Thursday, March 09, 2006

Generic hostility, Part 6

Randy's turn. The Alroy paper he's talking about is this one:

Alroy, J. 2003. Taxonomic inflation and body mass distributions in North American fossil mammals. Journal of Mammology 84(2):431-443.

And obviously he did relent and give me permission to post all of this.

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

Subject: RE: Piss and piss

Fair enough. Do you mind if I post it all on the blog? :-) Seriously.

Actually, I'd prefer if you don't. If only because my first email and your first response were rather vicious. I might be a raging lunatic, but I don't want it clear to see on the internet! You could paraphrase our arguments though.

Okay, so maybe that was a bit sweeping. But wouldn't it be interesting if there was even a single category of data that even for some vertebrates picked out what we normally label genera? I went a bit overboard last time.

But I don't see how you escape the circularity of saying it goes to the resolution of genus, and then using it to separate genera! As I said before, it is self-fullfilling. This is conflating definition and diagnosis (in the broadest sense, not necessarily cladistically). You can't define genera on osteocalcin sequences and then diagnose all genera on the same data - there needs to be separation. Plus, even the best character data method fucks up sometimes. That is why we use monophyly (natural grouping) to define our groups, and then we could use osteocalcin to diagnose the genera.

I don't think genera are based on nothing.

I'm not sure what you mean here. I agree that genera are built on phylogenetic data (even pre-cladistically). And certainly, phylogenetic systematists who use Linnean ranks make sure that there taxa (inc. genera) are monophyletic. However, I don't think that genera or species are anything real in terms of being a special level with emergent properties of all their own. For example, you can't say, oh, there is clearly a level of selection at the genus level that is unique to the genus level (it is just normal clade selection). So, why should we single out the clades as "genera". Just call them clades, cause they ain't special!

At least with fossil vertebrates, I think genera are more solid than species. People have argued for decades about how many species of Apatosaurus and Diplodocus there are, but no-one argues about whether a particular specimen belongs to Apatosaurus or Diplodocus. For what it's worth, fossil genera seem to correlate with recognizable chunks of morphospace, better than species anyway. I've talked to some mammalian paleontologists about this and gotten a similar response.

I think this is mostly a result of the distribution of characters among taxa. Genera these days are more often diagnosed with good hard autapomorphies, whereas species generally are diagnosed using unique combinations of character-states (without autapomorphies). This means that at the species level, it is easier to argue that your characters are a result of individual variation, etc. And this is where the lumpers and splitters continually fight.

I don't think the trend you see in sauropods is true for even all dinosaurs. It would be interesting, we should ask Mike Benton or Matt Carrano (who have large taxic databases) if they see more turnover in genera or species through paleontologist time. I've included an interesting Alroy paper on the subject.

Sorry, I'm not on board. I don't see why a method that is useful for identifying monophyletic groups in one clade shouldn't be used just because it doesn't pick them up in all clades.

That's because you're misunderstanding my argument! As I say later in my previous email, it's the difference between taxonomy & phylogeny. The method is VERY useful for understanding phylogeny, but not for defining your taxonomy (the act of ranking things, etc.).

But I don't think genera are going away, at least not anytime soon, and I'd rather have objectively delineated genera, even in one little corner of the tree of life, than nothing.

Maybe maybe not. But I don't think genera going away is a problem. People are already revising alpha taxonomy with the idea of making groups monophyletic. The only difference in the future is that we just won't call them "genera", they'll just be a low-level clade. And that's perfectly fine. The names won't change. SO I don't understand why people are so reticent to jettisoning the little prefix "genus". Of course, for the sake of nomenclatural validity, none of us will be doing this until the Phylocode comes into effect. Nevertheless, in a paper I have in press, I phylogenetically define and diagnose a genus - the reviewers had no problem with that.

If you disagree with any link of the chain, you're not going to like what I'm saying.

I disagree with (1) definitely in the sense that these genera are not "special" or consistent across clades (even within say Dinosauria). With (2), I definitely recognize that there is a patchy distribution across morphospace, but I don't think genera encapsulate it successfully. Only a phylogenetic taxonomy that is rankless does a good job. Why? Because the amount of empty morphospace between two sister genera is different, not consistent. And this is simply a reflection of what morphospace is occupied. Also, I never understand why people are so fast to haul out constraint when they talk about patchy morphospace. There definitely is constraint and modularity and all that jazz, but I think a lot of empty morphospace is mainly due to differential extinction!

And osteocalcin might not work as advertised, or maybe it only appears to work because it ticks over every 2 million years and just by chance all of the genera on which it's been tested so far are more than 2 million years old. There are a jillion things that could derail this. But IF it works, it will be interesting, and useful, and cool.

Let me reiterate. I'm not criticizing osteocalcin's potential for being a new character data set. I'm only skeptical of claims that it can "pick out" genera. I think it has wondrous potential for improving our phylogenetic view of extinct taxa.

Is there some better way to say it?

Not make the claim until you have better evidence? Sorry, I'm just jaded and cynical, and I want proof in the pudding before someone makes a statement about it.

P.S., yeah, they sequence the peptides.

Regards,
Randy

2 Comments:

Blogger Mike Taylor said...

I'm glad Randy has an ms. in press that defines a new genus phylogenetically ... but I can't help thinking that if he was as ritually pure as his posts on this thread imply, he wouldn't raise a genus at all. So, Randy, why not a clade?

Regarding the utility of phylogenetically defined genera, here is a brief observation from an ms. that seems to have been in review since forever. (Darn, why can't more editors be like Randy? :-)

Very few efforts have been made to convert existing genera into
clades, and these efforts have not always yielded pleasing
definitions. For example, Clarke (2004) converted the genus name
Ichthyornis into "a modified apomorphy-based name for the clade
stemming from an ancestor that possessed all of the morphologies
described by apomorphies 2, 5, 6, 7, 8 (in the
Ichthyornis dispar
Diagnosis) homologous (sensu Patterson, 1982) with those in YPM 1450
(holotype of
Ichthyornis dispar Marsh, 1872b) and more closely related
to YPM 1450 than to Aves" (Clarke, 2004:20). The five key apomorphies
are found on pp. 25-27, five pages away from the definition that
relies on them, and require 700 words of description. While the
rigour and explicitness of this definition are laudable, one cannot be
too surprised if it fails to win over Linnaean taxonomists.


Well, I guess I stand by that :-) I hope Randy's definition is less excruciating.

1:22 PM  
Anonymous Anonymous said...

Hey, I just saw your post. In response to your question, "why not a clade"?: I do define it as a clade, and think of it as such. The only reason I call it a genus is to satisfy the ICZN. I also provide a diagnosis for the same reason. Phylocode has yet to come into effect (I generally agree with it, but it certainly isn't perfect), so I have to work within the current system. Also, even when it does come into effect, I don't want my taxonomy thrown out as invalid by the traditional Linnean taxonomists just because I didn't follow the ICZN. Things will be much easier when the ICZN bites the dust :P

Regarding your MS excerpt, the definition you cite may not be euphonic, but at least it gets the job done. This isn't romantic literature, although I do agree that the apomorphies should have been spelled out instead of just listing the numbers. Nevertheless, I am not a fan of apomorphic definitons, because they confuse the difference between definition and diagnosis. Clades should always be defined with OTUs and diagnosed with characters. Its bad practice for many reasons to both define and diagnose your clades with characters. Stick to stem and node-based definitions folks! Plus, they're shorter, and not as complicated as the example you cite.

Randy

7:22 PM  

Post a Comment

<< Home