[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: Pisanosaurus and Effigia

Darned right and straight up!  In my (admittedly limited) experience
of checking the codings in other people's matrix, every single taxon
has multiple mistakes -- some of them clear-cut.

And that's not limited to dinosaurs. A paper will soon be submitted that consists of hardly anything but taking a published matrix on the origin of lissamphibians, checking it for errors, and looking how the tree changes when the questionable (and the outright wrong) scores are modified... I'm also working on another such matrix, which is also full of mistakes.

I am sure that's not because I am so much more awesome than they are
(:-) but simply because cladistic coding is difficult, tedious and
error-prone.


begin rant;

The "tedious" part seems to be the most important factor here. There are people who score taxa from memory rather than pulling every single description out of the shelf and checking. Others (...or even the same people...) find a statement "taxon X has state Y" in the literature and then code all members of X in their matrix as Y, without bothering to check if the character is perhaps unknown in some (or most) of them, or even if one or two are known to have a state other than Y and the statement in the literature was just a poorly worded generalization or an implicit reconstruction of the ancestral state.

And all this assumes that it's clear what the character is and where the limits between its states are. This is a major can of worms. Some people are very fond of character descriptions like "bone X large, plate-like, and with conspicuous process: absent (0), present (1)" and don't seem to have noticed that the default assumption must be that these are three characters, not one -- sure, they could be correlated, in which case they _must_ be treated as a single character, but this must be _demonstrated_ by means of a statistical study, rather than just tacitly assumed --, nor do they seem to have noticed that, unless large morphological gaps are evident, "large", "plate-like", and "conspicuous" _must_ be defined in quantitative terms if the coding is supposed to be reproducible, i. e., scientific.

endrant;