A specimen of Dicynodon trigonocephalus from the Madumabisa mudstones of Zambia is described. The jaw adductor musculature is reconstructed. It is concluded that two slips of the adductor externus medialis were present. A posterior adductor ran from the quadrate to the medial surface of the lower jaw and a forwardly running muscle may have inserted on the reflected lamina of the angular. The masticatory cycle is much as Crompton & Hotton (1967) proposed, but new interpretations are developed for the movable quadrate and the W-shaped jaw hinge. It is proposed that the horny jaw covering inserted into the longitudinal dentary sulcus. The reflected lamina of the angular is discussed. As well as being a possible site for adductor muscle insertion, it is also suggested that hyoid musculature attached to it. The nasal capsule and cranial nerves and blood vessels are described. The forelimb and pectoral girdle are described and the main muscles reconstructed. It is concluded that the principle movements of the forelimb were protraction and retraction with only a small degree of long axis rotation. Rotation of the glenoid to lengthen the stride may have occurred. Flexion and extension of the lower limb was a particularly powerful action, but in contrast protraction and retraction were not powerful. The pectoral girdle and forelimb are seen as providing a stable and strong base for the massive anterior part of the body, but not generating much locomotory thrust. In contrast the hindlimb is supplied with musculature (the ilio-femoralis) that generates a powerful locomotory thrust by retraction and long axis rotation. The caudi-femoralis is much reduced but the ischio-trochantericus muscle has assumed importance in pulling the femur postero-ventrally. The pubo-ischio-femoralis externus is in a less advantageous position to do this and has assumed a role largely of a ventral adductor. In the axial skeleton, rotation of the head may occur, to a limited extent, between the atlas neural arch and the odontoid. The neck is a flexible region but the neck vertebrae bear strong transverse processes for origin of the extensive musculature that supports the head. However, the trend more posteriorly is to limit flexibility of the vertebral column by more vertically oriented zygapophyses and by ligamentous and muscular ties between vertebrae. The insertions of the longissimus dorsi have been modified from the presumed primitive condition and occupy only the dorsal surface of the transverse process. There are five sacral vertebrae. The centra of the first four are fused. The fifth is free. The anterior and posterior zygapophyses are reduced. The neural spines become longer antero-posteriorly, but very narrow transversely. The sacral ribs are all expanded but not fused either to the ilium or to the centra. The caudal region consists of possibly thirteen vertebrae and is probably complete. Ribs are present up to the fifth caudal vertebra. From vertebra 6 backwards haemal arch facets are seen on the centra. It is concluded that Dicynodon trigonocephalus was a slow-moving herbivore, nevertheless capable of a sustained locomotory effort. It may have avoided predators by crepuscular or cryptic behaviour. The powerful masticatory mechanism of D. trigonocephalus would have allowed exploitation of tough and bulky food sources. The relationships of D. trigonocephalus are discussed. It is possible that it is a close relative of the Permian stock that gave rise to Lystrosaurus.