Using a modified version of Chicken, referred to as the Threat Game, Brams and Kilgour have developed a theoretical model of threat escalation and stability in superpower crises. More specifically, they derive explicit thresholds for sufficient pre-emption to stabilize a crisis before it escalates out of control. An appropriate degree of coercion thereby is designated for the response to an adversary, in order to deter further escalation. The overall objective of this study is to use data on superpower crises to test a modified version of the model developed by Brams and Kilgour. Mixed results emerge from testing the most general propositions. The choice of response level usually does not surpass the model's recommended value. However, stabilization is feasible in virtually all crises and, when an actor does meet the threshold, satisfaction with the outcome occurs more frequently than otherwise. Two more specific hypotheses also are tested, with partial support for the model once again being the result.