Using detailed microearthquake data, we present a stress tensor inversion scheme with new methods for selecting the fault planes and allowing for errors in the focal mechanisms. The nonuniqueness of earthquake focal mechanisms is accounted for in our inversion scheme through the introduction into the inversion of a range of well-fitting focal mechanisms for each event. The range of focal mechanisms significantly improves the quality of the estimated stress tensor. Relative localization of clusters of microearthquakes is used to obtain information about which nodal plane could be correct fault plane. The clusters frequently fall on a common fault plane, and if there are acceptable focal mechanisms where one nodal plane has orientation similar to the common plane, we assume this is the correct fault plane for the event. If there is no predefined fault plane, we utilize a simple Mohr-Coloumb failure criterion to obtain a physical choice of fault plane between the two nodal planes in the focal mechanism. The nodal plane with highest relative instability is chosen as the fault plane. Differences between the instability and the standard slip angle criterion are investigated. The new inversion scheme has been applied to microearthquake data from the ¿lfus area in the vicinity of the southwest Iceland triple junction. We estimate an oblique strike-slip state of stress, maximum horizontal stress at N30 ¿E, and minimum horizontal stress at N60 ¿W, with significant normal faulting influence. The instability fault selection criterion predicts very well the orientation of faults mapped by relative localization. ¿ 1999 American Geophysical Union