In many groundwater systems, fluid density and viscosity may vary in space and time as a function of changes in concentration and temperature of the fluid. When dense groundwater plumes interact with less dense ambient groundwater, these density variations can significantly affect flow and transport processes. Under certain conditions, gravitational instabilities in the form of lobe-shaped fingers can occur. This process is significant because it can lead to more rapid and spatially extensive solute transport. This paper presents new experiments carried out in a sand filled glass flow container under both fully saturated and variably-saturated conditions and focuses upon the processes that occur at the capillary fringe and below the water table, as affected by a dense contaminant plumes migration through the unsaturated zone. Source fluids stained with Rhodamine-WT were introduced at the upper boundary of the tank at a range of low and high densities. In addition to the fluid density gradients and porous medium permeability that determine the onset conditions for instabilities in fully saturated experiments, volumetric water content appears critical in the variably-saturated laboratory runs. Plume behaviour at the water table appears dependent upon the density of the fluid that accumulates there. For neutral and low density fluids, plumes accumulate at the water table and then spread laterally above it and the water table forms a barrier to further vertical flow as pore water velocities reduce with increasing water content. For medium and high density fluids, vertical movement continues as instabilities form at the capillary fringe and fingers begin to grow at the water table boundary and move downwards into the saturated zone. In these cases, lateral spreading of the plume is small. Despite these more qualitative observations, the exact nature of the relevant stability criteria for the onset and growth of instabilities in variably-saturated porous media presently remain unclear. All experimental results suggest, however, that the unsaturated zone and position of the water table must be considered in contaminant studies in order to predict the migration pathways, rates and ultimate fate of dense contaminant plumes. It is possible that the results of experiments presented in this paper could form a useful basis for the testing of variable-density (and variably-saturated) groundwater flow and solute transport numerical codes because they offer controlled physical laboratory analogs for comparison. They also provide a strong basis for the development of more rigorous mathematical formulations that are likely to be either developed or tested using numerical flow and solute transport simulators.