Safety is monitored by various proactive and reactive methods, including the investigation of adverse accidents and incidents, which are collectively known as safety investigations. In this study we demonstrate how accident and incident investigation reports can be useful to identify implicit safety views and accident investigation approaches. An analysis framework was developed based on contemporary safety literature. The framework incorporates aspects such as hindsight bias, judgemental approach, proximal or distal focus, and the application of systemic versus sequential accident causation models. The framework was piloted through the analysis of sixteen (16) accident investigation reports published by a Nuclear Power Plant (NPP). The comments of independent researchers lead to framework refinements that increased the inter-rater reliability substantially. The initial results were validated through interviews with the staff of the NPP. Afterwards, the framework was applied to 52 air accident reports published by the Dutch Safety Board (DSB) from 1999 to 2013. Frequency calculations revealed the extent of new safety thinking embracement from the DSB, and Fisher’s Exact Test showed that none of the modern safety aspects has changed over time. The framework can be used to analyse accident investigation reports published by various organisations as means to identify implicit safety views and evolution of accident investigation practices over time. Further research will explore the reasons for potential gaps between theory and practice and contribute to minimizing such distance. Safety is monitored by various proactive and reactive methods, including the investigation of adverse accidents and incidents, which are collectively known as safety investigations. In this study we demonstrate how accident and incident investigation reports can be useful to identify implicit safety views and accident investigation approaches. An analysis framework was developed based on contemporary safety literature. The framework incorporates aspects such as hindsight bias, judgemental approach, proximal or distal focus, and the application of systemic versus sequential accident causation models. The framework was piloted through the analysis of sixteen (16) accident investigation reports published by a Nuclear Power Plant (NPP). The comments of independent researchers lead to framework refinements that increased the inter-rater reliability substantially. The initial results were validated through interviews with the staff of the NPP. Afterwards, the framework was applied to 52 air accident reports published by the Dutch Safety Board (DSB) from 1999 to 2013. Frequency calculations revealed the extent of new safety thinking embracement from the DSB, and Fisher’s Exact Test showed that none of the modern safety aspects has changed over time. The framework can be used to analyse accident investigation reports published by various organisations as means to identify implicit safety views and evolution of accident investigation practices over time. Further research will explore the reasons for potential gaps between theory and practice and contribute to minimizing such distance.