The distribution of instrumental and historical seismicity in Wales is described, including two events in 1984 for which good hypocentral locations were determined. Most of central Wales is aseismic throughout the historic record, suggesting that this is not a feature of reporting bias or population distribution alone. There appears to be a relation between the epicentres and some of the major fault structures of Wales, including the Bala Fault and the Swansea Valley Disturbance. The relationship between faults and seismicity is examined in detail for the two recent events, and more generally for the rest of the data, which have much lower locational accuracy. Stress measurements for NW Europe are shown to give a consistent NW-SE direction for the maximum horizontal stress. This is appropriate for sinistral strike-slip on N-S vertical faults. First motion data for both 1984 events also give a nodal plane in this orientation. In the first case, the event can be clearly attributed to mapped N-S vertical faults SE of Newtown, and in the second, a deep N-S normal fault under the Lleyn Peninsula is inferred as the most likely fault plane orientation from regional geological considerations. The proposed activity of the Bala Fault and the Swansea Valley Disturbance can also be explained within this stress field by their thrust reactivation as low angle structures. The general difficulty of relating the remaining seismicity to geological structures is demonstrated by the fact that the second 1984 event could not be directly connected with any surface structure, or even with known structures at depth.