From 1982 to 1995, Merapi volcano (Java, Indonesia) experienced two main eruptive cycles, from 1984 to 1986 and from 1992 to 1994. Each cycle had a duration slightly larger than 2 years. Four major eruptive events occurred in 1984, 1986, 1992 and 1994. These eruptions were monitored by a telemetered seismic network. We present an overview of the seismicity observed between 1983 and 1995. From seismological observations, several types of characteristic events are recognised. Volcanotectonic earthquakes are standard microearthquakes, either shallow (<1.5 km below the summit) or deep (between 2.5 and 5 km). Volcanotectonic earthquakes seem to correspond to an injection of magma into the upper part of the volcano and to a high gas pressure regime. The explosive eruptions in 1984 and 1992 were both preceded by deep and shallow volcanotectonic earthquakes but the 1986 and 1994 eruptions were not. Multiphase events are associated with the formation of the dome and with the supply of magma and of volatiles to the dome. Their hypocentres are near the crater, beneath the dome. Multiphase events occur when the magma flow in the chimney is large; sometimes the dome grows without multiphase activity. Low-frequency events and tremor are associated with fluid resonance. Rock falls (guguran in Indonesian) dominate the seismic records by their large number. The number of multiphase events is correlated to the number of rock falls. Both are related to dome growth. An aseismic zone is situated between two seismic zones, at a depth between 1.5 and 2.5 km below the summit: we postulate that it is a small shallow magma reservoir where magma injected from below can be stored temporarily. We try to relate seismic observations during both eruptive cycles to the initial injection of deep magma to the superficial magma system. The first eruption in each case seems to involve magma enriched in gas that extrudes the content of the conduit and explodes the old dome. A regular outflow of magma follows the initial explosion and gradually a dome is building up. Then it stops and a period of repose is observed. Later a second episode of dome growth occurs. The height of the dome reaches a maximum corresponding to a state of lithostatic equilibrium. The time delay of more than 2 years between the beginning and the end of a cycle may result from crystallisation in a volatile rich magma and from the building up of pressure in the superficial magma pocket. Some important events, like the collapse of the dome on 22 November 1994, are not preceded by any seismic precursor. When it reaches a large volume, the dome is not stable and is likely to collapse by gravity, for instance after the sudden influx of matter or gas or after a heavy rain. On Merapi, tremors are observed but do not always precede a dangerous event.