The osteology of Ornithosuchus is described in some detail. This study is largely based on material discussed by previous workers, but also takes into account specimens hitherto undescribed. It is considered that the species O. taylori Broom 1913 is invalid, being based on larger individuals of the form previously named O. woodwardi by Newton (1894). Furthermore, evidence is presented to show that specimens previously described by Huxley (1877) and Walker (1961) as Dasygnathus longidens are also referable to Ornithosuchus. It is concluded that but one species is present in the material, the correct name for which is Ornithosuchus longidens (Huxley). At least eleven individuals are present with skulls ranging from about 50 to 450 mm in length. Observations on the smaller members of this series suggest that changes in the shape and proportions of the skull took place with growth in a similar manner to those seen in living crocodiles. In the light of new information concerning the osteology of Ornithosuchus it is considered that this reptile is a primitive carnosaur, occupying a position very close to the ancestry of the Jurassic and Cretaceous members of this group. An examination of descriptions and figures of Triassic reptiles referred by von Huene (1932) and later authors to the Carnosauria has led to the conclusion that the great bulk of this material is more properly allocated to the Prosauropoda, and that the only Triassic carnosaurs known at the present time, apart from Ornithosuchus, are Teratosaurus and Sinosaurus, these two names being here used in a restricted sense. It is suggested that Ornithosuchus is close to the common ancestry of both the Megalosauridae and the Tyrannosauridae, and the derivation of the skull patterns of these forms from that of Ornithosuchus is discussed. Additional fenestrations marginal to the preorbital fossa and in the surangular of advanced carnosaurs are held to be related to the development of the pterygoid musculature rather than to the need to lighten the skull. In the course of a brief review of the Carnosauria the view is put forward that Acrocanthosaurus is a Lower Cretaceous representative of the tyrannosaur group, using this term in a broad sense, and is perhaps related to the Cenomanian Spinosaurus and the Wealden Altispinax. As a result of examination of English megalosaurian material, the name Eustreptospondylus oxoniensis gen. et sp.nov. is proposed for the Oxford specimen previously known as `Streptospondylus' cuvieri, and Metriacanthosaurus gen.nov. for Megalosaurus parkeri. The incomplete cranium from Dives figured by Piveteau (1923) is made the type of a new species, E. divesensis, to which certain other carnosaurian material from Normandy is provisionally allocated. Following Chakravarti (1935), the endoskeletal elements included by Matley (1923) in the type material of the Indian nodosaur Lametasaurus are held to belong to one or other of the two carnosaurs Indosuchus and Indosaurus, but the numerous scutes are not thought to be carnosaurian and the name Lametasaurus is here restricted to them. It is suggested that Indosuchus from the? Turonian is a tyrannosaurid. On the other hand Ceratosaurus and Proceratosaurus are regarded as coelurosaurs. A modified classification of the Infra-order Carnosauria is given, in which two main groups are recognized; Superfamily Megalosauroidea to include the Megalosauridae, and Superfamily Tyrannosauroidea to include the Ornithosuchidae (restricted), Spinosauridae and Tyrannosauridae.