The interaction of liquid steel and an inclined impinging oxygen jet in an electric arc furnace (EAF) is of interest both commercially and scientifically. The bath activity in the EAF may vary from a surface splash ejected to high elevation to intense subsurface mixing. The system is appropriately modelled using water and nitrogen with scaled flowrates. In previous work, the gas/liquid contact was investigated by use of a geometrically similar 1/3rd scale three-dimensional water model. The cavity formed by the jet contacting the liquid surface was characterised by four modal regimes. These regimes were seen to depend on the lance angle, the height of the lance and the jet flowrate. To investigate the evolutionary mechanisms of the cavity regimes, a two-dimensional water model study was undertaken. The two-dimensional water model was a rectangular viewing tank with an inner movable glass wall that allowed a very thin slice of the system to be obtained. The shape of the cavity formed on the water surface was seen clearly along with the expected cavity oscillations. High speed video footage of the two-dimensional system allowed the cavity oscillation to be directly observed. The gas-liquid interaction produced a wave that travelled along the surface of the cavity until it reached the cavity crest where it was torn from the liquid surface and dispersed in a splash. It is the regular progression of the wave's nodes and antinodes along the cavity surface that makes the cavity appear to oscillate. When the wave reaches the crest of the cavity, it will either fall back into the path of the gas jet or be projected as a splash depending on the verticality of the cavity surface. The two-dimensional work, along with the initial three-dimensional investigation, has shown that the troublesome splash in the EAF is caused by how the lance is positioned directionally or azimuthally. By changing the lance angle or height the deleterious splashing of molten metal may be prevented, ameliorated or controlled. The frequency of the wave production was determined from the high speed video footage. The cavity oscillation was found to be a function of the size of the cavity, the inclined height of the lance and the modal regime being produced. Alterations to the flow through the lance had only a moderate effect on the frequency of oscillation indicating that velocity was not the major influential factor.