For example, several characters that are given as unambiguous synapomorphies of a major group, are also shared with other major groups. Iguaniamorpha, as far as I can tell is synonymous with Iguania under Conrad's taxonomy and are defined by what I consider to be mostly bad synapomorphies which include obliteration of the notochordal canal by centrum ossification, procoelous vertebrae, premaxilla fusion and possession of a ventromedial process of the pterygoids. All of these are shared by some major schleroglossan groups! Just about every lizard is procoelous. Gekkotans are the only ones that I can think of that have an unossified notochort, and that is probably due to the highly derived, paedomorphic state of the group rather than an ancestral condition, something that should have been apparent from comparison to Sphenodon sp. as the outgroup. A fused premaxilla turns up allot throughout squamate evolution: iguanians, varanids, mosasaurs, snakes, dibamus, some skinks, some geckos.
The ventromedial process of the pterygoid is especially problematic because it not only appears in other major groups, particularly the Scincophidia (tax. nov.) and the anguinae, but it is reversed several times back and forth within the Iguania. The Chamaeleontiformes, which includes the living acrodont iguanians and several fossil taxa is considered one of the major divisions in iguanians in other analyses has one reversal from the 'ancestral' condition and there is yet another reversal from the chamaeleontiform condition in two fossil taxa within that group. This happens again later in the Scincophidia. It supposedly united some derived skinks, dibamids, amphisbaenids and snakes, but dibamids don't actually have it. This isn't the only character that does this, but I'm using it as the example.
Intuitively, reversals from the ancestral character state shouldn't go back and forth several times in one lineage. If it does, then there must be a strong functional correlate to it and it is therefore a highly plastic and probably bad character. Also, characters like this popping up all over the tree would make it very hard to think of the systematics in terms of a dichotemous key. I realize that a dichotemous key doesn't actually reflect evolutionary relationships, but if you only have one or two characters that unite big groups, it aught to at least be close or show up at some point in the creatures development. It works for plants, birds, mammals and just about every really major group, so why not lizards?
The analysis has 363 characters and 222 ingroup taxa, so clearly this isn't a problem that is just going to be solved by throwing more characters or more taxa at it. So what other options are there? Character weighting or unidirectional characters? The snakes with legs are my preferred group for this example but other groups could be used. In the Conrad analysis, snakes fell out with other legless lizards in a clade supported by characters largely relating to enclosure of the braincase and modification of the palate. That by itself doesn't really pose a problem except that snakes with legs, some of the oldest known fossil snakes, fall out as derived macrostomatans. They are legged beasts nested deep within an otherwise legless group. I find it hard to believe that they either re-evolved seemingly vestigial legs or that every other group independently lost them. Possession of legs is such a big character that it implies that the polarity of other, more fine detail characters, is wrong. Do you make leg possession a unidirectional character? Weight it more than other characters? Both of these seem wrong since the natural polarity of a character should be supported by the other characters in an analysis.
The problem of convergence in snakes and legless lizards could be huge. The uropeltine snakes are a very bizarre group, specialized for a very specific mode of life. If they are a natural group, one would expect them to come out in a fairly basal position due to the sheer number of characters that they have in common with each other compared to any other member of the snakes. If there is a single charcter that would, say unite them with Cylindrophis and 5 that unite them with each other, then Cylindrophis might be pulled out and stuck with macrostomatans because of plesiomorphies. Establishing character polarity in a group is important but difficult, even in groups that you are certain are monophyletic. The basal split between horned lizards (phrynosomatidae) and all other iguanians seems similar. You have a very bizarre, highly specialized group, of unquestioned monophyly pulled out into a basal position on the phylogenetic tree. This kind of thing could be what causes all of the character reversals higher up in the tree.
4 comments:
I learned that you know a lot about lizards and I know nothing. No, that's not true, I know what I learned from your talk. Oh, Will, you're tho thmart!
My goal is not to overwhelm you but to instruct. I'm glad that you learned something, but I really could use constructive feedback about improving my talks. That's the second time I've given this talk in 4 years and both times people said it was good but couldn't give me constructive feedback.
I actually agree with a lot of what you said in your blog post and have many of the same concerns... The tree topology and distribution of synapomorphies is simply what I came up with given the data set.
Since publishing that paper, I've eliminated all of the composite codings from my data matrix, expanding it to around 415 taxa and 630 morphological characters. The tree is largely the same (there are, of course, some exceptions), although synapomorphies supporting some of the groups have moved around.
Some of the major problems are this: There is a long missing lineage for the earliest squamates and homoplasy is RAMPANT.
Keep up the good work,
Jack L. Conrad
Post a Comment