The present study sought to describe police officers' decision processes in simulated counter-terrorism events. Based on previous phase models of decision making and existing police policy in dealing with critical incidents of this nature, a descriptive SAFE-T (Situation Assessment, Formulate a plan, Execute a plan and Team learning) model was applied to team decision processes. Proximity and Lag Sequential analyses tested the occurrence of the predicted decision phases set out in the model; these results indicated that the model did not fully capture the complexity of the process. Specifically, further qualitative analyses (and comparisons to a 'Gold Standard' set of subject matter expert decisions, conducted with the benefit of removing ambient and time pressure stressors) illustrated that a complex combination of ambient (uncertainty), cognitive (accountability pressures) and organizational (an existing blame culture and lack of policy) factors 'derailed' officers from making 'save life' decisions. Instead, albeit in a minority of cases, they either made errors of omission by failing to make any decision at all or inappropriate choice deferrals (by insisting another agency made the decision or that the decision could be made later). The potential benefits of making the SAFE-T decision process and how derailment can occur explicit in police critical incident training events are discussed.