The identity of the ancestors and closest relatives of the turtle lineage was a longstanding scientific mystery, though new discoveries and better analyses in the early 21st century began to clarify turtle relationships. They had frequently been considered relatives of the
captorhinids, which also possessed an
anapsid skull configuration. During the 1990s, the consensus shifted towards Testudinata's placement within
Parareptilia, another "anapsid" clade. Analysis of fossil data has shown that turtles are likely
diapsid reptiles, most closely related either to the
archosaurs (crocodiles, birds, and relatives) or the
lepidosaurs (lizards, tuatara, and relatives). An early proponent of this scenario was
Goodrich (1916), who defended a diapsid origin of turtles based on morphological evidence. Genetic analysis strongly favors the hypothesis that turtles are more closely related to
archosaurs (crocodilians and birds) than to lepidosaurs (lizards, snakes and the
tuatara), though studies using only fossil evidence often continue to recover them as relatives of lepidosaurs or as non-diapsids. Some studies using only fossils, as well as studies using a combination of fossil and genetic evidence, both suggest that
sauropterygians, the group of prehistoric marine reptiles including the
plesiosaurs and the often superficially turtle-like
placodonts, are themselves stem-turtles. Lee (2001) found that forcing the turtle group to cluster with archosauromorphs resulted in
Rhynchosauria becoming Testudinata's sister clade. Forcing a relationship with lepidosaurs resulted in turtles being close relatives of sauropterygians within Lepidosauromorpha. The anapsid hypothesis was still better supported, although an archosauromorph affinity could not be rejected. Although morphology-based analyses usually do not support a turtle-archosaur clade (
Archelosauria), Bhullar & Bever (2009) identified a laterosphenoid bone, typical of
Archosauriformes, in the stem-turtle
Proganochelys. It may serve as a synapomorphy for this proposed clade. The cladogram shown below follows the most likely result found by an analysis of turtle relationships using both fossil and genetic evidence by M.S. Lee, in 2013. This study found
Eunotosaurus, usually regarded as a turtle relative, may be only very distantly related to turtles in the clade
Parareptilia. However, Lee also discusses other possibilities, including a potential compatibility between the parareptile and archosaur affinities. The cladogram below follows the most likely result found by another analysis of turtle relationships, this one using only fossil evidence, published by Rainer Schoch and Hans-Dieter Sues in 2015. This study found
Eunotosaurus to be an actual early stem-turtle, though other versions of the analysis found weak support for it as a parareptile. Bever
et al. (2015) redescribed the skull of
Eunotosaurus, identifying a lower temporal fenestra, with a juvenile specimen also having visible upper temporal fenestrae. This instigated a reinterpretation of this taxon as a diapsid instead of an anapsid. Their phylogenetic analyses strongly supported
Eunotosaurus's state as a stem-turtle and the placement of Pantestudines in Diapsida, though they couldn't determine a well-defined position within that clade. Sauropterygia and
Acerosodontosaurus also end up as possible stem-turtles in some of the trees. Benton (2015) compiled 2 synapomorphies of Ankylopoda (which would also include Sauropterygia,
Thalattosauria and
Ichthyosauria close to lepidosaurs): prootic-parietal contact and hooked fifth metatarsal. Time-calibrated phylogeny recovered by Shaffer
et al. (2017) dated the split of Pantestudines from its sister clade (the clade containing archosaurs and all tetrapods more closely related to archosaurs than to any other living animals) to mid-
Carboniferous. Laurin and Piñeiro (2017) placed turtles close to pareiasaurs among parareptiles once more. However, parareptiles were considered derived diapsids in this analysis. The authors interpreted these results as an indication that there might be no conflict between the hypotheses of a parareptilian origin and a diapsid origin. The cladogram below follows the analysis of Li
et al. (2018). It agrees with the placement of turtles within Diapsida but finds them outside of Sauria (the Lepidosauromorpha + Archosauromorpha clade). Gardner & Van Franken (2020) criticized the analysis by Li
et al., citing problems with the data set and observing that their proposed phylogeny was not supported once the issues were corrected. Lichtig & Lucas (2021) proposed
Pappochelys was related to sauropterygians,
Eunotosaurus was a
caseid synapsid, and turtles were derived pareiasaur parareptiles close to
Anthodon. According to this hypothesis, the turtle shell evolved from a fusion of the ribs to dorsal osteoderms.
Odontocheys, which lacked a carapace, is seen as a highly derived taxon instead of a representative of the ancestral state of turtles. The reliability of the molecular support for Archelosauria was also questioned. Contrary to this opinion, Simões
et al. (2022) found morphological support for Archelosauria. In their analysis,
Pappochelys is the basalmost stem-turtle but
Eunotosaurus is a basal neodiapsid instead of a stem-turtle, parareptile or synapsid. ==References==