The hyomandibular of Eusthenopteron was incorrectly described by Sternberg: it compares closely with that of Ectosteorhachis described by Romer. Sternberg's account of the oticoccipital part of the cranium is also inaccurate, because his material is badly crushed. His phylogenetic conclusions are therefore suspect. It is considered that Eaton's theory of the transformation of the Rhipidistian hyomandibular to the primitive tetrapod 'stapes' or columella is essentially correct. The processus opercularis becomes the processus tympanicus (extrastapes or extracolumella). The tympanic and quadrate processes of the columella (stapes) are homologous in Stegocephalia and Reptilia. The development of a freely movable neck-joint in early tetrapods was associated with considerable modifications in the hyobranchial apparatus, resulting in the loss of the operculum, allowing the formation of a primitive tympanic diverticulum, and leading also to the linear discontinuity of the hyomandibular and ceratohyal. The living Amphibia possess two proximal 'middle ear' elements--a 'plectrum' and an 'operculum fenestrae ovalis'--not found separately in any amniote. They may be recognized as the homologues of processes of the stapes of fossil labyrinthodont Stegocephalia, from which group (including their small relatives the 'Phyllospondyli') the living Amphibia were probably derived. In the whole group the chorda tympani was primitively post-tympanic, and only a dorsal tympanic diverticulum was present. In primitive Reptilia the region of the otic notch moved ventro-laterally. The dorsal diverticulum was still present but reduced in importance, and was supplemented by a recessus infrastapedialis which enveloped the quadrate process of the stapes. In typical Theromorpha a further downward movement of the skeletal roof of the tympanic diverticulum took place, and an additional extension of the cavity (recessus mandibularis) was formed, applied to the lower inner surface of the mandible and later accommodated in a special cavity bounded by the reflected lamina of the angular. In mammals these parts of the tympanic cavity have enveloped the posterior bones of the primitive mandible and the quadrate, and the old tympanic membrane has enlarged and migrated somewhat antero-ventrally. The relations of the mammalian tympanic membrane and cavity to skeletal elements, muscles and nerves are shown to be direct consequences of this theory. Broom's suggestion that the cavity below the reflected lamina of the angular in Theromorpha housed a salivary gland is discussed, and the possible position of the parotid and submandibular glands in cynodonts is indicated.