Rapid emplacement of a mass via pyroclastic flows, or edifice failure, generates volcanic tsunamis. Physical modelling demonstrates that the efficiency of tsuna-mi generation is influenced by the angle the mass enters the ocean. Efficiency de-creases with increasing slope angle from 20° to 60°, before increasing to a maxi-mum at 90°, which corresponds to a mass falling directly into the ocean without interacting with the slope (impact tsunami). Further, in the case of surging pyro-clastic flows or regressive failures, successive closely spaced events may generate larger tsunami waves than a single event of comparable volume. It is difficult to assess if physical model results are meaningful for real world tsu-nami events due to limited observational data. This paper compares numerical models developed from physical simulations with palaeotsunami deposits from two New Zealand palaeo-events – pyroclastic flows from Mt Tarawera and edi-fice failure at Whakaari (White Island) – which constrains numerical simulations of the source mechanisms. The Mt Tarawera event involved multiple pyroclastic flows entering a lake during the AD 1314±12 Kaharoa Eruption. The interaction of multiple closely spaced pyroclastic flows is necessary to generate the 6-7 m maximum wave height inferred from near source tsunami deposits. Tsunami de-posits in the Bay of Plenty, dated to 2962±52 BP, are consistent with edifice fail-ure at Whakaari. In this case a single event with a volume of 0.23 km3 is suffi-cient to account for the tsunami deposits. Hence, if the failure was regressive, the successive stages were sufficiently close together to be indistinguishable from a large single event.