The eruption that started in the Hekla volcano in South Iceland on 17 January 1991, and came to an end on 11 March, produced mainly andesitic lava. This lava covers 23 km2 and has an estimated volume of 0.15 km3. This is the third eruption in only 20 years, whereas the average repose period since 1104 is 55 years. Earthquakes, as well as a strain pulse recorded by borehole strainmeters, occurred less than half an hour before the start of the eruption. The initial plinian phase was very short-lived, producing a total of only 0.02 km3 of tephra. The eruption cloud attained 11.5 km in height in only 10 min, but it became detached from the volcano a few hours later. Several fissures were active during the first day of the eruption, including a part of the summit fissure. By the second day, however, the activity was already essentially limited to that segment of the principal fissure where the main crater subsequently formed. The average effusion rate during the first two days of the eruption was about 800 m3 s−1. After this peak, the effusion rate declined rapidly to 10–20 m3 s−1, then more slowly to 1 m3 s−1, and remained at 1–12 m3 s−1 until the end of the eruption. Site observations near the main crater suggest that the intensity of the volcanic tremor varied directly with the force of the eruption. A notable rise in the fluorine concentration of riverwater in the vicinity of the eruptive fissures occurred on the 5th day of the eruption, but it levelled off on the 6th day and then remained essentially constant. The volume and initial silica content of the lava and tephra, the explosivity and effusion rate during the earliest stage of the eruption, as well as the magnitude attained by the associated earthquakes, support earlier suggestions that these parameters are positively related to the length of the preceeding repose period. The chemical difference between the eruptive material of Hekla itself and the lavas erupted in its vicinity can be explained in terms of a density-stratified magma reservoir located at the bottom of the crust. We propose that the shape of this reservoir, its location at the west margin of a propagating rift, and its association with a crustal weakness, all contribute to the high eruption frequency of Hekla.