Monday, December 20, 2010

Herbivorous maniraptoriforms- Analysis criticism

Zanno and Makovicky just published a paper (from Zanno's thesis) presenting a quantitative and statistical analysis of the acquisition and distribution of traits associated with herbivory in coelurosaurs.  Basically they found that most maniraptoriforms except dromaeosaurids and (at least) derived troodontids weren't hypercarnivores.  No big surprise there, but it's an interesting concept to analyze.

On a basic level, I would have liked more evidence the traits used are actually associated with omnivory/herbivory.  For instance the bald eagle has several of the characters used-
- decurved anterior dentary, creating anterior gap between jaws.
- ventrally concave dentary.
- dentary exhibits tooth loss.
- premaxilla edentulous.
- mandibular symphysis fused.
- ischium over 66% of pubic length.
- opisthopubic pelvis.
- more than ten cervical vertebrae.
I think it's important to do a statistical study of dentary curvature/concavity, ischial length, cervical number, etc. in living birds to see if a correlation exists before assuming a correlation to study the diets of extinct taxa.  An elongate ischium is found in most non-maniraptorans except coelophysids, so its inclusion is especially confusing.

Yet even if we assume the traits correlate with herbivory, there are issues with the data used.  In figure 2 Yanornis is shown as having extrinsic evidence of herbivory while Confuciusornis is left with an ambiguous entry.  Yet both taxa are known from specimens preserving fish remains, as the supplementary information table S1 correctly indicates.  By their rules in table S1, Confuciusornis should be marked as having character 8 (evidence of carnivory: present only) since it has character 7 (vertebrate gut content: present only), which would then make it marked as lacking character 9 (extrinsic evidence of herbivory).  Similarly, Yanornis should be marked as having character 8, which would force 9 to be polymorphic since it also preserves direct evidence of herbivory.

There's also more of our friend the incompletely coded matrix.  Looking at Confuciusornis and Yanornis for instance since we're already dealing with them, neither is coded as lacking a U-shaped symphysis.  Yanornis isn't coded for its unserrated premaxillary teeth, densely packed teeth or short ischium.  Strangely, neither characters 1 nor 3 are coded for any taxon in the matrix. 

Also, Confuciusornis is miscoded as having a decurved anterodorsal dentary margin, lacking a ventrally displaced mandibular glenoid and having an ischium over 66% of pubic length.  It's also coded as lacking an inset dentary tooth row, conical anterior dentary and+or premaxillary teeth, elongate premaxillary teeth, unserrated premaxillary teeth, lanceolate cheek teeth, recurved teeth, ziphodonty, heterodont dentition, procumbant premaxillary teeth or having replacement waves between teeth, but all of these should be inapplicable since it lacks teeth.  Yanornis is miscoded as lacking tooth recurvature.

This is out of 31 characters, mind you.  So that's 21 wrong codings out of 62.  Note too there are unecessary characters such as "dentary exhibits tooth loss", when the list already contains "rostral dentary exhibits tooth loss" and "caudal dentary exhibits tooth loss."  Also "rostral teeth (premaxillary or dentary) conical to subconical" when it contains "premaxillary teeth conical or subconical (e.g., “incisiform”)" and "rostralmost dentary teeth conical: absent (0), present (1)." 

So while I don't doubt the general observation that maniraptoriforms were more herbivorous than most other theropods, I'm skeptical of the character distributions and statistics.

Reference- Zanno and Makovicky, 2010. Herbivorous ecomorphology and specialization patterns in theropod dinosaur evolution. PNAS Early Edition. 6 pp. doi 10.1073/pnas.1011924108


  1. After a quick preliminary reading of the paper, I remember that in the supplementary information, the authors were forced to encode the inapplicable characters as "0", since the program they used cannot work with unknown/inapplicable cells. This may explain the series of uncorrect codings in Confuciusornis and other taxa.

    I agree with your conclusion.

  2. Yet their matrix has tons of "-" codings which I can only assume are unknown/inapplicable codings. They're certainly distributed that way (e.g. Anserimimus is coded "-" for all cranial characters).