Rethinking the placoderm jaw apparatus
14 September 2026 · morphology, Placodermi
Every few months I return to the gnathal plates of arthrodires, and every few months I leave more convinced that our character coding treats them as a tidy series when the material itself is anything but. The superognathals and inferognathals of a single specimen can disagree with each other about what "cusped" even means, and across specimens the variation slides continuously between states that our matrices insist on separating.
The deeper problem is homology assessment at the level of the whole apparatus. We code individual plates as if they were independent characters, but the functional unit is the bite, and the bite is a configuration. Two taxa can share every plate "state" in a list and still bite in mechanically different ways. When a matrix rewards the list and ignores the configuration, the resulting tree is answering a question we did not ask.
What I want to try next, on a small pilot set of well-preserved Dunkleosteus-grade material, is coding the apparatus as a compound character with explicitly ordered states, and comparing tree length and support against the conventional atomized coding. If the compound coding collapses resolution, that is informative too: it would suggest the atomized characters were carrying phylogenetic signal that the configuration view discards.
A caution to myself before starting: the pilot must be scored from specimens or high-resolution figures, not from redescription chains. Half of the odd states in the published matrices trace back, eventually, to a single drawing made in the 1950s. The chain of custody matters more than the character count.