Thursday, January 28, 2016

Aubert's analysis of phylogenetic terminology, final part 8: two parallel classifications

Continuing the discussion of this paper from here, here, here, here, 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".
  • It would be preferable to have two parallel classifications, one of clades and one that includes taxa that are allowed to be non-monophyletic.

Why not have parallel phylogenetic and 'evolutionary' classifications?

Damien Aubert ends his paper with the suggestion to have two parallel classifications: The official system of living organisms should allow paraphyletic taxa, and all the mainstream systematists who want to make groups monophyletic should go away and build up a new, separate system of nested clades.

At this moment one might wonder how this suggestion can be consistent with earlier arguments that (a) a strictly cladistic classification is formally impossible and (b) clades can't be taxa, that is named groups of organisms, anyway. If Aubert believes his own arguments, the separate phylogenetic classification he suggests is not possible. If he believes his solution to be possible, those other arguments must be wrong. Presumably the idea would be that the cladistic system should not be called a classification, and that named groups of organisms are not necessarily taxa, but only if they are paraphyletic. Which doesn't make sense to me because even 'evolutionary' taxonomists have to recognise some monophyletic taxa.

Anyway, let's assume that a phylogenetic classification is possible, and let's also disregard for the moment that Aubert would like to reserve the official, widely used classification for what is currently a non-mainstream, minority position. Why not have two parallel classifications?

Of course we do already use parallel classifications all the time. We classify organisms by the food they eat into predators, herbivores, parasites and carrion eaters. We classify them by their life cycle into annuals and perennials. We classify them by their growth form into trees, shrubs, and herbs. We classify them by pollination syndrome into melittophilous, ornithophilous, anemophilous or myophilous. We classify them by dispersal syndrome into myrmecochorous, endozoochorous, exozoochorous or ballistochorous. We classify them by our interactions with them into pests, weeds, ornamentals, crops, lifestock, game, etc. We recognise other polyphyletic groups such as lichens. And, since Darwin, we understand that the one official, natural system of life should be organised by relatedness, which despite not having the concept of synapomorphies available Darwin himself already argued natural to mean in this context, and which is what phylogenetic systematists are working towards.

So if we can have just one phylogenetic classification next to all the non-phylogenetic ones, fine, count me in; that's all that phylogenetic systematists ever wanted.

But the problem I have is that the so-called 'evolutionary' classification that allows monophyletic and paraphyletic taxa does not appear to be good for anything. It does not appear to have any information content, and thus it doesn't have any use case. The classification would sometimes group by relatedness. But sometimes - if the difference in morphology is intuitively large enough (leaving open what large enough means) or, as Aubert argues, if there was a large enough shift in diversification rates (ditto), it would not group by relatedness. It would be inconsistent and unreliable.

If we want to know about groups of close relatives, for example for bioprospecting or to design a biogeographic study, we could not rely on Veronica to be an inclusive group of close relatives, because who knows if in that case the last monographer used the criterion of Hebe being shrubby to separate that subgroup out. So in that case a phylogenetic system would serve us better.

Aubert's paper constantly returns to phenetic information content ("naturalness means maximizing the overall similarity of organisms withing grous"), but if that was our concern we could not trust the 'evolutionary' classification either, because who knows if Veronica wasn't circumscribed as monophyletic by somebody who doesn't consider shrubbyness to be all that impressive a difference, given indistinguishable flowers and fruits? And who gets to say what counts into "overall", and how to weigh the different traits? So in that case several single-character phenetic systems that contain groups like 'herbs' or 'shrubs' would serve us better.

Yes, we could look up if Veronica is paraphyletic in the 'evolutionary' classification and to what other genera, or if the shrubs are included or not. But as I have argued before, the whole point of having a classification is that you do not have to look up the details. The whole point of having a category name is that it alone will already reliably and consistently provide some information on the content of the category. The whole point of the natural system is that a layperson can learn a genus name and assume that it refers to an evolutionarily meaningful unit instead of part of an evolutionarily meaningful unit.

Shrub has direct information content in a single-character phenetic classification based on growth form. Veronica has direct information content in a phylogenetic classification. But Veronica does not have any reliable information content in a so-called 'evolutionary' classification.

So even if the entire scientific mainstream were happy to leave the official classification to a small movement that appears to have lost the argument sometime in the 1970ies, I would still have no idea what an official classification with paraphyletic taxa would be good for. One idea might be to uphold a tradition, on the lines of making it easier to refer back to classifications from 1920. But that is not strong enough an argument to outweigh other considerations, nor is it what science is supposed to do. We don't uphold the traditions of using blood-letting or assuming special creation either.

Another argument appears to be the one advanced by Richard Zander and now Damien Aubert himself, that paraphyletic taxa are necessary to study macroevolution. Most evolutionary biologists, however, seem to do fine with only species and clades, and so I cannot help but wonder if the argument is not a bit circular: Macroevolution is first defined as one supraspecific taxon producing another such taxon at the same level, and then by definition there is no macroevolution if paraphyletic taxa aren't allowed.

None of the research topics listed here, for example (accessed 27 Jan 2016), - adaptive radiations, changes in biodiversity through time, genome evolution, mass extinctions, estimating diversification rates, change in rates of evolution, and the role of development in evolution, have anything to do with or require paraphyletic taxa. In fact in some cases any taxon above the species level that isn't a clade would be positively misleading.

Again, I personally don't see a use case for 'evolutionary' classification, and thus I have no idea why such a system needs to exist.

This concludes my series of posts on the paper. As always I am doing this because I am personally interested in the paraphyly discussion, hope to understand the arguments of the other side, and remain curious if I will ever come across that shows phylogenetic systematics to be undesirable or impossible.

In this case the most novel aspect, at least to me, was the idea of basing the delimitation of a paraphyletic taxon from a clade nested within it on an acceleration of the rate of evolution and/or the diversification rate in the nested clade. Unfortunately I still think that this makes the mistake of defining a group artifiically based on not having something that another group has (non-Russians are not a group equivalent to Russians), and in addition the idea seems like a non sequitur to me (WHY should such a shift define a taxon?). But it was a new idea, and so I found the reading rewarding.


  1. "all the mainstream systematists [...] should go away and build up a new, separate system of nested clades".
    I argue that we should work together on an integrated dual system, not two independent ones. Maybe I was not clear enough about that? Moreover, you argue yourself (right?) that we should discard ranks. So, you may also see the situation from another point of view, we go away and we bring our ranks with us. But all of this is just emotional blah blah. Both should be considered equally official.

    Arguably, cladist taxonomists are also a minority today in the systematics community. Most of the young phylogeneticists don't bother with taxonomy, they just want to resolve trees and name nodes. They do not care about the philosophical debate about naturalness or artificialness of holophyletic groups, naming nodes allows them to avoid making complex taxonomic decision. Codes have become cumbersome, and the names of the node, through often latinized to some degree, are becoming more and more informal, i.e. not validly published as taxa. There are many examples: Patterson's "stramenopiles", Cavalier-Smith's "neomura", AGP's "eurosids II" (!), Mast & Thiele's "/Cryptostomata", etc. My proposal is to formally recognize this situation.

    Concerning consistency, there are two points to consider. Firstly, it is not because I regard cladistic classification as a hollow and misleading enterprise that I should prevent cladists to pursue their vain goal. Every one is free. Secondly, although strict cladistic classification is impossible, one can build a relaxed version providing (1) paraphyly is allowed at the specific and infraspecific level, (2) unresolved polytomies are allowed, (3) clades are unranked (hence, it is not properly a "classification").

    Concerning information content, your metaphysical opposition between relatedness and phenetic is just misleading. Evolutionary systematists don't build their taxa with either the first or the second criterion, but always with both! So, simply stated, when one is dealing with an evolutionary taxon he knows that (1) the group is constrained to be of monophyletic origin, (2) all its species live in the same more or less broad adaptive zone or ecological niche depending on the rank (so, if he knows a sister taxon of the same rank, then this gives him an intuitive idea of the disparity), (3) two species from the ingroup have greater probabilities to be more similar to each other than to a species from another taxon. Example: species 1 and 2 belong to taxon A and species 3 to taxon B. I know that species 1 have trait "a" and species 3 have trait "b". Then likelihood is higher that species 2 have trait "a" but not trait "b". This degree of likelihood depends on the rank.

    Your notion of "information content" is far too narrow, restricted only to cladistic relatedness or to single-character keys. This is typological thinking, it is completely opposed to probabilistic thinking.

    "a layperson can learn a genus name and assume that it refers to an evolutionarily meaningful unit instead of part of an evolutionarily meaningful unit"
    That's why evolutionary systematics is more consistent, paraphyletic taxa can be evolutionary units.

    1. The point is, the 'young phylogeneticists' you speak of want to name clades, not partial clades which they would consider to be non-groups. I do not quite understand how you manage to convince yourself that you are at the same time arguing against naming only clades but not against anybody except 1970ies old school cladists in the very strictest sense.

      "Strictly cladistic classification is possible if..."

      (1) Paraphyly does not apply at or below the species level, and never has. (2) I have yet to learn which cladist ever demanded that the system be dichotomous; surely no phylogenetic classification ever had problems with several groups at the same rank. (3) I do not believe that having explicitly named ranks is any part of the definition of a classification. If I divide cars into trucks and personal vehicles I have a classification, although a very primitive one, but that's that. I conclude that a strictly cladistic classification is possible.

      all its species live in the same more or less broad adaptive zone or ecological niche

      Do you realise how vague this sounds? You can justify just about anything with that as long as you are willing to consider something as fuzzy like "living on dry land" an adaptive zone.

      This is typological thinking, it is completely opposed to probabilistic thinking.

      No, typological thinking is to classify organisms by their fit to an ideal type of a taxon - like the centroids in your phenetic clusters, for example. Classifying by relatedness is completely different.

      Probabilistic thinking in the same context as classification looks like a category error to me. We put stuff into the right place in a classification because we tentatively conclude it belongs in that place, not because we really prefer to rely on luck when extracting information from a name.

      No, paraphyletic taxa are not evolutionary units. Species are the highest rank in biology that are actually the subject of evolution, and a clade is what a species has turned into as it diversified. Dandelion and common lawn daisy, for example, do not share any common or related or correlated evolutionary destiny in any way whatsoever, although they are considerably more closely related than many species you would want to put into one box. (They may try to shade one another out though.)

    2. From the years 1990ies, phylogeneticists became somewhat Felsensteinian. You can't seriously say that Felsenstein is a cladist!

      I disagree with all of your 3 replies concerning strict cladistic classification, but I have already explained all of this somewhere else.

      "typological thinking is to classify organisms by their fit to an ideal type of a taxon" Then evolutionary classification is not typological by your definition since evolutionists don't rely on an ideal type. Centroids aren't considered meaningful by us. On the contrary, the nested set of synapomorphies that define clades are just nested types, "character essentialis" in Linnaeus' terminology.

      "not because we really prefer to rely on luck when extracting information from a name" So you prefer artificial typology. But nature does not work like that.

      "No, paraphyletic taxa are not evolutionary units." Yes, they are. You cannot explain intertaxic competition, or higher taxa selection without, or heritability of extinction and speciation rates without considering higher evolutionary units.

    3. The relationship between Felsenstein and the Willi Hennig Society is irrelevant to whether you will convince people to accept paraphyletic taxa.

      Your constant attempts to call phylogenetic systematics names do not impress me any more than they did a few days ago.

      If I need higher units clades will do just fine, thank you very much.

    4. I don't try to impress anybody. Words are important because concepts are important in science, as well as in philosophy of science.

  2. "the argument is not a bit circular: Macroevolution is first defined as one supraspecific taxon producing another such taxon at the same level, and then by definition there is no macroevolution if paraphyletic taxa aren't allowed." It's a definition, so it is not circular. You must not confuse macroevolution and large-scale microevolution. You can still argue that macroevolution doesn't exist and that it can be reduced to microevolution. Many think it is not the case.
    In the list you provide:
    - "adaptive radiations" How do you tell if your radiation is adaptive or not, if not recognizing adaptive zones and macroevolution? You can still avoid naming paraphyletic taxa but you have formally ackonwledged their evolutionary properties by contrasting a subclade to its parent clade.
    - "changes in biodiversity through time" How do measure biodiversity? You cannot use unranked clades to do this. You must define families for example and count them. But this is meaningless if your families are just arbitrarily ranked clades. Biodiversity cannot be estimated without a proper theory of ranking, hence the use of comparable disparity density and the exclusion of divergent autophyletic subclades.
    - "change in rates of evolution" When acknowledging this phenomenon you contrast a subclade to its parent clade, so you recognize that the parent clade without this subclade is a meaningful unit, whether you formally name it or not.

    "WHY should such a shift define a taxon?" I may reply provocatively, why not? And also why holophyly alone should define a taxon?

    "But it was a new idea, and so I found the reading rewarding." I am happy you had fun reading my paper, but I must admit the idea is not new at all. It is at least as old as Simpson's book (1944) "Tempo and Mode in Evolution".

    "I am personally interested in the paraphyly discussion, hope to understand the arguments of the other side". I hope you will, as I did!

    However, you may still disagree forever about this. But we must agree that the end user's interest is to have a clear system. The whole point of my conclusion is that it would be beneficial to clearly state that a taxon is known to be paraphyletic with for example an asterisk like Osteichthyes*, or that a clade is known to be dysphyletic (either mixophyletic, lumping too much, or schizophyletic, splitting too much) by dropping the capital letter. Phylogeneticists not interested in taxonomy could just name their nodes this way.

    You can still not see the point of using paraphyletic taxa, but you must not prevent people, who see an advantage in them (for example, it is more stable! See Ruggiero et al. 2015), to use them.

    This dual system is based on two principles: transparency and freedom.

    Really, what's the point of recircumscribing Osteichthyes* to include Tetrapoda where you could just coin a new word like euteleostomi to encompass them? We could endlessly debate to know which one is artificial and which one is natural, but there is no advantage in fighting for using the same name for different things.

    1. Okay, so it is a definition - one that virtually nobody accepts or takes seriously.

      Adaptive radiations - People easily recognise adaptive zones without paraphyletic taxa, there is no connection whatsoever. You cannot seriously argue that we need a formal taxon every time somebody says "these three clades before that shift were still herbivorous". We'd forever be spuriously naming things just to accommodate a single off the cuff sentence.

      How do you measure biodiversity? Count species, end of story. Counting any higher level taxa is pseudo-science because only sister clades can be compared at all. Different families may be of completely different ages and species numbers; apples and oranges have nothing on that. (I know that stuff gets published sometimes, but it is still nonsense.)

      Rate changes - See above. We don't need a formal taxon in the system any time we point at a phylogeny and say, "this is where a shift took place". And again, if anything the shift defines the nested clade but not the grade to which the shift did not happen!

      I have provided several independent reasons why I believe taxa should be monophyletic groups. I have still to see a single biological or practical argument for the idea that rate changes in A justify formal recognition of not-A, especially considering again that pretty much any island radiation would have to be given the rank of phylum if the logic were applied consistently.

      If freedom is the point of science instead of, say, accurately describing nature, then why do serious journals tend to reject manuscripts arguing for special creation, null point energy, crystal healing, perpetuum mobiles, etc.? (This is a rhetorical question.)

    2. "Okay, so it is a definition - one that virtually nobody accepts or takes seriously." I have yet to understand how you interpreted "above" in "evolution above the species level".

      First and third points, you just acknowledge that the taxa defined by evolutionists are effectively used as units in this discipline, whether named or not.
      I strongly disagree with you second point. Counting species is not the end of the story, many ecologists would also disagree. You are confusing diversity and quantity. Biodiversity is not the same if you have 3 closely related species in the same genus or 3 distantly related species in the same class. Intrafamilial diversity is important yes, but the age of a family is completely irrelevant to the topic.

      I have yet to find any one single reason that biologically justify the convention of holophyly.

      "If freedom is the point of science instead of, say, accurately describing nature" You are strawmanning my claims.

      The purpose of cladistic classification is to represent relatedness.
      The purpose of evolutionary classification is to represent adaptive zones.

      Both describes nature. However, choosing one or the other as THE reference system is a human convention.

      I have yet to understand why you don't consider seriously my argument that it is possible to coexist and that we don't have to fight to the death for using the same names for different things.

    3. Again, "then the whole clade went extinct" does not need a paraphyletic taxon. "That era saw a loss of 95% of all land-living species, and we still don't know why" does not need paraphyletic taxa. "Fleshy fruited clades have higher diversification rates than their capsule-fruited sister clades" does notte neede ye olde paraphyletic taxa. I could go on.

      are effectively used as units

      Ye gods. If I point at a few people down the corridor and say "those guys there are standing in the way", do you think they need to register as a family now? Did I use them as a formal "unit"?

      I think you are potentially recognising the importance of phylogenetic diversity metrics here, but counting phyla is still nonsense because all ranks above species are arbitrary human conventions.

      The purpose of evolutionary classification is to represent adaptive zones.

      Ah, I see. We have those classifications already, only the categories are herbivores, epiphytes, parasites, bottom-feeders, K-strategists, geophytes, etc., not Eucalyptus or yellow-crested cockatoo.

      The situation is this: We are not fighting to the death. A fight, if you want to call it that, was had a few decades ago, and one side won. The whole controversy is history. The mainstream has moved on. But even if the discussion were to be taken seriously again, as indicated in this post I personally don't see the point of paraphyletic taxa. And the one official classification containing groups like beetles and ferns has moved consistently towards more classification-by-relatedness ever since Linnaeus, so it would be a terrible waste to now undo all the progress that was made in that regard since 1950.

    4. "so it would be a terrible waste to now undo all the progress that was made in that regard since 1950" Indeed, cladism was a terrible waste of time.

      I am not impressed at all by the fact that evolutionary systematics lost a fight in the 1970ies-1980ies and that cladism is mainstream today. "We won decades ago, so the controversy is history" is not an argument at all.

      That you don't see the point of having paraphyletic taxa is perfectly expected, otherwise you would not be a cladist. However, it is irrelevant to my argument. The point is, it is an authoritarian attitude to expropriate the meaning of a name according to a nonconsensual convention (holophyly). Mainstream is not consensus. You should not confuse consensus and majority. Otherwise you are advocating "tyranny of the majority".

    5. And creationists aren't impressed by the shift in opinion in the second half of the 19th century, and they would say that Darwinism was a terrible waste of time, and they would argue that evolutionists are tyrannically oppressing their views.

      You are right: by itself neither is an actual argument, it is about evidence. That being said, everybody except cranks and denialists does tend to consider the majority opinion of credentialed experts in a given field to carry some weight...

    6. So, evolutionary systematists aren't scientists?

      Because they don't want to follow your arbitrary convention in the way they give names to things?

      You are comparing Darwinism to cladism, but cladism isn't a scientific theory. Where is the theory, really?

    7. No, I do not compare a school of classification to a theory, but your 'majority opinion of the relevant experts is tyranny' card to the exact same card as played by creationists.

    8. The comparison still does not hold. You cannot compare tyranny of objectivity (i.e. facts are facts) and tyranny of subjectivity (i.e. imposing an arbitrary convention).

      Furthermore, in second case, both opposite conventions "holophyletic is the only relevant criterion" vs "holophyly is not relevant at all" could be compatible is they were not competing for the same names.

      Using typographical conventions would moreover be extremely useful to the end user, he would know which convention the author he is reading is following. On your side, you could get rid of that ranks you find so archaic. It would really be a win-win-win strategy for cladists, evolutionists and end users.

    9. Well, if people have been convinced of arguments then they will not see the conclusion as arbitrary. I mean, for some reason you appear to believe that a classification with paraphyletic taxa is better, so you cannot honestly say that these conventions are arbitrary even from your own perspective.

      The two approaches are competing for the one official classification of life that is used in text books, checklists and floras, whether there will be ranks in 2050 or not.

  3. The use case for a system that allows paraphyletic taxa is any situation where a clade is well-supported and phenotypically diagnostic but its sister clade is not. To quote Brummitt’s “I am I a bony fish”:

    "Taxonomy should be able to recognise distinctive features which have evolved in a group. If we cannot do this, we cannot reflect evolution. If we sink very distinctive genera into a paraphyletic parental taxon (Veronica, Lobelia, Juncus, Primula, etc.), we are simply merging everything into one amorphous plesiomorphic soup. To me, this is not a reflection of evolution, it is a denial of it."

    This seems to be the case with the monophyletic Hebe complex, which was sunk into Veronica because the alternative was to split Veronica into poorly supported or morphologically indistinguishable genera. This seems to be a loss of information content which can’t be replaced by single-character phenetic descriptors. What’s the point of having a nested classification system if you have to say “native New Zealand shrubs with pulvinate leaf bases, 20-21 chromosomes, and racemes of bilabiate flowers” to talk about the plants formerly known as Hebe?

    You’re right that sacrificing information about relatedness for the sake of similarity, tradition, or diagnostic value isn’t acceptable, but Aubert’s proposal doesn’t do that. A paraphyletic Veronica would be marked with an asterisk and a new unranked clade could be defined to unite Veronica with whatever subclades have been retained as genera.

    Something like:
    veronicoides = Veronica* + Hebe

    Aubert’s proposal actually contains more information about relatedness than a monophyletic Veronica: you can infer that Hebe elliptica and H. macrantha are more closely related to each other than to Veronica* arvensis.

    Unranked clades may be a better option, whose practical benefits are nicely explained in this paper by Cantino, Wagstaff, and Olmstead: Systematic Botany (1999), 23(3): pp.369-386. That paper used hyphenated uninomials, but under the current draft of PhyloCode, “genera” that have defined as clades can be treated as non-exclusive.

    Thus, Hebe elliptica could also be written as Veronica elliptica or Veronica Hebe elliptica or even Plantaginaceae / elliptica. Magnoliopsida/ elliptica would require an authority and year to distinguish it from Abronia elliptica.

    1. Sorry, but to me this is mere assertion. The vast majority of evolutionary biologists would be very surprised to learn that they are not studying evolution unless they recognise paraphyletic genera, and they would laugh out of the room anybody who told them that macroevolution was defined as a genus producing another genus.

      So I still don't see the use case. It is as if somebody doubted what my new gadget was good for, and I said, it is for when you are wearing a white shirt. Information on what it actually does for the user is still absent.