In the nuclear age, and more especially since the Cuban missile crisis of 1962, international crises have become a major focus of study. The reasons are evident: crises between nuclear powers confront whole societies with the spectre of their sudden destruction and require decision makers to contemplate choices of intolerable consequence, sometimes under extreme time pressure. From an earlier era, the example of Sarajevo stands as an awesome warning of the way in which governments can lose control over events. Analysts have examined many crises, drawing on a wide range of theoretical approaches, and there have been some outstanding works. However, the literature on crises shows serious limitations. Many studies are narrowly based, being restricted to the decisions of a single actor; relatively few crises have been studied in depth; and persisting differences over theoretical assumptions stand in the way of agreed findings. Above all, there is no accepted answer to the question: under what conditions do crises lead to war, and when are they resolved peacefully? Since the diversity of international conflicts may preclude any general answer, however, the starting point for the present inquiry is a reformulation of the question: to what extent does the outcome depend on the gravity of the underlying conflict – on what may be termed ‘structural’ factors – and to what extent on crisis diplomacy – on decisions and interactions during the crisis itself?