Such expressions as that famous one of Linnæus, and which we often meet with in a more or less concealed form, that the characters do not make the genus, but that the genus gives the characters, seem to imply that something more is included in our classification, than mere resemblance. I believe that something more is included; and that propinquity of descent,—the only known cause of the similarity of organic beings,—is the bond, hidden as it is by various degrees of modification, which is partially revealed to us by our classifications (Darwin, 1859, p. 413f).

Sunday, 10 May 2009

The Science of Systematics

ResearchBlogging.orgIn a recent post, by John Wilkins (Evolving Thoughts), there is a quote by Borgmeier (1957):
    "As the science of order ("taxonomy"), Systematics is a pure science of relations, unconcerned with time, space, or cause. Unconcerned with time: systematics is non-historic and essentially static; it knows only a simple juxtaposition of different conditions of form. Unconcerned with space: geographical factors are not primary criteria in the definition of taxonomic units. Unconcerned with cause: systematics has no explanatory function as far as the origin of the system is concerned; it is merely comparing, determining, and classifying" (Borgmeier, 1957: 53).
On further reading we find this:
    "Systematics is independent of the theory of descent. This is admitted today [1957] even by convinced evolutionists. The reasons are as follows. (1) Systematic methods provide definite results without reference to the idea of evolution; phylogenetics has no special methods, it is essentially the interpretation of systematic facts. (2) Systematics is a science; phylogeny is a hypothesis of a historical process containing a fundamentally unverifiable element (Thompson) and can therefore never be the foundation of a science. (3) Systematics is [an] investigation of facts; phylogenetics is often 'a dangerous play with mere possibilities' (Hennig); Kant called it 'a daring adventure of the mind'

    Of course, any systematist is free to speculate on the probable phylogeny of certain species or genera, on the basis of systematic facts" (Borgmeier, 1957: 54-55; see also Williams & Ebach, 2009).
Notice the date – 1957. Cladism was gestating at the height of the Modern Synthesis. Mayr had already named his enemy – 'typology' – and created the essentialist myth. What was Borgmeier playing at? Was he an early cladist (Wilkins pers comm., 2009), or, was he someone, like Candolle (1813), who understood the importance of classification over inference?

The insistence that systematics and phylogeny should be treated separately would ally Borgmeier with Naef (1919) and Systematic morphology, rather than with the emerging numerical methods that would later predominate Hennigian cladistics. Borgmeier was not a cladist, but someone who understood the difference between phylogenetic inference and systematic classification, two fields that were as confused then as they are now. Borgmeier's message was aimed at readers of Systematic Zoology, namely evolutionary biologists and phylogenetists who insisted that their phylogenies (trees) were classification schemes (cladograms).

Borgmeier's points above may appear slightly heretical to modern day evolutionary biologists. After all 2009 is Darwin Year and a time to celebrate the achievements of evolutionary biology, rather than to dissect them. But if we do wield our scalpel at the underbelly of evolutionary biology, what do Borgmeier's three points mean for present day phylogenetics?

Decoding Borgmeier's Points for the 21 Century Phylogenetist

Point 1 "Systematic methods provide definite results without reference to the idea of evolution; phylogenetics has no special methods, it is essentially the interpretation of systematic facts." This quote consists of three parts. Take the relationship A(BC) for example. The result is definite in the sense that it states a relationship (i.e., homology, monophyly). This relationship is part of a classification, not a genealogical or phylogenetic lineage. That is, A does not necessarily have to be an ancestor of of either B or C. In fact A could be an extant mammal where as B and C could be trilobites (now extinct). This means there is no notion of time or transformation in cladograms (e.g., trilobites did not evolve from mammals). Cladograms are classifications, which depict systematic relationships that may include numerous hypothesized genealogical or phylogenetic lineages. If we skip to the third part of Borgmeier's point, namely '... it is essentially the interpretation of systematic facts', we see that phylogenetic inference comes from systematic relationships, not the other way around. In a modern context, we can hypothesize genealogical or phylogentic lineages once we discover cladograms and not the other way around (i.e., evolutionary taxonomists hypothesize lineages prior to finding classifications, hence the creation and acceptance of paraphyletic groups).

The second part to Borgmeier's point, '... phylogenetics has no special methods' is more relevant today than it was in 1957. Phylogenetic 'methods' are based only on inference. This means phylogenetic 'methods' are confused with phylogenetic models. Since models are immune to testing hypotheses, from a historical context, they fail. That is, models themselves are hypothetical and not based on actual observations. Phylogenies therefore remain hypothetical whether they fit the model or not. We will never know if A is actually ancestral to B or C through scientific methodology (i.e., testing). We can however only hypothesize which model is most 'likely', 'parsimonious' or 'similar'.

Point 2 "Systematics is a science; phylogeny is a hypothesis of a historical process containing a fundamentally unverifiable element (Thompson) and can therefore never be the foundation of a science." Point 2 proposes an interesting problem - Systematics as a science and phylogenetics as a hypothesis. Given this, the term Phylogenetic systematics appears to be an oxymoron. At one level taxa are treated in a systematic way (i.e., no concept of time or transformation), whereas on the other, characters and their states are treated as phylogenies (i.e., transformation, reversals, dating nodes etc.). Borgmeister may have seen the flaw in Hennig's system, however he didn't refer to it directly. Separating systematics as 'a science' from phylogenetics as 'a historical process' muddies the waters of cladistics. Hennigian cladistics happily confuses the two, whereas pattern cladists (sensu Brady, 1982) treats them separately (see Ebach et al. 2008). Moreover, Borgmeier's statement today would read differently; phylogenetics is now considered as a science and systematics its method. The confusion still continues.

Point 3 "Systematics is [an] investigation of facts; phylogenetics is often 'a dangerous play with mere possibilities' (Hennig)" I will not go into what 'facts' are, but for the purpose of this argument we may refer to systematics as an investigation of relationships. Point 3 is a more concise rephrasing of Point 1. Any systematic relationship may contain multiple hypothetical phylogenetic or genealogical lineages. The 'play with possibilities' becomes 'dangerous' once we use systematics to choose between them. Considering that systematics is silent about time, transformation and descent, it is impossible to use cladograms to choose one possible phylogeny over another. Other evidence is needed. After all, it is not the goal of systematics to find or propose lineages or find ancestors.

Borgmeier's three points are still relevant today. Whether phylogenetists will understand the dangers of confusing systematics with phylogenetics is another matter. The literature on this topic is readily available, but many do not realize that there is a problem. Understanding the nature of systematics and phylogenetics, their role in our research and their limitations, has more to offer than just another computer algorithm.

Malte C. Ebach & David M. Williams

References

Candolle de, A. P. (1813). Théorie élémentaire de la botanique ou exposition des principes de classification naturelle et de l'art de décrire les végétaux. Paris.
Borgmeier, T. (1957). Basic Questions of Systematics Systematic Zoology, 6, 53-69
Brady, R.H. (1982) Theoretical issues and 'pattern cladistics'. Systematic Zoology 31: 286–291.
Ebach, M.C., Morrone, J.J. & Williams, D.M (2008). A new cladistics of cladists. Biology & Philosophy 23: 153-156.
Naef, A. (1919). Idealistische Morphologie und Phylogenetik (zur Methodik der systematischen). Verlag von Gustav Fischer, Jena.
Williams, D.M & Ebach, M.C. (2009). What, Exactly, is Cladistics? Re-writing the History of Systematics and Biogeography. Acta Biotheoretica DOI:10.1007/s10441-008-9058-5.

8 comments:

David Marjanović said...

Considering that systematics is silent about time, transformation and descent, it is impossible to use cladograms to choose one possible phylogeny over another.I don't understand.

Cladograms show which phylogenetic hypotheses are the most parsimonious ones given the same dataset. The principle of parsimony is one of the two parts of the scientific method, and outright falsification is hardly ever possible in phylogenetics, so parsimony is the best we have...

Nomenclature and classification, and therefore taxonomy and systematics, are not sciences. Phylogenetics is one -- it is now one, thanks to a list of people from Hennig to Swofford and Goloboff; it wasn't one in 1957 outside of East Berlin.

(In fact, classification is wholly unnecessary if you use both phylogenetics and phylogenetic nomenclature. Why bother trying to "translate a cladogram into a classification"? Just publish the tree and tie labels to it in defined places.)

Yes, I know that Hennig saw phylogenetics as a means to an end, and classification as that end. But, in the fields I'm familiar with, nobody agrees with him on this anymore.

unverifiableThere is no such thing as "verify" or "prove" in science.

who insisted that their phylogenies (trees) were classification schemes (cladograms).Oh no, a cladogram is not a classification scheme! A cladogram is 1) the outcome of a cladistic = phylogenetic analysis and therefore 2) a phylogenetic hypothesis (well tested for parsimony against lots of alternatives).

Cladograms [...] depict [...] relationships that may include numerous hypothesized genealogical or phylogenetic lineages.In principle, yes, but in practice, anything that has autapomorphies is most parsimoniously interpreted as not being an ancestor of any other taxon in the same cladogram.

Taxa without known autapomorphies (metataxa) that are older than the oldest known representatives of what the cladogram shows as their sister-group enter into consideration for being the ancestor of that "sister-group", but that's it. Unless there's reason to believe that the fossil record of that particular group from that particular geological time is so complete that it would be unparsimonious to assume a ghost lineage, that's all we can say. Again, science cannot prove, only disprove.

phylogenetics is now considered as a science and systematics its method.I've never come across such a claim.

=========================

Evidently, Borgmeier did not describe cladistics. He described phylogenetics as it was done before that, and that was indeed an art and not a science, driven by scenarios about how it would make sense (to the author) that evolution would have worked on two or three characters, rather than by total evidence and the scientific method.

Roberto Keller said...

(i.e., evolutionary taxonomists hypothesize lineages prior to finding classifications, hence the creation and acceptance of paraphyletic groups)Studying some taxonomic monographs made by evolutionary taxonomists I rather get the impression that many paraphyletic groups result from the practice of erecting some taxa base on overall similarity, for example, and then arranging them phylogenetically either as internal or terminal nodes on a tree.

For example, you first create genera A, B, C and D and then go on to say that A is ancestral to B, from which C and D in turn descended. As such, from an evolutionary taxonomist's point of view, classification is first and phylogeny is added later.

David Williams & Malte Ebach said...

Considering that systematics is silent about time, transformation and descent, it is impossible to use cladograms to choose one possible phylogeny over another. I don't understand.Phylogeny is about ancestor-descendant relationships; cladograms are about character distributions. A number of phylogenies are subsumed in any one cladogram; choice among the phylogenetic trees is thus problematic.

Cladograms show which phylogenetic hypotheses are the most parsimonious ones given the same dataset. The principle of parsimony is one of the two parts of the scientific method, and outright falsification is hardly ever possible in phylogenetics, so parsimony is the best we have...Cladograms show character distributions. Parsimony might be understood as a particular algorithm or a way of looking at data. It is not one thing, as some would have us believe.

Nomenclature and classification, and therefore taxonomy and systematics, are not sciences. Phylogenetics is one -- it is now one, thanks to a list of people from Hennig to Swofford and Goloboff; it wasn't one in 1957 outside of East Berlin.Does this really mean:

Nomenclature = taxonomy
Classification = systematics

If so, it is a rather unusual way of looking at things. I suppose, then,

Phylogenetics = phylogenetics = science

Hennig, Swofford and Goloboff have all contributed many good things. The latter two work on algorithms. We are interested in notions, if you like. While some may see a logical connection between Hennig’s work and any particular algorithm, we do not see that as necessary.

(In fact, classification is wholly unnecessary if you use both phylogenetics and phylogenetic nomenclature. Why bother trying to "translate a cladogram into a classification"? Just publish the tree and tie labels to it in defined places.)Phylogenetic nomenclature? The Phylocode. OK, that’s for another time. Classification is unnecessary? To us, that is odd. After all, I can talk of monocots and vertebrates, and so on.

Yes, I know that Hennig saw phylogenetics as a means to an end, and classification as that end. But, in the fields I'm familiar with, nobody agrees with him on this anymore.Nobody? That’s a large crowd.

unverifiableThere is no such thing as "verify" or "prove" in science.

who insisted that their phylogenies (trees) were classification schemes (cladograms).Oh no, a cladogram is not a classification scheme! A cladogram is 1) the outcome of a cladistic = phylogenetic analysis and therefore 2) a phylogenetic hypothesis (well tested for parsimony against lots of alternatives).
See above but, if you like, ‘phylogenetic hypothesis’ = classification or the other way around.

Cladograms [...] depict [...] relationships that may include numerous hypothesized genealogical or phylogenetic lineages.In principle, yes, but in practice, anything that has autapomorphies is most parsimoniously interpreted as not being an ancestor of any other taxon in the same cladogram.

Taxa without known autapomorphies (metataxa) that are older than the oldest known representatives of what the cladogram shows as their sister-group enter into consideration for being the ancestor of that "sister-group", but that's it. Unless there's reason to believe that the fossil record of that particular group from that particular geological time is so complete that it would be unparsimonious to assume a ghost lineage, that's all we can say. Again, science cannot prove, only disprove.
Again, see above on trees and cladograms.

phylogenetics is now considered as a science and systematics its method.I've never come across such a claim.It seems to be the claim you are making above.

=========================

Evidently, Borgmeier did not describe cladistics. He described phylogenetics as it was done before that, and that was indeed an art and not a science, driven by scenarios about how it would make sense (to the author) that evolution would have worked on two or three characters, rather than by total evidence and the scientific method.We understand Borgmeier and his colleagues to be somewhat more scientific than you give them credit. Odd that scientific means only “total evidence and the scientific method”. One might see more than a little phenetics in that comment.

Malte C. Ebach said...

Roberto said: "Studying some taxonomic monographs made by evolutionary taxonomists I rather get the impression that many paraphyletic groups result from the practice of erecting some taxa base on overall similarity, for example, and then arranging them phylogenetically either as internal or terminal nodes on a tree. For example, you first create genera A, B, C and D and then go on to say that A is ancestral to B, from which C and D in turn descended. As such, from an evolutionary taxonomist's point of view, classification is first and phylogeny is added later."I am not certain if higher taxa (genera and above) are solely erected on overall similarity. Such taxa are erected in all manner of ways. What is the problem is when the description and name have priority over any systematic evidence. What if we discover via cladistic analysis, for example, that this genus is non-monophyletic? Then the taxonomy needs to be revised. This is why paraphyly is a taxonomic (and not a phylogenetic) problem - and this is why paraphyly = bad taxonomy (see Ebach et al., 2006). The phylogeny comes in after classification, that is taxonomy and systematics:

Systematic way:

Descriptive Taxonomy -> Systematics [to test for monophyly]- > Revisionary taxonomy [if group is not monophyletic] -> Phylogenetic hypothesis.

Evolutionary taxonomic way:

Descriptive Taxonomy -> Phylogenetic hypothesis [based on taxonomy] -> Systematics [used to choose which phylogenetic model fits best].

Paraphyly is rejected the using the former, where as paraphyly is chosen as the phylogenetic model in the latter. This means evolutionary taxonomists do not see cladistics or systematics as a test, but as an inference of phylogenetic models.

For instance, evolutionary taxonomists would mostly likely use the cladogram A(BC) to choose A -> B - > C over A -> C -> B. Again, they confuse trees with cladograms.

Ebach, M.C., Williams, D.M. & Morrone, J.J. (2006). Paraphyly is bad taxonomy. Taxon, 55: 831-832.

John S. Wilkins said...

I think it is unnecessary to force Borgmeier into the post-1966 debates. His view of classification, however, as a scientific activity, seems to me rather rigid: classification may come as a prelude to hypotheses about history, or it may come out of these hypotheses. There is no single, set pathway for methodology.

From the eighteenth century on, one sees people arguing about what is hypothetical (and at various times that means "unsupportable", sensu Cuvier) and what is "scientific" in the context of taxonomy. Since pretty well all generalisations are hypothetical, we should be past that by now.

Gustavo A. Ballen said...

But John, we are not forcing him to such a debate, since his main points are seriously conflciting with modern epistemology (post-Popper), and the notion of science, even the existence of historical sciences and nomothetic ones, are notions that were then known. I would agree with you in that "There is no single, set pathway for methodology," if and only if you're equating "pathway for methodology" with "knowledge," since science is but one strategy for generating knowledge, yet a powerful one when considering Hume's problem of induction. Those philosophical concerns were known by the date of Borgmeier's publication.

Also, you should be more clear when referring to "hypothetical" and "scientific," I think you (or people today) are confounding two completely different types of explanation. Even when a metaphysical statement or explanation qualifies as an explanation, Popper left us with a way for deciding whether we were doing science or metaphysics: the usage of empiricism or the potential of a hypothesis for being tested as a main concern when considering a scientific hypothesis (or explanation, it's the same) as such.

We need to be explicit in science, and distinctions as those outlined above are to me, important and seriour concerns in cladistics today, when philosophical topics appear to vanish between superfluous operational aspects.

John S. Wilkins said...

With respect Gustavo, Popper is not the last word on what counts as a hypothesis and scientific, and his demarcation criterion has been roundly criticised. I do not think his ideas work for the bulk of science and it in particular fails badly when applied to classification and phylogenetic systematics.

But that is for another forum.

Gustavo A. Ballen said...

Well, I hope you're not misunderstanding my point, and recognizing the importance of Popper's ideas is not the same as accepting them as "the last word", please note that it would be inconsistent with a scientific approach to knowledge, and I try to be consistent with that. Second, there appears as if confusions of the type outlined earlier are the main source of criticisms to the pertinence of Popper's logics to systematics. For example, and just as a little one, the popular idea concerning the testing power of the congruence test. I'd agree with Lars Vogt (Cladistics, 2008) in that congruence is by no means a popperian test; my criticism with his point is when he equates the epistemic component of claditics with the congruence test... it is not a popular view in these days, but congruence is not the testing phase in a cladistic analysis but a priori tests of homology what qualifies as the testing phase in cladistics. Kluge (several papers, 1990-2007) is right when emphasizes the importance of recognizing phylogenetic systematics as a historical science (as opposed to a nomothetic one)... this fundamental difference may be what causes you to think that popperian logics are not pertinent to cladistics. Again, this is but a point of view, and it is interesting to discuss others.