Both the occurrence and behaviour of the Vaiont landslide have not been satisfactorily explained previously because of difficulties arising from the assumption that the failure surface was ‘chair’ shaped. It is now known that there was no ‘chair’, which means that the 1963 landslide could not have been a reactivated ancient landslide because the residual strength of the clay interbeds would have been insufficient for stability prior to 1963. Furthermore, the moderately translational geometry reduces the influence of reservoir-induced groundwater and hence of submergence. Standard stability analyses now show that prior to 1960, the average shear strength must have significantly exceeded the peak shear strength of the clay interbeds known to have formed the majority of the failure surface. Three-dimensional stability analyses confirm these results and show that at the time of the first significant movements in 1960, the rising reservoir level had a negligible effect on the Factor of Safety. According to these results, the Vaiont landslide was most likely initiated by pore water pressures associated with transient rainfall-induced ‘perched’ groundwater above the clay layers, in combination with a smaller than hitherto assumed effect of reservoir impounding, then developed by brittle crack propagation within the clay beds, thus displaying progressive failure. Further, very heavy rainfall accelerated the process, possibly due to reservoir-induced groundwater impeding drainage of the rainwater, until the limestone beds at the northeast margin failed. With the shear strength suddenly reduced to residual throughout, the entire mass was released and was able to accelerate as observed.