It has been suggested that the sustainability of turbidity currents is derived from self-sustainment of the flow. This self-sustainment can be realized through the process of acceleration of the current as it increases its own density due to the incorporation of sediment eroded from the substrate. Although self-sustaining turbidity currents have been predicted theoretically, experimental turbidity currents to date have all been net-depositional decelerating flows. We report here the results of experiments on self-sustaining turbidity currents. In order to produce non-depositional flows, we modeled the sediment using two types of plastic particles with densities that were much lower (1.3 and 1.5 g/cm 3) than that of siliciclastic sands. After the bed was covered with sediment, a mixture of sediment and water was injected to produce a turbidity current at the upcurrent end of the flume. As a result, some experimental flows showed acceleration, and measurements using siphons revealed that the sediment concentration of the flow increased downcurrent, so indicating a trend toward self-sustaining turbidity currents.