In 1999, an unusual earthquake swarm at the 85°E/85°N volcanic centre on the ultraslow spreading Gakkel Ridge, Arctic Ocean, was detected teleseismically. The swarm lasted over 9 months and counted 252 events with mb ≥ 3.1. It represents the strongest and largest ever recorded mid-ocean ridge earthquake swarm, and it occurred at a site where spreading rates are only about 10 mm yr–1. We relocated the earthquake swarm comparing the performance of three different localization algorithms: (1) the absolute least-squares routine HYPOSAT, (2) the absolute probabilistic routine NonLinLoc and (3) the relative least-squares routine Mlocate. The epicentres as calculated by each algorithm mostly did not agree within their error ellipses. Thus, the choice of location algorithm proved more critical than, for example, the choice of a local velocity model.We compiled a set of well-localized events which closely agree in at least two routines, mostly Mlocate and NonLinLoc. We conclude that the earthquake swarm of 1999 was related to a spreading episode and shows a complex interplay of tectonic and magmatic events. Our geological interpretation revealed three phases in swarm activity: In the first phase from January 17 up to February 1 fracturing of the crust took place, either as a result of or enabling magmatic intrusion. Seismicity in the second phase from February 2 to April 6 expanded along- and across axis. It showed signs of magmatic interaction, but a clear dyke migration pattern is absent. At the beginning of the third phase, a distinct break in the event rate suggested a change in the physical process, either an adjustment of the stress field to the new regime or a transition to an effusive stage.