		Cladistics - a Critique

Page Back

Page Next

Unit Up

Unit Back

Unit Home

Unit Next

Systematics Glossary

References

Cladistics - a Critique

Cladistics

(Phylogenetic Systematics)

Cladistics - also called Phylogenetic Systematics or Phylogenetic Taxonomy - is a method of classifying organisms by common ancestry, based on the branching of the evolutionary family tree. Cladistics is currently the most popular paradigm of phylogenetic classification in biological taxonomy.  Based strictly on determining branching points in the ancestry of organisms, it establishes groups based on their shared, derived features (synapomorphies), while ignoring primitive features (plesiomorphies) inherited from ancestors. Organisms that share common ancestors (and therefore have similar features) are grouped into taxonomic groups called clades. Like Evolutionary Systematics (which it has currently supplanted) Cladism is a method of classification based on the evolutionary history of organisms, dividing organisms into meaningful groups and subgroups. It was developed by Willi Hennig, an entomologist, in 1950, but was not really accepted until the 1980s.

Clades can be represented in terms of a cladogram. A cladogram is a branching diagram that depicts species divergence from common ancestors. They show the distribution and origins of shared characteristics.

Cladistics is based on three principles:

1. Groups of organisms are descended from a common ancestor.
2. At each node (divergence of a population), there are two branching lines of descendants.
3. Evolution results in modifications of characteristics over time.

Cladistic Phylogenetic Systematics acknowledges only Monophyletic groupings as valid.  Paraphyletic groups (accepted in the Linnean hierarchy and in Evolutionary Systematics) and Polyphyletic groups are rejected as invalid, as is the whole Linnean hierarchy above genus rank (although sometimes taxa such as family etc are used in a more limited context)

Cladists present cladograms are testable hypotheses of phylogenetic relationships. In this way, cladistic methodology can even be used to predict properties of yet-to-be discovered organisms.

Cladistics have the advantage over Linnean and Linnean-based systems, in that getting rid of higher taxa it also does away of the arbitrariness of whether a higher taxon is, say an order or a class.

A problem with the cladistic definition is determining which features are unique to that group (diagnostic in other words) and possessed by the most recent common ancestor, which are reversals (secondarily primitive, following the loss of advanced characteristics), and which evolved later, or independently.

To give an example.  When cladistics first came out there was a huge reaction against palaeontology.  The fossil record was considered irrelevant to deducing the history and relationships of organisms.  Some bright spark then came upon the idea that because birds and mammals have a number of metabolic and anatomical features in common (these being necessary in order for them to function as endotherms - warm-blooded animals) they must have descended from a single Most Recent Common Ancestor that was unrelated to all other (cold-blooded) reptiles.  So the clade Haemothermia ("warm blooded") was coined.  This strange concept appeared in some of the respectable scientific literature and was considered quite respectable for some time, although it received a humorous write up in New Scientist (I don't recall the volume number unfortunately).

What this example proves is that depending on which characteristics we use, we have a different common ancestor.  If metabolism is used birds and mammals are close cousins and crocodiles unrelated to each (the Haemothermia hypothesis).  If skull and heart structure is used birds and crocodiles are close cousins, and mammals unrelated to each (the standard hypothesis).  This proves that despite being an attempt to construct a more "objective" system than the Linnean one, cladistics still has to fall back on the same subjectivity and arbitrariness.

So the problem here is which characteristics are weighted (considered more important)

One way around this is to measure as many characteristics as possible. So even if some characteristics, like endothermy in the above example, is shared by both, other characteristics of the skull, skeleton, and soft body indicate separate evolution. Add to that things like evo-devo biology, molecular evolutionary analysis and the fossil record and we have a pretty good idea of the general picture. However, there is still the problem of more detailed small-scale phylogenies, and that is where there are often many competing cladograms, regarding which it is difficult to determine which is to be preferred. Change the features that are considered more important, or intrioduce a newly discovered species, and the whole picture may change radically.

Another illustration of difficulties facing cladistic methodology are given by vertebrate palaeontolgist Philippe Janvier (himself a dedicated cladist) in his book Early Vertebrates.

"Until the 1970s, discussions on interrelationships of major taxa were relatively simple to sum up. Since they were based on the consideration of a small number of characters. Most of them were 'pet characters' of one or another authority, and the 'minor' characters that contradicted them were regarded as homoplasies or simply not considered. With the advent of phylogenetic systematics, or cladistics, an increasingly large number of characters were taken into consideration for phylogeny reconstruction, but it was still possible to compare and discuss contradictory cladograms, since these were few. Since efficient cladistic computer programs came on the scene, the phylogenies, and the characters on which they are based have become so numerous that the task of comparing in detail the merits of respective phylogenies is now virtually impossible at the level of a general book such as this. This method of analysis of the characters, which no doubt will be in general use for a long time, is based on parsimony, and the only point that can be discussed by the average reader is the criterion of choice among a number of equally parsimonious trees. The consensus tree of these equally parsimonious trees produced by the computer is often the only result published by phylogeneticists, and it is generally not very informative. There are even instances of authors who have published only one of the 400 or 500 equally parsimonious trees they obtained, but without telling the reader! Moreover, the options Offered by current cladistic computer programs are numerous and the same data matrix may produce widely different sets of trees, depending on character weights, ordering of character states, etc. Less than ten years after the publication of the first computer-generated phylogenies of the major craniate groups, the discussion of the various competing phylogenies, character by character, has become quite difficult."

Another of the problems I find with the cladistic system is it's disregard of the sequence of fossils in the stratigraphic record.  This is based on the fact that cladists seem to assume it is enough simply to know all the characteristics of the representative members of a group in order to work out their precise phylogenetic relationships.  If the resulting cladogram agrees with the fossil record for the group (the "basal" (= primitive) taxa appearing first, the most "derived" (=advanced) taxa last, well and good.  If it doesn't the fossil sequence is rejected.  The reasoning seems to be that since the fossil record is so incomplete in any case the missing taxa simply weren't preserved, but lived elsewhere.  The result is a series of hidden or "ghost" lineages that are base don absolutely no evidence other than assumptions drawn from (possibly unreliable) cladograms.

A good illustration of the vagaries of (either on or both) the cladistic method and the fossil record is shown in the diagrams to the left, from Philippe Janvier's book Early Vertebrates.  At the top are cladograms of two families of  closely related early Devonian Ostracoderms (armoured jawless fish - in this case Osteostraci), the Kiaeraspididae (A) and the Boreaspididae (B).  Both groups of organisms are found in the same locality (the Wood Bay Formation of Spitzberg) and the animals when alive presumably lived in a very similar environment and had similar habits (lagoonal bottom-dwelling filter-feeders).  At the bottom is shown the stratigraphic range of each of these taxa. These creatures were geographically localized (endemic) and lived in conditions that favored preservation.  In addition their hard exoskeletons (head shields) were easily preserved.  So we have pretty optimal conditions here.

Now, here is something interesting.  The Kiaeraspidid (A1) cladistic analysis matches their stratigraphic occurrences (A2).  The most basal or primitive species (a) also appears earliest (at the bottom of the stratigraphic chart).  And the most advanced or specialized ("derived") species (g) and (h) occur last (top of time range diagram).

But the exact same cladistic analysis applied to the Boreaspidids (B1) does not match their stratigraphic occurrences (B2).  Although the basal taxon (i) is also among the earliest, some of the most advanced types (r) are equally ancient.

So we have a situation where apparent "ancestors" (well, the grand-uncles rather than the grand-fathers, because cladistics is all about the descendents of the common ancestor, never the common ancestor at the base of the node) appear after their "descendents"

The cladistic assumption is simply that the fossil record is incomplete.  These creatures evolved elsewhere, or if they evolved in that locality they were not preserved.  However, it is possible to reconstruct the "ghost lineages" - e.g. the dotted line in the Sigurdfjellet formation leading to the j/k/l/m cluster that appears on the next formation up (more recent).

I would offer an alternative and I believe much simpler (though obviously equally one-sided) explanation.  The cladistic analysis of the Boreaspidids may simply not be as accurate as the Kiaeraspidid analysis.   The j/k/l/m cluster, rather than being a basal group, may actually be an advanced one, like the s/t grouping, and it only appears primitive (perhaps through degeneration or loss of characteristics).

It is of course impossible to tell which of these two alternatives is the correct one.  Lacking a functioning time machine, one cannot travel back and actually observe these creatures evolving, making stopover observations at, say, 50,000 year intervals, while hopefully avoiding the grandfather paradox!  In that way phylogenetics is not and cannot be a hard science in the way chemistry, physics, or even neontological (study of extant species) biology is.  Between science and metaphysics there is not such a gap after all, a sobering realization that nevertheless gives a perceptive insight into the way our understanding of the world works.

I should point out that Professor Janvier, despite his critiques, still fully supports and uses the cladistic system, and is totally innocent of my misappropriation of his text and images. Indeed his Early Vertebrates, without doubt the best book in its field, and a classic without peer, is organized completely around a cladistic arrangement. It would be totally unfair therefore to somehow imply that he is criticizing or refuting the cladistic system. Similarily in my case, despite my concerns and misgivings as expressed here, I see cladistics as a superb tool for mapping out the "family tree of life". It is just that I don't consider it the absolute, fundamentalistic answer. Science deals with hypotheses, not religious or ideological truths.

The positive side of things though is that this method of considering character-state distributions prevents - to some extent - writers from falling in the pitfall of 'pet characters'. Second, the rapidity of such character analyses enables a large number of research workers to propose phylogenies based on various sets of characters.  Over a period of less than ten years, it appears that the trees produced for the same (mainly recent) taxa are more and more stable, that is, better and better corroborated.  Of course we have to remember that a consensus is not necessarily the truth.  Science can indeed be seen as a progression of more and more useful metaphors, but as Thomas Kuhn has shown it is not necessarily an inexorable march from ignorance to truth.

illustration from Philippe Janvier Early Vertebrates (Clarendon Press, Oxford, 1996), p.286

Links

Proceedings of a Mini-Symposium on Biological Nomenclature in the 21st Century - suggests replacing the Linnean system with a cladistic phylogenetic system of nomenclature.

Cladistics - notes and glossary

The Compleat Cladist: A Primer of Phylogenetic Procedures by E.O. Wiley, D. Siegel-Causey, D.R. Brooks, and V.A. Funk - first edition - available free for download in Adobe Acrobat format. (648 kb)

Glossary of Phylogenetic Systematics with a critic of mainstream cladism - Günter Bechly

Re: Quick cladistics question - Adam Yates makes some pertinent observations on difficulties arising in cladism through the use of Linnean terms.  From the Dinosaur mailing group archives

Much ado about species (and genus) level taxonomy - how the species concept fits in with cladism. A distinction is made "between lineages (a series of entities forming a single line of direct ancestry and descent; [e.g. a species]) and clades, clans, and clones (all paths or lines of descent from a given ancestor). Clades, clans, and clones are monophyletic, whereas lineages may be paraphyletic." by Thomas R. Holtz; from the Dinosaur mailing group archives

images not loading? | error messages? | broken links? | suggestions? | criticism?
contact me