At 00:02 GMT (03:02 local time) on 17 August, 1999 a magnitude 7.4 (M-s) earthquake occurred 100 km east of Istanbul causing extensive destruction. The event was expected and several scientists have published and attempted to publicize the danger. A paper on stress interactions for NW Turkey (J. Geophys. Res. 103 (1998) 24466-24469) concluded that "by combining the stress change map with the map of active faulting, likely locations for the occurrence of future earthquakes can be refined; faults in the Izmit Bay area, the western part of Biga Peninsula, the Saroz Gulf and a part of western Sea of Marmara must be regarded as posing a specific hazard". An extension of that study is described here. It is shown that the Izmit (1999) earthquake loaded faults both to the east and west of the Izmit rupture. About three months after the lzmit event an M 7.2 earthquake occurred with an epicenter at Duzce extending the lzmit rupture to the east. In the Marmara Sea, west of Izmit, faults have been loaded by between 1 and 5 bar; 5 to 30 % of typical earthquake stress drops in the region suggesting the likelihood of a future event. The risk of a major event on a fault depends not just on stress increases associated with an individual earthquake, but also on the longer-term earthquake history and on tectonic loading. The roles of both are examined over two time periods from 1900 to 1999 and 1700 to 1999. Whatever interpretation we place on the data we conclude that one or two events as great or greater than the recent one is likely to occur within the next few decades near to the northern coast of the Marmara Sea. (C) 2001 Academie des sciences / Editions scientifiques et medicales Elsevier SAS.