Please see Nick Gardiner’s comment and my reply below, as he attempts to discredit this day’s theme with traditional thinking and I reply with an explanation of my niche in paleontology by following this day’s theme.
Grandcolas et al. 2001 push for the use of
“as many characters as possible should be included in the cladistic analysis.” The risk, as seen by Grandcolas et al. not to do so is to, “bias the analysis.” The risk in my opinion is you’ll never run the analysis if you keep looking for more and more characters. Their number, if not infinite, is certainly some multiple factor of legion. In certain circles we call this sort of encouragement snipe hunting.
The authors’ wish/suggestion/push runs counter to my arguments
for the large reptile tree (LRT), and its overlapping satellites, the large pterosaur tree (LPT) and the therapsid skull tree (TST). I have argued that the minimum number of characters and character states should be employed that will separate each and every taxon and each and every node apart from one another, with high Bootstrap scores. After all, the goal of every analysis is to model and replicate actual evolutionary events to the best of our ability, given our detachment from taxa in deep time and the reduction of data to only skeletons or partial crushed, and broken skeletons in most cases. If most of the characters used are general in nature and are set in place before the addition of any included taxon, then no a priori bias can be said to exist with regard to included characters. Case in point: the LRT has successfully used intended generalized reptile traits to nest birds, mammals and fish with high resolution.
Frankly, the authors’ arguments go over my head throughout
as they argue semantics, possibilities and a priori issues; nothing specific. They point out many theoretical errors reported by prior authors. Thankfully they sum it all up neatly with a statement near the conclusion, “Not to use available and logically suitable characters is like suppressing evidence.”
On that note, the authors should have remembered
that every judicial case has a court date, a moment in time when evidence is no longer sought and gathered, but used. Theory is one thing. Practice is another. At one point or another, you simply have to run the analysis.
The authors conclude,
“Phylogenetics must propose refutable hypotheses: characters should not be included or excluded from the analysis because of a priori ideas regarding their evolution.” I think we can all agree to that.
However, let’s remember, we all lead busy lives.
Seeking characters ad infinitum sooner or later leads to decreasing value for the incremental and extended effort. Seeking just enough characters to recover a fully resolved tree that documents a gradual accumulation of derived traits at every node will still leave you time to eat, sleep, drive, work and brush your teeth.
Adding taxa
is still the best way to add value to a phylogenetic analysis if the LRT, LPT and TST are any indication. All use a large enough character list with variations within to separate virtually all taxa from one another in a way that appears to echo micro-evolutionary processes in deep time from jawless fish to blog readers.
Thanks to Neil for bringing this paper to my attention.
References
Grandcolas P, Deleporte P, Desutter-Grandcolas L and Daugeron C 2001. Phylogenetics and Ecology: As many characters as possible should be included in the cladistic analysis. Cladistics 17:104–110.
The Pterosaur Heresies 
If I notice an unusual feature on a bone I’m inspecting, and realize that it is also found on a specific other taxon, how do I test whether the feature (character) is homologous or homoplastic? Unless it is polymorphic, ontogenetically variable, or otherwise unsuitable for taxonomic purposes, I would put it in a phylogenetic analysis in an attempt to reject (or not reject) the null hypothesis that it points to close relations between taxa. What other quantitative method can be used to translate that physical information? You can probably sympathize with this, as you point out shared characteristics between taxa all the time.
The difference arises when you assume that a character that requires inspection of a specimen is somehow less important than one which only requires inspection of a skeletal, and skeletals are by nature simplified inferences or representations of specimens. As an example, the LRT has one character specific to the tibia: its length relative to the ilium. An isolated tibia is useless in your eyes, but tibiae possess many helpful features (ridges, crests, proximal or distal shape) which can elucidate relationships. Adding in characters is not an arbitrary ad infinitum quest with no reward, it is a way to test phylogenetic hypotheses which may arise from looking at bones. Before cladistics, people had to speculate wildly about homology and homoplasy, but now we can just plug in the features we see.
The LRT does have high bootstrap scores, yes. From a distance, its taxa do resemble each other, yes. But looking closely, at the individual bones, the distribution of characters (“gradual evolution of traits”) is all over the place. All phylogenetic analyses look at gradual evolution of traits, that’s what Most parsimonious trees are. So claiming that others don’t fulfill your standards of trait evolution is incongruent with reality. And claiming that the LRT does, despite its taxonomy adding hundreds of steps with others’ wider character sampling, is hypocritical. Another point you like to bring up when arguing for the LRT’s methodological soundness is bootstrap scores. High bootstrap scores are no guarantee of validity, as miscodes or other methodological issues can be responsible for apparently stable links. Teraterpeton + phytosaurs seemed stable until it was corrected. Luperosuchus as an erythrosuchid was also supported only through iffy character codings and formulations. And next, I’ll be looking into PVL 4597, the Gracilisuchus specimen which you retain as a separate taxon as a result of a couple miscodes.
You are concerned about wasting time with adding characters, as if it’s some form of bureaucratic filibustering which restricts scientific advancement. But paleontologists see analyzing fossils and their characters as a form of scientific advancement in its own right. I certainly get a kick out of it, which is one of the few reasons I stick around to critique you. You consider yourself a journalist or entrepreneur, getting the word out about some new discovery. You want your ideas to be heard. But science isn’t about doing something fast, it’s about doing something right. And we want to do our work slowly and methodically so that we can minimize mistakes. Maybe you should slow down as well. You don’t need to post every day, you could spend that time rereading the literature closely, or understanding tibia evolution, and most of all, discussing your methods with myself and others. I don’t want misinformation on the internet, I want to minimize the damage that could be caused by reckless actions.
Neil, congratulations. You have blackwashed positive attributes and made them negatives. BTW, if you think working on the LRT, LPT and TST for eight going on nine years is fast, well… there you go. I appreciate that you don’t want misinformation on the Internet. Does that mean you are actively writing to other authors as well?
I’ve archived this page both on my personal desktop and multiple archiving tools (Wayback Machine, Archive.is, etc). If anyone ever wonders (including you), why I don’t regard your LRT as science, you’ve answered it all yourself here:
“However, let’s remember, we all lead busy lives.
Seeking characters ad infinitum sooner or later leads to decreasing value for the incremental and extended effort. Seeking just enough characters to recover a fully resolved tree that documents a gradual accumulation of derived traits at every node will still leave you time to eat, sleep, drive, work and brush your teeth.”
You are fundamentally disinterested in putting in the hard work of examining fossils (whether in person, via CT, microscopy, etc.) and comparing characters and sorting through versus looking at traits you can tell superficially (either from skeletals, photographs, etc) and when someone calls you on it, you claim you just don’t have the time to put the work in.
Consider this scenario, Nick. I have chosen to look at more taxa and fewer characters. Others, like you, choose to look at fewer taxa and more characters. Those workers are cited in this blog so that those who want to dig deeper and repeat observations may do so.
When you look at more taxa you get ‘the big picture’ how a longer list of taxa are related to one another. When you look at fewer taxa you see each one in greater detail, but forego the big picture, the evolution of one form into another.
My niche is to consider as many taxa as possible, and to make suggestions to workers who have omitted taxa recovered by my studies. Sometimes that pans out. Others have published on relationships that were first reported years before here in this blog.
When Carroll 1988 ( Vertebrate Paleontology and Evolution. WH Freeman and Co. New York) published on all the vertebrate taxa known at the time do you think he visited every taxon he listed? I agree with you, and Carroll, I don’t have the time or travel expenses to run around the world looking at every specimen, just like Carroll. If workers have published material that is not trustworthy, that’s on them. When I see something untrustworthy, I call them on it. When someone calls me on a mistake, I correct it when validated, as I have done tens of thousands of times over the past 8 years, mostly on my own reviews.
On the other hand, when I have visited fossils in person and reported on them in greater detail than prior workers, often enough the work was rejected by reviewers claiming I didn’t spend enough time with a specimen. When I have been published, the work has been ignored. That’s why pterosaur experts continue to say we have no idea how pterosaurs got their wings. So, to your point, I’ve been down that traditional alley.
Regarding your demeaning term ‘superficially’… defined as, “as to the outward appearance only; on the surface” that’s going to cover virtually all the data I use in my cladograms. I will leave histology and cranial nerve openings to others.
Regarding “don’t regard your LRT as science,” let’s put your scenario into effect. Let’s imagine that I double the number of characters. That’s still not enough. I missed some, someone will say. Double that number. Still missed some, someone else will say. Ad infinitum. Adding characters is not necessary when the number of characters and their multiple scoring options are sufficient to lump and separate at virtually every node. I was surprised to see that over time and taxa as you are skeptical of it.
Now let’s turn that around. I say, based on the LRT, currently at 1592 taxa, someone excluded a few pertinent taxa in their more focused study. Or they made a chimaera of two taxa. Or they included a taxon that should not be in a clade based on the LRT results. Those are specific fixes that, based on my own experience, can be repaired posthaste. That’s what I’m here for. That’s my niche. Someone had to do it. I raised my hand.