Saturday, March 26, 2016

Discordant paraphylies executed

Strangely, editors of systematics journals do not appear to tire of opinion pieces rehashing the discussion about paraphyletic taxa that had already been laid to rest in the 1970s. The newest example of such a publication showing up in my alerts is Seifert et al. 2016 in the journal Insectes Sociaux, who  take issue with suggested taxonomic changes in ants.

As usual I would like to tease apart the argumentation, examine it for its merits, and consider if there is anything new in it that constitutes a good reason to accept paraphyletic taxa. But first, the title of the piece, which is (sorry to say) one of the most unwieldy titles I have ever seen on a scientific publication:
Banning paraphylies and executing Linnaean taxonomy is discordant and reduces the evolutionary and semantic information content of biological nomenclature
If I may attempt to rephrase a bit, I think the authors mean the following:
Banning paraphyletic taxa is incompatible with Linnaean taxonomy and reduces the evolutionary and semantic information content of biological nomenclature
...although that is still too long, and it is not clear what is meant with evolutionary and semantic in this context. At any rate, this already suggests what the two main arguments in favour of paraphyly might be. Both of them are not exactly new and have repeatedly been dealt with at length, but of course the hope is that the present paper acknowledges the cladist responses and provides additional counter-arguments instead of ignoring them.

As mentioned above, the contribution starts by considering a recent case in ant systematics where apparently several apomorphic segregates were sunk into larger genera. In particular, several highly specialised socially parasitic ant genera were sunk into the larger genera they are phylogenetically nested in. The immediate complaint is this:
Establishing the new system of Ward et al. will have severe consequences: practitioners studying biology of whole ant groups, comparing traits between related groups of species, studying mutualistic relations between these or simply making biodiversity studies in ecosystems or nature conservation will suffer from this reductionism. Being confronted with a multidimensional reality, they would be forced to use a language developed by a logical system knowing only two dimensions: time an phylogenetic splitting.
Talk of dimensions will become even more frequent in later parts of the paper, so at this moment I will just mention that I am confused in what sense a phylogenetic system has two dimensions. On the one hand, I think that the concept of dimensions does not apply to a nested hierarchy; at best, we are dealing with a strained analogy. But if we accept the analogy we have to conclude that a phylogenetic classification has only a single dimension, relatedness. And that is how it should be. Any useful system or classification must be based on a single criterion, otherwise it is inconsistent, confusing, and without use value.

(In this context I have previously used the Dewey Decimal System of libraries as an example. Its use value lies precisely in its property of employing the topic of a book as the sole criterion as opposed to mixing it up with cover colour, author nationality, or some other additional criterion in a misguided attempt to provide more "information content".)

More importantly, however, the quoted section makes several rather strong claims without supporting any of them. And in many cases I just do not have the foggiest idea how the endeavours in question would be harmed by a phylogenetic system. What difference does it make to the study of mutualistic "relations" if a species is called Tetramorium whatever instead of Teleutomyrmex whatever? And comparing traits between related groups of species or "making" biodiversity studies is best based on scientifically accurate classifications, whereas paraphyletic groups would actively misinform and mislead exactly those undertakings.

After a paragraph providing the ever-popular argumentum ad populum by citing fourteen papers arguing in favour of paraphyletic taxa (while ignoring the thousands of papers happily implementing phylogenetic classification that have appeared over the same time period) we are presented with what I like to call the "but they look so different" gambit; alternative names might be the Argument From Ignoring Intermediate Ancestors or the Argument From Denying That Evolution Is Gradual. In this case, the socially parasitic ants have undergone adaptive changes that make them look and behave rather differently from their closest relatives in the larger genus paraphyletic to them.

Also, there may be some confusion about autapomorphies, synapomorphies and convergent characters as well as their implications for classification. Taking the text at face value, the authors could potentially be taken to think that the second are the first, and the third the second.

Next follows a reiteration of the Argument From Misrepresenting Punctuated Equilibrium, which admittedly I am now seeing only for the third time. However, this is also where I start to find things a bit scary. I mean, I am certainly not an expert on rates of change and I have read only a few discussions in books, if memory serves two encyclopedia articles, and the odd evolutionary biology teaching material website about Punctualism. But even with that little background knowledge I can immediately tell that the authors completely mischaracterise it:

Punctuated Equilibrium is about most change happening at speciation events as opposed to between speciation events, while what the authors are talking about are large differences in rates of change on different branches of the tree of life as opposed to equal rates across all branches. Those are two completely different topics.

Note that I do not at all doubt that the latter differences exist; instead the points are that (a) I consider this fact irrelevant to the discussion about phylogenetic systematics, (b) evolution is still the gradual change of allele frequencies from generation to generation and thus not nearly as saltationist as it would need it to be for the argument to work, and, most to the point for present purposes, (c) such a complete misunderstanding of Punctualism does not exactly strengthen the overall argumentation.

Imagine I were to argue that Canberra is the capital of Australia because that is what the Soviet Politburo decided in 1964. Although my conclusion accidentally happens to be true one would be forgiven for taking everything I say on 20th century history with a bucket of salt after hearing my justification of the conclusion. Same here for any argument invoking patterns of evolution.

At this stage the paper has already presented several arguments at length, but it now says its arguments are yet to come:
Having briefly described the situation and the problem, we present in the following three lines of argumentation why the original generic names of social parasites are needed in the scientific language. The first argument focuses on functionality and semantic content of biological nomenclature, the second on a severe disaccord of the Hennigian system with Linnaean classification and the third on the evolutionary information content of paraphyletic taxa.
Functionality and semantic content

This section starts with a mere assertion that genera should be circumscribed based on similarity instead of relatedness. "The logic of constructing determination keys requires this, and further, teaching, nature conservation and science in general need these names for strong operational, semantic and mnemonic reasons." I do not understand where any of that comes from.

Identification keys should be written completely independently from the classification and be entirely based on what characters are easy to see and distinctive, otherwise you are doing it wrong. Seriously, I get a fit whenever I run into a key where the taxonomist uses some obscure, hard to define and hard to see character in the first couplet just because that divides the genus into the two subgenera they recognise. And especially so if a systematically useless but beautifully diagnostic character like yellow versus white bract colour would divide the species 50/50!

Teaching and science, for that matter, would benefit greatly from communicating, and amazing people with, the information that this ant over here is a Tetramorium that has evolved to be parasitic instead of misinforming them by claiming that it is not a Tetramorium. (Just think, birds are dinosaurs that evolved a bill and feathers! Isn't that an awesome realisation? Should that information be obscured and confused by pretending that birds are a group at the same level as and outside of 'reptiles'?)

As far as I can tell, the semantic and mnemonic angle is left entirely unexplained.

The second paragraph is an appeal to tradition: we shouldn't change names because "species lists would become invalid and all students of ants would have trouble reading old literature". This can only mean any and all scientific progress in biological classification is to cease immediately.

In related news, we should go back to calling Australia New Holland, because otherwise people will be confused by old maps.

"Disaccord" of Hennig and Linnaeus

I am not going to spend a lot of time on this because it is once more Richard Brummitt's argument: (1) If we want to apply Linnean ranks, and (2) if we have an ancestor to classify, and (3) if we are making an asynchronous classification, then the ancestor will need a genus name, and that genus will be paraphyletic to all other genera descended from that ancestral species. Thus if the ancestor of, say, all land animals were found as a fossil, all land animals would have to be in the same genus. Note the profusion of "ifs".

This is indeed the strongest argument against phylogenetic systematics that I have ever seen. (Not that that is saying much given its competition.) However, the problem can be solved trivially by relaxing the requirement to place a species at all Linnean ranks, or by having separate synchronous classifications, or by classifying all ancestors as terminals, or by giving ancestors a unique genus name based on the clade that is descended from them and then realise that a single species cannot be paraphyletic anyway given the definition of that word, or by just scrapping Linnean taxonomy as the pre-Theory of Evolution relic that it is.

The last solution is clearly unacceptable to the authors of the present piece, although one would have wished for a more explicit explanation why that is so. But as the list of alternatives shows such a radical change is not even necessary. There are many phylogenetic systematists who happily use Linnean ranks. Perhaps one might productively wonder how they manage to do so if the "disaccord" is supposedly so great. I would just like to point out that a hypothetical answer on the lines of "they are ideologically biased to such a degree that they just do not want to admit that it doesn't work" would reveal more about the person advancing such a claim than it would about phylogenetic systematics.


This is the oddest part of the paper. Taking the relevant text and the two figures at face value, the argument appears to be that we need to accept paraphyletic taxa because the two-dimensional projections of cylinder, cone, and sphere are all circles if projected along just the right axis. One is left to wonder what this has to do with a nested hierarchy; remember that the Linnean system, which the authors use themselves, has precisely as many dimensions as the PhyloCode - none if taken literally, one if we accept what I consider to be a rather tortured analogy. So whatever the shape of the object, it will invariably end up 'one-dimensional' when grouped into such a nested hierarchy.

There is a reference to a paper by Elvira Hörandl, so it is presumably her argument about the information content of classifications that they were going for. The argument is that whereas a phylogenetic system contains only information on the relatedness of taxa, an 'evolutionary' system with paraphyletic taxa contains that plus other information, especially on the degree of divergence between two taxa.

Although I can see where this is coming from, I am not convinced. First, the degree of divergence is just another way of saying "but they look so different" - an approach based on this logic once more ignores the gradualism of evolution and would thus be thrown into confusion by any newly found intermediate fossil. Second, whereas using similarity as a criterion in classification would by itself be unproblematic, the 'evolutionary' approach is to use it at the same time as the totally different criterion of relatedness. This means that the end user of a classification, i.e. anybody who reads or says the word "Tetramorium", won't be able to know if that group they just mentioned is a clade or a phenetic cluster. The 'evolutionary' classification consequently has zero reliable information content as far as the end user is concerned.

Which brings us to the third point, which is that the argument is build on confusing the fact that a lot of information was considered by the taxonomist when they built a classification with the ability of the end user to extract that (or any) information. Only the latter is what matters in practice, and here a single-criterion classification will always be superior, be it an ecological or a phylogenetic one.

Reticulate evolution

The last section surprisingly adds yet another previously unmentioned argument, although only in passing. Supposedly "botanists" - it is implied most of them, but only two are cited - are much more open to paraphyletic taxa because they encounter more reticulate evolution. I would be surprised to learn that hybrid speciation happened across such phylogenetic distances as to make phylogenetic systematics impossible. Excessive reticulation would also make taxa non-paraphyletic (one might say non-phyletic), so as an argument for 'evolutionary' systematics it is self-defeating. What is more, the authors seem to conflate systematics-relevant lines of descent with systematics-irrelevant occasional horizontal gene transfer. Even more confusingly, apomixis is mentioned in the same breath, although as asexual reproduction it actually makes things easier for cladism.


As so often I am unable to follow where a pro-paraphyly paper sees problems with phylogenetic classification. "They look so different": There were intermediate ancestors between the two extremes we see today - what about them? Semantic content: The argument here seems to boil down to an aversion to name changes, but the only alternative is to arrest scientific progress. Discord between Linnaeus and Hennig: Seems to be a non-issue to most practising systematists. Information content: Yes, phylogenetic classification reflects only a single criterion, but 'evolutionary' classification is the worst possible solution because by design it cannot follow either of its two contradictory criteria consistently. It is good for nothing. Reticulate evolution: If it happened as often and across as large distances as required to make phylogenetic systematics impossible, it would also make the 'evolutionary' approach impossible. But it doesn't anyway.

And sadly, I did not see cladist counter-arguments being acknowledged and addressed anywhere. That's a pity, because in this way we will only ever be going around in circles.

No comments:

Post a Comment