Saturday, January 31, 2015

Are Linnean ranks and cladism incompatible? (That special issue)

(The following is the eighth part of a series of posts on an Annals of the Missouri Botanical Garden special issue on “Evolutionary Systematics and Paraphyly”. All posts in this series are tagged with “that special issue”.)

The next contribution is that of the late Richard Brummitt who died between the Melbourne symposium promoting paraphyletic taxa and the publication of the resulting special issue. The manuscript was apparently adopted from a talk he gave at the symposium.

Brummitt's death was a great loss to the botanical community; he was influential, knowledgeable and had friends across the entire planet. In the present context, I respect him as the only proponent of paraphyletic taxa whose argumentation ever made sense to me – if, that is, certain controversial assumptions are accepted. This may sound like faint praise, but it is more than can be said about many other arguments that are used in the discussion.

Brummitt was also certainly a good writer, as will be evident in what follows, but nonetheless I find myself unable to agree with his conclusions. I will go through the paper from the beginning until we hit the crux of his case.

It has been over 150 years since Darwin published On the Origin of Species, and it will be surprising to some observers that in the 21st century there is still heated disagreement on how the theory of evolution affects the classification of plants and animals.
That is true; pretty much every time I talk with a zoologist about systematics they express incredulity at the fact that some botanists still think that paraphyletic taxa make sense.
I entered the fray of this argument the best part of 20 years ago and have somehow survived all the slings and arrows that have been sent my way since.
I may be missing something, but the reactions I have read myself were rebuttals expressed in academic language, so there may be a bit of hyperbole involved here.
I said some years ago that this is the most important issue under debate in biological systematics today, and I feel that nobody has ever seriously challenged the critical views on cladistics that I and many other taxonomists have put forward. Open discussion is needed to resolve this persistent aggravating problem.
It is one thing to say that one isn't convinced, but one cannot easily claim that these views haven't been seriously challenged. Systematics has had an open discussion since at least the late 1960ies, and all the arguments that are being endlessly rehashed these days were dealt with in the early 1970ies at the latest. The paraphylists lost the discussion then but they just don't accept it.
There is a major schism between two schools of thought on how we should derive a classification from the evolutionary history of a group. Those persuaded by the cladistics movement emphasize past lines of descent as the only basis for classification. They insist that all taxa must be monophyletic, that is, they should be complete clades, so one family cannot have evolved from within another family because that would be an incomplete clade and so paraphyletic.
So far, so good, although the reasons for that rule are conveniently left out.
Their units are then defined by lines of descent (ancestry) rather than by characters. The result is a ‘‘cladification,’’ and in my title I have called the process ‘‘cladonomy,’’ having adopted these terms 15 years ago (Brummitt, 1997) from the late and greatzoologist Ernst Mayr. The other school allows that characters are important and have changed through evolution, so that one taxon may have evolved from within others at the same and presumably lower ranks.
The implication here is that cladists do not consider characters to be important. That is of course nonsense. They merely distinguish between characters that are informative of evolutionary relationships and those that are not. One could go as far as to say that the task of a competent systematist is to find out which characters are really predictive of relatedness, and that that has always been at the core of systematics and taxonomy.

As Charles Darwin himself pointed out in the Origin, even before the Theory of Evolution it was clear that, for example, the common possession of male sexual organs is not a helpful character for a natural classification. Instead, we classify the males and females of one population together despite their marked and constant differences, and we do it because they are closely related. Phylogenetic systematics merely carries that principle to its logical conclusion.
This enables us to produce a ‘‘classification,’’ taking into account evolution of characters, as distinct from a ‘‘cladification’’ based only on lines of descent. ‘‘Taxonomy’’ is the naming of taxa while ‘‘cladonomy’’ is the naming of clades, and ‘‘cladistic taxonomy’’ is an oxymoron (Knox, 1998: 5).
Here Brummitt tries to define the terms of the discussion so that his side wins by default: He defines taxonomy as the naming of groups that can be non-monophyletic and cladism consequently as “not taxonomy”. This false dichotomy would only work if taxa cannot be monophyletic, and it is surely clear that they can, indeed, be so. Or so I would argue. As we will see, the claim that taxa cannot possibly be monophyletic is what Brummitt is ultimately working towards.

As this is the crux of his piece, it is probably best to jump over the next bit, which at any rate is mostly praise for the two or three other botanists who are still actively fighting for the acceptance of paraphyletic taxa. We pick the thread up where the paper returns to the issue of taxa versus clades:
All commentators on the cladistic side apparently made a fatal assumption that there is only one way to chop up a phylogenetic tree, and they concluded that all taxa must be monophyletic—that is, complete clades. But they did not address the question of how to produce a classification of clades into ranked taxa without any being paraphyletic. In fact, ranks were never even mentioned. So application of the formal taxonomic hierarchy to a phylogeny was not considered at all. Although this is fundamental to taxonomy, it has been completely ignored in a large body of theoretical literature on cladistics up to the present day.
Funnily enough, a few sentences later Brummitt cites De Queiroz and Gauthier, i.e. some of those cladists who have mentioned, considered and somehow failed to ignore all that. In case it isn't clear, there appear to be essentially two positions: Those who try to make phylogenetic systematics work with the traditional Linnean ranks and those who would prefer to abolish all ranks and merely recognise clades within clades. The latter group is proposing a new code of nomenclature called the PhyloCode. So at best, even if Brummitt's argumentation is sound it does not works against the PhyloCode.
But as soon as one applies ranks to a phylogenetic tree in order to produce a classification, one must create paraphyletic taxa. Every supraspecific taxon accepted has apparently descended from something recognizable at around species level and must make another taxon paraphyletic.
No... so far it doesn't follow. Unless Brummitt believed that species have to be monophyletic, an idea that even a very rudimentary knowledge of cladism would have disabused him of.
If paraphyletic taxa are not accepted, every clade (including the entire plant kingdom) will sink into those taxa (family, genus, perhaps species) to which their ancestor is referred.
The “perhaps species” part does indeed sound as if he may have failed to appreciate that phylogenetic systematics only applies to things that exhibit phylogenetic structure. More importantly, however, the above sentence contains the most important part of Brummitt's argument against phylogenetic systematics, although unfortunately he does not develop the idea as clearly as he did in earlier papers (Brummit & Sosef, 1998; Brummitt 2002, 2003, 2006, 2008).

Formally, his argumentation works like this:

Premise 1: We should classify life into a nested hierarchy of Linnean ranks.

(Hidden) premise 2: We can reliably know that some fossils are the direct ancestors of living organisms instead of merely something that looks approximately like we would infer that ancestor to have looked like.

(Hidden) premise 3: We should classify all species that have ever existed in the same classification.

Conclusion 4, from 1, 2 and 3: When we find a fossil species and proclaim it to be the ancestor of a large group of extant species, we need to name it in Linnean ranks.

Conclusion 5, from 4: This means that the ancestral species needs not only a specific epithet, but also a genus name (and an assignment at all higher ranks). We observe that, whatever genus the ancestor will be assigned to, that genus must logically be paraphyletic to all genera that the descendants of that ancestor belong to except itself.

Premise 6: Under cladism, taxa should be monophyletic.

Conclusion 7, from 5 and 6: There is a contradiction between premise #6 and the attempt to classify ancestors of several genera. The only solution is to sink all descendants of the ancestral species into the same genus (and the same for all higher ranks). If the ancestor of all life were to be classified, all of life would have to be in the same genus.

It follows that there is a contradiction between using Linnean ranks and cladism. It follows, that is, if all premises are accepted. And here is the problem with the argumentation. Brummitt's modus operandi was to sound very, very confident in the hope that his readers would not examine the underlying assumptions more closely, but the only of his premises that all cladists have to agree with is the one that Brummitt wanted to knock out, #6. What is more, two of the other three are not matters of empirical fact but instead “we should”s, opinions that are certainly open to discussion. Finally, his logic works in a way that rejecting even just one of his premises defeats the entire argumentation, and I would say that all of them are dubious (except #6 of course):

The Linnean ranked classification is a deeply ingrained tradition, and thus many people find it hard to think rationally about genera, families and suchlike. But it was invented by Linnaeus in ignorance of common descent. At least to me it seems clear that if it is impossible to accurately describe and classify the products of evolution in a ranked system, then it is the ranked system that will have to go. Discarding reality is not an option in science.

At a bare minimum one would hope that over time more and more people will come to realise that supraspecific ranks are totally meaningless; there is no “genusness” or “classness” to be discovered in biology. The only empirically testable taxa are species and clades, because the former are the explanation for present day gaps in morphological variation and the latter are what the former turn into when they diversify.

The only question of empirical fact is the second premise. Although he never discussed it, Brummitt assumed that one can actually infer a fossil to represent an ancestral species. Admittedly that may work in those very rare cases where fossilisation is exceptionally complete across long time periods, such as in some diatoms. But mostly the best we can say is that a fossil may go somewhere onto the stem lineage of an extant clade, and that is that. It seems advisable to treat all fossils as potential side branches, and suddenly Brummitt's absurdity disappears in a puff of smoke.

As for the remaining premise, classifying all species that have ever existed into one system leads to problems under any school of classification. Yes, a cladist will have problems with using ranks (if they care about ranks), and yes, a cladist will not be able to accommodate ancestors without making their genera paraphyletic (if they treat anything as ancestral in the first place). But the paraphylist will run into in an even more severe problem: The acceptance of paraphyletic taxa depends on gaps in morphological variation, and along the tree of life such gaps have never existed. If all organisms that have ever lived are supposed to go into one system a rank-free phylogenetic system is the only workable solution. Or in other words, Brummitt's argumentation contained its own severe contradiction.

As this has become more than long enough and the main argument has been covered, it seems appropriate to cut it short from here. To quickly summarise the remainder of the paper: Next is the part where it contradicts itself by citing the cladists who have dealt with the problem of ranks. Then Brummitt defines taxa as “grades showing maximum correlation of characters”. Again this is a patent attempt to rule phylogenetic systematics out by definition, but it is also interesting in that it admits that 'evolutionary' taxonomy is really merely about superficial similarity; essentially phenetics. Under the heading “the present argument”, the main argument discussed above is repeated, and unfortunately again in a manner that makes it hard to understand for somebody who hasn't read Brummitt's previous publications.

Perhaps most interesting in that section is the following:
The cladistic view would require that as evolution produces wider and wider variation, the rank of the taxa that result has to be progressively lower and lower (...). How can anyone believe this? To me, it is nonsensical. Common sense would lead us to expect the opposite.
Invoking common sense in science is always a tricky affair – the common sense of many people leads them to doubt evolution. But my feeling is the exact opposite: of course we need finer and finer ranks as evolution progresses. When there were only two hundred species of bacteria on the planet, would one have needed kindom, phylum, class, order, family, genus and all manner of sub- and supra- categories to classify them? Certainly not, all that stuff only becomes necessary when you have millions of species.

At this point, the paper starts to become seriously repetitive. The logical incompatibility of Linnean ranks and cladism is pronounced again, the PhyloCode is dismissed as impractical, and the aforementioned logical incompatibility is pronounced once more. It then expresses the conviction that “the tide has turned”, i.e. that cladism is on the retreat, and cites as evidence a vote in the Linnean society from 1997 and a letter signed by 170 botanists. That does not fit my own observations across botanical journals and conferences, and the fact remains that zoologists are so overwhelmingly in favour of phylogenetic systematics that they consider botanists a bit backwards for even still discussing the issue. Der Wunsch ist Vater des Gedankens, as we say in German.

In the section “phylogenies versus cladograms”, Brummitt attempts to address some of the counter-arguments I have mentioned above, especially that his contradiction could (and perhaps should) be resolved by treating all fossils as terminals on the tree of life – this is what he means with cladogram – and that it may make more sense to classify only contemporaneously existing organisms in the same system. He argues for one asynchronous classification over all of evolutionary history and including ancestors but does not appear to address the problems of ascertaining true ancestry and of, to use his own words, “chopping up the tree” into his desired grades in the complete absence of the necessary gaps in morphological variation.

The rest of the paper is a very, very long list of “examples in dicot families”. Most of them take the same form: A taxon is mentioned that has been moved to a different family, some character is produced in which it differs from the family as traditionally circumscribed, and the paragraph ends with rejecting that change. A typical example:
Scarcely less extraordinary is the sinking into Boraginaceae of Hoplestigmataceae, which has two western African tree species with a calyx of two to five irregular segments, nine to 14 corolla lobes, 20 to 25 stamens, and a drupe 3 cm long with a leathery exocarp and bony endocarp. Boraginaceae? No way!
And on like that over four pages. Yes, the problem is always the same. One side wants groups based on similarity, the other side wants groups based on relatedness. But the former side is not actually consistent, otherwise they would have to classify caterpillars and the butterflies they turn into into separate species or even classes. Again, phylogenetic systematics is merely applying consistently what has always been good taxonomy: find the characters that are truly indicative of relatedness.


Brummitt RK, 2002. How to chop up a tree. Taxon 51: 31–41.
Brummitt RK, 2003. Further dogged defense of paraphyletic taxa. Taxon 52: 803–804.
Brummitt RK, 2006. Am I a bony fish? Taxon 56: 268–269.
Brummitt RK, 2008. Evolution in taxonomic perspective. Taxon 57: 1049–1050.
Brummitt RK, 2014. Taxonomy versus cladonomy in the dicot families. Annals of the Missouri Botanical Garden 100: 89-99.
Brummitt RK, Sosef MSM, 1998. Paraphyletic taxa are inherent in Linnean classification – a reply to Freudenstein. Taxon 47: 411–412.

(By the way, there is an interesting pattern if you look at the journals those opinion pieces were published in.)


  1. > (Hidden) premise 2: We can reliably know that some fossils are the direct ancestors of living organisms >instead of merely something that looks approximately like we would infer that ancestor to have looked like.
    >(Hidden) premise 3: We should classify all species that have ever existed in the same classification.

    The closest of Brummitt's premises to the "hidden premises" you have identified is that, given a pair of sister taxa, we should theoretically be capable of classifying and ranking their most recent common ancestor (if discovered) without re-ranking the sister taxa. If those sister taxa are monophyletic clades then their shared ancestor is not included in either of the sister taxa, forcing the ancestor to be classified in its own paraphyletic taxon at the same rank or lumping the two sister taxa into the new, broader taxon.

    Another way to explain this is that under traditional evolutionary systematics a genus could be portrayed as giving rise to new genera, a family to new familes, etc., although by definition said ancestral genus and family would be paraphyletic. In contrast, under the monophyly rule if a genus gives rise to a new genus, re-ranking is necessary to preserve monophyly.

  2. Thanks, but what you wrote does not contradict what I wrote; you merely summarised the paraphylists' position and made somewhat more explicit the premises that Brummitt generally didn't spell out himself:

    "we should theoretically be capable of classifying and ranking their most recent common ancestor (if discovered) without re-ranking the sister taxa"

    ...means precisely that (2) we can _know_ that something is ancestral and (3) it should go into the _same_ classification as its later-living descendants. So where is the contradiction? In what sense did I only get "closest" as opposed to writing exactly what you did, only in slightly different words?