A second extrusive phase of the currently ongoing 1995–2003 eruption of Soufrière Hills Volcano (SHV), Montserrat, commenced in mid-November 1999 following ∼19 months during which no fresh lava had reached the surface. By mid-March 2000, a new andesite lava dome constructed within a collapse scar girdled by remnants of the 1995–1998 dome complex had attained an estimated volume of ∼29±3 million m 3 (Mm 3). On 20 March 2000, during a period of heavy rainfall on the island, a significant collapse event ensued that removed ∼95% of the new lava dome (∼28±3 Mm 3) during ∼5 hours of activity that generated ∼40 pyroclastic flows and at least one magmatic explosion. The associated ash cloud reached an altitude of ∼9 km and deposited ash on the island of Guadeloupe to the southeast, and a number of lahars and debris flows occurred in valleys on the flanks of SHV. A large quantity of observational data, including contemporaneous field observations and continuous data from the broadband seismic network on Montserrat, allow a detailed reconstruction of this dome collapse event. In contrast to most of the large dome collapses at SHV, the 20 March 2000 event is distinguished by a lack of short-term precursory elevated seismicity at shallow depths beneath the lava dome. Broadband seismic amplitude data recorded during the event are used to infer the cumulative volume of collapsed dome as the collapse progressed. These data indicate that the high-velocity pyroclastic flows observed at the climax of the event removed by far the largest portion (∼68%) of the lava dome at peak discharge rates (estimated from the seismic record) of ∼2×10 4 m 3 s −1. Following the 20 March 2000 collapse, lava dome growth recommenced immediately and continued without significant interruption until another, larger dome collapse occurred on 29 July 2001. The 29 July 2001 event also coincided with heavy rainfall on Montserrat [Matthews et al. (2002) Geophys. Res. Lett. 29; DOI:10.1029/2002GL014863] and lacked precursory elevated seismic activity. We attribute the initiation of the 20 March 2000 collapse to a prolonged spell of heavy rainfall on the lava dome prior to and during the event. The precise causal mechanism remains controversial, though some combination of mechanical erosion and/or destabilization of a critically poised face of the lava dome, the action of pressurized steam or water on potential failure surfaces within the dome, rapid cooling of hot lava and small phreatic explosions seems likely. Anecdotal evidence exists for other rainfall-induced activity on Montserrat, and the triggering of explosive or pyroclastic flow activity by rainfall has been noted at dome-forming volcanoes elsewhere, including Merapi, Indonesia [Voight et al. (2000) J. Volcanol. Geotherm. Res. 100, 69–138], Unzen, Japan [Yamasato et al. (1997) Papers Meteorol. Geophys. 48], Santiaguito, Guatemala [Smithsonian Institution Global Volcanism Network Bull. 15 (1990)] and Mount St. Helens, USA [Mastin (1994) Geol. Soc. Am. Bull. 106, 175–185]. Hazard mitigation plans at dome-forming volcanoes would therefore benefit from the inclusion of meteorological forecasting and rain monitoring equipment, particularly in the tropics.