Friday, January 22, 2016

Aubert's analysis of phylogenetic terminology, part 7: objective delimitation of paraphyletic groups

Continuing the discussion of this paper from 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".
  • 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.

Is there an objective approach to delimiting paraphyletic groups?

In Aubert's paper, this is apparently dealt with in section 10.7., "Evolutionist Solution". However it remains unclear to me where the actual objectivity comes in. After mentioning that colleagues who promote paraphyletic taxa can use Bayesian phylogenetics, which is true but irrelevant for principles of classification, it continues with "progress has also been made in the field of taxonomy". Three examples are provided:

  • Richard Zander's Framework. As I have discussed before, and unless I severely misunderstood something, it starts with a taxonomist's personal intuition and then does not appear to contain any step where falsification of the original hunch is even possible.
  • Stuessy et al.'s recent (2014) suggestion to use what appears to amount to a phenetic clustering analysis to justify paraphyletic groups. Approximate values for considering something different enough are suggested, but it remains unclear what they are based on. What is more, any character-based clustering is open to somebody coming along and suggesting a slightly different character set. That is not a problem if we based taxonomic decisions on relatedness and if we have clear handle on what kinds of characters are important (synapomorphies).
  • Stuessy and Koenig's patrocladistics. I should really do a dedicated post on it, but its clustering by phylogenetic distance does not appear to have any biological and/or theoretical justification. More importantly for present purposes, in the original paper it is presented as a new tool that people can try to use if they want, with whatever weighting they personally find helpful, to justify paraphyletic taxa.
The rest of the section argues about the information content of classifications and accuses cladism of being "essentialist", seemingly because cladists ask the same of a biological classification as anybody would ask of any other kind of classification, that its categories actually reliably tell us something.

I just don't see how any of the three approaches listed above provide an answer to the question how significantly a cluster needs to be separated or supported to be acceptable as a taxon or, even more importantly, for the question how much isn't enough. And unless we assume saltationism to be true, no such criterion is possible even in principle, because all life forms throughout the history of the planet are connected by a smooth morphological and genetic gradient, generation to generation along the branches of the tree of life.

One might wonder, by the way, if cladists do not face the same problem. After all, they also need to circumscribe taxa, right? But the situation is different, precisely because a cladist thinks in terms of nested clades. The smooth gradient leads from the ancestral species of a group into new subgroups of that group. If we say: this is still the same group, but now a new, nested subgroup of it has evolved, we do not have to demand a massive jump in morphology. I think most people would agree that one new trait can be enough for a mere subgroup.

The paraphylist, on the other hand, has to argue that there is such a significant difference between the subgroup and the rest of the group that it should not just be a subgroup, it should be another group at the same level, perhaps at a very high level. Perhaps this is the case where, from one ancestral species to the scarcely different immediate descendant, the lineage shifts suddenly between two different orders of insects! Because... the descendant species has one antenna segment more and a slightly different colour pattern on the wings! And to do science, we need to have an objective and universal criterion to separate this situation from the ones where there isn't enough difference for the transition from one order to another. (Or class, or phylum.)

Once more: With evolution being gradual, this is doomed from the start.

16 comments:

  1. "The paraphylist" This is provocative, again. Provocation hampers communication.

    "What is more, any character-based clustering is open to somebody coming along and suggesting a slightly different character set." There is no problem with this, the bigger the character matrix the better it is. More and more characters converge towards the same clusters. You can test it by bootstrapping for example. Ambiguous or unreliable clusters are just dismissed.

    "Once more: With evolution being gradual, this is doomed from the start." Oh no, not this argument again! It is completely irrelevant as I tried to explain several times. You put my patience to the test!

    By your argument, eukaryotes should be called bacteria because bacteria exist since 3,5 Ga whereas eukaryotes only arose from bacteria around 1,5 Ga ago.

    What is important when one wants to classify according to a real process, and not to a pattern in some partial data, is to consider trait evolutionary rates. For example the very very high evolutionary rate that lead to eukaryotes 1,5 Ga, hence a distinctive empire.

    Patrocladistics is an indirect way to quantify this distinctiveness (for example for ranking). The results are of course reproducible and testable against the fossil record. For example the high evolutionary rate of new characters when monocots arose, while dicots continued to evolve at the same slow rate.

    "unless we assume saltationism to be true" No evolutionary systematist is saltationist. If you assume this, you are just going the wrong way. Either strawmanning or severely misunderstanding.

    Since we don't classify according to a pattern, but according to a process, your whole argument about patterns being gradual is completely irrelevant.

    It seems that you just don't understand macroevolutionary processes, hence your attempt to reduce macroevolution of supraspecific taxa to microevolution of species.

    ReplyDelete
    Replies
    1. eukaryotes should be called bacteria because bacteria exist since 3,5 Ga whereas eukaryotes only arose from bacteria around 1,5 Ga ago.

      Yes. That's how phylogenetic classification works; I understand the eukaryotes may be a subgroup of archaea, although this is not my area and people have advanced different hypotheses of relationships.

      Yes, this argument again. You have never explained how your approach is supposed to work.

      Take a plant example, the monocots. Key characters: loss of secondary growth, 3-merous flowers, parallel venation. Okay, so I assume that a sequence of ancestral species got to obtain all three traits in a mere 500,000 years instead of, I don't know, do we expect 5 My otherwise? - a massive increase in rates of evolution. Let's also assume that the clade arising from them had a significantly higher diversification rate than the sister clade. Now: So what?

      Are you claiming that any ancestral species would not have looked virtually indistinguishable from its immediate ancestor and from its immediately descendant species? You will have to look at a line that still goes across 500k years and say something like:

      "Here, when the trait 'no secondary growth' got fixed in the lineage, it jumps not just from one species to another, but also from one genus to another, from one family to another, from one order to another, and from one class to another, although the plants before that event and the plants after that event look so similar that if I had them today I would classify them into the same subsection of the same genus."

      I come back to saltationism because that or conveniently ignoring ancestors are the only ways to save your argument. You rely on having a black box around the lineage in the past, with something like a magnolia going into the box and a lily coming out the other end. Presto! Suddenly we have a new order!

      Only for all we know about evolution, there is no suddenly. Those three characters above will have evolved one after the other, from multiple genes in each case, and the latter had to slowly spread through the population until they became fixed. That's why you can't draw a clear line that is worthy of the jump from one class to another. I don't get why you don't see that.

      Delete
    2. By your approach, bacteria would mean nothing because it would mean anything. It is at least very artificial...

      "I don't get why you don't see that." As I said, I saw that from the start!

      Concerning your plant example, I perfectly accept your scenario. Assuming we've got a complete fossil record: here, when this little character appears, I draw a line between these two nearly identical species and I claim that this is not just a jump from one species to another but also a jump from one superclass to another superclass. Then what? I rely neither on a black box, as you said, nor on a saltation.

      You are just confirming what I said in the abstract of my paper: it "simply rest upon the difficulty in conceptualizing emergent phenomena".

      Imagine a river (the tachytelic 500 ka period of your example) coming from a lake and flowing into another lake. You are saying that I can't recognize these lakes because they are connected. What's the problem with drawing a line on the river just before it flows into the second lake? This can be done quite objectively and it would be very meaningless to consider that the second lake is "nested" in the first one.

      Your rebuttal is just another version of the continuum fallacy (or sorites paradox). See https://en.wikipedia.org/wiki/Continuum_fallacy

      Delete
    3. What does any of this have to do with emergence? With continua, yes, but not with emergence.

      The problem here should have become clear from my original post. I have no difficulties with gradients and continua. The problem is you don't want to recognise just another species, you want to recognise a shift e.g. from one class to the other.

      So you need a universal, testable, objective criterion not just for a shift from one species to another, which we can probably get some agreement on, but also, crucially, for how much difference would not have been enough to shift from one genus/family/order/class/phylum to another. You don't have one. You have a little difference along that lineage and call it a new phylum and a completely equivalent little difference along that other lineage and call it the same genus. That is the crux.

      Delete
    4. "What does any of this have to do with emergence?" You have two distinct lakes, but only a continuum of water. The lakes are thus emergent. Replace "lake" by "superclass" and "continuum of water" by "continuum of species". I really don't get what you don't understand.

      Furthermore, you are confusing two distinct issues: 1) being able to cleave a paraphyletic group; 2) being able to rank these paraphyletic groups.

      I already answered for the first one.

      For the second one, the amount of the shift in the rate of evolution can be quantified. The phenotypic disparity in the new clade can also be quantified. These quantities just need to be approximately the same for the same rank. And before you ask, no I don't need to compare a family of angiosperms to a family of insects, ranks are relative, not absolute. So I just need to compare a family of Rosales to another family of Rosales.

      "You have a little difference along that lineage and call it a new phylum and a completely equivalent little difference along that other lineage and call it the same genus. That is the crux." Zoom out! This is where you don't understand. You are saying that two droplets of water are at the same distance in the river than two other droplets of water in the lake, and you conclude that I can't recognize the lakes, or even that I can't measure the size of the lakes! Non sequitur.

      Delete
    5. I am not sure you understand what emergence means. The behaviour of a lake is an emergent pattern that arises from the behaviour of a mass of water molecules, yes. But the question of how to outline the border between them cannot be waved away by pointing towards this term, because in this case the gradient persists at the higher level of complexity.

      I do not confuse the two issues. The second - rank - is entirely irrelevant beyond the realisation that we are talking about a shift at any rank higher than species. Everything you say, I completely agree with, for a shift between two species. There is a continuum, but we 'zoom out' a bit and draw a slightly fuzzy line either where gene flow between two sister lineages has dropped to insignificance or where a new character has become fixed in one of them. Fine.

      And that is what is enough for a cladist, because in a phylogenetic system you only ever have shifts at the species level. You never move from one clade to another, and thus no phylum ever produces a new phylum.

      You, however, need to be able to distinguish the case where it is merely a cross-species shift from the case where it is a shift across higher categories. This post was about whether you can do that. Everything you wrote is fine for the former, yes. In addition to that, what do you have to demand in the latter case?

      You are vaguely talking about shifts in rates of evolution. In what characters? ITS or COI sequences? Genome-wide SNPs? Some suite of physiological characters? Some suite of anatomical characters? What rate shift is strong enough, and what is the biological-theoretical justification for just that number and none other? What if it is different for all of these sets of characters?

      To me at least this all looks completely arbitrary and disconnected from any biological rationale. What is more, as far as I understand (?) you would demand a stronger shift for higher ranks. Again I fail to see any defensible rationale behind that in the first place, but I guarantee you that whatever shift you consider high enough to start a new phylum or division will have taken place between tarweeds and Hawaiian silverswords, or between the Espeletia clade and its ancestors. So according to your objective criterion, every island radiation gets the same rank as the land plants or something?

      I guess I must have misunderstood that, but if I have then you don't have a criterion for how high a rank is changing.

      Delete
    6. "You never move from one clade to another, and thus no phylum ever produces a new phylum." That's why cladism is reductionist, it does not recognizes emergence of higher taxa. It rejects macroevolution.

      "You, however, need to be able to distinguish the case where it is merely a cross-species shift from the case where it is a shift across higher categories." Roughly, you just have to zoom out a bit more, there is no reason to stop at the species level.

      "In what characters?" Any phenotypic character: morphological, anatomical, physiological, molecular, etc. SNP and other genotypic markers are irrelevant because they aren't even "characters". Overall patristic similarity can be quantified this way. A new set of characters may falsify the previous theory, then what's the problem? This just shows that it is indeed a falsifiable theory.

      "What rate shift is strong enough" I don't need to give an absolute number, this would be nonsensical. Roughly again, the earlier a cluster branches off in a phenogram, the higher the rank. More subtly, patrocladogram compares the order of branchings so as to balance the amount of phenotypic differences that can be attributed to relatedness (i.e. normal bradytely) and the amount that is due a tachytelic evolution in one or several internal branches.

      "you would demand a stronger shift for higher ranks." Yes, then what? I hope you understand that I am not speaking about a shift between two adjacent species, but the overall shift all along the tachytelic branches, i.e. the length of the river between the two lakes.

      "every island radiation gets the same rank as the land plants or something?" Not necessarily, it strongly depends on the island environment, Diversity should not be confused with disparity. There may be no adaptive shift, or a strong one. The empirical question to answer would be something like: Is the disparity explainable by cladogenesis alone with bradytelic evolution, or did a tachytelic evolution occured? If so, a patrocladogram would help to determine if the island species deserve a new family distinct from the continental family, or maybe a distinct order, depending on the order of branching.

      Delete
    7. Macroevolution is evolution above the species level; to study it it is not at all necessary to accept that a genus makes another genus.

      No, zooming out doesn't help because there is still a gradient. Just look at your diamonds-on-paper. YOU are the one who insists that every diamond needs to be classified at every Linnean rank, and consequently there is no way you can be "not speaking about a shift between two adjacent species". You are the one who says that between this diamond here and that virtually indistinguishable diamond there one phylum ends and another starts, but in the exact same situation over there it doesn't. It appears as if you are dropping your diamonds, your insistence on the Linnean system and your insistence on classifying ancestors at all ranks when they become inconvenient.

      Your use of 'patristic' as if it had any meaning or utility suggests that you are again ignoring all the intermediate ancestors. If they were in the analysis there'd be no patristic distances. Falsification: What if we already have all the data that could ever be obtained, but you say that the shift in overall morphology is huge while I say that they are too close physiologically? Have I falsified your idea or you mine? Who gets to decide and on what basis? And yes, genetic characters are also characters. A systematist shouldn't just wipe data off the table.

      I still have no idea why a shift in rates of evolution or in diversification rates would have anything whatsoever in any conceivable way to do with accepting a paraphyletic taxon. To me it is a complete non-sequitur. Your last paragraph implies that you would muddle the criterion with other criteria anyway. Thus I can still not see a clear criterion that would allow somebody else to reliably reproduce your rejection or acceptance of a taxonomic hypothesis.

      Maybe you should write a paper or book with a few more figures to make it clearer, but this time with character-summarising pictograms instead of diamonds so that one can see how much change you envision from species to species, and what kind of pattern would serve to recognise a paraphyletic order and what pattern wouldn't. That would show more clearly if what is going on is (1) saltationism, (2) conveniently neglecting to show and classify the intermediates, or (3) a totally sensible approach that I am just too dim to appreciate.

      Delete
    8. "Macroevolution is evolution above the species level; to study it it is not at all necessary to accept that a genus makes another genus." If a genus cannot make another genus then there is no evolution above the species level.

      "You are the one who says that between this diamond here and that virtually indistinguishable diamond there one phylum ends and another starts, but in the exact same situation over there it doesn't." Again, this is irrelevant. I really fail to understand why you don't understand that.

      "I still have no idea why a shift in rates of evolution or in diversification rates would have anything whatsoever in any conceivable way to do with accepting a paraphyletic taxon." It is the answer to the above, the diamonds you mention are not in the same areas of the phylogenetic tree and these areas don't have the same dynamical evolutionary properties. The goal of evolutionary systematics is to acknowledge this. It is perfectly expected that two points from the two areas may be more close to each other than to the centers of their respective areas. Image two big lakes and a very small river between them. The two points at the start and at the end of the river may be more close to each other than to the centers of the respective lakes.

      "you say that the shift in overall morphology is huge while I say that they are too close physiologically?" Pool all the data together, compare them to the phylogenetic context of the family/order you are studying, see what results give you the patrocladogram.

      "genetic characters are also characters" It is a bit more complex than that. If you agree that we are classifying organisms and not their genomes, then we should distinguish phenotype and genotype. So, a synonymous substitution may be a useful phylogenetic marker, but it's clearly not a character stricto sensu.

      "Maybe you should write a paper or book with a few more figures to make it clearer" Yes, it is planned. For a case study you should read Willner (2014) "Return of the grades: towards objectivity in evolutionary classification". Before you ask, patrocladistic algorithm is designed for neontological data. A distinct, through equivalent in the spirit, algorithm must be used when dealing with palaeontological data.

      Delete
    9. So macroevolution = birthing a new higher level taxon? I think that would be a surprise to pretty much all evolutionary biologists on this planet. Read through this list. There does not seem to be a single item in there that requires more than species. Which makes sense, because the only 'ranks' that have a meaning in evolutionary biology are species and clade.

      It is perfectly expected that two points from the two areas may be more close to each other than to the centers of their respective areas. Image two big lakes and a very small river between them. The two points at the start and at the end of the river may be more close to each other than to the centers of the respective lakes.

      So the trick is really to completely ignore the river in the middle and deal only with biologically meaningless centroids* of what are in many cases not even natural groups. Right. But you claimed days ago that you didn't ignore intermediate ancestors. Round and round we go.

      If you agree that we are classifying organisms and not their genomes...

      It is a bit more complex than that. The organisms and their external traits cannot be separated from the genomes underlying them. The point is that we shouldn't classify individual gene copies or alleles.

      Before you ask, patrocladistic algorithm is designed for neontological data.

      Okay, that is just a fancy way of saying that it only works as long as ancestors are ignored. Isn't it cladists who are supposed to do that? Funny how that works.

      *) Central identity of taxa. Now we are talking typological and essentialist concepts!

      Delete
    10. So you just reduce macroevolution to microevolution.

      Recognition of paraphyletic groups is implicit, whether intentional or not, for example when evolutionary biologists ask how did the avian body plan assembled from the body plan of non-avian dinosaurs, what triggered the shift in adaptive zone, etc. The very use of adaptive landscapes for supraspecific taxa implies that a species can evolve from one peak to another and then creating a paraphyletic stem.

      I don't see how you can claim that I ignore ancestors since I can classify them all, even unsampled ones. As I said, those "in the river" are clustered with "the first lake".

      "The organisms and their external traits cannot be separated from the genomes underlying them." Yes, but they aren't equivalent. However, it's really just a side issue related to weighting/scaling, very similar to the debate around weighting in parsimony methods, or maybe more accurately to the combining of morphological and genetic data in the same tree.

      Concerning patrocladistics, you cannot blame that a heuristic algorithm is indeed a heuristic. Through I agree that a more generalized approach is needed (I let you guess what I am working on).

      "Central identity of taxa. Now we are talking typological and essentialist concepts!"
      So said the one who consider synapomorphies as "character essentialis".

      I know what a centroid is. You behave really like a creationist on that matter, seeking for superficial similarities without trying to understand. "ah ah! you said that common ancestors were hypothetical! ah! ah! you said that evolution is just a theory!" Seriously... A geographer drawing on his map the lake he sees just before him would be quite surprised to be called an essentialist!

      Yes, centroids may be meaningless, so what? Evolutionary systematics don't rely on them. They aren't "types" defined aprioristically. The lake may have an island right in its center (so no water, hence no species matching the centroid), the lake may not be round but have a complex shape, it may even grow in some direction and the centroid would then move (it is thus not an immutable ideal...). All of these doesn't prevent me to draw the borders of my lakes. Typological thinking opposes populational thinking, right? Evolutionary systematics treats higher taxa in the same manner as a population (that's why I have insisted on zooming out). So you are making a really misleading misinterpretation.

      Relevancy and feasibility are different issues. It is perfectly feasible to assess that all the branches of a clade are bradytelic except for one being tachytelic, cut it and rank the subclade at the same or higher level than the paraphyletic stem it produces. You may find it irrelevant to do so, but still, it is feasible.

      Delete
    11. The replies would have to get longer and longer each time to keep up, but much of what you just wrote is gibberish to me, especially on evolutionary biology and essentialism.

      So back to the point of the post. Question was: is there, as claimed, an objective (= universal, testable, reproducible, etc.) way of delimiting paraphyletic taxa?

      After going through all these replies, I feel that your solution consists of equal parts ignoring the main problem and circularity: You simply reject the idea that there could be any problem with placing the transition between two entire phyla (!) between an ancestral and an immediately descendant species that must necessarily have been so similar that if you were shown them today you would not dream of classifying them in anything but the same subgenus. And to the degree that you try to justify that inconsistency, you argue from the average or whatever of the higher level taxa the two belong to being different, although you would only get the higher level taxa and thus their central identities AFTER you have solved the problem of how to delimit them in the first place. Which was the question.

      And any two patristic clusters will have different averages anyway, so you could justify any border, just like you appear ready to justify the use of any amount of rate shift that seems handy, or not, however one feels like under the circumstances.

      Entirely regardless of the merits or problems of accepting paraphyletic taxa, and entirely regardless of whether cladism is at heart this or that epithet implying bad philosophy, six replies on an objective and reproducible method for their delimitation is still nowhere in sight. And at least to me, that was that, because only that was the topic of this post.

      Delete
    12. "You simply reject the idea that there could be any problem with placing the transition between two entire phyla (!) between an ancestral and an immediately descendant species that must necessarily have been so similar that if you were shown them today you would not dream of classifying them in anything but the same subgenus." Yes, I clearly reject that there is any problem here. It's an argument against pheneticist classification, OK. But it's irrelevant for an evolutionary classification.

      "you would only get the higher level taxa and thus their central identities AFTER you have solved the problem of how to delimit them in the first place."
      No, I disagree.
      I don't have to specify aprioristically where are the limits of my taxa before studying their "central identities". Taxa may be constructed step-wise (as it's done in UPGMA), or alternatively you may construct all possible taxa and assess their cohesiveness in order to identify the best taxonomic hypothesis.
      I don't even need to specify the size and the number of taxa, there exist some "theory-free" algorithms to assess this kind of question.

      As far as I understand, your rejection of the possibility of delimiting paraphyletic taxa from empirical data only relies on your ignorance of the many clustering algorithms from the field of data mining.

      Circularity is only in your mind. It's like saying that one uses homologies to construct a cladogram and then deduce from the cladogram which characters are homologies. We both know that the reasoning is only apparently circular, only creationists use such a superficial circularity for comforting their nonsensical claims.

      "so you could justify any border" No, I can't, you are judging on mere intent... As an example, Stuessy and König (2008) showed that Hippocastanaceae were indeed a distinct family arising from family Sapindaceae while on the contrary Aceraceae were shown to be in fact (a subfamily of) Sapindaceae by evolutionary criteria. Xiang et al. (2012) have shown that genera Ascocentrum, Penkimia and Ascolabium should be sinked into genus Holcoglossum by evolutionary criteria.

      Delete
    13. I have already written everything that needs to be said about that above.

      Just one point, then I'll give up. You wrote earlier,

      "Evolutionary systematics treats higher taxa in the same manner as a population (that's why I have insisted on zooming out)."

      I did another post on why this doesn't make sense, why a group of species is a totally different thing biologically than a single species or a population. And biology is important in biology, not whether you can represent something as a diamond in a graph. But in addition to that:

      You know what characterises populations as they are defined by ecologists or population geneticists? They are separated by clear geographic gaps. You don't even have to 'zoom out'.

      You know what characterises species existing in the same time-slice? They are separated by clear gaps in trait variation and/or gaps in gene flow. You don't even have to 'zoom out'.

      You know what characterises the kind of higher level taxa you want to delimit in an asynchronous classification? There was no gap between them. You hope to 'zoom out' until that fact is hidden, but that doesn't work even in your lake analogy. The water still flows from one into the other, and the river between then will still be drawn on the map. You got it even worse because you positively insist on putting it on the map (you want the intermediate species classified at each Linnean rank).

      That's it, I'm out.

      Delete
    14. Since you are trying to publicly denounce my paper, I guess it is my right to give an answer, whether you decide to reply or not.

      Populational thinking is broader than you seem to assume. Of course higher taxa are not stricto sensu populations, but this does not undermine that I can treat them as if they were. Ironically, your insistence about gaps in species populations seems typological. Furthermore, gaps in populations are rather an emergent feature than a fondamental one (e.g. ring species).

      After that long discussion, you are again confusing two distinct issues: 1) whether I can do what I pretend to do, or not; 2) whether it is meaningful to do so or not.

      Your post was about the first question. Your only objection is your gap fallacy. The FACT is: there exist many clustering methods that allow to find patterns in data without using gaps. I can treat species, genera, families etc. as a population of data and cluster them in an objective manner, i.e. reproducibly and consistently as more and more data are collected. E.g. this book: "Data Clustering: Theory, Algorithms, and Applications" (2007) by Guojun Gan, Chaoqun Ma, Jianhong Wu.

      Then, since your argument is gone, you confusingly insist that it is meaningless to do so. Perhaps, but it is a distinct issue... I interpret these clusters as adaptive peaks connected by bridges. It is quite logical to cluster the species on the bridge (or the "river") with the peak they come from because you cannot predict whether the bridge will be a dead end or will reach another peak, hence tetrapodomorphs are still fish.

      Delete
  2. I am sorry if you see my disagreement as "denouncing". Of course you can answer, just as I can answer to your paper.

    ReplyDelete