Therocephalia

Therocephalia is an extinct clade of eutheriodont therapsids (mammals and their close relatives) from the Permian and Triassic periods. The therocephalians ("beast-heads") are named after their large skulls, which, along with the structure of their teeth, suggest that they were carnivores. Like other non-mammalian synapsids, therocephalians were once described as "mammal-like reptiles". Therocephalia is the group most closely related to the cynodonts, which gave rise to the mammals, and this relationship takes evidence in a variety of skeletal features. Indeed, it had been proposed that cynodonts may have evolved from therocephalians and so that therocephalians as recognised are paraphyletic in relation to cynodonts.

The fossils of therocephalians are numerous in the Karoo of South Africa, but have also been found in Russia, China, Tanzania, Zambia, and Antarctica. Early therocephalian fossils discovered in Middle Permian deposits of South Africa support a Gondwanan origin for the group, which seems to have spread quickly across Earth. Although almost every therocephalian lineage ended during the great Permian–Triassic extinction event, a few representatives of the subgroup called Eutherocephalia survived into the Early Triassic. Some genera belonging to this group are believed to have possessed venom, which would make them the oldest tetrapods known to have such characteristics. However, the last therocephalians became extinct by the early Middle Triassic, possibly due to climate change, along with competition with cynodonts and various groups of reptiles — mostly archosaurs and their close relatives, including archosauromorphs and archosauriforms.

Anatomy and physiology
Like the Gorgonopsia and many cynodonts, most therocephalians were presumably carnivores. The earlier therocephalians were, in many respects, as primitive as the gorgonopsians, but they did show certain advanced features. There is an enlargement of the temporal opening for broader jaw adductor muscle attachment and a reduction of the phalanges (finger and toe bones) to the mammalian phalangeal formula. The presence of an incipient secondary palate in advanced therocephalians is another feature shared with mammals. The discovery of maxilloturbinal ridges in forms such as the primitive therocephalian Glanosuchus, suggests that at least some therocephalians may have been warm-blooded.

The later therocephalians included the advanced Baurioidea, which carried some theriodont characteristics to a high degree of specialization. For instance, small baurioids and the herbivorous Bauria did not have an ossified postorbital bar separating the orbit from the temporal opening—a condition typical of primitive mammals. These and other advanced features led to the long-held opinion, now rejected, that the ictidosaurs and even some early mammals arose from a baurioid therocephalian stem. Mammalian characteristics such as this seem to have evolved in parallel among a number of different therapsid groups, even within Therocephalia. Several more specialized lifestyles have been suggested for some therocephalians. Many small forms, like ictidosuchids, have been interpreted as aquatic animals. Evidence for aquatic lifestyles includes sclerotic rings that may have stabilized the eye under the pressure of water and strongly developed cranial joints, which may have supported the skull when consuming large fish and aquatic invertebrates. One therocephalian, Nothogomphodon, had large sabre-like canine teeth and may have fed on large animals, including other therocephalians. Other therocephalians such as bauriids and nanictidopids have wide teeth with many ridges similar to those of mammals, and may have been herbivores.

Many small therocephalians have small pits on their snouts that probably supported vibrissae (whiskers). In 1994, the Russian paleontologist Leonid Tatarinov proposed that these pits were part of an electroreception system in aquatic therocephalians. However, it is more likely that these pits are enlarged versions of the ones thought to support whiskers, or holes for blood vessels in a fleshy lip. The genera Euchambersia and Ichibengops, dating from the Lopingian, particularly attract the attention of paleontologists, because the fossil skulls attributed to them have some structures which suggests that these two animals had organs for distributing venom.

Classification
The therocephalians evolved as one of several lines of non-mammalian therapsids, and have a close relationship to the cynodonts, which includes mammals and their ancestors. They are broadly regarded as the sister group to cynodonts by most modern researchers, united together as the clade Eutheriodontia. However, some researchers have proposed that therocephalians are themselves ancestral to cynodonts, which would render therocephalians cladistically paraphyletic relative to cynodonts. Historically, cynodonts are often proposed to descend from (or are closest to) the therocephalian family Whaitsiidae under this hypothesis, however a 2024 study instead found support for a sister relationship between cynodonts and Eutherocephalia. The oldest known therocephalians first appear in the fossil record at the same time as other major therapsid groups, including the Gorgonopsia, which they resemble in many primitive features. For example, many early therocephalians possess long canine teeth similar to those of gorgonopsians. The therocephalians, however, outlasted the gorgonopsians, persisting into the early-Middle Triassic period as small weasel-like carnivores and cynodont-like herbivores.

While common ancestry with cynodonts (and, thus, mammals) accounts for many similarities between these groups, some scientists believe that other similarities may be better attributed to convergent evolution, such as the loss of the postorbital bar in some forms, a mammalian phalangeal formula, and some form of a secondary palate in most taxa. Therocephalians and cynodonts both survived the Permian-Triassic mass extinction; but, while therocephalians soon became extinct, cynodonts underwent rapid diversification. Therocephalians experienced a decreased rate of cladogenesis, meaning that few new groups appeared after the extinction. Most Triassic therocephalian lineages originated in the Late Permian, and lasted for only a short period of time in the Triassic, going extinct during the late Anisian.

Taxonomy
Therocephalia was first named and conceived of by Robert Broom in 1903 as an order to include what he regarded as primitive theriodonts, based primarily on Scylacosaurus and Ictidosaurus. However, his original concept of Therocephalia differed strongly from the modern classification by also including various genera of gorgonopsians (including Gorgonops) and dinocephalians. From 1903 to 1907 Broom added more therocephalian genera, as well as some non-therocephalians, to this group, including the anomodont Galechirus. The latter's inclusion highlighted Broom's view of therocephalians as 'primitive' and ancestral to other therapsids, believing anomodonts to be descended from a therocephalian-like ancestor such as Galechirus. However, by 1908 he considered its and some other non-therocephalian's inclusions to the group to be doubtful. In 1913, Broom reinstated Gorgonopsia as distinct from Therocephalia, but for many decades after there was still confusion from him and other researchers over which genera belonged to which group. The group's rank also varied from order, suborder and infraorder depending on authors' preferred therapsid systematics.

At the same time, the small 'advanced' therocephalians now classified under Baurioidea were often regarded as belonging to their own subgroup of therapsids distinct from therocephalians, the Bauriamorpha. Bauriamorphs were classified separately from therocephalians for many decades, though were often inferred to have evolved from therocephalians in parallel with cynodonts, each typically from different therocephalian stock. The inclusion of baurioids under Therocephalia was only firmly established in the 1980s, namely by Kemp (1982) and Hopson and Barghusen (1986).

Various therocephalian subgroups and clades have been proposed since the group was named, although their contents and nomenclature have often been highly unstable and some previously recognized therocephalian clades have turned out to be artificial or based upon dubious taxa. This has led to some prevalent names in therocephalian literature, sometimes in use for decades, being replaced by lesser-known names that hold priority. For example, the Scaloposauridae was based on fossils with mostly juvenile characteristics and is likely represented by immature specimens from other disparate therocephalian families.

In another example, the name 'Pristerognathidae' was extensively used for a group of basal therocephalians for much of the 20th century, but it has since been recognised that the name Scylacosauridae holds precedent for this group. Furthermore, the scope of 'Pristerognathidae' was unstable and variably was limited to an individual subgroup of early therocephalians (alongside others such as Lycosuchidae, Alopecodontidae, and Ictidosauridae) to encompassing the entirety of early therocephalians. Similarly, various names have been used for therocephalians corresponding to the family Adkidnognathidae in 20th century literature, including Annatherapsididae, Euchambersiidae (the oldest available name) and Moschorhinidae, and members have often had a confused relationship to whaitsiids. Consensus on the name and contents of Akidnognathidae was only achieved in the 21st century, asserting that a family-level group is established on the oldest referable genus and thus Akidnognathidae takes precedent for this group of non-whaitsioid eutherocephalians.

On the other hand, some groups previously thought to be artificial have turned out to be valid. The aberrant therocephalian family Lycosuchidae, once identified by the presence of multiple functional caniniform teeth, was proposed to represent an unnatural group based on a study of canine replacement in early therocephalians by van den Heever in 1980. However, subsequent analysis has exposed additional synapomorphies supporting the monophyly of this group (including delayed caniniform replacement), and Lycosuchidae is currently considered a valid basal clade within Therocephalia. However, most genera included in the group have since been declared dubious, and it now only includes Lycosuchus and Simorhinella. Modern therocephalian taxonomy is instead based upon phylogenetic analyses of therocephalian species, which consistently recognises two groups of early therocephalians (the Lycosuchidae and Scylacosauridae) while more derived therocephalians form the clade Eutherocephalia. Some analyses have found scylacosaurids to be closer to eutherocephalians than to lycosuchids, and so have been united as the clade Scylacosauria, while others have suggested they are each other's sister taxa. Within Eutherocephalia, major clades corresponding to the families Akidnognathidae, Chthonosauridae, Hofmeyriidae, Whaitsiidae are recognised, along with various subclades grouped under Baurioidea. However, while individual groups of therocephalians are broadly recognised as valid, the interrelationships between them are often poorly supported. As such, there are few higher-level named clades uniting the multiple subclades, with the exceptions of Whaitsiioidea (uniting Hofmeyriidae and Whaitsiidae) and Baurioidea.

Phylogeny
Early phylogenetic analyses of therocephalians, such as that of Hopson and Barghusen (1986) and van den Heever (1994), recovered and validated many of the therocephalian subtaxa mentioned above in a phylogenetic context. However, the higher-level relationships were difficult to resolve, particularly between the subclades of Eutherocephalia (i.e. Hofmeyriidae, Akidnognathidae, Whaitsiidae and Baurioidea). For example, Hopson and Barghusen (1986) could only recover Eutherocephalia as an unresolved polytomy. Despite these shortcomings, subsequent discussions of therocephalian relationships relied almost exclusively on these analyses. Later analyses focused on the relationships of early cynodonts, namely Abdala (2007) and Botha et al. (2007), included some therocephalian taxa and supported the existence of Eutherocephalia, but also found cynodonts to be the sister taxon to the whaitsiid therocephalian Theriognathus and thus rendering Therocephalia paraphyletic.

Later phylogenetic analyses of therocephalians, initiated by Huttenlocker (2009), emphasise using a broader selection of therocephalian taxa and characters. Such analyses have reinforced Therocephalia as a sister clade to cynodonts, and the monophyly of Therocephalia has been supported by subsequent researchers.

Below is a cladogram modified from an analysis published by Christian A. Sidor, Zoe. T Kulik and Adam K. Huttenlocker in 2022, simplified to illustrate the relationships of the major recognised therocephalian subclades. It is based on the data matrix first published by Huttenlocker et al. (2011), and represents the broad topologies found by other iterations of this dataset, such as Sigurdsen et al. (2012), Huttenlocker et al. (2014), and Liu and Abdala (2022). An example of the lability of these relationships is demonstrated by Liu and Abdala (2023), who recovered an alternative topology with Chthonosauridae nested deeply within Akidnognathidae.

Below is a cladogram modified from Pusch et al. (2024) analysing the relationships of therocephalians and early cynodonts. Their analysis focused on including endocranial characteristics to help resolve the relations of therocephalians and cynodonts to supplement previous analyses that relied almost entirely on superficial cranial and dental characteristics that are subject to convergent evolution, and as such only includes taxa with available applicable data. Of these, only four therocephalians could be included. However, they each represent four major groups within therocephalian phylogeny: the two 'basal therocephalians' Lycosuchus (Lycosuchidae) and Alopecognathus (Scylacosauridae) and two derived members of Eutherocephalia, Olivierosuchus (Akidnognathidae) and Theriognathus (Whaitsiidae).

Notably, their analyses consistently found cynodonts and eutherocephalians to be sister taxa, with the basal therocephalians Lycosuchus and scylacosaurids in a more basal position, rendering therocephalians as they are traditionally conceived paraphyletic. This differs from previous proposals of a paraphyletic Therocephalia which typically regarded cynodonts as being closest to derived whaitsiid therocephalians.