This study addresses the initiation mechanisms of mass failures on clinoform foresets. Previous studies have created mass flows by releasing dense water–sediment mixtures into standing water, thus imposing the initial conditions for the mass failures rather than allowing them to form on their own. Para-meters such as the density, composition and initial momentum of the failures are pre-determined, precluding observation of the factors that set them initially. This study uses a new experimental method that allows a range of mass failures to self-generate. Building a clinoform using a cohesive mixture of walnut-shell sand and kaolinite allows the foreset to build up and fail episodically, generating mass failures. Slopes undergo a series of morphological changes prior to failure, creating a concave shape that becomes exaggerated as deposition continues. This morphology leaves the slope in a metastable state. Once the slope is destabilised, failure is initiated. This study investigates the effect of clinoform progradation rates on failure size and frequency by conducting experiments over a range of water and sediment discharge rates. Neither failure size nor failure frequency changes with discharge rate; instead, increases in sediment supply are taken up by changes in the partitioning of sediment between the steep upper foreset and the more gradual delta-front apron (toeset) below. Sediment is delivered to the delta-front apron by a form of semi-continuous slow creep along the foreset. This slow creep is a failure mode that has not received sufficient attention in the submarine mass-flow literature. The independence of failure size and frequency from sediment supply rate suggests that the presence of mass-failure deposits does not provide information on the rate of sediment delivery. If these relations hold at field scales, this would imply that individual mass failures are relatively insensitive to changes in water and sediment supply.