A complete description is given of all the known material attributable to Anthracosaurus russelli Huxley. Apart from one skull specimen from Usworth Colliery, Washington, Tyne and Wear, England, all known specimens come, or are thought to come, from the Blackband Ironstone of Airdrie, near Glasgow, Scotland. Both horizons are of Coal Measure age, the former Westphalian A, the latter Westphalian B. Complete preparation of the holotype allows a reconstruction of the skull roof and occiput and, together with the Usworth specimen, the palate and a series of skull sections. Two lower jaw specimens are described for the first time. The cranial anatomy of Anthracosaurus is that of a massive embolomerous anthrocosaur which differs from eogyrinid embolomeres, notably Eogyrinus attheyi, in a number of important features. These include loss of the skull table/cheek kinetism and presence of an orbital lacrimal, biramous tabular horns and a wide dorsal exposure of the quadrate. Features of the palate are correlated with the distinctive dentition of relatively few very large marginal teeth and long palatal tusks: they include broad plate-like pterygoids, whose anterior rami meet in a median suture and extend laterally to the palatine tusks (thus largely concealing the palatine bones from ventral view) and whose quadrate rami lack a deep descending flange. The lower jaw is unique in having confluent meckelian fenestrae. Vertebrae attributed to A. russelli are of normal embolomerous type except that the notochordal canal is occluded by a bony plug in the centra. An interclavicle of distinctive shape is also attributed. Several North American embolomere specimens are presently placed within the family Anthracosauridae. Of these the most complete is Eobaphetes kansensis Moodie consisting of three blocks, the original holotype and paratype and a third block not before described. As was first demonstrated by Dr Donald Baird these comprise a single skull, here described in its entirety. Description of Eobaphetes is preceded by a discussion of its provenance. This was originally recorded as 'Coal Measures', Washington County, Kansas, which is improbable but not more so than its current attribution to the Namurian of Washington County, Arkansas. Coal and spore analysis, supported by palaeogeography, suggest a late Pennsylvanian, probably Stephanian, horizon. In most respects the anatomy of skull roof and lower jaw of Eobaphetes suggests relationships with Eogyrinus rather than Anthracosaurus. This is also the case with the closely related Leptophractus obsoletus from Linton, Ohio, to which species 'Anthracosaurus lancifer', also from Linton, is referred. However, Eobaphetes and Leptophractus, together with Neopteroplax conemaughensis, share a number of distinctive features. These include a flat-topped surangular crest to the jaw and, importantly, a dentition of massive palatal tusks and anterior marginal teeth but numerous small posterior teeth in maxillary and dentary. Tooth shape is also distinctive. It is thus proposed to retain the family Anthracosauridae for A. russelli alone, while placing the American species within the Eogyrinidae, but as a new subfamily the Leptophractinae, distinct from the Eogyrinidae. A diagnosis of each of these taxa is given.