Wednesday, January 6, 2016

Aubert's analysis of phylogenetic terminology, part 4: cladism impossible?

Continuing the discussion of this paper from here, here, and here, and working through the main claims of the paper as I see them:
  • The various definitions provided in the paper are in some way better than the ones that are currently accepted.
  • There is no relevant difference between the systematics-relevant relationships and structures existing at any level of the diversity of life. (E.g. mother > daughter is completely equivalent to bony fish > land animals - they can all be drawn as diamonds and arrows, right?)
  • A strictly phylogenetic classification is formally impossible.
  • Cladism is part of structuralism and therefore characterised by "anti-realism and a metaphysical way of thinking".
  • Cladism is built on biologically unrealistic assumptions that have been empirically falsified.
  • There exists an objective approach to delimiting paraphyletic groups.
  • It would be preferable to have two parallel classifications, one of clades and one that includes taxa that are allowed to be non-monophyletic.
So today is going to be about the claim that a strictly cladistic classification is formally impossible.

Staying at first at the abstract level, imagine a group Aides. Its ancestral species A split into species B and C. Subsequently, B split into species D and E, while C split into F and G. The following is a strictly cladistic classification, and it does not look impossible:

Aides = { A, Bides = { B, D, E }, Cides = {C, F, G } }

Realistically, we will probably not know the ancestral species in most cases, so in practice we will be dealing with synchronous classifications like this:

Aides = { Bides = { D, E }, Cides = {F, G } }

Again, this is a strictly cladistic classification, and it is not self-evidently impossible. So what is going on?

What is going on is this: A strictly cladistic classification is impossible if and only if all the following premises are accepted:

1. We have fossils that we can characterise sufficiently well that we can really talk about 'species', as opposed to, say, merely pollen fossils that belong somewhere into our group.

2. We want an asynchronous classification of all species that have ever existed, as opposed to one classification for every major time-slice.

3. We insist on classifying the fossil species as ancestral if they look the way our analyses imply the ancestor would perhaps more or less have looked like, as opposed to classifying them as another terminal.

4. We insist on using the Linnean ranked categories, and here in particular binominal species names, as opposed to finding a different solution that is less artificial and pre-evolutionary.

The key is #4, because binominal species names consist of genus name and specific epithet. So if an ancestral species needs a genus name, that genus will automatically be non-monophyletic to all its descendant genera unless all descendants are collapsed into it. This is the same as Richard Brummitt's argument, so nothing new.

Note, however, that for the argument to work, all four premises have to be accepted; reject a single one, and it will fail.

#1: Most systematic work on the planet has to make do without any good fossils, so the problem never arises in the first place. That, however, is admittedly the strongest link in the chain as additional fossils could always turn up.

#2: For practical purposes, most humans care only about extant species except in a "dinosaurs are cool" sense, so it is not immediately clear that we should worry about accommodating fossils in the same classification as we use for managing weeds, food plants and protected species. Similarly, one could easily produce a classification for the flora and fauna of the Cretaceous without worrying about the descendants of some of the species in it. It is at least not self-evident that an asynchronous classification is required.

#3: Most cladists and, I would argue, also most practising systematists who are agnostic about cladism, treat all fossils as side branches. In fact phylogenetic analyses are generally designed to work that way, and it would be difficult to do it any differently without begging the question. This is where Aubert argues that cladists are anti-realists for excluding the possibility of a fossil being a real ancestor. More on that perhaps another time, but I would argue it is a case of pragmatism and prudence instead of anti-realist philosophy.

#4: In practice, the creators of modern classifications have become very pragmatic about leaving groups unassigned at some of the Linnean ranks, and I see no problem with that. Conversely, many researchers increasingly use informal clade names such as Euasterids II, not least because there are simply not enough Linnean ranks. There is no good solution yet for leaving species unassigned at genus rank, but that is only because of the weight of tradition. Some are ready to seriously consider moving beyond the state of the art of the 1750ies.

What I want to stress is that while e.g. the late Brummitt and here Aubert simply assume(d) all four premises to be obviously indisputable, many practising systematists would not agree with that assessment, and consequently they have no reason to believe that a phylogenetic classification is impossible.

But even to those who do insist that all of them are indisputable the question must at some point arise whether the supposed impossibility of phylogenetic classification is the fault of an ill fit of cladism to empirical reality, as the paraphylists believe, or the fault of an ill fit of the artificial Linnean system to reality.

Or to look at it from another angle, at some point there is no way around the realisation that any asynchronous classification that tries to treat species as static diamonds-in-jelly will produce absurdities. The reason is simply that ancestral species were not actually static but have over time gradually evolved and diversified. The descendants aren't more diamonds-in-jelly that were specially created to fit into somebody's graph, but they are what the ancestors turned into.

And here is the rub: Cladists and, whatever other downsides their attempt may have, the people behind PhyloCode, are asking frankly how we need to change our approach to classification to accurately model and describe the products of evolution. On the other hand, 'evolutionary' systematists and, whatever other advantages the traditional system may have, the people who take it for granted that we should never under any circumstances relax the demands of Linnean ranks and binominals, seem to be asking themselves how best to force a recalcitrant reality into abstract, fixed categories that were invented still under the assumption of special creation.

To repeat what I wrote in a reply under my last post on the issue, the latter approach looks a bit too much like trying to hammer a square peg into the round hole of an ideal. We can't do anything about the peg, because changing reality is not an option. Perhaps we should try a different hole.


  1. In your first cladistic set you have two trichotomies, which is forbidden by the very cladist dogma that insist on dichotomies. So you have in fact written an impossible example and you corrected it by eliminating the ancestors.

    1. Fossils belong to species. So it indisputable that in principle we must classify fossils as species. And we have sometimes good evidence that a particular fossil belong to a well characterized species.

    2. Your are advocating that a bunch of many many synchronous classifications is preferable to a single comprehensive one, only to save your holophyletic principle. It is the same as Podani's pitfall in his papers about this very topic. Furthermore, I would argue that synchronous paraphyletic species do exist, so considering only synchronous classifications doesn't escape the problem (but I will develop that argument in your previous post).

    3. See refs in my paper, fossils can sometimes very confidently be characterized as ancestral to another one. It is of course anti-realism to ignore this fact just because it doesn't fit the round hole of your ideal.

    4. This is also my solution. Cladists aren't interested in taxonomy and evolutionists aren't interested in cladonomy. So we could have two separated systems, one with ranks, one without them, and we would cooperate on our common interest about phylogeny.

    1. What you call cladist dogma entails that supraspecific taxa should include their commonly ancestral species and all its descendant species. That's bascially it. Cladists disagree about many other things, including whether to accommodate to Linnean ranks or whether to drop them, whether we can know that a species is ancestral or not, whether to use Hennig's internodal species concept or the Composite one.

      So you are either straw-manning or at a minimum picking a very tiny fraction of the cladist spectrum to represent all of it.

      But really, where do you get the idea that there can only be dichotomies in the classification? Are you confusing this with the idea that species will generally arise from lineage splits or with the way phylogenetic analyses work? Look at any modern phylogenetic classification and you will find situations like twelve subfamilies in the same family. Definitely not a dichotomy. Nobody says that we have to recognise every single clade as a taxon, only that that only clades may be supraspecific taxa.

      1.-3. I should have made clearer that I am not advocating for any of these. I do believe that we can have fossils that can - as always: tentatively - be considered to be ancestral, and that an asynchronous classification is desirable. My own disagreement is with #4, and that is where your argument falls down for me, personally. Just wanted to point out that lots of people do not have the problem you discuss in your paper for other reasons.

      4. Taxonomy is naming of groups, so cladists are interested in taxonomy. I still have to understand what an 'evolutionary' classification is good for; its is uninformative and has no use-case. What is more, a cooperation on phylogeny would not be possible because 'evolutionary' classifications misrepresent and misinform about both phylogeny and patterns of evolution.

    2. You cannot both complain about linnean ranks and saying that there are not enough of them (besides, you can add as many as you want), and then saying that thanks to them you are obliged to represent misleading clades. One of the main argument of cladism is that cladification is a direct translation of cladogram whereas evolutionary classification is supposed to be more subjective. So, how do you objectively decide which clades should be taxa? I guess the answer is related with most significant synapomorphies, most distinctive clades and macroevolution, so more or less the same criteria than evolutionary taxonomists use to determine paraphyletic taxa. So, cladists who advocate that all clades should be taxa are definitely more self-consistent. Not being interested in self-contradictory cladists is not really straw-manning, quite the contrary in fact.

      Taxonomy is about naming taxa, i.e. groups of organisms having an affinity. Cladists are just naming clades and they think these are taxa. If you reject the linnean/darwinian concept of affinity, and thus ranks, it is consistent for you to prefer cladonomy. But concepts of taxa and clades should not be conflated or confused, which is too often the case.

      Why wouldn't phylogeneticists from different schools cooperate? Both are reconstructing cladograms the same way, only formal classification differs. Many sciences are using different kinds of clusterings for different purposes, I don't see why systematics should be the only science to use only one classification for every purposes. You can think that evolutionary classification misrepresent evolution, there is no problem as long as you let your colleagues that find it useful using it. That's why I advocate two distinct classifications for distinct purposes. And then just use and amend the one you wish without interfering with the other one.

    3. You cannot both complain about linnean ranks and saying that there are not enough of them (besides, you can add as many as you want), and then saying that thanks to them you are obliged to represent misleading clades.

      Unfortunately I do not understand the end of this sentence. But if I understand the Linnean system correctly, you cannot just add more ranks, although perhaps the relevant Taxonomy Sessions of the international congresses could. I suspect it would get confusing quite soon though.

      cladification is a direct translation of cladogram

      The classification should not contradict the tree of life. That is not the same.

      So, how do you objectively decide which clades should be taxa?

      You are confusing two different issues. The question is how to know whether we can recognise a taxon, not how to know whether we absolutely need to recognise it. Because cladism is based on relatedness, it is always reproducibly testable through phylogenetic analysis, and clear given the best phylogeny we can come up with at the moment. Because 'evolutionary' systematics relies on 'important' characters or long branches, the answer depends on the subjective hunch of an individual taxonomist about what is important or long enough. Unless the answer is that 'E'S allows ANY paraphyletic taxon to be valid, in which case we'd have utter chaos anyway.

      cladists who advocate that all clades

      I have yet to meet such a cladist.

      More importantly, your beef is with taxa having to be monophyletic, and your paper operates on the dichotomy of cladist vs. 'evolutionist' throughout. It follows that everybody who publishes papers in PSE, Taxon, BJLS etc. trying to recircumscribe genera as monophyletic sensu Hennig is a cladist as per your paper. It then follows that under that definition the vast majority of cladists do not advocate this, end of story.

      Even paraphylists will have at least one clade in any of their classifications - dividing the study group up in a way that precludes this would only be possible with polyphyletic taxa. So it seems a bit odd if not self-defeating for you to claim that a clade cannot be a taxon by definition.

      Funny how under your solution you and perhaps a couple dozen others tops would get to have the established official classification and the >95% of colleagues who work towards a phylogenetic classification are supposed to go away and make a new one...

    4. Then evolutionary classification does not contradict cladograms either. Groups are just mapped another way on them. Yes, any paraphyletic taxon is potentially valid under evolutionary systematics since topology (besides forbidding polyphyly) is not considered a criterion. However, to decide which one to select you have to pay attention to biological and historical criteria. Long branches and important characters can be evaluated algorithmically to make groups the most homogeneous and the gaps the sharpest. So it is perfectly reproducible and testable against fossil record when available (for example to verify that gaps indeed matches true leaps, or that long branches matches revolutionary periods of high rates of evolution).

      Of course I never said that evolutionists will never have clades in their classifications. Clades are quite good at representing evolution when evolutionary rates are constant. It is only under this condition that they maximize information content.

    5. Yes, they do, whenever there is a paraphyletic group. I thought you had acknowledged that there were no long branches in reality.

      Fine, can we then drop that nonsense about clades not being taxa? Note once more your subsequent argument only works if phenetic information is assumed to be the only relevant information.

    6. No they don't. As long as you can map a group on a tree there is no contradiction between the classification and the cladogram. You are confusing "being isomorphic to" and "being consistent with". Only polyphyletic groups are inconsistent with phylogenetic trees.

    7. E.O. Wiley has discussed the matter in various publications in greater detail than I can afford at the moment, and Hennig's reply to that Mayr paper in the early 1970ies is another good examination of the problem.

  2. Hrm, ah, well, I'm not disagreeing with your larger point, but as a paleontologist (who works on animals) I'd point out that we'd get a lot of grief if we didn't work in the same system of taxonomy as modern biology. And that means having to account for things like extant genera who are morphologically indistinguishable from fossils from the Cambrian (pterobranch genera, if you were wondering).

    As for ancestors, well, in some well-sampled groups it is common to discuss ancestor-descendant relationships. Quantifying them is another story, but very recent advances in phylogenetics have made sampled-ancestor phylogenetics a real thing:

    Which means soon we will have fossil + extant phylogenies with OTUs actually placed (or not) as ancestors.

    1. Thanks for the comment and reference! As I replied above, I should have made clearer that I do not necessarily agree with all four points myself.