The order Nectridea of the subclass Lepospondyli is of Palaeozoic age. Within this order, later members of the Keraterpetontidae developed hyper-extended tabular horns, so that in plan view the skull is boomerangshaped. Many unsuccessful attempts have been made to explain this shape in functional terms. The two genera that show the greatest development of these horns. Diplocaulus Cope and Diploceraspis Romer, are analysed experimentally in a low-speed wind-tunnel. Differences between the two genera include that of size: the former is about twice the size of the latter genus and it has a prominent, ventrally directed flange on the quadratojugal-squamosal region not seen to the same extent in Diploceraspis. In Diplocaulus the otic notch, which has come to lie on the ventral side of the skull, is large and extends proportionally further towards the tip of the horn than in Diploceraspis. The otic notches may have supported pharyngeal pouches, developed as accessory respiratory organs. A full-scale model of the Diplocaulus skull was made from information obtained from published illustrations and casts. It was mounted in the wind-tunnel so that angles of incidence varying from - 10 to + 25 degrees were possible, and a fixed body was modelled so as to account for interference effects. The Diploceraspis condition was simulated by removing the prominent quadratojugal flange. Four conditions were investigated: (1) the Diplocaulus model; (2) the Diploceraspis model; (3) the investigation of the effect of roughness of the surface of the model, to simulate the labyrinthodont condition of the dermal bones; and (4) an investigation of the effect of mouth-opening on the behaviour of the model. In addition, a flow-visualization test was carried out. All the experiments were carried out at a speed corresponding to the animal moving at 1.65 m/s in water. Coefficients of lift, drag and pitching moment were measured, over the range - 10 to + 25 degrees, at two degree intervals. The significance of the results lies in the behaviour of the lift and pitching moment coefficient curves. The position of the centre of pressure does not move with the change of these two parameters and therefore the point of action of the centre of lift is fixed with respect to the occipital condyles. Forces exerted on the head are proportional to the deflexion of the head. Small but significant differences are seen when the curves for the two genera are compared. In Diplocaulus the lift and pitching moment curves cross the zero-line very close to the origin, but in the other genus they cross considerably to the left of the origin. Roughening of the surface in two stages causes lift coefficients to be diminished and drag increased. It is concluded that the surface of the living animals was smooth or almost so. With the smooth model, it was noted that when the mouth was opened that neither lift nor drag coefficient was significantly altered. It is concluded that when taking prey these animals must have suffered little deceleration. The flow visualization tests show that two flow regimes operated on the upper surface of the Diplocaulus model. To begin with, the flow was streamlined and parallel to the midline over all the head area, but after a critical angle of incidence the central region of the head stalled and the flow over the horns became stabilized as laterally directed vortices. From all of the results it is believed that the Keraterpetontidae were active midwater feeders preying on small fish, larval amphibia, aquatic arthropods and gastropods. They used the unique physical properties of the head to effect steep climbing ascents from the lake or stream bed to attack their prey, before returning to the bottom. The differences between the two genera are related to their contrasting environments and the possible course of evolution giving rise to these extreme adaptations is considered. Dorso-ventral flattening is seen as an aid to an active mode of life in the two genera under discussion here, and the same possibly applies to the Labyrinthodontia as a whole.