Written by SOURCE: The Journal of Safety Research
In this 2026 study, Bills, Costello, and Cattani present a rare inside look at how 12 leading government Safety Investigation Authorities (SIAs) actually analyse major aviation accidents.
Drawing on written submissions and interviews with senior investigators from agencies including the ATSB, NTSB, BEA, AAIB and TSB, the study maps the investigation models and analytical frameworks used in practice by investigators and compares them with both ICAO guidance and contemporary academic debate.
The researchers collected structured responses to four research questions and conducted semi-structured interviews, which were thematically analysed. This allowed them to examine not only which methodologies were used, but why they were chosen, how they were combined, and what investigators saw as best practice .
The findings show that no SIA relies on a single accident model. Instead, all use multiple complementary approaches. Many draw on Rasmussen’s AcciMap (a systems-based model that maps contributory factors across different levels of a socio-technical system), Reason-based approaches such as the Swiss Cheese Model, Tripod Beta or HFACS (which classify latent and active failures), and barrier analysis (which examines the effectiveness of safety defences). Several use Leveson’s CAST (a systems-theoretic method that analyses safety control structures and feedback loops), Hollnagel’s FRAM (which models how everyday performance variability can resonate to produce unexpected outcomes), BowTie analysis (which visualises threats, top events and consequences with associated barriers), or bespoke in-house systems such as the ATSB’s AIMS or the TSB’s ISIM .
A key finding is that methodology choice is typically driven by accident complexity, scope, available resources, and investigator expertise. Some agencies adopt a structured organisation-wide framework; others allow flexibility at investigator-in-charge level. Across agencies, there is strong emphasis on systemic, non-blaming approaches that move beyond simple “root cause” thinking. Investigators value methods that help structure reasoning, support hypothesis testing, visualise system interactions, and translate analysis into defensible safety recommendations.
The study also highlights practical constraints. Highly detailed systemic methods such as CAST can be resource-intensive and require specialist training. Agencies therefore favour scalable, adaptable “toolbox” approaches that balance rigour with timeliness and usability. Importantly, the research challenges claims in some academic “New View” literature that accident investigation lacks systemic sophistication. In practice, SIAs are using a wide range of contemporary, system-oriented models—often in combination—and adapting them pragmatically to operational realities .
Overall, the paper demonstrates that modern aviation accident investigation is methodologically pluralistic, system-focused, and more advanced than many assume. It provides valuable insight into how theory is operationalised in real-world safety investigation and has implications for ICAO guidance, cross-agency learning, and other high-risk industries seeking effective approaches to analysing complex socio-technical failures.