A review of the literature showed that size-by-size partition curves for dense medium cyclones using both stable and unstable media manifested a feature referred to as the "pivot phenomenon", which was attributable to the entrainment of neutrally buoyant particles in the medium pulp. The potential advantage of this discovery was that the coordinates of the pivot point, referred to as the pivot parameters, identified the flow split and density of the medium pulp in the separation zone of cyclones using conventional unstable media, information which otherwise could only be inferred from the properties of the cyclone product stream pulps.The pivot parameters represented a common point on the partition curves for all particle sizes pertaining to one particular separation. This fact made it mathematically convenient to incorporate the pivot parameters explicitly into the representation of size-by-size density separations. It also meant that the separation density could be dispensed with as a curve parameter, and this was justified on the basis that interpretation of cyclone performance using the separation density was confusing due to the fact that its value responded to changes in both separation efficiency and the medium pulp properties.A phenomenological dense medium cyclone model was developed which required the prediction of four partition curve parameters: the pivot partition number, Yp, the pivot density, ρp, and two parameters describing the variation of Ep with particle size, k and n, as follows:Yij = 1 / { 1 + exp[ In(Yp-1-1) + 1.099(ρp-ρij)/Epi ] } (2.13)where: Epi = k.djnPilot plant experimental work conducted using 100mm and 200mm cyclones confirmed that the pivot density and pivot partition number were approximately equal to the feed medium pulp density and flow split to underflow of the medium pulp respectively, for low feed medium solids concentrations and for stable cyclone operation. As the the feed medium solids concentration increased, the value of the pivot parameters systematically departed from their expected values. This was attributed to the constriction of the separation zone to a specific location within the cyclone caused by the development of a high density zone in the apex region. The pilot plant experiments involved the independent variation of pulp density and viscosity over the ranges 1400 to 2700kg.m-3 and 20x10-3 to 150x10-3 Nsm-2. The magnitude of the Ep parameter, k, was shown to be directly proportional to the pulp viscosity, and was not affected by the pulp density. The value of the Ep parameter, n, was relatively constant at -1.3.Industrial surveys conducted using 200mm and 400mm cyclones manifested the pivot phenomenon and the same dependence of the Ep on pulp viscosity. However, large density differentials were associated with poor separation efficiencies (large Eps), and this phenomenon was particularly evident for coarse particles. This effect was attributed to the internal recirculation of excessive quantities of near-gravity material within the cyclone, and the consequent deleterious effects of inter-particle interference. The data indicated that for density differentials below approximately 400kg.m-3, or for feed ores containing very little near-gravity ore, internal recirculation was not a problem, a conclusion reached by other workers and reported in the literature.The effect of cyclone diameter on performance was examined using pairs of 100mm/200mm and 200mm/400mm cyclone data for which the only significant variation In operating conditions was cyclone diameter itself. The data indicated that there was a possible conflict between the benefits of constraining a separation to smaller radii, such that particles experienced higher centrifugal separation forces, and the diminution of centrifugal head within the cyclone due to lower Reynolds number operation. It was concluded, in keeping with data reported in the literature, that an improvement in the separation efficiency of small particles resulting from the use of smaller cyclones was not guaranteed, and instead depended on the prevailing cyclone Reynolds number.Medium solids sedimentation and classification data from the 200mm pilot plant cyclone, operating at feed medium densities ranging from 1400 to 2700kg.m-3, were compared in order to determine the relative contribution of each mechanism to the existence of the cyclone density differential. It was concluded that classification was the dominant mechanism, and classification models were developed which accurately predicted the value of the underflow density. For both the pilot plant and industrial cyclones, values for the cyclone underflow density could also be directly predicted using measurements of the medium pulp viscosity, reflecting the fact that this variable had a significant effect on the classification of the medium solids.