Attached is a recent publication from my lab group that shows morphology provides poor resolution for darter phylogenetic inferences. I am excited to post this here, because I think it may be the first citation of Chris Scharpf's checklist in a peer-reviewed scientific publication. So congrats Chris. We will hopefully be citing your checklist in our future work.

Edited by TomNear, 05 November 2009 - 11:17 AM.

Tom, Are all 135 morphological characters treated equally in this analysis? Seems like some would be naturally more "plastic" than others, and treating them all equally might make morphology look worse (for reconstructing phylogeny) than if the great darter-heads of the world like you pared down the list to exclude characters that are excessively plastic, or weighted the least plastic (good) ones.

Gerald-Good observation. In the Bayesian (model-based) analysis the variance among characters with regard to rate of change is corrected for in the tree estimation. If you were to exclude the fastest evolving characters, my guess is that the trees would be even less resolved than with the full dataset. Bottom line is that it is really hard to estimate a phylogeny of more than 20 species, and for these fish species discretely coded morphology is not going to provide well-resolved phylogenetic trees. This may not sit well with traditionalists, but it is what the data is telling us.

Tom, Are all 135 morphological characters treated equally in this analysis? Seems like some would be naturally more "plastic" than others, and treating them all equally might make morphology look worse (for reconstructing phylogeny) than if the great darter-heads of the world like you pared down the list to exclude characters that are excessively plastic, or weighted the least plastic (good) ones.

Tom, I think you and your lab people are very brave with this line of analysis -- much of the observed phenotypic variation could be due to pleiotropy, probably even dicier than plasticity for parsing out. Alleles with loci on different chromosomes shaping the same phenotypic character aren't under the same selection forces, and observed variation could result from relatively stochastic processes. I don't know if saying this makes me a traditionalist, or not, but the idea goes back to Mayr in the early 60's, anyway.

Bruce, I agree with your point about the multi-locus approach to thinking about phenotype. I am not sure how this is germane to the discussion of whether phenotypic characters will give a well resolved phylogeny. Can you elaborate?

Tom, I think you and your lab people are very brave with this line of analysis -- much of the observed phenotypic variation could be due to pleiotropy, probably even dicier than plasticity for parsing out. Alleles with loci on different chromosomes shaping the same phenotypic character aren't under the same selection forces, and observed variation could result from relatively stochastic processes. I don't know if saying this makes me a traditionalist, or not, but the idea goes back to Mayr in the early 60's, anyway.

Assuming any phenotypic variation is the result of pleiotropy from unlinked loci with multiple alleles at each locus, the variation may not be a clear phylogenetic signal since the interactions of different alleles could in principle produce a convergent phenotype. That sounds very lawyerly. But it makes major assumptions about the degree of genetic diversity within as well as between the various species which, frankly, I have no idea exists. I know that such work exists on commercially significant species like some salmonids; I would be in shock if it's been done in darters.

For phylogenetic work it's largely mtDNA being sequenced and analyzed with some nuclear sequences. But how much functional genetics research is done on darters? I say this not to question anyone's work, but from my own ignorance/curiosity. At this rate I'll be proposing a pleiotropic species concept.

With all the handwaving I truly am impressed by work that undertakes rigorous analysis of morphological variation to examine possible phylogenies. Life would be simpler if it matched.

Bruce, I agree with you completely. The future of darter phylogenetics, and phylogenetics in general is the use of DNA sequences from multiple genes that are analyzed in the context of variation among the gene trees that is due to "ancestral polymorphism" that results from the stochastic processes of sampling (e.g. drift). This is called "species tree" estimation and is an exciting direction at the cutting edge of phylogenetics. In our lab we are moving away entirely from using mtDNA for big picture darter phylogeny, because a large fraction of darter species have a heterospecific mtDNA genome as a result of hybridization with other darter species. This pattern has been reported in at least three papers, but it will be addressed generally in our big darter phylogeny paper.

The point that Nicole and I tried to make in this paper is that the best morphological dataset ever assembled for darters does not give a tree, or set of trees that are resolved and well-supported. That is not to say morphology cannot get it right when it comes to phylogeny, it just says that for the characters that are scored to estimate phylogeny in fishes are not that variable among darter species (e.g. osteology, myology, etc). To obtain a dataset that I would feel comfortable with, given that there are approximately 240 darter species, would take more than 700 morphological characters! This is just not going to happen. To resolve nodes in the phylogeny you need character changes, there are 240-1 nodes in the darter tree and roughly three character changes along a branch are needed to resolve a node. It is simple, for a clade as species-rich as darters there just is not enough observable morphological characters to obtain resolution.

Why I think that this is important is that many of the traditional practitioners of darter systematics accept groups/subgenera that have no phylogenetic basis. One cannot say that charcters X, Y, and Z define a clade unless those characters are scored along with a substantial set of other characters in the context of broad taxon sampling. Then one performs an analysis (e.g. parsimony) to determine if indeed 1) the subgenus is monophyletic and 2) it is defined by characters X, Y, and Z. It is not good practice to paint the characters on predefined groups. The fact that molecular analyses run counter to these pre-phylogenetic classifications is not a surprise.

I think that "morphology" in darters is consistently conflated between evidence for monophyly of groups and evidence of differences for species delimitation. These are two very different things and the types of morphological characters used for these different purposes are often different as well.

So when I present a phylogeny that shows the Percina subgenus Alvordius is not monophyletic this does not conflict with "morphology" as no one ever, and I mean ever, presented phylogenetic evidence from morphology that Alvordius is a clade.

I don't think what we are doing is brave, it is just true to the patterns emerging from the science. I will accept bold, but brave is certainly outside of our morphospace!

Edited by TomNear, 05 November 2009 - 06:56 PM.

I haven't had a chance to read much, but all I have to say is this was part of an undergrads work?!?! I know it's an Ivy school and all, but wow did that make me feel a little less intelligent. Kudos. The oppurtunities for an undergrad to publish, let alone participate on something like this, are too few.

Yes, this was Nicole's "senior essay" project, a requirement of all students in the College. I have had the benefit of mentoring eight senior essay students in my three years here. It has been the most rewarding part of my experience as a faculty member at Yale. More papers from these undergrads will be published soon. These students are really good and provide a wonderful talent pool to recruit future ichthyologists/evolutionary biologists.

I haven't had a chance to read much, but all I have to say is this was part of an undergrads work?!?! I know it's an Ivy school and all, but wow did that make me feel a little less intelligent. Kudos. The oppurtunities for an undergrad to publish, let alone participate on something like this, are too few.

I don't think what we are doing is brave, it is just true to the patterns emerging from the science. I will accept bold, but brave is certainly outside of our morphospace!

OK, I was trying not to say "bold"... I think my Mayr-ian thoughtcloud agrees with your intent. I was being literal-minded at first but now that I don't have a class in 14 minutes I more clearly see what you're saying. Maybe someone will come up with a 240 x 700 matrix for a completely obsessive morphological analysis? It is a staggering thought!

I think that it is more likely that someone will come up with a 240 X 20 matrix, where the 20 is the number of nuclear genes sequenced. This IS the future for darter phylogenetics. To quote a friend of mine, Prof. Mike Bell, "Morphology is too damn interesting to be wasted on phylogeny."