The 2003 Bingol earthquake (M-w=6.4) occurred very close to a region along the east Anatolian fault zone which was identified in 2002 as posing a particularly high seismic risk. This damaging earthquake occurred on a conjugate right-lateral blind fault that was inconsistent with the stress-change field calculated for preceding large earthquakes in the region. In this paper, four reasons which might be responsible for this stress discrepancy are identified and investigated individually. Firstly, co-seismic stress changes are considered. The time frame of the previous stress calculations is extended to include the large earthquakes in the 1780s which were not included in the earlier study. A sensitivity analysis is then conducted on the more recent events since 1822 to examine the effect of errors in their location and sizes. The possibility of the occurrence of a small magnitude event close to the 2003 epicentre is considered. We argue that, barring the occurrence of a low-probability, unmodelled local event, the Bingol earthquake was unlikely to have been triggered by co-seismic stress transfer from any known sequence of previous earthquakes. Finally we examine and modify the secular loading model used in the 2002 study and show that loading which is properly constrained by regional GPS data produces a positive stress change on the 2003 rupture. As a result of our examination of the stressing history of the Bingol hypocentre we argue that it is through a combination of historical seismology, guided and constrained by structural geology, directed paleoseismology in which the locations and extent of historical events are confirmed, and stress modelling which has been informed by detailed GPS data, that an integrated seismic hazard program might have the best chance of success. (c) 2005 Elsevier B.V. All rights reserved.