Gerobatrachus hottoni

Blogging on Peer-Reviewed Research

It's another transitional form, this time an amphibian from the Permian that shares characteristics of both frogs and salamanders — in life, it would have looked like a short-tailed, wide-headed salamander with frog-like ears, which is why it's being called a "frogamander".

i-3085796f01e5d09425d897b5cd6092e2-gerobatrachus.jpg
Complete specimen in ventral view, photograph (left) and interpretive outline drawing (right). Abbreviations: bc, basale commune; cl, cleithrum; cv, clavicle; dm, digital elements of the manus; dt3, distal tarsal 3; fe, femur; h, humerus; ic, intercentrum; il, ilium; is, ischium; op, olecranon process of ulna; pc, pleurocentrum; r, radius; sr, sacral rib.

For those of you with a more technical bent, here's the list of diagnostic characteristics that bridge both amphibian groups.

Amphibamid temnospondyl with 21 tiny pedicellate teeth on the premaxilla, and 17 presacral vertebrae; shares with crown group salamanders a basale commune (combined distal tarsals 1 and 2) and tuberculum interglenoideum ('odontoid process') on atlas; shares with salientians and caudates an anteroposteriorly reduced vomer; shares with Triadobatrachus and crown group frogs a rod-like, laterally directed palatine; shares with Karaurus, Triadobatrachus and crown group frogs a broad skull, shortened presacral vertebral column; shares with most temnospondyls, frogs and basal salamanders a pedal phalangeal formula of ?-2-3-4-3; shares with frogs, Amphibamus, Doleserpeton, Platyrhinops and Eoscopus a large otic notch closely approaching the orbit; shares with frogs, salamanders, caecilians, Amphibamus, Tersomius and Doleserpeton pedicellate teeth; shares with Amphibamus, Doleserpeton and Platyrhinops a foreshortened supratemporal; shares with Amphibamus, Doleserpeton, frogs and salamanders a foreshortened parasphenoid basal plate with wide lateral processes.

Now maybe some of you paleontologically-minded people can help me out with something. Molecular evidence places the divergence point between frogs and salamanders at some point in the late Carboniferous, between 308 and 357 million years ago. This animal is from the Early Permian, so it's more recent than that. Strangely, the authors claim that the discovery of Gerobatrachus places a lower bound on the divergence time, which I don't see at all; Gerobatrachus could be a late representative of a significantly earlier common ancestor. Am I missing something here?


Anderson JS, Reisz RR, Scott D, Fröbisch NB, Sumida SS (2008) A stem batrachian from the Early Permian of Texas and the origin of frogs and salamanders. Nature 453(7194):515-518.

.

More like this

a, b, Cranial reconstruction in left lateral (a; shaded area indicates the unpreserved portion) and dorsal (b) views. adc, anterodorsal concavity; al, anterior lamina; an, angular; aof, antorbital fenestra; d, dentary; dg, dentary groove; emf, external mandibular fenestra; en, external naris; if,…
Now this is an interesting beast. It's a 220 million year old fossil from China of an animal that is distinctly turtle-like. Here's a look at its dorsal side: a, Skeleton in dorsal view. b, Skull in dorsal view. c, Skull in ventral view. d, Body in dorsal view. Teeth on the upper jaw and palatal…
A lovely new dinosaur fossil from China is described in Nature today: it's named Xiaotingia zhengi, and it was a small chicken-sized, feathered, Archaeopteryx-like beast that lived about 155 million years ago. It shares some features with Archaeopteryx, and also with some other feathered dinosaurs…
First of all, here are some temnospondyls. This composite image was compiled by repositioning the reconstructions provided by DeFauw (1989): looks neat, doesn't it? So, I recently returned from the 69th Society of Vertebrate Paleontology meeting, this year held in Bristol, UK: in other words,…

PZ says
Molecular evidence places the divergence point between frogs and salamanders at some point in the late Carboniferous, between 308 and 357 million years ago..... the authors claim that the discovery of Gerobatrachus places a lower bound on the divergence time, which I don't see at all...

I agree. They could have found this frogamander dating from 100MYA and it still wouldn't necessarily change the estimated divergance point. It would simply mean that a member of this branch, having a common ancester with both salamanders and frogs, had croaked and gotten itself all fossilized. While frogs and salamanders went their separate ways, this branch may have retained characteristics from the common ancester.

J

This is obviously another example of "Homo diluvii testis", one of the witnesses of the Flood. Notice how his hands are protecting his "witnesses". A fine example. Thanks for posting this PZ.

By Stephanurus (not verified) on 25 May 2008 #permalink

More evidence for biological evolution! Any sane person regards biological evolution as a fact, given that there's so much evidence for it. But the religious nutjobs won't see that, will they? Feckin' edjits!

By Richard Harris (not verified) on 25 May 2008 #permalink

It's obviously just a fox with mange.

By ellindsey (not verified) on 25 May 2008 #permalink

Hi all!
first of all, Greetings to PZ! I always like the posts related to paleontology! Give us more such stuff!
Second: dear all, could you link me some fancy webpages in paleo topic (specially about "transitional" forms), i'm in debate with some local IDiots, i need some ammo :)
To avoid flooding please send the links here: bs@c2.hu
Thanks in advance!

what a beautiful wee beastie

Aww - coulda been a salliwog...

Clearly a forerunner of the Deep Ones. Very pretty.

Am I missing something here?

Yes...molecular "evidence" is worthless.

"I wished I had you in carrickfergus,
Only for nights in ballygrand,
I would swim over the deepest ocean,
The deepest ocean to be by your side.

But the sea is wide and I can't swim over
And neither have I wings to fly.
I wish I could find me a handy boatman
To ferry me over to my love and die.

My childhood days bring back sad reflections
Of happy days so long ago.
My boyhood friends and my own relations.
Have all passed on like the melting snow.

So I'll spend my days in endless roving,
Soft is the grass and my bed is free.
Oh to be home now in carrickfergus,
On the long road down to the salty sea.

And in kilkenny it is reported
On marble stone there as black as ink,
With gold and silver I did support her
But I'll sing no more now till I get a drink.

I'm drunk today and I'm rarely sober,
A handsome rover from town to town.
Oh but I am sick now and my days are numbered
Come all ye young men and lay me down."

As a member of Temnospondyli, it would be a separate order from Lissamphibia, which gave rise to anurans and caudates. I don't think this fossil has much to do with the split between frogs and salamanders then, as the split between Temnospondyli and Lissamphibia occured before that.

I have only seen the Science Daily account. If correctly quoted, Robert Reise is simply wrong in saying that finding any fossil tells you when that group originated. If you have good dating, you know that the fossil organism was around at that time, but that is all you know. You do not know when its speciation event occurred; if or when it gave rise to other lineages, or whether it became extinct without issue.

I don't know enough about the proposed classification, but I suspect I might not agree with it. If it were alive today, it likely would be considered the sister group of frogs + salamanders, and probably included in the same order.

By Jim Thomerson (not verified) on 25 May 2008 #permalink

But but but... It's kind hasn't changed! It's still a frog-salamander-fish-looking thing. So clearly it's not a transitional fossil!

By BoxerShorts (not verified) on 25 May 2008 #permalink

May I be the one to ask the Dumb Question?

I know that when a baby frog is growing up, there is a time when it has not yet lost its tadpole tail. How do we know this isn't the case with this frogamander? I'm sure there is a way to know this, I just don't know what it is.

Noni

By Noni Mausa (not verified) on 25 May 2008 #permalink

You know, just because creationists have faulted us for not having these transitionals, doesn't prove that these transitional forms imply evolution at all.

See, if they didn't exist, that would mean that evolution did not occur.

If they do exist, that means nothing about evolution happening.

You just have to think like the godly (apologies to the godly who do not think like this--I'm not targeting you), you know, inconsistently and one-sidedly. If you don't think like they do, it means you're prejudiced.

Glen Davidson
http://tinyurl.com/2kxyc7

This animal is from the Early Permian, so it's more recent than that. Strangely, the authors claim that the discovery of Gerobatrachus places a lower bound on the divergence time, which I don't see at all; Gerobatrachus could be a late representative of a significantly earlier common ancestor. Am I missing something here?

Unless they mean something that I don't see here, yes, I don't see what they mean by that. Here's another account, which suggests that this little beastie is a descendent of the more truly transitional type:

Carroll noted that some of Gerobatrachus's features crop up in older fossils, suggesting that the animal may be descended from an ancestral group that branched out into frogs and salamanders at an earlier point.

"I would say Gerobatrachus was a relic at least in relationship to the question of the ancestry and point of division between frogs and salamanders," Carroll said.

Bolt, the Field Museum expert, cautioned that it is difficult to say for sure whether this creature was itself a common ancestor of the two modern groups, given that there is only one known specimen of Gerobatrachus, and an incomplete one at that.

"At this point I would say it is by no means certain that this is representative of a common ancestor to frogs and salamanders, although it might be," Bolt said.

"The fact that something this remarkably like the modern amphibians is present [during this period] suggests that, really, you could have had an even earlier example of something that was ancestral to modern frogs and salamanders."

news.nationalgeographic.com/news/2008/05/080521-frog-fossil_2.html

Glen D
http://tinyurl.com/2kxyc7

Evolution could be completely seamless, every box checked, all the t's crossed, all the i's dotted, and the creationists would simply move the goalposts again, claiming that transitional forms were created as transitional forms.
Or, the ultimate YEC argument -- that the transitional form fossil was created as a fossil, in order to fool people.
In fact, this is the ultimate end of all creationist arguments -- "God made it that way". How . . . simpleminded.

Off topic, but not entirely irrelevant: It's Towel Day, the annual commemoration of humorist Douglas Adams, author of The Hitchhiker's Guide to the Galaxy, radical atheist (his own self-characterization), and acolyte of Richard Dawkins (says Dawkins: "my only convert").

Hug a towel in his honor today. Or dry yourself with one. It's all good.

I know that when a baby frog is growing up, there is a time when it has not yet lost its tadpole tail. How do we know this isn't the case with this frogamander? I'm sure there is a way to know this, I just don't know what it is.

I don't know that it is fully grown. But it doesn't look like a frog that hasn't lost its tail, which indeed looks like a frog with a tail.

The point is not that it has a tail, but that it has specific features that frogs now have that modern salamanders do not have, and it has features that salamanders do have and modern frogs do not have. To put it another way, a salamander tadpole is significantly different from a frog tadpole. Plus, this is obviously not a tadpole.

The differences are listed in the blog, but most of us aren't going to understand most of the terms used. Suffice it to say that specific skeletal features mark salamanders off from frogs and vice-versa, and this fossil combines skeletal features found in both animals.

Glen D
http://tinyurl.com/2kxyc7

Noni, you wanted to know how we recognize this fossil as a distinct species instead of a juvenile amphibian. First, the size of the creature matches up with what we'd suppose the adult forms would look like. Further, the comparative anatomy indicates that this species has traits it shares with both frogs and salamanders. A tadpole for example, would not have certain structures that would be found in amphibians. If you read through the diagnostic characteristics, you'll notice the comparison of this creatures anatomy to those of other amphibians. For example "shares with crown group salamanders a basale commune (combined distal tarsals 1 and 2) and tuberculum interglenoideum ('odontoid process') on atlas". As you can clearly see here, a tadpole would not share these characteristics with a salamander. Conversely, "shares with frogs, Amphibamus, Doleserpeton, Platyrhinops and Eoscopus a large otic notch closely approaching the orbit". In this case, a juvenile salamander would not have these traits, or at the very least, the precursor to these traits. Also, as we have expanded fossil evidence, we have found those of juvenile amphibians and we could further compare these together. I hope this helps, and if it sounds even more confusing, that's probably why I'm not an instructor in the subject.

By Helioprogenus (not verified) on 25 May 2008 #permalink

TWO MORE GAPS!

By arachnophilia (not verified) on 25 May 2008 #permalink

Hmm. I cringe a little when I hear words like frogamander. They convey the idea that transitional species should be the merging of colloquial 'kinds' and lead to creationist canards that demand transitional 'ratbats', 'catdogs', and 'creation scientists' when that's just not how evolution works.

However, I suppose the terms probably are useful for laypersons and journalists who generally 'get it' and have some understanding of the history of life but are pretty fuzzy on the details and won't know when the Carboniferous or the Permian were.

Now maybe some of you paleontologically-minded people can help me out with something. Molecular evidence places the divergence point between frogs and salamanders at some point in the late Carboniferous, between 308 and 357 million years ago. This animal is from the Early Permian, so it's more recent than that. Strangely, the authors claim that the discovery of Gerobatrachus places a lower bound on the divergence time, which I don't see at all; Gerobatrachus could be a late representative of a significantly earlier common ancestor. Am I missing something here?

1) "Lower" means "younger", as in "lower numbers".
2) Stop worrying. The authors overlooked this here. All older molecular divergence dates are too old because they are miscalibrated.

Notice how his hands are protecting his "witnesses".

These are his feet protecting the tip of his tail...

As a member of Temnospondyli, it would be a separate order from Lissamphibia

The very point of the paper is that there is no Lissamphibia: frogs and salamanders are temnospondyls, and caecilians are lepospondyls. Don't be misled into thinking that "order" actually means anything. :-)

(I personally reserve some doubt about this, however. A larger phylogenetic analysis is necessary, and a longer description wouldn't hurt either -- no surprise there, Nature is an extended-abstract publication, and the beastie is worthy of Nature.)

If correctly quoted, Robert Reis[z] is simply wrong in saying that finding any fossil tells you when that group originated.

It's not that simple. When you have a good fossil record, absence of evidence is evidence of absence. One reason why Precambrian rabbits would be such a surprise is that no lagomorphs older than Paleogene are known at all.

I don't know enough about the proposed classification, but I suspect I might not agree with it.

There is no classification. There's only a phylogenetic hypothesis, which says that Gerobatrachus is the sister-group of frogs + salamanders together.

I know that when a baby frog is growing up, there is a time when it has not yet lost its tadpole tail. How do we know this isn't the case with this frogamander?

Tadpole tail vertebrae never ossify. Or even chondrify (become cartilage) as far as I know. There's just the notochord, and that's it.

That said, the paper makes clear that Gerobatrachus is not fully grown, although it is postmetamorphic.

Here's another account, which suggests that this little beastie is a descendent of the more truly transitional type:

This is because Carroll feels (...and I fear "feels" is the right word here...) that frogs and salamanders are not sister-groups, and that instead salamanders are branchiosaurid temnospondyls, while frogs are amphibamid temnospondyls like Gerobatrachus. This requires an earlier split between frogs and salamanders.

By David Marjanović, OM (not verified) on 25 May 2008 #permalink

PZ: First, molecular clocks are only occasionally accurate, and are always calibrated using fossils. If the fossil says one thing and the molecular clock another, chances are strongly in favour of the fossil being correct.

As for using this specimen to establish divergence timing, I usually expect palaeontologists to use fossils to establish earliest divergence times, using as examples fossils that are the earliest known representations of given groups. In this case, the assumption is that the lineages did not diverge until after this point, the lemma here being that Gerobatrachus is ancestral to all extant frogs and salamanders. It is odd that the possibility of this being a relic (a Palaeozoic "living fossil", if you will) existing alongside already-diverged frog and salamander lineages, has been discounted. I imagine that these researchers will have had to defend such an assertion, but I cannot infer what that would be, myself.

#10: Arguing that frigs and salamanders are members of the order Lissamphibia, and not of the order Temnospoldyli, misses the point. What this fossil suggests is that frogs and salamanders are temnospondyls. The rank given to a given taxon (Lissamphibia, Temnospondyli, etc.) is ultimately meaningless. Meanwhile, if caecilians are in fact related more closely to lepospondyls, the group that we call Lissamphibia becomes polyphyletic -- with (in this case) two different unrelated ancestors, and therefore a biologically useless concept.

First, the size of the creature matches up with what we'd suppose the adult forms would look like.

This argument was not used -- because it's the worst one ever. Adult size is... relatively volatile.

Hmm. I cringe a little when I hear words like frogamander. They convey the idea that transitional species should be the merging of colloquial 'kinds'

The funny thing, though, is that, in some features, that's precisely what Gerobatrachus is. The basale commune and the interglenoid tubercle are only shared with salamanders, not with frogs. Yet, Gerobatrachus is (the paper finds) the sister-group of frogs and salamanders together, not of salamanders alone. We are thus looking either at convergence or at a secondary loss in frogs. Conversely, the shape of the palatine is only shared with frogs, not with salamanders. We are thus looking either at convergence or at a secondary loss in salamanders.

(Personally, I also find the basale commune a bit small and misplaced compared to the situation in salamanders, but I haven't seen enough salamanders yet, so I could be wrong.)

transitional 'ratbats', 'catdogs', and 'creation scientists'

:-D :-D :-D

By David Marjanović, OM (not verified) on 25 May 2008 #permalink

Actually... it's true that Gerobatrachus does not establish a younger bound on the confidence interval for the split between frogs and salamanders either. It establishes such a bound on the confidence interval for the split between frogs + salamanders on the one hand and their sister-group (Gerobatrachus itself) on the other.

By David Marjanović, OM (not verified) on 25 May 2008 #permalink

Well, the issue here is that Gerobatrachus, much like a number of other amphibamids, shows a number of derived "froglike" characters. We see this in the forward-directed ilium, the pedicellate teeth, the exaggerated otic notch, and a number of other features, described in Eoscopus, Georgenthalia, Plemmyradytes, Doleserpeton, etc. If we were looking at Gerobatrachus in a vacuum, we'd have more of a problem, but the temnospondyl origins of Caudates + Anurans was not in question prior to the discovery of Gerobatrachus. Carroll has supported it, Milner has supported it, Anderson has supported it, and a number of other workers have supported it. The question is a matter of whether lissamphibians form a monophyletic clade within Amphibamus-like amphibamids, or whether anurans specifically are derived from amphibamids, and that caudates (salamanders) are derived from a very closely related group, such as branchiosaurs or micromelerpetontids. Caecillians are a mess; there's one putative fossil caecillians from the Triassic, and the lack of caecillian fossils is expected, given the poor taphonomic environment where caecillians generally live (rainforest leaf litter). Carroll has suggested that they may be derived from the lepospondyl Rhynchonkos, but that's really pending more data; there are a variety of microsaur taxa (lysorophians, for instance) which may offer a potential solution to the caecillian problem.

There's a third problem with temnospondyls, and that's paedomorphosis, but I'm not going to get into that right now.

shows a number of derived "froglike" characters.

Yes, but also a number of derived salamanderlike characters -- and these are not always the same.

the temnospondyl origins of Caudates + Anurans was not in question prior to the discovery of Gerobatrachus. Carroll has supported it, Milner has supported it, Anderson has supported it, and a number of other workers have supported it.

Laurin 1994 (unpublished thesis), Laurin & Reisz 1997, Laurin 1998a, b, Laurin & Reisz 1999, Vallin & Laurin 2004... I'm not sure if I've got all the refs right from memory, but their number is right. A thesis will be defended next Monday that has, though tentatively, come to the same conclusion.

Besides, Carroll's only cladistic analysis of this question ever (2007: fig. 77 -- not fig. 78!) does not support lissamphibian polyphyly. It does support lissamphibians being temnospondyls -- but it also supports lepospondyls being temnospondyls... have you seen the character list for the data matrix?

Milner has never done a cladistic analysis of this topic. His 1988 and 1993 papers are just trees that come out of nowhere, plus lists of apomorphies that support each internode -- but he made no attempt to test if the tree was the most parsimonious explanation for his data!

a monophyletic clade

This message brought to you by the Department of Redundancy Department.

Caecillians are a mess;

The caecilians (one L) are a mess, but so are frogs and salamanders. And Doleserpeton. And pretty much everything else... :-}

there's one putative fossil caecillians from the Triassic,

I can't see what there is putative about Eocaecilia. Have you read its redescription (Jenkins, Walsh & Carroll 2007)?

Also, you forgot Rubricacaecilia, the Early Cretaceous sister-group of the caecilian crown-group (Evans & Sigogneau-Russell 2001, in Palaeontology). While fragmentary, it is an interesting animal. For example, it has an interglenoid tubercle, like Eocaecilia and the salamanders (...and the lepospondyls).

there are a variety of microsaur taxa (lysorophians, for instance) which may offer a potential solution to the caecillian problem.

Not just the caecilian problem, but the entire lissamphibian problem, possibly. See all the Laurin refs above. But then, the lysorophians are paedomorphic, and we have no idea of their ossification sequences, and so on...

BTW, they have never been considered microsaurs, though they may well be nested within them, as Laurin and coauthors have consistently found.

There's a third problem with temnospondyls, and that's paedomorphosis, but I'm not going to get into that right now.

Then let me start. Only the biggest, most mature specimen of Amphibamus has vomerine fangs. Oh, how lissamphibianlike of Doleserpeton and Gerobatrachus to also lack vomerine fangs... well, G. is immature, and D. probably also is (Bolt mentioned this in at least some of his papers on the beast, but it doesn't help that no decent-sized description of the amazingly abundant material has ever been published). Both should probably be scored as "unknown" rather than as "lacking vomerine fangs". Off the top of my head, the exact same holds for the palatine fangs, though it's almost 3 at night, so I won't look that up before going to bed...

By David Marjanović, OM (not verified) on 25 May 2008 #permalink

I suggest that comment threads of any future Pharyngula posts on fossil tetrapods be given over entirely to Marjanović.

By a Marjanović fan (not verified) on 25 May 2008 #permalink

Wow, methinks David M. needs a cigarette and a box of tissues in Paris. I'm sure the neighbors heard the moans.

Many years ago, when I got my MS in paleontology, lower meant older and upper meant younger. Upper Permian was younger than Lower Permian. Principle of Superposition, you know. Been out of the field a long time. Has this useage changed?

By Jim Thomerson (not verified) on 25 May 2008 #permalink

Yes, but also a number of derived salamanderlike characters -- and these are not always the same.

Right. But then, the question depends partly on where branchiosaurs fall out, and partly on some material that really could use a significant redescription. Schoch's recent description of supposed trophic morphs in Micropholis doesn't really help things too much; we really have to ask whether those are in fact one species, and if so, what the hell that means for the various amphibamids from C/P Euramerica. Branchiosaurs and Micromelerpetontids, too; there certainly are a lot of juvenile specimens out there, and the phylogeny might be reflecting that. It might also be reflecting degree of terrestriality, or something else entirely.

Laurin 1994 (unpublished thesis), Laurin & Reisz 1997, Laurin 1998a, b, Laurin & Reisz 1999, Vallin & Laurin 2004... I'm not sure if I've got all the refs right from memory, but their number is right. A thesis will be defended next Monday that has, though tentatively, come to the same conclusion.

And a lot of that conclusion is based on the holospondylous vertebrae, a feature which arose a number of times in a variety of Paleozoic vertebrate lineages. We certainly lack the empirical data to make a judgment call there, but at the same time, Laurin's arguments have become progressively weaker.

Besides, Carroll's only cladistic analysis of this question ever (2007: fig. 77 -- not fig. 78!) does not support lissamphibian polyphyly. It does support lissamphibians being temnospondyls -- but it also supports lepospondyls being temnospondyls... have you seen the character list for the data matrix?

Yes, I've seen that. I interpreted the entire mess as being evidence of insufficient data and an over-reliance on highly homoplasious characters. This is sort of my impression on most of the larger-scale tetrapod phylogenies. I'm really unconvinced that minimizing the treelength for post-parietal width vs. length is a meaningful endeavor. I could be completely wrong, here, though, but I suspect that you can't reject any of the debated trees with that dataset and statistical methods.

Milner has never done a cladistic analysis of this topic. His 1988 and 1993 papers are just trees that come out of nowhere, plus lists of apomorphies that support each internode -- but he made no attempt to test if the tree was the most parsimonious explanation for his data!

No offense, but if you think that parsimony is even a reasonable criterion when we have such a preponderance of homoplasy, then there may be bigger problems here. One of the biggest problems that paleo-based phylogenetics suffers from is the abuse of parsimony-based analyses on data which cannot reasonably be assumed to follow a minimum evolution model. There is no attempt to compare trees in a statistical framework of a Templeton or Kishino-Hasegawa test, and these are really easy to do with a couple of constraint trees and a few minutes in PAUP. I don't disagree that "just because I said so" is a poor justification for a tree, but I fail to see how most of the phylogenetic analyses conducted in vert paleo today differ significantly from "just because I said so."

This message brought to you by the Department of Redundancy Department.

I'll call bullshit there. We recover a hell of a lot of clades in the average parsimony analysis. At the same time, certainty should not be judged on the basis of bootstrap alone. The entire point of a phylogenetic analysis is to test how well we can infer the monophyly of a clade, not simply plug a nexus file into PAUP and hopefully get a tree.

The caecilians (one L) are a mess, but so are frogs and salamanders. And Doleserpeton. And pretty much everything else... :-}

Well, the caecilians are much more susceptible to phylogenetic mishap due to the presence of a long branch between the most basal caecilian and any putative caecilian outgroup. 70-80 Ma of unpreserved evolution, in a clade (either lepospondyls, temnospondyls, or lepospondyls+temnospondyls) which is highly-represented between the Bashkirian and Kungurian of Euramerica, and then essentially unrepresented after the Roadian Gap, either due to changes in the preserved faunal province or changes in faunal composition, is an optimal circumstance for a phylogenetic analysis to fail, especially if we're looking at features which may in fact be highly homoplasious.

As far as Doleserpeton goes, there are several amphibamids at Richard's Spur. There's Doleserpeton, there's a "Tersomius", and there's something else, also in that side-branch that includes Tersomius. I haven't a damned clue whether these disparate specimens represent different species or different ontogenetic stages, though.

I can't see what there is putative about Eocaecilia. Have you read its redescription (Jenkins, Walsh & Carroll 2007)?

Also, you forgot Rubricacaecilia, the Early Cretaceous sister-group of the caecilian crown-group (Evans & Sigogneau-Russell 2001, in Palaeontology). While fragmentary, it is an interesting animal. For example, it has an interglenoid tubercle, like Eocaecilia and the salamanders (...and the lepospondyls).

Somehow missed those. I've been primarily finishing up some lungfish work, so I'm only dabbling in tetrapods right now. Regardless, my point is that there is not a rich gymnophionan record in the Permian and Triassic.

Not just the caecilian problem, but the entire lissamphibian problem, possibly. See all the Laurin refs above. But then, the lysorophians are paedomorphic, and we have no idea of their ossification sequences, and so on...

BTW, they have never been considered microsaurs, though they may well be nested within them, as Laurin and coauthors have consistently found.

If I remember correctly, Laurin is not the only one who found lysorophians within the Microsauria, but then, I could be misremembering. I've looked at too many lysorophians in the last four years, and I've learned two things: Westoll's attempt to make every lysorophian except Pleuroptyx a member fo the genus Brachydectes was a bit on the hasty side, and, oh yes, I absolutely hate lysorophians.

Then let me start. Only the biggest, most mature specimen of Amphibamus has vomerine fangs. Oh, how lissamphibianlike of Doleserpeton and Gerobatrachus to also lack vomerine fangs... well, G. is immature, and D. probably also is (Bolt mentioned this in at least some of his papers on the beast, but it doesn't help that no decent-sized description of the amazingly abundant material has ever been published). Both should probably be scored as "unknown" rather than as "lacking vomerine fangs". Off the top of my head, the exact same holds for the palatine fangs, though it's almost 3 at night, so I won't look that up before going to bed..

Well, we did find vomerine fangs on Plemmyradytes, even though the material was pretty beat up, and I'm under the impression that vomerine fangs are pretty ubiquitous among the Tersomius-like amphibamids, although the distinction between these two clades (Tersomius-like and Amphibamus-like amphibamids) might actually be an ontogenetic rather than phylogenetic signal.

But this is a larger-scale problem than that. Placement of the dvinosaurs, the monophyly of many major euskelian clades, and placement of the stereospondyls all suffer from potential ontogenetic bias, either due to paedomorphosis or due to the number of juvenile specimens within the analysis. Once again, we see optimal circumstances to recover a significantly false tree with astounding certainty.

Many years ago, when I got my MS in paleontology, lower meant older and upper meant younger. Upper Permian was younger than Lower Permian. Principle of Superposition, you know. Been out of the field a long time. Has this useage changed?

Not to my knowledge. I'm sure it was a mistake. Lower refers to earlier strata, Upper refers to later strata. The quibble about strata is important, if you're talking about time, you simply use "Early" and "Late."

I think I figured out the answer to my own question:

A frogamander is a different kind of entity that matter and energy cannot produce.

I know far more than your average joe about vertebrate evolution, but I admit, Lissamphibia ain't an area I know too much about.

Reading this thread makes me realize what I must sound like to so many, as I am mostly clueless here. It all looks like, blah blah blah, frog, blah blah blah temnospondyli...

It's a lovely fossil... but there are still plants around from before the split of monocotyledons and dicotyledons, even thought the split happened way back when.

What it does say, I think, is that frogs and salamanders had a common ancestor closer than their common ancestors with other groups. If that was in doubt, it's a strong piece of evidence.

And it is a data point. Fossilization of delicate creatures is rare, but probability suggests these are what was common then. Perhaps the molecular evidence needs to be re-calibrated.

Can we get back to Don's Dump Quarry and dig for a few more specimens? Maybe something else will turn up from the same ecosystem.

/reads comments above

Me feel stupid. Me want pet Gerobachrach... Gerobatch... me want pet frogamander. Me no like so few amphibians in Earth.

arachnophilia, "TWO MORE GAPS!" No, it's only one more gap, there already having been one wider gap, into which this critter's been inserted.

What I'd like to know is, when did this frogamander croak?

By Richard Harris (not verified) on 25 May 2008 #permalink

"The frogamander is in no sense a transitional form. It is one more demonstration of my PEH, an animal which combines the features of two "prescribed" morphologies."

But there are no forms that combine features in the way you imply. When you see a Tiktaalik, or an Archaeopteryx, you see what evolution predicts. When we see M. gui, we know certain traits should be found, and other traits never are to be found.

"Is a shark, with a tue placenta, a transitional form beteen sharks and placental mammals?"

This is a false usage of language to make a point. The blood vessels involved are common to both, and we simply call any fetal exchange mechanism a "placenta" when similar connections occur. They need not be exact, or show unique similarities. They only need to fall under accepted definitions given to the term, which is defined by people,.

Simply put, we could have called it a, "Blingy blu babble"...it will never been the same structure seen in placental mammals. It only utilizes common structures that are basal to both in ways that increase surface area so exchange of materials can occur.

just because creationists have faulted us for not having these transitionals, doesn't prove that these transitional forms imply evolution at all.

See, if they didn't exist, that would mean that evolution did not occur.

The issue of "transitional fossils is a red herring.
Clearly, evolution has occurred. What is not clear is the mechanism. Is there any evidence that random, undirected processes can lead to the emergence of highly organized structures, processes an systems? I think not.

Many years ago, when I got my MS in paleontology, lower meant older and upper meant younger. Upper Permian was younger than Lower Permian. Principle of Superposition, you know. Been out of the field a long time. Has this useage changed?

No, it's just that the lower bound of a confidence interval, in mathematics, is the one with the lower numbers. It's confusing either way around: either you confuse the paleontologists/geologists, or you confuse everyone else...

----------------

Right. But then, the question depends partly on where branchiosaurs fall out, and partly on some material that really could use a significant redescription. Schoch's recent description of supposed trophic morphs in Micropholis doesn't really help things too much; we really have to ask whether those are in fact one species, and if so, what the hell that means for the various amphibamids from C/P Euramerica. Branchiosaurs and Micromelerpetontids, too; there certainly are a lot of juvenile specimens out there, and the phylogeny might be reflecting that. It might also be reflecting degree of terrestriality, or something else entirely.

I agree completely.

Papers on Apateon, Branchiosaurus and Micromelerpeton are in the works (see SVP meeting abstracts of 2007).

And a lot of that conclusion is based on the holospondylous vertebrae

This is not true. The vertebrae are just one character, out of 161 (in Vallin & Laurin 2004). Furthermore, Eocaecilia retains intercentra, giving it vertebrae pretty similar to those of Doleserpeton (and, of course, a number of "microsaurs", but not the holospondylous lysorophians, nor even the holospondylous brachystelechids).

I suspect that you can't reject any of the debated trees with that dataset and statistical methods.

All datasets published so far are too small (in both the number of taxa and the number of characters), and at least some of them are full of mistakes (pers. obs. and the mentioned thesis that will be defended next week).

So, I think that all three hypotheses -- lissamphibians as temnospondyls, lissamphibians as lepospondyls, polyphyletic "lissamphibians" as both -- remain in the race for now.

I'm really unconvinced that minimizing the treelength for post-parietal width vs. length is a meaningful endeavor.

On its own it isn't. Phylogenetics with a single character is not possible. Total-evidence approach.

No offense, but if you think that parsimony is even a reasonable criterion when we have such a preponderance of homoplasy, then there may be bigger problems here.

What else? Maximum munificence?

No, really. What other method is there?

Yes, Bayesian analysis. That should be tried. But it's not tremendously different; it's just parsimony with extra sophistications and a method to get support values without having to perform extra analyses.

I like reweighting, BTW. Ruta & Coates (2007) tried it.

One of the biggest problems that paleo-based phylogenetics suffers from is the abuse of parsimony-based analyses on data which cannot reasonably be assumed to follow a minimum evolution model.

Isn't minimum evolution a phenetic algorithm like neighbor-joining? That would explain why it hasn't been used for at least 10 years!

Parsimony is not the assumption of the minimization of convergence. It is the minimization of the assumption of convergence. I forgot who said that first, probably Paul Sereno.

And indeed, large parsimony analyses find very large amounts of convergence/reversals. A tree without convergence would have a consistency index of 1 -- the result of the so far largest analysis of tetrapod phylogeny (102 taxa, 333 parsimony-informative characters), Ruta & Coates (2007), has a CI of 0.22.

The entire point of a phylogenetic analysis is to test how well we can infer the monophyly of a clade

Sorry for the misunderstanding. I was only talking about terminology here: a clade is by definition monophyletic. "Clade" and "monophylum" are synonyms. The definition of "clade" is "an ancestor and all its descendants".

Regarding your actual point, what's wrong with bootstrapping?

Well, the caecilians are much more susceptible to phylogenetic mishap due to the presence of a long branch between the most basal caecilian and any putative caecilian outgroup. 70-80 Ma of unpreserved evolution [...]

Yes, long-branch attraction is a possibility here that must not be ignored. And indeed, many of the similarities between crown caecilians and other lissamphibians are most likely convergent -- Eocaecilia retains postparietals, jugals (like the albanerpetontids), tabulars or supratemporals...

References:

Farish A. Jenkins, Jr., Denis M. Walsh & Robert L. Carroll (2007): Anatomy of Eocaecilia micropodia, a limbed caecilian of the Early Jurassic. Bulletin of the Museum of Comparative Zology [of Harvard] 158(6): 285 -- 366.

That's the full-length description of Eocaecilia. No pdf seems to exist, and the journal is hard to get at least outside the USA. Ask the authors for a reprint -- they have probably distributed more reprints than the journal has sold copies.

Susan E. Evans & Denise Sigogneau-Russell (2001): A stem-group caeciian (Lissamphibia: Gymnophiona) from the Lower Cretaceous of North Africa. Palaeontology 44(2): 259 -- 273.

The description of Rubricacaecilia.

I've looked at too many lysorophians in the last four years, and I've learned two things: Westoll's attempt to make every lysorophian except Pleuroptyx a member fo the genus Brachydectes was a bit on the hasty side, and, oh yes, I absolutely hate lysorophians.

Cool. Is there a paper in the works?

I'm under the impression that vomerine fangs are pretty ubiquitous among the Tersomius-like amphibamids

Sure -- among adults, if we can generalize from Amphibamus and Eoscopus (Daly 1994).

But this is a larger-scale problem than that. Placement of the dvinosaurs, the monophyly of many major euskelian clades, and placement of the stereospondyls all suffer from potential ontogenetic bias, either due to paedomorphosis or due to the number of juvenile specimens within the analysis. Once again, we see optimal circumstances to recover a significantly false tree with astounding certainty.

I agree. The value of the following paper cannot be underestimated.

John J. Wiens, Ronald M. Bonett & Paul T. Chippindale (2005): Ontogeny discombobulates phylogeny: paedomorphosis and higher-level salamander relationships. Systematic Biology 54(1): 91 -- 110.

---------------------

Reading this thread makes me realize what I must sound like to so many, as I am mostly clueless here. It all looks like, blah blah blah, frog, blah blah blah temnospondyli...

Yeah, I've never had such an insider discussion on Pharyngula before... :-) (And I still haven't even figured out who JDP is! Some insider am I! -- Oh, wait! Jason D. Pardo, the third author of Plemmyradytes! Nice to meet you!)

---------------------

What I'd like to know is, when did this frogamander croak?

About 275 to 270 million years ago apparently.

By David Marjanović, OM (not verified) on 26 May 2008 #permalink

I wrote:

So, I think that all three hypotheses -- lissamphibians as temnospondyls, lissamphibians as lepospondyls, polyphyletic "lissamphibians" as both -- remain in the race for now.

I should have mentioned, though, that every single molecular analysis I'm aware of has found a lissamphibian clade that excludes the amniotes. Because the lepospondyls are more closely related to the amniotes than to the temnospondyls (every analysis since Carroll 1995* has found this), this contradicts the polyphyly hypothesis. Perhaps that's just an especially pernicious case of long-branch attraction... but I doubt it :-)

So, the polyphyly hypothesis remains in the race, but IMHO it has considerably fallen behind. We'll see if that changes.

* His first and last cladistic analysis till 2007. Sarcastic people say he tried cladistics, didn't like the results, and turned away...

By David Marjanović, OM (not verified) on 26 May 2008 #permalink

A side question: isn't a thesis always* published if a student finishes the degree, at least to the university's library?

*(Certain creationists excepted, who appear to be still updating their theses)

It is deposited in the university library, yes. But that's it normally. This does not count as publication. Publication usually requires some drastic reworking (often splitting into several papers just for space reasons) and often involves updating. Unpublished theses are usually difficult to get and are, usually, for most purposes, treated as if they didn't exist. They can be cited, though.

By David Marjanović, OM (not verified) on 26 May 2008 #permalink

Yes, the frogamander is great, but the fishapod Tiktaalik still rules! Your Inner Fish was recently chosen for the University of Pennsylvania's Penn Reading Project. All incoming freshman are being mailed a copy, and are expected to take part in discussions about it during Orientation.

Neil Shubin was formerly a professor at Penn, and will be speaking to the freshman. All this is part of Penn's sponsorship of the Year of Evolution.

By william e emba (not verified) on 26 May 2008 #permalink

every single molecular analysis I'm aware of has found a lissamphibian clade that excludes the amniotes

phew...as a neontological herpetologist who has made the claim of probable lissamphibian monophyly to classes for many years, this helps me breathe a bit easier.

By Sven DiMilo (not verified) on 26 May 2008 #permalink

jeez, old-troll city. JAD his bad(-smelling) self, and I think "Bishop Pontoppodan" smells like Charlie Wagner.

By Sven DiMilo (not verified) on 26 May 2008 #permalink

Clearly, evolution has occurred. What is not clear is the mechanism. Is there any evidence that random, undirected processes can lead to the emergence of highly organized structures, processes an systems? I think not.

That's definitely a Charlie Wagnerian sort of thing to say.

It's kind of depressing that in all these years, he's never paid attention to the basic science of biology, or opened a basic biology book, or read some of the excellent basic online information.

Say, who is "Bishop Pontoppodan", anyway? Google suggest that it should be "Bishop (Erik) Pontoppidan" (note spelling), of Denmark, who described the Kraken in 1752.

Oh, very funny Charlie.

By Owlmirror (not verified) on 26 May 2008 #permalink

This is not true. The vertebrae are just one character, out of 161 (in Vallin & Laurin 2004). Furthermore, Eocaecilia retains intercentra, giving it vertebrae pretty similar to those of Doleserpeton (and, of course, a number of "microsaurs", but not the holospondylous lysorophians, nor even the holospondylous brachystelechids).

Wasn't aware of that about Eocaecilia. As far as the phylogenetics goes, I'm aware that there are other characters associated as well (mostly palatal, if I remember correctly).

All datasets published so far are too small (in both the number of taxa and the number of characters), and at least some of them are full of mistakes (pers. obs. and the mentioned thesis that will be defended next week).

So, I think that all three hypotheses -- lissamphibians as temnospondyls, lissamphibians as lepospondyls, polyphyletic "lissamphibians" as both -- remain in the race for now.

I wouldn't disagree as far as the phylogenetics go. It's going to require more than "more taxa and more characters" though.

On its own it isn't. Phylogenetics with a single character is not possible. Total-evidence approach.

A phylogenetic analysis really is only as good as the characters sampled. There's this idea that a parsimony analysis is assumptions-free, but that's absolute bunk. A parsimony analysis requires that you meet a number of assumptions with your dataset.

And, just FYI, a phylogenetic solution can be reached with a single character, given you're working with a 4-taxon case. The question really is, can you trust it?

What else? Maximum munificence?

No, really. What other method is there?

Yes, Bayesian analysis. That should be tried. But it's not tremendously different; it's just parsimony with extra sophistications and a method to get support values without having to perform extra analyses.

Posterior probability isn't parsimony, although it will give similar results in most cases. The optimality criterion is different, and it is calculated in a different manner. It has a completely different set of assumptions, and it is somewhat easier to change the assumptions of a posterior probability analysis than a parsimony analysis.

Not as good as an ML, but ML isn't available for morphology (yet).

I like reweighting, BTW. Ruta & Coates (2007) tried it.

Reweighting is great in theory, problematic in practice.

Isn't minimum evolution a phenetic algorithm like neighbor-joining? That would explain why it hasn't been used for at least 10 years!

Minimum evolution, when used as a term to denote a specific method, is indeed an algorithmic method. However, parsimony meethods are also minimum evolution methods. What else do you think the minimum tree length/maximum parsimony criterion is?

Parsimony is not the assumption of the minimization of convergence. It is the minimization of the assumption of convergence. I forgot who said that first, probably Paul Sereno.

The assumption that minimizing convergence will give you the true phylogeny is in fact an assumption that one must make when approaching a phylogenetic analysis. There are certainly characteristics that we could look at and say "yes, this is probably far too plastic a character to have phylogenetic value" and we'd probably be right, even though this flies in the face of parsimony-based "total evidence approaches."

And indeed, large parsimony analyses find very large amounts of convergence/reversals. A tree without convergence would have a consistency index of 1 -- the result of the so far largest analysis of tetrapod phylogeny (102 taxa, 333 parsimony-informative characters), Ruta & Coates (2007), has a CI of 0.22.

Well, the question once again is, can we make a statistical distinction between a treelength of, say 130 steps vs 131 steps? Since any given sample of characters is just that, a sample, we have to assume some degree of error when assessing differences in treelength between hypothetical trees. We're not really talking about how good the best tree is, but rather how many trees there are that can't legitimately be rejected in favor of the best tree.

Sorry for the misunderstanding. I was only talking about terminology here: a clade is by definition monophyletic. "Clade" and "monophylum" are synonyms. The definition of "clade" is "an ancestor and all its descendants".

Right, but we're looking at a hypothesis of phylogeny, rather than an actual phylogeny. The hypothesis includes clades, which we then test. When we get right down to it, every criterion-based phylogenetic method out there forms a series of constraint trees and compares them to find the best possible. So we're looking at possible clades in this series of constraint trees and determining whether or not we can infer a monophylum from them.

Regarding your actual point, what's wrong with bootstrapping?

It's a resampling of your original matrix and then a calculation of consensus on the basis of each replicate. So, to a degree, it's a measure of the homogeneity of your dataset. At the same time, it won't tell you if your sample is itself poor, or if your sampled characters are sampled in a biased manner, and it certainly won't tell you if the basic assumptions of character evolution that you enter your analysis with are or are not correct. This is really the most severe problem.

Yes, long-branch attraction is a possibility here that must not be ignored. And indeed, many of the similarities between crown caecilians and other lissamphibians are most likely convergent -- Eocaecilia retains postparietals, jugals (like the albanerpetontids), tabulars or supratemporals...

One of the greatest things that Felsenstein did was establish the idea that there are certain conditions where, a priori, we should expect the existent methods to break down. Wiens et al. did the same with their 2005 paper (that you cited and I nearly did), as well as work by a number of other authors. The fact that we can more or less expect caecilians to fall out wrong on the phylogeny due to the a priori nature of the data is quite useful, all things considered.

Cool. Is there a paper in the works?

Maybe. It depends on whether or not I can justify the time expenditure. We have a bunch of material from the same locality as Plemmyradytes, as well as a species of lungfish and a very nice little microsaur. I've been hopping around from museum to museum looking at lungfish material, and I've been keeping an eye out for lysorophians while I'm at it, and I've got a bunch of notes, but I haven't a clue whether I'll ever formalize it all. There's a bunch of diversity there, though. As it is, though, I'm probably going to be dedicating the next few years towards theoretical work, so I'm not sure when that lysorophian material will ever come to fruition.

Sure -- among adults, if we can generalize from Amphibamus and Eoscopus (Daly 1994).

Well, once again, the question is one of ontogeny. We don't seem to have any larval amphibamids, but we do seem to have adults and subadults. The Tersomius-like branch of amphibamids seems to be almost entirely composed of adults, so they all seem to be well-equipped with vomerine tusks or fangs. At the same time, the Amphibamus-like branch seems to be more variable, but these animals are also more varied in ontogenetic stage. Which to me seems to beg the question as to whether there's an ontogenetic component to the phylogenies that we're not recovering.

I agree. The value of the following paper cannot be underestimated.

John J. Wiens, Ronald M. Bonett & Paul T. Chippindale (2005): Ontogeny discombobulates phylogeny: paedomorphosis and higher-level salamander relationships. Systematic Biology 54(1): 91 -- 110.

Absolutely. I nearly cited that one myself.

Yeah, I've never had such an insider discussion on Pharyngula before... :-) (And I still haven't even figured out who JDP is! Some insider am I! -- Oh, wait! Jason D. Pardo, the third author of Plemmyradytes! Nice to meet you!)

Indeed. Technically, second author. We switched authorships around at some point in review and it wasn't caught. Good to meet you as well!

A phylogenetic analysis really is only as good as the characters sampled. There's this idea that a parsimony analysis is assumptions-free, but that's absolute bunk. A parsimony analysis requires that you meet a number of assumptions with your dataset.

And, just FYI, a phylogenetic solution can be reached with a single character, given you're working with a 4-taxon case. The question really is, can you trust it?

I agree completely. I have a paper in review on a matrix with coding problems.

And, yes, ML would be nice -- except: where do we get the probabilities from? With molecular data we can find out how much more probable transitions are than transversions. With morphological data, I can't see such a thing.

However, parsimony meethods are also minimum evolution methods. What else do you think the minimum tree length/maximum parsimony criterion is?

That's just not how I've ever seen the term used.

The assumption that minimizing convergence will give you the true phylogeny is in fact an assumption that one must make when approaching a phylogenetic analysis.

No, minimizing the assumption of convergence gives one the most defensible phylogenetic hypothesis, where "defensible" can be quantified by an objective criterion. Science. Ockham's Razor. Truth is for philosophers. :-)

There are certainly characteristics that we could look at and say "yes, this is probably far too plastic a character to have phylogenetic value" and we'd probably be right, even though this flies in the face of parsimony-based "total evidence approaches."

It doesn't. There are ways to measure the phylogenetic signal of a character.

Michel Laurin (2004): The evolution of body size, Cope's rule and the origin of amniotes. Systematic Biology 53(4): 594 -- 622.

F. Robin O'Keefe & Peter J. Wagner (2001): Inferring and testing hypotheses of cladistic character dependance by using character compatibility. Systematic Biology 50(5): 657 -- 675.

Of course, neither method has been used anywhere near as often as it should have been. None of the analyses of tetrapod phylogeny has used one.

Well, the question once again is, can we make a statistical distinction between a treelength of, say 130 steps vs 131 steps? Since any given sample of characters is just that, a sample, we have to assume some degree of error when assessing differences in treelength between hypothetical trees. We're not really talking about how good the best tree is, but rather how many trees there are that can't legitimately be rejected in favor of the best tree.

Of course. That's why good papers check what happens when you look at trees that are 1, 2, 3... steps longer, and also check how many steps must be added to get a competing hypothesis.

So we're looking at possible clades in this series of constraint trees and determining whether or not we can infer a monophylum from them.

We're looking at possible clades in this series of constraint trees and determining whether or not we can infer a probable clade from them. "Clade" and "monophylum" are still synonyms. To the extent even that only the latter exists in German, for example -- "clade" wasn't invented by Hennig, and neither was "cladistics".

At the same time, it won't tell you if your sample is itself poor, or if your sampled characters are sampled in a biased manner, and it certainly won't tell you if the basic assumptions of character evolution that you enter your analysis with are or are not correct.

That goes without saying. Nothing is resistant to "garbage in, garbage out".

The fact that we can more or less expect caecilians to fall out wrong on the phylogeny due to the a priori nature of the data

Is it that dramatic? Especially now that we do have Eocaecilia and Rubricacaecilia?

The redescription of Eocaecilia also mentions at least one feature that, given our present knowledge of lepospondyl and dissorophoid ears, bolsters Lissamphibia: Eocaecilia has an operculum in the ear. But already Milner (1988) suggested that the extant caecilians should be coded as inapplicable, rather than as lacking the operculum, because the opercular apparatus requires a shoulder girdle that they don't have in the first place.

is quite useful, all things considered.

Well, it tells us what we already knew: inadequate taxon sampling can lead to problems, even when the taxon sampling is only constrained by the imperfect fossil record or our even more imperfect knowledge of it.

We don't seem to have any larval amphibamids, but we do seem to have adults and subadults.

I thought there were lots of Amphibamus larvae, and even at least one late larva of Platyrhinops so Schoch (2001 or 2002 -- the other shows A.) was able to illustrate its metamorphosis? Also, an SVP meeting abstract from last year promises the description of a Platyrhinops larva that had so far been referred to Branchiosaurus. -- But of course that's the Amphibamus-like branch...

By David Marjanović, OM (not verified) on 27 May 2008 #permalink

And, yes, ML would be nice -- except: where do we get the probabilities from? With molecular data we can find out how much more probable transitions are than transversions. With morphological data, I can't see such a thing.

Well, the transition/transversion probabilities in a GTR or HKY-85 model are estimated a priori. So, while there may be some degree of research on transition/transversion rates, the real issue is that we're looking at additional parameters when modeling substitution. As far as morphology goes, the same must ultimately apply; we can't expect to have perfect models, but even an imperfect model should give us a good deal of information.

No, minimizing the assumption of convergence gives one the most defensible phylogenetic hypothesis, where "defensible" can be quantified by an objective criterion. Science. Ockham's Razor. Truth is for philosophers. :-)

Occam's Razor has been misapplied far too many times in the attempt to justify parsimony as a method in and of itself. Occam's razor applies only to ad hoc hypotheses. We know how most of these morphological changes came about, and those we don't know, we can figure out soon enough and easily enough. For the same reason, I don't think homoplasy can really conceivably be considered an ad hoc hypothesis either, which gets us into trouble trying to justify phylogenetic inference via Occam's Razor. As far as I can tell (and I may be wrong, I'm still rather young and brash) parsimony is a moderate to middling approximation of maximum likelihood, and so, in situations where ML is completely impractical due to computational limitations, parsimony is an acceptable shortcut.

t doesn't. There are ways to measure the phylogenetic signal of a character.

Michel Laurin (2004): The evolution of body size, Cope's rule and the origin of amniotes. Systematic Biology 53(4): 594 -- 622.

F. Robin O'Keefe & Peter J. Wagner (2001): Inferring and testing hypotheses of cladistic character dependance by using character compatibility. Systematic Biology 50(5): 657 -- 675.

Of course, neither method has been used anywhere near as often as it should have been. None of the analyses of tetrapod phylogeny has used one.

Well, note that O'Keefe and Wagner (2001) were looking at specifically character dependence, that is, how one can determine whether multiple characters in an analysis are not in fact independent. Character independence is an assumption of parsimony analysis, so establishment of character independence is certainly one of a number of things that must first be established before one can even approach an analysis.

Of course. That's why good papers check what happens when you look at trees that are 1, 2, 3... steps longer, and also check how many steps must be added to get a competing hypothesis.

Well, when looking at phylogenetic studies, "good" papers not only quantify number of steps, they also quantify statistical significance of the difference in tree length. Given a particular set of data, as much as, say, 50 steps may in fact not be significantly different.

We're looking at possible clades in this series of constraint trees and determining whether or not we can infer a probable clade from them. "Clade" and "monophylum" are still synonyms.

I'd argue that "clade" generally refers to a group inferred from a cladogram, whereas a monophylum refers to the actual phylogeny. Cladograms and phylogenies are, of course, not the same thing, even though it's tempting to confuse the two.

Is it that dramatic? Especially now that we do have Eocaecilia and Rubricacaecilia?

Well, the long branch is still there. The question is whether it's long enough to be causing significant problems.

Well, it tells us what we already knew: inadequate taxon sampling can lead to problems, even when the taxon sampling is only constrained by the imperfect fossil record or our even more imperfect knowledge of it.

Not inadequate; uneven. There is a difference.

I thought there were lots of Amphibamus larvae, and even at least one late larva of Platyrhinops so Schoch (2001 or 2002 -- the other shows A.) was able to illustrate its metamorphosis? Also, an SVP meeting abstract from last year promises the description of a Platyrhinops larva that had so far been referred to Branchiosaurus. -- But of course that's the Amphibamus-like branch...

I'll have to take another look through the literature, then, but I'm still under the impression that it's not entirely clear. There are some larvae from Nyrany, and there's definitely Platyrhinops from Nyrany, but, if I remember correctly, Schoch's ontogenetic sequence was a bit incomplete. Regardless, it will be good to see more larval and adult material described.

Well, the transition/transversion probabilities in a GTR or HKY-85 model are estimated a priori.

No, they are usually estimated from the data.

parsimony is a moderate to middling approximation of maximum likelihood

Almost.

Kolaczkowski & Thornton (2004, in Nature) kicked loose a literature battle over whether parsimony sometimes fares better than ML and Bayesian methods. That's because ML and Bayes squeeze all characters into a small number of rate categories (usually just four, because computation time increases drastically with more categories -- and four categories are clearly a very crude approximation to reality in most cases). Parsimony does not assume any correlation between the rates of evolution of any two characters, and this is, at least sometimes, a more realistic assumption. (It equals putting each character into its own rate category.) The paper confirms once again that parsimony is more prone to long-branch attraction, but this is not always the greatest danger.

Unfortunately I still haven't read any of the subsequent back-and-forth papers.

I'd argue that "clade" generally refers to a group inferred from a cladogram, whereas a monophylum refers to the actual phylogeny.

I've never seen this usage before.

Not inadequate; uneven. There is a difference.

Sorry, I didn't define "inadequate". "Uneven" is a subset of that. -- As far as I can tell, taxon sampling that is even but too small still leads to problems.

but, if I remember correctly, Schoch's ontogenetic sequence was a bit incomplete.

Oh yes. There are no remotely complete sequences like what we almost have for Apateon. Schoch's 2001 and 2002 papers only compare "before metamorphosis" and "after metamorphosis".

By David Marjanović, OM (not verified) on 28 May 2008 #permalink

No, they are usually estimated from the data.

Right, but you estimate these prior to the phylogenetic analysis on the basis of prevalence data.

Almost.

Kolaczkowski & Thornton (2004, in Nature) kicked loose a literature battle over whether parsimony sometimes fares better than ML and Bayesian methods. That's because ML and Bayes squeeze all characters into a small number of rate categories (usually just four, because computation time increases drastically with more categories -- and four categories are clearly a very crude approximation to reality in most cases). Parsimony does not assume any correlation between the rates of evolution of any two characters, and this is, at least sometimes, a more realistic assumption. (It equals putting each character into its own rate category.) The paper confirms once again that parsimony is more prone to long-branch attraction, but this is not always the greatest danger.

Unfortunately I still haven't read any of the subsequent back-and-forth papers.

Well, the more specific your model, the more accurate it will be. The computation limits that existed for ML 10 years ago when PAUP was programmed are not such a big problem today, and they'll continue to shrink in importance. It should be intuitive that the more specific your model, the better your analysis, but I fail to see how parsimony is systematically better-equipped to handle rate variation.

Sorry, I didn't define "inadequate". "Uneven" is a subset of that. -- As far as I can tell, taxon sampling that is even but too small still leads to problems.

I generally view an inadequate sample to be one which lacks enough data to give sufficient resolution. "Uneven" falls into the category of "biased samples" not inadequate samples.

Oh yes. There are no remotely complete sequences like what we almost have for Apateon. Schoch's 2001 and 2002 papers only compare "before metamorphosis" and "after metamorphosis".

That's what I was thinking.

Sorry I didn't come back for so long.

It should be intuitive that the more specific your model, the better your analysis,

Unless it makes the analysis converge on a wrong solution, of course!

but I fail to see how parsimony is systematically better-equipped to handle rate variation.

By not putting artificial and extremely low limits on the number of rate categories. ML assumes that all nucleotides in the same category evolve with the exact same speed. Parsimony makes no assumption about the speed of evolution at all, except that it always chooses the slowest defensible one -- for each character separately, not for four or six categories.

By David Marjanović, OM (not verified) on 05 Jun 2008 #permalink