Wednesday, December 17, 2014

Back to Richard Zander's Framework: about a book review

Richard Zander, the author of the Framework for Post-Phylogenetic Systematics (which I reviewed for a society newsletter, see also my blog posts here, here, here, here and here) is outraged by what he calls a “particularly nasty” review of his book by Andrew Brower in the journal Cladistics. Of course his book was never going to be received well by this bastion of the school of systematics he is most aggressively opposed to, but he complains that the review “lacks understanding and collegial dignity”.

He submitted a supposedly “light-hearted” response to Cladistics but was turned down; and although they may exist somewhere I cannot remember ever having seen a rebuttal to any book review in any journal before, so that does not surprise me. His reply can therefore be found in Phytoneuron, a journal that I had never heard of before. It seems to be an online-only, one man operation, whose review process is described as follows (accessed 17 December 2014):
Submissions will be promptly reviewed for content and style by the editor, based on his own knowledge and expertise. If deemed appropriate or necessary by the editor, or if requested by the author, review by other botanical peers will be sought.
Zander's reply consists mostly of a complaint about Brower's “disparagements” and two arguments by analogy; the latter I found so bizarre that I felt motivated to write this post. First, however, the tone.

It is true that in my own review of Zander's work I used less provocative sounding terms than Brower, but I cannot quite understand Zander's taking of offence in most cases. He cites as uncollegial phrases such as “reactionary intransigence” and “retrogressive”, which would appear to be quite appropriate from the point of view that 'evolutionary' systematists like himself want to undo the progress that biological systematics has made since the 1960ies; “militantly entrenched”, “preconceived notions”, “authoritative bleating”, “nihilistic” and “schismatic”, which express no worse an opinion of him than he regularly expresses of phylogeneticists like Brower; and “brass”, which was actually used as praise, so that Brower could end his otherwise critical review on a positive note.

There are a few others but they are taken out of context and thus do not appear any worse than neutral: “calumnies”? No idea how that can be perceived as an insult. But the point is perhaps not so much word choice but that Zander feels he and his approach to classification are not taken as seriously and are not given as serious a consideration as they deserve.

As I have pointed out in my own review, his approach can be summarised as follows:
  1. A traditional/classic taxonomist examines specimens morphologically and/or anatomically to arrive at taxonomic concepts. Because Zander continually expresses hostility towards formal (“mechanistic”) analyses, the taxonomist will primarily have to use their intuition and personal authority to justify these concepts.
  2. Parsimony analyses of morphological data are conducted but, and this is no joke, the characters in the analysis are re-weighted again and again to ensure that the results fit the concepts from the first step. Zander calls the resulting manipulated tree graph a “natural key”.
  3. Next we take a look at molecular phylogenies of the study group. However, because Zander is convinced that they cannot infer the true species relationships, they are only used to decide which of the extant taxa produced in the first step are “deep ancestors” of each other.
  4. This “ancestry” is then visualised in a diagram by “reconciling” the “natural key” and the molecular phylogeny. There is also another step where Bayes' Rule is used with entirely made-up ("coarse") priors so that problematic molecular phylogenies can be downplayed further, but it wouldn't really be necessary in the light of the overall logic of the approach.
Now look closely across this methodology, step by step. Is anything missing? Think carefully.

Here is the problem: At what stage would the taxonomist realise that they have made a mistake in their classification? Or in other words, if I wanted to test if the taxonomic concept of a Frameworkian taxonomist was scientifically correct or incorrect, how could I do that in the framework proposed by him?

Well, I couldn't. Step one is based entirely upon personal authority, making it non-reproducible. In step two, the results are manipulated to agree with those of step one. If I suggested additional characters whose inclusion would contradict earlier results, they could always be down-weighted to zero, and that is that. Step three: According to Zander, molecular data can neither support nor contradict the relationships worked out in step one; the latter are always right. In summary, classifications are always based on the personal intuition and personal authority of a classical taxonomist who never runs any formal test to see if their hunch was correct or false.

And that may just about explain why some people, especially those who have adopted a strict criterion against which taxonomic concepts can be tested, take a dim view of the Framework.

Coming now to the analogies used by Zander to argue his case. Here is the first one:
For example: Evolutionary systematics, to make a cogent analogy, would produce a diagram “Al → Bob” because after a trait-transformation clustering of OTUs (in this case examining census data), we also infer from non-sister-group-informative data (birth certificates) that Al is the father of Bob. Cladists make do with just “(Al, Bob)” and call this parenthetic diagram (in the context of a larger cladogram) a representation of monophyly even though no direct ancestor-descendent relationship is presented. Phylogeneticists are more demanding, however, and insist on the model “unknown shared ancestor → (Al, Bob).” Evolutionary systematists can infer who their immediate patristic ancestors are, which, in the case of Bob, is “Al.” Cladists apparently do not care, and phylogeneticists are confused, assuming, possibly, a stork? This analogy, though intraspecific, is demonstrative of the relationships promoted by the three schools of systematics.
The thing about analogies is that they actually have to work as analogies. This one is like saying “see, the sun is just like a light bulb, it also runs on electricity”.

The whole point (!) of phylogenetic systematics is to make a distinction between intraspecific tokogenetic relationships such as the father-son relationship between Al and Bob on one side and phylogenetic relationships between separately evolving lineages on the three of life on the other. Yes, Al is the father of Bob, and no cladist would ever have any problem with Al → Bob.

But every child should know by now that you cannot simply say chimpanzee → human. Ye gods, this is something we laugh about when creationists make that beginner's mistake. The chimps are not our ancestors because the chimps exist now, contemporaneously with us, and have evolved just as much since our lineages diverged as we humans did. Instead, we share ancestors with the chimps, and thus “unknown shared ancestor → (chimps, humans)” is precisely correct.

The same for all other species. No, today's fish aren't the ancestors of the land animals, instead these groups share ancestors. No, today's “basal angiosperms” are not the ancestors of the oaks and roses, instead these groups share ancestors. This is not rocket science.

Now the second analogy:
The reviewer avers that support for clades should be empirical, not warm and fuzzy group favoritism. So, another analogy: Consider a salesman who wants to visit several cities by the shortest route and needs to know how much money to take for gasoline. The optimal route is the solution to the classical traveling [sic] salesman problem, and support for the shortest route may be based solely on distances (how much better the shortest route is than the next shortest, and so on). Thus, support is measured by information already in the data set. But suppose information not relevant to the distance between cities is also available, for instance, roads connecting various cities are sometimes closed when you show up in your car but you know what the probability of closure is for each road and you thus know which roads to avoid. Optimality judged only from the data set on road lengths can be more precise than total evidence involving uncertainty with other information, but accuracy based on total evidence can be critical in decisions even when there is more uncertainty when all data are examined. Cladograms are precise but you can run out of gas.
What is this even about? Does Zander argue that he should be allowed to make taxonomic groups up and ignore empirical data because phylogenetic trees need to include information on road closures? I do not even... what? What aspect of phylogenetic relationships is the road closure an analogy for? I mean, if he has any evidence that is more “total” than what the cladist has, why can't the cladist take that additional evidence and enter it into the same analysis they did before? And what would Zander then have left except, once more, a gut-feeling-based rejection of the results of that analysis?

Anybody care to explain? Help?

No comments:

Post a Comment