In a recent Nature, R. John Ellis, author of How Science Works, takes exception to Eugenie Scott’s review and says this about her use of “homology”:

The word was invented in 1843 by anatomist Richard Owen to mean “the same organ in different animals under every variety of form and function”. But Owen did not believe in evolution and interpreted the observation of homologies as the preferences of a supernatural agent.

Modern biologists define homology differently, to mean the occurrence of similar traits in different organisms that are evolutionarily related by descent. To cite the modern definition as part of the evidence for evolution is a classic example of a circular argument, as pointed out by evolutionary biologist Mark Ridley in his textbook Evolution (Blackwell, 2003).

The solution is to avoid using ‘homology’ when discussing the evidence for evolution, and instead use ‘similarity’, the meaning of which is intuitively obvious but implies no particular interpretation. Homology can then be used to describe one result of evolution.

Ellis is repeating a view that is widely held, but is, I think, fundamentally mistaken – and in a fight over what the history of evolutionary biology shows, I would tend on general principles to side with Genie, but not this time; here’s what she wrote:

Similarly problematic is Ellis’s idiosyncratic treatment of homology. He argues against defining it as traits derived from a common ancestry because that definition is based on a circular argument, but he does not offer a better one. So the reader is left with the wrong impression that homology is merely anatomical similarity.

More below the fold.

It is true that Owen defined homology in 1843, although similar ideas were around prior to that, particularly in the writings of Lorenz Oken (in his Physiophilosophy) and Geoffroy St Hilaire (who used the term analogie): basically it simply meant an “agreement” (the meaning of the Greek) between parts of organisms, and the famous example is that of Pierre Belon noting the agreement between the skeleton of a pigeon and a human in 1555:


The relation here was not, however, one of similarity, because in many respects (particularly the forelimbs of birds and other tetrapods) the similarity has been erased by extreme modification, and nevertheless anatomists still treat parts as homologs.*

As Ellis notes, there was a prior meaning before evolutionary biology. I am highly skeptical of Owen’s appeal to divine agency here: that doesn’t match the Owen I have read. He was more likely to assign secondary causation, and did for species as a kind of crystallisation,** but his practice showed that he identified homologies on the basis of their developmental congruence, not on any divine agency.

Ellis is right that we should not define a term on the basis of a theory that we want it to support. But there is a deeper issue here than finding an operational and pre-evolutionary meaning (which is not “similarity”, to which I shall return). It is the basic mistake of confusing the explanation for the phenomenon to be explained: the explanans for the explanandum. “Homology” is something that was observed, and later used both to bolster evolution by common descent, and then to draw up phylogenies. It is wrong to say homology is a trait  shared with a common ancestor, because we don’t observe that, given that we do not observe ancestor-descendent relations; we infer them. But we do observe homologies.

Fast forward with me now to the period just as cladistics was getting going in the later 1960s and early 70s. The school of Mayr and Simpson known now as “evolutionary systematics” was duking it out with the school called “numerical taxonomy” and which we now call “phenetics” (from the Greek word phaneros, to seem). The question was whether one might have a “theory-free” or “operational” way of identifying taxa, or not. Phenetics held that one could, and it was an attempt to use any character at all, and just run cluster analyses. Evolutionary systematics, on the other hand, was tied to classical comparative anatomy to a degree (Simpson was a paleontologist and thus simply had no other data to work with), and so it would exclude similarities that were what E. Ray Lankester had called “homoplasies”, which were agreements that were convergences, or analogies. The problem of identifying homoplasies was a methodological issue, but so far as we could, we should, because only homologies are informative. Only they give us, as it were, a “phylogenetic signal”.

However, now we have a circularity: if phenetics won’t work (it doesn’t, since if you don’t exclude the homoplasies you get contradictory groupings; hence you need to be “theory-dependent”. However, the algorithmic techniques of cluster analysis survived the death of phenetics itself), and if you need to use evolutionary history to exclude homoplasies, don’t you have a viciously circular argument in each case? You are determining evolutionary history in terms of evolutionary history?

Along came cladistics, or as it is properly called, phylogenetic systematics.† Based on the ideas of Willi Hennig, an East German specialist in dipterans (flies), this held that one can recover the evolutionary history by identifying the homologies (he called them synapomorphies; for those interested in the full panoply of cladistic terminology, go here). Group them into a schema, and you will generate a tree-shaped classification he called a “cladogram” (literally, drawing of the branches). Hennig thought, and many still do, that the cladogram was an evolutionary tree.

But a number of cladists thought that this move was too quick, and that there are a large number of actual histories, that is, of evolutionary trees, that were consistent with any given cladogram of sufficient complexity, and so the cladogram only tested these hypotheses of history. To these theoreticians, a cladogram was a summary of the overall homologies, a “synapomorphy scheme”, so the nature of homologies was critical. Niles Eldredge suggested a distinction of homology into taxic homology, which was used to identify taxa, and transformational homology, which was a matter of comparative developmental biology. Colin Patterson, one of the “pattern cladists” who thought that a cladogram was not, ipso facto, an evolutionary tree suggested that synapomorphy and homology were the same.

Taxic homologies are assertions of relations between taxa:

Assertions such as “simple perforation plates are a synapomorphy” or “laticifers are homologous” are meaningless by themselves. However, “simple perforation plates are a synapomorphy of the two-taxon clade in Fig. 2” or “axially oriented laticifers characterize Moraceae” specify unique groups of the phylogenetic hierarchy. [From Olson 2005:510]

The notion of homology is complex, and as we recently saw when I asked about the use in mathematics, it has a slew of other meanings, but the one that seems to me to be consistent across all uses is this: a homology is a mapping or “agreement” of parts of organisms with other parts of organisms. A mapping relation is not a similarity, and it is not the explanation of the relation (such as evolutionary common ancestors, which are proposed to explain the homology). It is an identity relation: this is the same as that. The identity may be an identity of place, of sequence, of developmental process, or just of a shared name, but what it is not is similarity or common ancestry. Similarity may be how we identify homology (and what kinds of similarity depends on what we use), and common ancestry may be how we explain homology, but in both cases homology is the relation itself.

So I agree with neither Genie nor Ellis…

[Hat tip: Roberto Keller]

* The English writers of the early 19th century, under the baleful influence of Coleridge and his school, added Francophilic –es to many words, but the American usage is older and to be preferred. Its a programme of mine.

** Likewise, -ize is to be rejected as unsightly.

† The term “cladistics” is an insult by Mayr, who contrasted it with “classification”. This will become significant, dear reader…

7 thoughts on “Homology

  1. “Cladistics” and “Phylogenetic Systematics” have at some point diverged in meaning: Now, “cladistic” analyses, which can produce “phylogeny,” are a form of “systematic” recovery, and is mathematical in form: A cladistic analysis recovers a potential phylogeny using various algorithms to find a best fit model with a series of operation taxonomic units (OTUs) and a series of numbers representing predefined sequences (DNA, or a character matrix), and using the overall math to match the numbers among OTUs the best way (parsimony, jackknifing, etc.). “Phylogenetic systematics” is the process of assessing “total data” (the principle of non a priori exclusion of pertinent data) and is more phenetic than it is mathematical, although it can be quantified using “cladistics.” The end result is that PS uses Cladistics to formulate hypotheses, and is thus a thing to observe — an object — while PS is the observing medium — the way or process, the seeing.

    I may be wrong.

    1. Yes, but I think this is because the cladistic revolution was not properly incorporated in modern thought before a consensus, which I call the Felsenstein Syndrome after Joe Felsenstein’s influential views on phylogenetics, took hold and became orthodoxy.

  2. I’m only very lightly read in Owen, so I might be off a bit, but I suspect Ellis’s attribution of the position that the homologies are to be explained by divine preference arises from a misreading of Owen’s later Platonistic musings, which had nothing to do with ‘preferences’ but types, and wasn’t put forward to explain homologies but simply (if I remember correctly and am not accidentally Whewell-izing him) to explain, speculatively, why that general kind of explanation turned out in fact so useful.

  3. Genie didn’t come up with that way of putting it herself, rather it reflects pretty well what the professors teach at Berkeley, e.g. Kevin Padian, Brent Mishler, et al. Genie was just doing a book review so it’s not like the whole thing could be explained, but the full picture includes a distinction between ontology and epistemology, i.e. definition and diagnosis, and someone like Mishler would say homology is defined as a character sharing common ancestry, but is diagnosed through similarities like shared relative position, similar development, etc. These are used to create a hypothesis of homology, i.e. each character is something you believe to have a higher probability of being due to common ancestry than not. Then you assemble all of these characters and do the phylogenetic analysis, and inevitably some turn out to be homoplasies, but (you hope) the law of large numbers applies, and the dominant congruent signal will be the phylogenetic one, a conclusion which itself can be tested in various ways…

    Defining homology based on shared similarity, development, etc. is ok as a first approximation, but eventually leads to homology-bogglers, i.e. when something about the development shifts but the final structure remains the same. E.g. the eyes of frogs obviously all trace back to a common ancestor, but in some species the eyes develop out of a different tissue than normal, leading to cries of eyes not being homologous among frogs from certain excitable people, cries assiduously picked up and distorted by people like Jonathan Wells. There is a similar story with dinosaur/bird fingers and the Birds Are Not Dinosaurs people (BANDits).

    So initially I didn’t like defining homology based on common ancestry, but now I’ve been convinced/brainwashed that the Mishlers of the world have got it right…

Leave a Reply