When frontline staff do not adhere to policies, protocols, or checklists, managers often regard these violations as indicating poor practice or even negligence. More often than not, however, these policy and protocol violations reflect the efforts of well intentioned professionals to carry out their work efficiently in the face of systems poorly designed to meet the diverse demands of patient care. Thus, non-compliance with institutional policies and protocols often signals a systems problem, rather than a people problem, and can be influenced among other things by training, competing goals, context, process, location, case complexity, individual beliefs, the direct or indirect influence of others, job pressure, flexibility, rule definition, and clinician-centred design. Three candidates are considered for developing a model of safety behaviour and decision making. The dynamic safety model helps to understand the relationship between systems designs and human performance. The theory of planned behaviour suggests that intention is a function of attitudes, social norms and perceived behavioural control. The naturalistic decision making paradigm posits that decisions are based on a wider view of multiple patients, expertise, systems complexity, behavioural intention, individual beliefs and current understanding of the system. Understanding and predicting behavioural safety decisions could help us to encourage compliance to current processes and to design better interventions.