[This started as a discussion of the debate mentioned below. It got lost somewhere, and became me riffing on my favourite topics. Sorry.]
I love it when people I know have a barny* in public, but it presents some delicate choices and sensibilities to be honoured. The case in point today is between Malte Ebach and David Williams in the red corner, and Joseph Felsenstein in the blue. I'm not the referee - I'm just the seasoned journalist in the front row...
The issue is what counts when we classify in biology, and why. Malte and David argue that there are some notions of classification that are pre-Darwinian and valid. I won't rehearse them here, you can go read their book, or the blog itself, but this all hinges on the meaning of a "natural group".
Malte and David (hereafter M&D) write:
Fundamentally, monophyly is defined as a "natural group or classification". What this means is that it is a "natural group or classification". That is all. We may choose to interpret monophyly in different ways. It could be "a group that includes a most recent common ancestor plus all and only all of its descendants" (Kitching et al. 1998: 210). It could be a group that may have a ".. general typical organization" (Agassiz, 2004: 182). In either case, the fundamental definition remains universal despite the explanation given.
In case you are not aware, the Louis Agassiz volume here is a reissue of his Essay on Classification published in 1856. In their book, they provide an extensive, and one of the best I have seen so far, historical narrative of the notion of natural classification. But I wonder if in fact Agassiz' notion is in any way the same as the post-Hennig cladistic sense.
A comment of Plato's that is often repeated by taxonomists is this from the Phaedrus 265d-266a:
Soc. The second principle is that of division into species according to the natural formation, where the joint is, not breaking any part as a bad carver might. [Jowett’s translation]
But is this really the same as "natural classification in Agassiz and others? Plato means here, by "nature" everything, including social terms like justice and the good, which are logical species, not biological. So the sense of "natural" changed somewhat between the Greek era and the modern era.
People were seeking "natural systems" from the beginning of natural history in the 16th century. At first, it was probably an undefined goal, one that was obvious. We want to know what kinds of plants there are, for such purposes as herbals, agriculture, and so on. Prior to this development, the listing of plants was local, undisciplined, and often imposed by other considerations (such as the diseases one used plants for - herbals were pharmacopeias).
Linnaeus' system was often, later, called "the natural system", which is ironic as he himself considered what we now call the Linnaean system, based solely on fructification organs in plants, to be a matter of convenience for learning and "data storage". He tried a "natural system" based on many characters, but did not get far. And the work of Michel Adanson, in Familles des Plantes (1763-4) is important also:
It was necessary to seek in nature for nature’s system, if there really was one. With this aim, I examined plants in all their parts, without omitting one, from roots to embryo, folding of leaves in the bud, anner of sheathing, development, position, and folding of the embryo and radicle in the seed relative to the fruit; in a word, a number of features to which few botanists pay attention
Adanson tried, therefore, to use as many characters as he could, although he was hardly a pheneticist in the sense of Sneath and Sokal in the 1970s. Adanson was later claimed, somewhat illicitly, as a precursor to the so-called “phenetics” school of classification (Winsor 2004), but it is not the case that he held that all characters have equal weight in classification, but rather that there should be no a priori specification of what characters were to be used, as Linnaeus had done (Stafleu 1963, 184). He noted that “[w]hat is sufficient to constitute the genera of certain families is not sufficient for other families, and neither the same parts nor the same number of these parts invariably furnish these [constituent] parts in each family” (Morton 1981, 305).
So, is "natural system" a term independent of its explanations, like "spoon", as M&D suggest? Is it functionally the same thing in Adanason, Linnaeus, Agassiz and modern cladistics? I do not think so, and here's why.
One of the things that happens when you read historical documents is a constant temptation to treat terms the same way as they are now. This is referred to as "presentism" or "whiggism" in history. Science is, as a matter of necessity, presentist; it is obvious that the current state of research is the one that is most adequate in nearly every field. But this tends to mean that historical precursors get treated as heroes or villains, depending on how they are interpreted with respect to the currently favoured view.
But the notion of a natural classification in the twentieth century was rather different to that in the eighteenth. Adanson and Linnaeus assumed there was a way to bolster classifications naturally; but we just hadn't yet worked out the details. By the time Agassiz wrote, his view had hardened into the notion that God's plan was the target (a view that Linnaeus occasionally, but not emphatically, expressed); so we should, according to him, expect that all taxonomic ranks were natural. He had moved from Linnaeus' own view that the "natural system" was a matter of convenience to the view that classes, orders and kingdoms were real objects. He said:
There is a system in nature … to which the different systems of authors are successive approximations. … This growing co-incidence between our systems and that of nature shows … the identity of the operations of the human and the Divine intellect …
(Agassiz 1859, 31)
He also, however, thought that these were ideal systems, not in the sense of being abstractions, but in the Platonic sense - the realities to which individual organisms answered were Platonic Ideas:
Species then exist in nature in the same manner as any other groups, they are quite as ideal in the mode of existence as genera, families, etc., or quite as real. …
In short, he was a Platonist. His "natural system" was a system of thoughts in the mind of God.
So already we have a major shift in what counts as "natural". Linnaeus and Adanson held that the natural system would be diagnosable by using what amounts to a multivariate cluster analysis. But Agassiz held that this would never give more than a rough approximation to the clarity of the true forms and taxa.
Darwin of course and his contemporaries held a variety of views. There were the various "numerological" systems of Macleay and Swainson. Darwin famously argued that classification was genealogical (which was not, in itself, novel; it was widely held that species were genealogical objects, and Buffon had earlier held that larger groups were devolved or degraded forms of a genealogical ancestral stock. Darwin's novelty was in the extension to all organisms, at least to the kingdom level). But it is with Haeckel and the early German paleontologists of the 20th century that phylogenetic relations become the core of classification, and we all know, of course, that Hennig defined a natural group as a monophyletic group.
But there remains a problem for this sense of natural group or classification. Cladists do not, in their own right, have a taxonomic unit. That is, there is no phylogenetic notion of species that applied to all and only species. There are roughly three ways out of this: one is that you might identify or diagnose a species or nodal taxon independently. The second is that you might take some phylogenetic criterion and apply it to organismal lineages. And the third is that you might simply say there are no nodal taxa, just specimens.
Each of these have been proposed. Many use some kind of cluster analysis (which is, I think, what a phenetic group is) to identify your nodes; the group at the species level is whatever is roughly similar, or at some percentage of agreement, either genetic or eco-morphological. Some rely, as Hennig himself seemed to, on a Dobzhansky-Mayr notion of species as independent breeding populations. This is the "diagnosable cluster" definition first proposed by Donn Rosen, I believe, and later by Norman Platnick and Gareth Nelson in their "blue book".
The second approach is to treat species as those that are "reciprocally monophyletic", which as far as I understand it means that any two organismal or populational lineages will share a common ancestor with any other lineage/population of that species; or in neutral terms, that they will share a unique combination of character states. The problem with this is that we often find species that do not share these states or ancestors, and they are sometimes called "paraspecies" for that reason.
And some, like Gareth Nelson, say to treat species as if they were any other taxa, and not privilege the rank at all. There will be diagnosable clusters at the levels of population, haplotype group and so on, and they should be diagnosed. All that you need for a cladogram is a specimen.
Now, these different conceptions indicate that what counts of "natural" is pretty diverse. It does not seem to me that cladistic classification is in possession of a notion of taxon that grounds its classifications, although the naturalness of a clade (a monophyletic group) is at the very least one important kind of "natural object". As has been said, I think by Sober, a clade is a segment of a tree that is formed by a single cut - this makes clades objective parts of a larger object (the tree of life).
Words do not remain static, nor do they always turn on a given moment. It is true that the notion of "natural" that Darwin and his contemporaries used is not radically different from those who immediately preceded them. In fact, my slogan is that naturalists were not stupid before Darwin and did not get preternaturally smart after him. He and Wallace are just one moment in the development of taxonomic ideas. However, I do think that classification changes over time in what it means by terms like "natural classification".
There has to be a tension between the use of history by scientists to merely backup what they were going to think anyway ("Darwin represented the birth of the modern blah") and the deprecation of everyone because no one person is wholly responsible for the way ideas and terms have changed. Take "evolution" for example, and M&D do. It originally had nothing to do with what we call evolution today. It meant, and this is why it was applied to supra-specific change initially, the orderly developmental sequence of growth from conception. The early evolutionary views were based on the great chain of being, and it was thought particular species underwent a process of change from their spontaneous generation through to their "senescence" and possible death. The reason Darwin is so reluctant to use the term is because of this connotation. He lost that battle, of course, but terminological wars are a marked feature of the development of science. Often, who controls the word controls the debate (think of the way Mayr tried to redefine "monophyly").
Darwin and Wallace were not mere historical ephemera instantiating the ineluctable progress of science. Nor were they the crucial individuals in every aspect of the development of future biology, lone geniuses who discovered what could not otherwise have been. They are neither dwarfs on the shoulders of history, nor giants in a featureless plain with neither precursors nor peers. History is not that simplistic, and scientific language evolves in many ways and rates, just as do the taxonomic objects they refer to.
* Technical Australian, and probably IrishCockney, term for a bar brawl.
Late note: David and Malte have responded, graciously and without eviscerating me, here.
Agassiz, Louis. 1859. An essay on classification. London: Longman, Brown, Green, Longmans and Roberts and Trubner.
Morton, A. G. 1981. History of botanical science: an account of the development of botany from ancient times to the present day. London; New York: Academic Press.
Stafleu, Franz Antonie. 1963. Adanson and the «Familles des plantes». In Adanson: The bicentennial of Michel Adanson's «Familles des plantes», edited by G. H. M. Lawrence. Pittsburgh PA: Carnegie Institute of Technology.
Winsor, Mary Pickard. 2004. Setting up milestones: Sneath on Adanson and Mayr on Darwin. In Milestones in Systematics: Essays from a symposium held within the 3rd Systematics Association Biennial Meeting, September 2001, edited by D. M. Williams and P. L. Forey. London: Systematics Association.
Barney def:- Argument, generally British English, probably middle 19th century, etymology unknown.
Good meander through some interesting stuff, agree almost 100% with your comments on the use and abuse of history of science.
Oooh! I love that last paragraph ... which is why it may go missing in the middle of the night.
which is why it may go missing in the middle of the night.
Hands off I saw it first!
See that thing in grey to the left? That's a Creative Commons license - you can use anything I write on this blog, so long as you attribute it and don't modify it...
Oh come on Wilkins! What is the point of borrowing something as good as your last paragraph if one cannot claim to have written it oneself?
A response to this post can be found on the Systematics and Biogeography blog.
John, your comments are interesting but you say of the argument between Ebach/Williams and me that
"The issue is what counts when we classify in biology"
and then you give an interesting discussion of the history of classification. Ebach and Williams were originally upset at the history chapter in my book, "Inferring Phylogenies" and they keep using the term "phenetic" to refer to my position. In the book I wanted to avoid the terms "phenetic" and "cladistic" when discussing not classification, but the inference of phylogenies. I disclaimed any interest in discussing classification, and was at pains to argue that I wasn't talking about classifying. They want to use the term "phenetic" for some (most, I think) positions people take on reconstructing phylogenies. I don't see where your erudite and interesting historical discussion addresses this. (Nor does their more recent post, intended by him to "illuminate the differences of opinion between Joe Felsenstein and ourselves.")
How your post and theirs are relevant to the argument escapes me.
Joe, I know I failed to address the debate - I said as much in the disclaimer. But this barn[e]y has me thinking and I'm going to write a post on what I increasingly think of as Two Dogmas in Classification (a resonance of a philosophy paper of some renown). It will take me a while.
I only addressed the post Malte and David wrote, which is very far from the claim that your views are phenetic. I don't know myself why they think that - do you use cluster analysis in a manifold? I haven't read your book carefully (it's on the list).
I know you are of the It Doesn't Matter Very Much school with regards to classification; what do you think the purpose of phylogenies is? I expect you will say (and I would agree) that they are the basis for further inference, but I truly don't know.
I've put a link to this post up at Linnaeus' Legacy.
I only noticed the disclaimer after I posted my complaint. Oops. Anyway, my understanding is that if you take some aligned DNA sequences and use any ordinary phylogeny method (say parsimony, for example), with the intent of reconstructing evolution but not with the intent of classifying, Williams and Ebach will say you're being a pheneticist. That recruits an awful lot of people to the fold of pheneticism. Or am I wrong in this interpretation?
(And to answer your question, yes, you can use phylogenies to make further inferences about evolutionary mechanisms, whether or not you also use them to classify.)
I think it's not phenetics to do that. But it is a large difference between your approach and Malte's and David's. For them, as I understand it, reconstructing evolutionary history is something apart from drawing up phylogenies (Malte, feel free to correct me on this). The so-called pattern cladist approach assumes only that there are relationships between taxa, and this acts as a test of any historical phylogeny. In Gary's and Norman Platnick's "blue book" they make the now-uncontroversial point that for any cladogram there are an exponential number of possible consonant historical trajectories. So at best a cladogram constrains only the space of possible histories, but it doesn't ipso facto provide one.
How this is phenetic escapes me. I hope Malte and David expand on this. I can't find it in their book.
On classification, I think that you don't use a cladogram to classify - it already is a classification. In my forthcoming post, I will aregue that there are two kinds of classificatory activities - one by identity (or homology), and the other by similarity. Both offer equivalence classes: cladification by monophyly, and gradification (my own term) by overall similarity.
Cladification is strict and allows inductive inference across clades. Gradification is open-ended, and permits only deductive inference across the similarity metric.
My 2 cents...
Still puzzled: is what I proposed doing (using an alignment, a parsimony algorithm, and making an estimate of the unrooted tree) the same as making a cladogram (in their view, or in yours)? Making a phenetic classification? None of the above? If it's something else, what is it? I would say it's estimating an unrooted phylogeny, but do they define that as something else?
An example of a cladogram versus a given a number of possible lineages can be found in Chapter 1 (1.2 Cladistic Analysis) (Williams & Ebach 2007: 9 - 11).
A response to the comments above can be found here.
I don't have the book to look up the example, and I don't see any answer to my questions in the "response to the comments", which has many nice things to say, but about other issues.