Stephanospondylus pugnax (Geinitz and Deichmuller 1882) Early Permian ~290 mya was originally considered a rarely studied diadectid, and therefore a pre-amniote, not part of the reptile clade. Here it nests between bolosaurs, like Bolosaurus, and pareiasaurs like Anthodon.
Larger and distinct from Belebey, the skull of Stephanospondylus had long sharp teeth. The nares were oriented anteriorly. No lateral fenestra pierced the cheek.
The neural spines were inflated. The dorsal ribs had expanded costal plates, as in Odontochelys and Diadectes. The scapula had an acromion process, as in turtles. The pelvis had a tall ilium and a separated pubis and ischium. The pubis included a distal attachment to a presumed ventral plastron.
The limbs were robust and short. The femur had a distintly angled neck, as in turtles.
Williston (1917) classified reptiles based on their temporal fenestra or openings and so placed turtles with primitive reptiles without temporal openings.
Broom (1924) reported, "Those who regard the structure of the temporal region of the skull as the safest guide to affinity will naturally place the chelonianans either with the primitive mammal-like reptiles or the cotylosaurs (primitive reptiles without temporal fenestration); those who hold that more reliance can be placed on the structure of the girdles and limbs will be more impressed with the affinities to the primitive diapsids such as Sphenodon."
Rieppel and DeBraga (1996) considered turtles to be diapsids close to Placodus in which the temporal fenestra had resealed. Unfortunately they also considered the squamosal to be the quadratojugal and the supratemporal to be the squamosal in the turtle Proganochelys. Further confusing the issue, they nested Placodus within the Lepidosauriformes following Younginiforms.
Gregory (1946) compared
living and fossil turtles with placodonts, "cotylosaurs" (captorhinids, pareiasaurs,
procolophonoids, and diadectomorphs, all considered amniotes at the time), and
with seyrnouriamorphs. He concluded that Testudines were derived from Paleozoic "cotylosaurs," and that among those, pareiasaurs approached Triassic turtles more
closely than the geologically older diadectids. Although placodonts. especially
Henodus, had evolved an amazingly turtle-like appearance, Gregory concluded
that they were not related to turtles and that convergent evolution, especially related
to dermal armor, causes a serious problem in recognizing testudine relationships.