Are morphological taxonomists just going to ignore molecular data?

Myrmicine ants are a mess. The subfamily Myrmicinae holds about half of all ant species and includes fire ants, harvester ants, leafcutter ants, turtle ants, and many more. Attempts to arrange the diversity into workable subgroups have been difficult and largely unsatisfying. Many species stubbornly show traits of several groups at once. Molecular trees (Moreau 2006, Brady et al 2006) have reflected this uncertainty, recovering only a few reliable clades among a statistically uncertain base.

Consider the tribe Stenammini (sensu Bolton 2003) laid over Corrie Moreau’s 2006 molecular chronogram of the subfamily:


There is certainly no molecular signature for Stenammini as a group. As we can see, the four included species do not group together. What’s more, the morphological traits defining Stenammini- like a 12-segmented antenna with a 3-segmented club- are mostly just the pleisiomorphic base states for ants. So morphology doesn’t give us compelling reasons to accept Stenammini either. If we accept this pattern, Stenammini appears to be a random draw of myrmicine ants that retain an arbitrary set of ancestral characters.

I bring this up because of Boltonidris, a new fossil ant genus described a couple weeks ago by Radchenko and Dlussky and painstakingly placed in the Stenammini:

Abstract: The new extinct ant genus and species, Boltonidris mirabilis, are described from the late Eocene Rovno Amber (Ukraine). This genus belongs to the tribe Stenammini of the subfamily Myrmicinae. It possesses the plesiomorphic characters of the tribe Stenammini, e.g. 12-segmented antennae with 3-segmented apical club, characteristic structure of the clypeus and frontal lobes, absence of gastral shoulder, but it has a series of autapomorphies, e.g. modified mandibles with the only two teeth on the masticatory margin, well developed longitudinal medial groove on the head dorsum, somewhat depressed areas lateral to the frontal carinae (like “vestigial” antennal scrobes), and finely swollen postero-lateral area of head, close to the occipital corners. Additionally, it has two short blunt teeth on the pronotum.

Here’s a pic:


The descriptive work in the paper is solid, and it looks to my untrained eye like this ant deserves the new taxon. So the paper is a valuable contribution.

Yet Radchenko and Dlussky spend much of their discussion relating Boltonidris to other Stenammini, with no mention that the two most ambitious attempts to resolve myrmicine relationships to date (Moreau 2006 and Brady et al 2006) both concluded Stenammini didn’t exist. Thus, theirs reads like a discussion to determine which pinhead is most likely to have to staged dancing angels. Pointless.

If taxonomists aren’t going to read each other’s research, why even bother publishing?


Bolton B (2003) Synopsis and classification of Formicidae. Memoirs of the American Entomological Institute, 71: 1–370.

Brady SG, Fisher BL, Schultz TR, Ward PS (2006) Evaluating alternative hypotheses for the early evolution and diversification of ants. Proc Natl Acad Sci USA 103:18172–18177.

Moreau CS, Bell CD, Vila R, Archibald SB, Pierce NE (2006) Phylogeny of the ants: Diversification in the age of angiosperms. Science 312:101–104.

Radchenko A, Dlussky GA (2012) Boltonidris Gen. Nov., the First Extinct Stenammini Ant Genus (Hymenoptera, Formicidae) from the Late Eocene Rovno Amber. Annales Zoologici 62: 627–631.

17 thoughts on “Are morphological taxonomists just going to ignore molecular data?”

  1. A couple of questions for you (because I truly do not know much about such studies) and some observations.

    How many of these type of molecular studies used a large (n > 10) sample of within colony, separate colony & widely separate range replicates within species ? How do we know the nuclear DNA segment used in these studies is relevant and appropriate for use in phylogenetic reconstruction ? How much variation (statistically significant or not) in these DNA segments is there within and between populations in each of the species studied ?

    Assuming the answer to Q1 = 0, Q2 = we don’t, and Q3 = unknown, why would any of us be surprised if morphologist’s ignore or discount molecular studies for cause. The impression I get from what I little I do know about molecular studies is that participants regularly employ stats when it suits them and ignore the requirements of stats when it does not — but correct me if I have it wrong in the case of all ant researchers.

    One thing is certain. The study of species populations and genetics are profoundly based on statistical processes. I cringe when I see inadequate sample size or even as sad as ONE per species in such studies. I see nothing wrong with small sample size as long as we recognize the massive limitations on inference that results.

    1. Sample size should be evaluated relative to the question, I’d say. That’s a valid point. I’d not want a sample like that shown above for delineating species or genera.

      In my experience, DNA sequence of nuclear genes is largely invariant (0-5% variable sites) for most loci within an ant genus. Thus, including many representatives of each species wouldn’t provide much additional information at the level needed to answer questions about ant tribes. The sample could certainly stand to be larger, but there’s enough here to be able to say that a monophyletic Stenammini would make the observed genetic data nearly impossible.

      What really bugs me, though, is that this study didn’t even cite the earlier papers, even though both were by prominent ant systematists (the Brady 2006 team includes several morphologists, too), and both were in high-profile journals. It’s one thing to take in the molecular data and disagree with the conclusions. It’s another to not even acknowledge the earlier research exists. The bibliography reads like a C.V. of the authors.

      1. Thanks for the response, Alex, that was helpful. 0 – 5% variable sites is kinda what you would expect with closely related groups. We supposedly share 50% with bananas and 99% DNA sequences with chimpanzees.

        The replicates within species and between species would be NOT performed to provide more info for tribe differentiation. Rather it provides info that would say some loci are more or as variable within species as between species, genera or tribes. Thus allowing the elimination of loci/segments that should not be used to characterize larger clades. What exactly are the investigators characterizing in that DNA segment ? Irrelevant non-coding random variability left behind by some million year old viral infection or some characteristics showing the road-map of ant evolution ?

        What bothers me with small sample size is that the nature of variability within the DNA segment selected are ignored. How does the scientist know the variability represent some characteristics that vary within the species more than between species, genera, families, say as the result of current race formation, some new protein, or simply just some random set of somatic mutations of the particular individual tested. These are the basics that one would expect to be eliminated by adequate sample size, particularly if the function of loci on the DNA segment compared is largely unknown which I suspect is the case here.

        I agree that scientists should be well read in their field and incorporate discussion of previous relevant conclusions in their paper but parochialism by anyone should not surprise too many of us.

    2. So Alex responded earlier, but I was going to say this, in response to BioBob:

      Q1: it’s not important to have deep intra-species coverage for this type of study because that’s not the subject. Here it’s the deeper nodes (genera, subtribe, tribe, subfamily…) that are studied. So, really, if the single specimen per species is well identified, then it gives the information that is wanted, and sampling more specimens of the same species would just increase costs astronomically for little scientific benefit. Large intra-species sampling in important when studying populations, large intra-genus species is important when studying species, etc. (and see Q2).

      Q2: We know from the support of nodes on the phylogeny. If support is low, then either the segment is not appropriate, or it is appropriate but not enough bases have been sampled.

      Q3: I don’t know, (and Alex has answered) but it is often low for genes that are suitable for those deeper nodes.

      1. You are missing the point. Basic sampling and stats protect the scientist from making stupid errors whatever their aim. If your one sample is the mutant from hell ant, with the ant equivalent of downs syndrome, polyploidy, with a twist of lemon, and you conclude EUREKA !, what does that do for you ?

        Replicates save you from yourself, since we ALL make mistakes. Oops, sample taken from the wrong beast, samples switched, sample contaminated, whatever.

        By support, I assume you are talking about numbers of loci in concordance. That is also my point but on the basis of sample size. As I said, scientists sometimes use stats when they want, but ignore them when inconvenient.

        On Q3, how low, how often, and how would you know with a sample size of 1 ?
        With a sample size of 1, all errors and variance are essentially INFINITE.

        1. Yes, the number of loci in concordance does increase support of nodes, and loci that are not in concordance generally lower it, but that is not what I mean by node support. The calculations for support also include statistics. I mean bootstrap, Jackknife and Bayesian posterior probabilities, they all resample the data.

          In PCR, primers don’t anneal with anything, they’re often specific, so contamination would have to come from an organism relatively close phylogenetically. It does happen, but not as often as you imply, and that’s why sequences are checked by the researchers. “Odd” sequences are usually excluded and the sample is re-amplified, or re-extracted and re-amplified.

          You are entirely right about basic sampling, but it does not follow that basic sampling should be at the species level. Here for example, four genera in the tribe gives a good idea that Stenammini isn’t supported by molecular data. Four individuals, from different genera is enough to say that. It’s even possible to constrain the tribe as monophyletic and check support for that hypothesis. I’m saying that the basic sampling at the species level should be replaced by basic sampling at the genus level to maximize useful information when deeper nodes are the subject of study. Finally it’s important to be realistic in our expectations in terms taxon coverage; as costs go down, we’ll come to expect more coverage.

  2. In my experience, you can usually recognize a good molecular study by thinking about whether its results make sense from a morphological viewpoint. Since Stenammini is not defined by any convincing synapomorphies, it makes sense that this “taxon” is not resolved in molecular studies because the morphological evidence was never very strong to begin with.
    With groups such as ants, for which there exists a good knowledge base of both morphology and molecular data, the data are expected to match, and if they don’t there can be mistakes both in the molecules and in the morphology; mistakes that need to be tracked down and found. It can be as boring as sequence contamination, or it can be as interesting as the discovery of novel character systems or even underlying synapomorphies. The molecular revolution is brilliant for taxonomy, because there is finally several independent ways of thinking about the same problems.
    Neither molecules or morphology is an inherently better method, and both need to be subject to the highest scientific scrutiny. As Alex rightly points out, the molecular people need to read and understand the morphology people’s arguments, and vice versa.

    1. Very thoughtful, Gunnar. I only wish we were further along and had better understanding of variability based on adequate sampling as well. I look forward to exciting times ahead as we understand more about insect genetics and tie loci to form, function and behavior.

      It seems to me that since so many ants are essentially chemical automata, the study of their chemical evolution would be another potentially useful research avenue.

  3. This is completely off topic: for some reason it bothers me a lot more when I can not access such specialized article in such a specialized journal due to a pay-wall than, say, an article in journals such as Science or Nature (my institution does have subscriptions to these latter ones, though). You can count with your fingers the number of people in the whole World who are interested in this paper by Radchenko and Dlussky (ok, maybe you need your toes too).

    Anyway, on topic, I cannot comment on a paper I haven’t read, but Alex’s concern works both ways: most paper publishing on molecular data to bear on phylogenetic/taxonomic questions fail to include morphology into consideration (into the analysis), and limit themselves to discuss disagreements with the standing classifications.

    Granted, in the case of Myrmicinae classification, worker morphology has been ok to create working hypothesis about delimitation of genera (while also creating major paraphyletic taxa), but it is too poor to resolve relationships among those genera.

    1. I second Roberto. The two molecular studies fail in including morphology nor is the data easily accessible to redo the analyses, so they are not complete and as much agnostic as the morphological study only. Since the molecular study does not include morphology, the two data sets cannot be combined without the effort of one side to contribute all the morphological data. I thought this was part of the grant Ward et al got to do the phylogeny in the first hand?!

      Furthermore, publishing in journals that are not open access is like hiding results from the “anonymous” reviewer. There is no excuse anymore not to make content open access.

  4. I can’t say that I can add much to this discussion, but at least for the Barcode Region of COI, the more I sample, the more haplotypes I find, the greater within “species” genetic variation seems to be present, and the more the elusive magic percentage difference between species recedes. I do try to both sample several within a site and several sites that are maximally distant (although only a few hundred kilometers apart). Still, the haplotypes show little geographical partitioning.

  5. Great post Alex and you raise an important point. However, I must say that your title does suggest that it is only the morphologists who are guilty of this, which is obviously not the case. I have read many molecular studies that have failed to cite important morphological works and vice versa. From the example you give it would appear that the morphological data alone is enough to determine that Stenammini is not monophyletic and that this is further corroborated by the molecular data. In this case the authors are guilty not only of ignoring the molecular studies, but also of ignoring the morphological evidence and continuing to define a taxon based solely on plesiomorphies. This is as much bad morphological taxonomy as it is anything else and not really a case of morphology versus molecules.

    1. This is a can of worms I probably shouldn’t have opened, but my instincts are to be harder on the morphological taxonomists when they disregard molecular evidence than vice versa. Morphologists are usually the ones who introduce nomenclatural changes that future workers must address, whereas poor-quality molecular work can just be dismissed without consequence. Taxonomists in my mind have somewhat more responsibility.

      The converse is that molecular folks often fail to follow through on the taxonomic implications of their work, of course, and that’s problematic though probably less so.

      1. Don’t get me wrong, I agree with what your saying. I just wanted to point out that the authors of the article in question ignored not only molecular data but also sound morphological evidence. By defining a clade based entirely on plesiomorphies their error was first and foremost one of sloppy comparative morphology. Of course morphologists should be chided when they disregard molecular data, but ultimately they do so at their own peril as their classifications will eventually be revised in a more integrative fashion. Fortunately, I expect (hope?) that the number of morphological systematists that ignore/dismiss molecular data is on the decline.

  6. Pingback: Links 1/18/13 | Mike the Mad Biologist

Leave a Reply