This debate has highlighted both a degree of consensus and a degree of continuing controversy within the thorny subject of the predictability of earthquakes. In terms of the four levels of prediction of seismicity I introduced at the start of this debate, a consensus has emerged that at least some form of time-dependent seismic hazard can be justified on both physical and observational grounds. The phenomenon of earthquake triggering leads to a transient, local increase in probability of future earthquakes, for example as aftershocks, but also sometimes in the form of subsequently larger events. In fact warnings based on such clustering are already in use in California (Michael, week 2). On the other hand, all of the contributors to this debate who expressed an opinion agree that the deterministic prediction of an individual earthquake, within sufficiently narrow limits to allow a planned evacuation programme, is an unrealistic goal.