The standard chicken game is a popular model of certain important real scenarios but does not allow for the escalation behaviour these are typically associated with. This is problematic if the critical, final decisions in these scenarios are sensitive to previous escalation. We introduce and analyse, theoretically and by experiment, a new game which permits escalation behaviour. Compared with an equivalent chicken game, Pareto-suboptimal outcomes are significantly more frequent. This result is inconsistent with our rational choice analysis and possible psychological roots are explored.